Browse the corpus

X 2 % periods (n = 295; Supplementary Table S2) for the two ICUs. During the baseline period CHX 0.20 and 0.12 % were used for oral care in hospitals A and B, respectively, without evidence of oral lesions in any patient. All other procedures related to oral care remained identical in each hospital during both periods. Amongst the CHX 2 % treated patients, occurrence of side effects was associated with male gender, APACHE II score, length of stay in the ICU, and duration of mechanical ventilation, suggesting a dose-response relationship, with increasing risks of oral mucosal lesions for the more severely ill patients, undergoing mechanical ventilation, and receiving CHX 2 % for longer periods (Table 1). This hypothesis is supported by the localization of the lesions in the oral cavity; most lesions occurred where stasis of the mouthwash might have occurred—despite suctioning after administration—such as below the tongue and in the buccal pockets.Table 1 Baseline characteristics of CHX 2 % treated patients with and without adverse events

Introduction Microdialysis is unique in that it allows the chemistry of the extracellular interstitial fluid to be monitored continuously. Since its conception by Ungerstedt and Pycock in the 1970s [1] and its introduction into clinical practice approximately 25 years ago [2], it has been applied to study the tissue chemistry of several human organs. Most experience has been acquired in the setting of neurocritical care. In this arena, microdialysis has been applied to patients with several conditions, and in particular traumatic brain injury (TBI) and subarachnoid hemorrhage (SAH). There is no doubt that this technique has increased our understanding of the pathophysiology of these disease processes [3]. Furthermore, microdialysis has evolved into a clinical tool for the management of patients on an individual intention-to-treat basis. In neurocritical care, microdialysis data is typically collected together with intracranial pressure (ICP) [allowing calculation of cerebral perfusion pressure (CPP)] and brain tissue oxygen tension (PbtO2). Microdialysis complements these techniques by providing additional information on substrate delivery and metabolism at the cellular level. It thus provides the most direct means to monitor the fundamental process of “energy failure”. Of critical importance is that such measurements can be made in real time at the bedside.

rodialysis complements these techniques by providing additional information on substrate delivery and metabolism at the cellular level. It thus provides the most direct means to monitor the fundamental process of “energy failure”. Of critical importance is that such measurements can be made in real time at the bedside. In 2003, a group of experts met to review the status of microdialysis as a clinical monitor. This culminated in the publication of a consensus statement in 2004 [4] providing guidance on the use of the technique in TBI and SAH patients. More recently, the role of microdialysis has been evaluated by participants of the International Multidisciplinary Consensus Conference on Multimodality Monitoring [5].

onitor. This culminated in the publication of a consensus statement in 2004 [4] providing guidance on the use of the technique in TBI and SAH patients. More recently, the role of microdialysis has been evaluated by participants of the International Multidisciplinary Consensus Conference on Multimodality Monitoring [5]. In April 2014, an international forum was convened in Cambridge, UK, with the aim of reviewing evidence for the clinical application of microdialysis in neurocritical care and producing a revised and updated consensus statement [4]. Since the original consensus statement, ~680 articles have been published on microdialysis in neurocritical care. With this increased experience, there was a need to update the 2004 consensus statement. Although there was some overlap between the objectives of this meeting and that of the International Multidisciplinary Consensus Conference on Multimodality Monitoring, i.e. to review the evidence for using microdialysis to guide clinical care, the principal objective of the International Forum in Microdialysis differed in that we aimed to combine literature review with expert opinion to produce practical guidance for the use of cerebral microdialysis as a clinical monitor and to help guide future clinical studies utilizing cerebral microdialysis.

nical care, the principal objective of the International Forum in Microdialysis differed in that we aimed to combine literature review with expert opinion to produce practical guidance for the use of cerebral microdialysis as a clinical monitor and to help guide future clinical studies utilizing cerebral microdialysis. Methods The senior authors selected specific ‘key speakers’ to review a particular area of the literature. These individuals were selected based on their experience and contribution to the literature on a particular aspect of microdialysis monitoring. See Appendix 1 in the supplementary material for a list of key speakers and for the topics they each reviewed. The other participants of the meeting were identified through literature review and by correspondence with the key speakers who were able to identify other clinicians and scientists active in using microdialysis in neurocritical care patients. At the meeting, the literature was presented to the whole group followed by discussion to allow consensus generation. After the meeting, the recommendations were circulated to all participants allowing further discussion and revision.

inicians and scientists active in using microdialysis in neurocritical care patients. At the meeting, the literature was presented to the whole group followed by discussion to allow consensus generation. After the meeting, the recommendations were circulated to all participants allowing further discussion and revision. In addition, for the purposes of the consensus statement, we performed a PubMed database search using the term microdialysis plus one of the following terms: ‘traumatic brain injury’, ‘brain injury’, ‘trauma’, ‘subarachnoid hemorrhage’, ‘stroke’, ‘epilepsy’, ‘intracerebral hematoma’ and ‘cost effectiveness’. We restricted our review to using articles published in the English language. Where recommendations are based on published observational data, the relevant references are given although formal grading was not performed. Where references are not provided, the recommendations are based on expert opinion. Discussion Advances since the 2004 consensus statement Over the past 10 years, there have been significant advances in the clinical utility of microdialysis in neurocritical care. Evidence from large numbers of patients on how brain chemistry relates to clinical outcome means that we can better define pathological thresholds for microdialysis values. In addition, there is increasing evidence of how different therapeutic manoeuvres can improve chemistry. For a summary of the main advances since the 2004 consensus statement, please see Table 1.Table 1 Summary of advances since the consensus statement by [4]

ter define pathological thresholds for microdialysis values. In addition, there is increasing evidence of how different therapeutic manoeuvres can improve chemistry. For a summary of the main advances since the 2004 consensus statement, please see Table 1.Table 1 Summary of advances since the consensus statement by [4] 2004 consensus statement [4] Current consensus statement Microdialysis methodology Monitoring of small molecules using standard 10-mm 20-kDa catheter Advances in monitoring of large molecules, with experience of using 100-kDa membrane and colloid for perfusate [13–20] Focus on microdialysis metabolites as a marker of ischemia and cell damage Novel applications of microdialysis for monitoring and understanding brain pathology following TBI and SAH Core data reporting information Not defined Details are given of the essential information required to interpret and compare microdialysis data Reference values Not defined Pathological thresholds defined for glucose, lactate and the LP ratio [6, 51, 53–55, 68–73, 79, 80] Tiered approach to microdialysis metabolites for clinical application Not defined Glucose and LP ratio more clinically useful than glutamate and glycerol in TBI and SAH patients Guidance for microdialysis-directed management Not given Suggested therapeutic interventions for when glucose is low (<0.2 mM) and for when the LP ratio indicates ischemia ± tissue hypoxia Monitoring in TBI Guidance on catheter placement in focal or diffuse injury Guidance on single or multiple catheter placement based on whether the injury is focal or diffuse and based on the aims of microdialysis monitoring Monitoring in SAH Guidance on catheter placement in the tissue at risk Two principal indications for microdialysis monitoring are defined: 1. As a primary monitoring device in mechanically ventilated patients 2. As a monitor of patients with a secondary neurological deterioration

on the aims of microdialysis monitoring Monitoring in SAH Guidance on catheter placement in the tissue at risk Two principal indications for microdialysis monitoring are defined: 1. As a primary monitoring device in mechanically ventilated patients 2. As a monitor of patients with a secondary neurological deterioration Most attention has been directed at the clinical utility to monitor TBI and SAH patients: see Table 2 for a summary of how brain chemistry relates to different aspects of the care of patients with TBI and SAH. Microdialysis has also been used in other neurological conditions including intracerebral hemorrhage [6], acute ischemic stroke [7–9], hepatic encephalopathy [10] and epilepsy [11, 12]. However, there is insufficient evidence at present to specifically incorporate the application of microdialysis in these conditions into the consensus statement.Table 2 Summary of the evidence for how brain chemistry relates to different aspects of the management of patients with TBI and SAH

[10] and epilepsy [11, 12]. However, there is insufficient evidence at present to specifically incorporate the application of microdialysis in these conditions into the consensus statement.Table 2 Summary of the evidence for how brain chemistry relates to different aspects of the management of patients with TBI and SAH How microdialysis monitoring can be used in neurocritical care Traumatic brain injury Subarachnoid hemorrhage Outcome and prognostication [51, 53, 78] [67, 79, 81] Early warning system of secondary insults [26, 27] [28, 29, 80, 82] Monitoring and treatment of low cerebral glucose; guiding systemic glucose management and insulin use [56, 61, 62, 64, 65] [56, 63, 83, 84] Monitoring during CPP-augmentation/reduction [48, 85, 86] [54, 87] Monitoring during neurological wake-up test (tolerating moderate rises in ICP) [25, 88] Deciding on transfusion thresholds [89] Evaluating the effect of body temperature on cerebral chemistry [90] [91] Monitoring after decompressive craniectomy [92] [93] In addition to recent advances as a clinical monitor, microdialysis continues to be a powerful research tool with numerous, varied and several novel applications that provide insight into various aspects of cerebral biology and pathophysiology. For a summary of on-going and future research, see Table 3. Overall, further research should be directed at the integration of brain chemistry and other clinical monitoring data to better define targets for the individualized goal-directed management of the brain-injured patient.Table 3 A summary of on-going microdialysis research applications

and future research, see Table 3. Overall, further research should be directed at the integration of brain chemistry and other clinical monitoring data to better define targets for the individualized goal-directed management of the brain-injured patient.Table 3 A summary of on-going microdialysis research applications Investigating the concept of lactate as a substrate as opposed to a metabolic by-product in select patients [79, 94] Use of 100-kDa microdialysis membranes to measure larger molecules including cytokines [15, 16, 18, 19, 95] Use of 13C-labelled substrates to interrogate metabolic pathways in more detail, e.g., the fate of glucose metabolism (glycolysis vs. pentose phosphate pathway) and the fate of lactate as a substrate [94, 96, 97] Monitoring drug penetration across the blood–brain barrier and the effect of drugs on brain chemistry [98, 99] Clinical use in pediatric practice [100–102] Monitoring of the ionic component of the interstitial space [103] Monitoring of biomarkers [18, 19, 104–111] Development of microfluidic based on-line assays that give continuous neurochemical information in real time [23, 24, 112] Advances in microdialysis methodology The technique of microdialysis is well established. For details on technique and on the factors that affect relative recovery, i.e. how the substance measured in the dialysate is related to the free concentration in the tissue interstitial space, please see supplementary material.

Investigating the concept of lactate as a substrate as opposed to a metabolic by-product in select patients [79, 94] Use of 100-kDa microdialysis membranes to measure larger molecules including cytokines [15, 16, 18, 19, 95] Use of 13C-labelled substrates to interrogate metabolic pathways in more detail, e.g., the fate of glucose metabolism (glycolysis vs. pentose phosphate pathway) and the fate of lactate as a substrate [94, 96, 97] Monitoring drug penetration across the blood–brain barrier and the effect of drugs on brain chemistry [98, 99] Clinical use in pediatric practice [100–102] Monitoring of the ionic component of the interstitial space [103] Monitoring of biomarkers [18, 19, 104–111] Development of microfluidic based on-line assays that give continuous neurochemical information in real time [23, 24, 112] Advances in microdialysis methodology The technique of microdialysis is well established. For details on technique and on the factors that affect relative recovery, i.e. how the substance measured in the dialysate is related to the free concentration in the tissue interstitial space, please see supplementary material. Microdialysis is used clinically to estimate extracellular interstitial concentrations of small molecules, but can also be used to recover much larger molecules such as inflammatory mediators from the interstitial fluid. Instead of the standard 20-kDa nominal molecular weight cut-off membrane, which recovers glucose, pyruvate, lactate, glycerol, glutamate, and other small hydrophilic molecules, a 100-kDa membrane is used to also recover larger molecules including cytokines. The recovery of small molecules does not differ between the two membrane types [13]. Increased experience in using microdialysis for large molecules less than 100 kDa has been achieved in the past 10 years. Importantly, the use of colloid in the perfusate (e.g., albumin or dextran) significantly improves the relative recovery of these large molecules [14–16]. However, in some situations, colloid perfusate can cause net influx of fluid into the catheter potentially dehydrating the interstitial space, and dextrans of molecular weight 40–250 kDa may leak through the microdialysis membrane potentially disturbing the interstitial microenvironment [14, 15, 17]. These problems may be overcome by using higher molecular weight dextrans, such as 500-kDa dextran, as colloid in the perfusate [18–20]. A useful alternative colloid to dextran is human serum albumin (HAS), which has been shown to improve recovery for the majority of cytokines compared to crystalloid perfusate without significantly dehydrating the interstitial space [16].

lar weight dextrans, such as 500-kDa dextran, as colloid in the perfusate [18–20]. A useful alternative colloid to dextran is human serum albumin (HAS), which has been shown to improve recovery for the majority of cytokines compared to crystalloid perfusate without significantly dehydrating the interstitial space [16]. Most experience of microdialysis in neurocritical care has been obtained with hourly measurements although more frequent sampling is possible [21–25]. Hourly sampling appears sufficient to detect the metabolic changes that can sometimes precede episodes of intracranial hypertension in TBI and symptomatic delayed ischemia in SAH [26–29]. Hence, microdialysis has the potential to be used as an early warning system of secondary insults. However, dynamic changes in brain chemistry, for example due to spreading depolarization [21–23] or observed during aneurysm surgery [24, 30], may not be detected with hourly measurements, so there is potentially scope for improved technology with more frequent microdialysis readings in future, which may lead to better warning of adverse events.

brain chemistry, for example due to spreading depolarization [21–23] or observed during aneurysm surgery [24, 30], may not be detected with hourly measurements, so there is potentially scope for improved technology with more frequent microdialysis readings in future, which may lead to better warning of adverse events. Clinical application in intensive care The clinical application of microdialysis in neurocritical care has focused on the delivery of glucose and its metabolism via glycolysis to pyruvate, which under oxidative conditions feeds into the tricarboxylic acid (TCA) cycle. Under hypoxic conditions, or if mitochondrial function is compromised, pyruvate is metabolized to lactate. Hence, the LP ratio is used as a marker of aerobic versus “anaerobic” metabolism not requiring oxygen [31, 32]. Glutamate is measured as a marker of hypoxia/ischemia and has been considered as an indicator of excitotoxicity [31–34]. Glycerol is regarded a marker of hypoxia/ischemia and cell membrane breakdown [32, 35–37].

ce, the LP ratio is used as a marker of aerobic versus “anaerobic” metabolism not requiring oxygen [31, 32]. Glutamate is measured as a marker of hypoxia/ischemia and has been considered as an indicator of excitotoxicity [31–34]. Glycerol is regarded a marker of hypoxia/ischemia and cell membrane breakdown [32, 35–37]. Safety profile The technique of cerebral microdialysis is safe. Several published series of patients studied with microdialysis, which include non-brain-injured patients, have not reported adverse events related to microdialysis catheter insertion [29, 38–40]. Cerebral microdialysis has a safety profile at least equivalent to that of intra-parenchymal pressure sensors owing to the catheter’s greater flexibility and small diameter [41]. In most circumstances when an adverse event occurs, it relates to the insertion technique rather than the catheter itself. Cerebral microdialysis has mostly been used as a tool for observational studies. Further evaluation of microdialysis as a clinical monitor should include assessment of potential harm caused by microdialysis-directed interventions.

erse event occurs, it relates to the insertion technique rather than the catheter itself. Cerebral microdialysis has mostly been used as a tool for observational studies. Further evaluation of microdialysis as a clinical monitor should include assessment of potential harm caused by microdialysis-directed interventions. Cost-benefit analysis No cost effectiveness studies evaluating microdialysis in neurocritical care have been performed. One study compared ICP monitoring alone against multimodal monitoring, which consisted of transcranial Doppler, jugular venous oxygen saturation and/or PbtO2 monitoring but not microdialysis [42]. Albeit a small study, it demonstrated that increased upfront costs due to consumables and equipment was offset by better clinical outcomes, which meant that multimodal monitoring was cost effective. In TBI patients, there are indications that aggressive management, which includes invasive monitoring, improves outcomes and is cost effective [43–46]. However, these studies have not examined microdialysis monitoring per se. Recommendations from the 2014 International Forum on Microdialysis––the 2014 consensus statement Methodology Catheters should be inserted according to local institutional protocols either by twist drill hole, transcranial bolt, or at craniotomy. The first hour of microdialysate collected should not be used for clinical monitoring due to unreliable results caused by insertion trauma and the pump flush sequence.

Recommendations from the 2014 International Forum on Microdialysis––the 2014 consensus statement Methodology Catheters should be inserted according to local institutional protocols either by twist drill hole, transcranial bolt, or at craniotomy. The first hour of microdialysate collected should not be used for clinical monitoring due to unreliable results caused by insertion trauma and the pump flush sequence. To monitor glucose, pyruvate, lactate, glycerol and glutamate catheters with a 20- or 100-kDa cut-off are available (100-kDa catheters are not yet FDA-approved, although they are CE marked for use in Europe). A flow rate of 0.3 μL/min with hourly sampling is recommended, which is the flow rate most commonly used in the cerebral microdialysis literature. Publication of microdialysis data should include the following information (core data reporting):catheter type catheter location based on post-insertion imaging flow rate membrane length perfusion fluid composition time from ictus to monitoring Interpretation of cerebral microdialysis Microdialysis monitors substrate delivery and metabolism at the cellular level. Chemistry should be interpreted in the context of the clinical condition of the patient and in conjunction with other monitored parameters including ICP, CPP, PbtO2, cerebrovascular pressure reactivity (PRx) and systemic parameters, in order to determine the likely cause of perturbed metabolism. For example, a rise in LP ratio associated with a fall in CPP and loss of cerebrovascular reactivity (i.e., a high PRx) indicates that the likely cause of disordered chemistry is ischemia.

cerebrovascular pressure reactivity (PRx) and systemic parameters, in order to determine the likely cause of perturbed metabolism. For example, a rise in LP ratio associated with a fall in CPP and loss of cerebrovascular reactivity (i.e., a high PRx) indicates that the likely cause of disordered chemistry is ischemia. Microdialysis is a focal technique. The heterogeneity of brain injury means that brain chemistry varies in different regions of the brain. In TBI, peri-lesional brain demonstrates more perturbed chemistry, in particular a higher LP ratio, compared to other areas of brain [47–52]. Therefore, brain chemistry should be interpreted according to catheter location in relation to focal injury based on CT/MRI imaging. Glucose Glucose is the main substrate for brain metabolism. Periods of low glucose (<0.8 mM) are observed in TBI and SAH. Low brain glucose is associated with unfavorable outcome [53–57]. There is also evidence that high brain glucose is associated with unfavorable outcome indicating that there is an optimal range for brain glucose, although, there is currently insufficient data to define this range [51, 58]. Serum glucose concentration and glycemic control influence brain glucose although this relationship may be lost in injured brain [56, 59–65]. Brain glucose can be reduced rapidly by secondary insults such as spreading depolarization [22, 66]. Lactate/pyruvate ratio A high LP ratio is associated with unfavorable outcome [6, 51, 53, 54, 57, 67–73]. The LP ratio is a marker of cellular redox status. The LP ratio is a quantitative measure (independent of relative recovery).

Brain glucose can be reduced rapidly by secondary insults such as spreading depolarization [22, 66]. Lactate/pyruvate ratio A high LP ratio is associated with unfavorable outcome [6, 51, 53, 54, 57, 67–73]. The LP ratio is a marker of cellular redox status. The LP ratio is a quantitative measure (independent of relative recovery). An increased LP ratio may result from a failure of oxygen delivery (ischemic hypoxia) or from non-ischemic causes (e.g., mitochondrial dysfunction) [74, 75]. The absolute lactate and pyruvate concentrations should be considered when interpreting a high LP ratio. Ischemia and mitochondrial dysfunction are two ends of a spectrum of factors that increase the LP ratio. An increase in the LP ratio in the presence of low pyruvate (and low oxygen) indicates ischemia. An increase in LP ratio in the presence of normal or high pyruvate (and normal oxygen) indicates mitochondrial dysfunction. Glutamate Glutamate is an excitatory amino acid and neurotransmitter. Excess levels are thought to be an additional injurious mechanism and may exacerbate injury in TBI and SAH. Excess glutamate release is observed in ischemia [8, 31, 33, 76] and seizures [11, 12, 76, 77]. There is a described association between glutamate levels, clinical course and outcome in TBI and SAH [29, 78]. Measuring cerebral glutamate is an option and may be useful in estimating prognosis. Glycerol Glycerol is a marker of cell membrane breakdown. It is a potential marker of oxidative stress.

Excess glutamate release is observed in ischemia [8, 31, 33, 76] and seizures [11, 12, 76, 77]. There is a described association between glutamate levels, clinical course and outcome in TBI and SAH [29, 78]. Measuring cerebral glutamate is an option and may be useful in estimating prognosis. Glycerol Glycerol is a marker of cell membrane breakdown. It is a potential marker of oxidative stress. Glycerol has limited specificity; brain glycerol concentrations are influenced by systemic concentrations. Systemic glycerol concentrations reflect a stress response and/or administration of glycerol-containing substances. There is no definitive evidence of a relationship between glycerol and outcome. Cerebral glycerol is an option as a marker of cerebral injury. Guidance for use of microdialysis in traumatic brain injury and subarachnoid hemorrhage—catheter location, reference values and interventions Traumatic brain injury In diffuse TBI, we recommend placing the catheter in the right (non-dominant) frontal lobe. In focal TBI, there are different options for catheter placement that depend on whether the goal is to monitor tissue at risk or normal brain, e.g., to guide systemic glucose treatment. Where there is a focal lesion, we recommend, if feasible, catheter placement ipsilateral to the lesion and in radiographically normal brain. Multiple catheters are an option in focal TBI.E.g., placed at craniotomy for a focal lesion into peri-lesional brain with a contralateral ‘bolt’ catheter in radiographically normal brain. Stereotactic placement is an option but rarely practical.

Where there is a focal lesion, we recommend, if feasible, catheter placement ipsilateral to the lesion and in radiographically normal brain. Multiple catheters are an option in focal TBI.E.g., placed at craniotomy for a focal lesion into peri-lesional brain with a contralateral ‘bolt’ catheter in radiographically normal brain. Stereotactic placement is an option but rarely practical. Subarachnoid hemorrhage There are two principal indications for the insertion of microdialysis in SAH patients:As a primary monitoring device in mechanically ventilated (‘poor-grade’) patients. As a monitor of patients with a secondary neurological deterioration. As a primary monitoring device, we recommend catheter location in the watershed anterior cerebral artery–middle cerebral artery (ACA–MCA) territory (frontal lobe) on the same side as the maximal blood load seen on CT or the ruptured aneurysm. If the blood load is symmetrical, we recommend non-dominant frontal lobe placement. In patients with a secondary neurological deterioration, catheter location should be guided by local practice to identify tissue at risk (e.g., CT perfusion scanning or trans-cranial Doppler). Multiple catheters are an option in SAH. Reference values and interventions It is currently difficult to define absolute normal or abnormal values based on the literature. Different groups have used different threshold values to relate microdialysate values to outcome. Furthermore, some authors have used a combination of values to relate microdialysis to clinical outcomes.

es and interventions It is currently difficult to define absolute normal or abnormal values based on the literature. Different groups have used different threshold values to relate microdialysate values to outcome. Furthermore, some authors have used a combination of values to relate microdialysis to clinical outcomes. The trend is as important or possibly more important than point values or threshold values. It is important to distinguish between normal values, which have been reported in the awake and anesthetised brain of patients undergoing surgery for benign intracranial lesions, and values that characterize pathophysiological disturbance of brain chemistry. We propose the following pathological thresholds (one or two stages), for microdialysis at 0.3 µl/min, based on observational studies that have explored statistical differences in outcomes in relation to thresholds of microdialysate values. Microdialysate values observed beyond these thresholds indicate that the area of brain being monitored is ‘at risk’. We propose clinical interventions that may be appropriate in response to disturbed brain chemistry. Further research is needed to elucidate whether these thresholds can be applied to both peri-lesional and to radiographically normal brain and to identify whether interventions directed by these thresholds improve clinical outcomes.Glucose: <0.2 and <0.8 mmol/L [53–55, 73].

ate in response to disturbed brain chemistry. Further research is needed to elucidate whether these thresholds can be applied to both peri-lesional and to radiographically normal brain and to identify whether interventions directed by these thresholds improve clinical outcomes.Glucose: <0.2 and <0.8 mmol/L [53–55, 73]. If brain glucose is low (<0.2 mM), a trial of increasing serum glucose (by intravenous or enteral administration and/or loosening glycemic control) should be considered. Factors to consider when deciding whether this is an appropriate intervention include baseline serum glucose concentration and whether other parameters indicate cerebral ischemia. If baseline serum glucose concentration is high, further increasing the glucose concentration is likely to increase the risk of both neurological and systemic complications from hyperglycemia. The precise definition of blood sugar thresholds for safety is beyond the scope of this manuscript, but frank hyperglycemia should be avoided. If other parameters, such as the LP ratio and PbtO2, indicate ischemia, interventions directed at improving cerebral perfusion should be considered first-line. Lactate: >4 mmol/L [51, 73, 79, 80]. Lactate/pyruvate ratio: >25 and >40 [6, 51, 53, 54, 68–73]. If the LP ratio indicates ischemia, i.e. an increase in the LP ratio in the presence of low pyruvate, CPP augmentation is a therapeutic option.

If brain glucose is low (<0.2 mM), a trial of increasing serum glucose (by intravenous or enteral administration and/or loosening glycemic control) should be considered. Factors to consider when deciding whether this is an appropriate intervention include baseline serum glucose concentration and whether other parameters indicate cerebral ischemia. If baseline serum glucose concentration is high, further increasing the glucose concentration is likely to increase the risk of both neurological and systemic complications from hyperglycemia. The precise definition of blood sugar thresholds for safety is beyond the scope of this manuscript, but frank hyperglycemia should be avoided. If other parameters, such as the LP ratio and PbtO2, indicate ischemia, interventions directed at improving cerebral perfusion should be considered first-line. Lactate: >4 mmol/L [51, 73, 79, 80]. Lactate/pyruvate ratio: >25 and >40 [6, 51, 53, 54, 68–73]. If the LP ratio indicates ischemia, i.e. an increase in the LP ratio in the presence of low pyruvate, CPP augmentation is a therapeutic option. If the LP ratio is increased in the presence of low brain tissue oxygen, interventions that improve oxygen delivery, such as judiciously increasing the cerebral perfusion pressure, increasing PaCO2, increasing inspired concentration of oxygen and/or correcting anemia, should be considered. However, all of these interventions have potential side effects, and the choice of intervention will depend on the pre-intervention levels of any given variable, and a consideration of the side effects of the intervention. Thus, for example, in patients with significant hypocarbia, an increase in PaCO2 might be the most appropriate intervention, but may be difficult to achieve due to increases in intracranial pressure.

epend on the pre-intervention levels of any given variable, and a consideration of the side effects of the intervention. Thus, for example, in patients with significant hypocarbia, an increase in PaCO2 might be the most appropriate intervention, but may be difficult to achieve due to increases in intracranial pressure. Tiered approach to the clinical value of substances Accumulating evidence since the last consensus statement indicates that the value of the metabolites can now be considered in a tiered fashion (tier 1 being most robust and useful) for their clinical application as follows. This hierarchy is based on the larger volume of observational data linking glucose and LP ratio with outcome compared to glutamate and glycerol and on the greater potential for glucose and LP ratio to direct clinical interventions. Tier 1: glucose and LP ratio. Tier 2: glutamate. Tier 3: glycerol.

Tiered approach to the clinical value of substances Accumulating evidence since the last consensus statement indicates that the value of the metabolites can now be considered in a tiered fashion (tier 1 being most robust and useful) for their clinical application as follows. This hierarchy is based on the larger volume of observational data linking glucose and LP ratio with outcome compared to glutamate and glycerol and on the greater potential for glucose and LP ratio to direct clinical interventions. Tier 1: glucose and LP ratio. Tier 2: glutamate. Tier 3: glycerol. Summary and future directions Cerebral microdialysis is a reliable and safe technique that is used in the clinical management of neurocritical care patients and in particular those with severe TBI or SAH. In addition, there are several research applications that are important for developing our understanding of brain physiology, pathophysiology and drug development. Since the 2004 consensus document, there have been significant advances in our understanding of how microdialysis can be used. There is now evidence from large numbers of patients on how abnormal brain chemistry relates to clinical outcome. The measurement of glucose, lactate and the LP ratio are now considered more useful than glutamate and glycerol. The LP ratio, interpreted in the light of absolute pyruvate concentrations and PbtO2, can be used to differentiate ischemic from non-ischemic causes of energy dysfunction. Importantly, there is increasing evidence of how different therapeutic manoeuvres influence brain chemistry. Microdialysis is well placed to help guide the management of patients in an individualized and targeted fashion. For its effective use, microdialysis should be integrated into brain multi-modal monitoring systems and interpreted with knowledge of catheter location and clinical context. Future clinical research should focus on assessing the clinical effectiveness of decision-making based on microdialysis, as part of multi-modality monitoring of acute brain injured patients, and its integration into treatment paradigms in neurocritical care.

ted with knowledge of catheter location and clinical context. Future clinical research should focus on assessing the clinical effectiveness of decision-making based on microdialysis, as part of multi-modality monitoring of acute brain injured patients, and its integration into treatment paradigms in neurocritical care. Electronic supplementary material Supplementary material 1 (PDF 2447 kb) The participants of the 2014 International Microdialysis Forum are listed in the electronic supplementary material (134_2015_3930_MOESM1_ESM).

ted with knowledge of catheter location and clinical context. Future clinical research should focus on assessing the clinical effectiveness of decision-making based on microdialysis, as part of multi-modality monitoring of acute brain injured patients, and its integration into treatment paradigms in neurocritical care. Electronic supplementary material Supplementary material 1 (PDF 2447 kb) The participants of the 2014 International Microdialysis Forum are listed in the electronic supplementary material (134_2015_3930_MOESM1_ESM). We gratefully acknowledge financial support for participants as follows: P.J.H.—National Institute for Health Research (NIHR) Professorship and the NIHR Biomedical Research Centre, Cambridge; I.J.—Medical Research Council (G1002277 ID 98489); A. H.—Medical Research Council, Royal College of Surgeons of England; K.L.H.C.—NIHR Biomedical Research Centre, Cambridge (Neuroscience Theme; Brain Injury and Repair Theme); M.G.B.—Wellcome Trust Dept Health Healthcare Innovation Challenge Fund (HICF-0510-080); L. H.—The Swedish Research Council, VINNOVA and Uppsala Berzelii Technology Centre for Neurodiagnostics; S. M.—Fondazione IRCCS Cà Granda Ospedale Maggiore Policlinico; D.K.M.—NIHR Senior Investigator Award to D.K.M., NIHR Cambridge Biomedical Research Centre (Neuroscience Theme), FP7 Program of the European Union; M. O.—Swiss National Science Foundation and the Novartis Foundation for Biomedical Research; J.S.—Fondo de Investigación Sanitaria (Instituto de Salud Carlos III) (PI11/00700) co-financed by the European Regional Development; M.S.—NIHR University College London Hospitals Biomedical Research Centre; N. S.—Fondazione IRCCS Cà Granda Ospedale Maggiore Policlinico.

the Novartis Foundation for Biomedical Research; J.S.—Fondo de Investigación Sanitaria (Instituto de Salud Carlos III) (PI11/00700) co-financed by the European Regional Development; M.S.—NIHR University College London Hospitals Biomedical Research Centre; N. S.—Fondazione IRCCS Cà Granda Ospedale Maggiore Policlinico. Conflicts of interest M Dialysis, Stockholm, Sweden, provided financial support for the 2014 International Microdialysis Forum although neither honorariums nor speaker fees were received by any of the authors. M Dialysis were neither involved in designing the structure of the meeting nor in selecting the participants. H.M. received a travel grant from M Dialysis to attend the meeting. The other authors have no conflicts of interest in relation to this manuscript. An extended conflict of interest statement for all participants is provided in the supplementary material (see Appendix 2).

Dear Editor, Oral care using a chlorhexidine solution is commonly used as an infection prevention measure in European ICUs [1]. The preventive effects of different decontamination strategies, one of which is mouthwash with chlorhexidine digluconate 2 % (CHX 2 %), on the incidence of ICU-acquired bacteremia with multidrug-resistant bacteria is being investigated in a multicenter cluster-randomized study in 13 ICUs in six European countries [www.clinicaltrials.gov NCT02208154] (see Supplementary Material for detailed methods). An unexpected high incidence of oral mucosal lesions was observed in ICU patients receiving CHX 2 %. Oral mucosal lesions, including erosive lesions, ulcerations, white/yellow plaque formation, and bleeding mucosa were observed in 29 of 295 patients (9.8 %) that had received CHX 2 % in the first two hospitals testing this intervention (Supplementary Table S1, Pictures 1–4). The median time to onset of oral lesions was 8.0 days (IQR 4.5–11.0) in the 24 patients in whom duration of exposure could be ascertained. CHX 2 % was discontinued prematurely in 16/29 cases and oral mucosal lesions disappeared after cessation of CHX 2 % in all patients. Patient characteristics were comparable for the baseline (n = 310) and CHX 2 % periods (n = 295; Supplementary Table S2) for the two ICUs. During the baseline period CHX 0.20 and 0.12 % were used for oral care in hospitals A and B, respectively, without evidence of oral lesions in any patient. All other procedures related to oral care remained identical in each hospital during both periods.

supported by the localization of the lesions in the oral cavity; most lesions occurred where stasis of the mouthwash might have occurred—despite suctioning after administration—such as below the tongue and in the buccal pockets.Table 1 Baseline characteristics of CHX 2 % treated patients with and without adverse events Adverse events (N = 29) No adverse events (N = 266) Pearson Chi square/indep. t test Male gender 23 (79.3 %) 161 (60.5 %) P = 0.047 Admission type P = 0.155 Medical 13 (44.8 %) 153 (57.5 %) Trauma 5 (17.2 %) 20 (7.5 %) Surgical 11 (37.9 %) 93 (35.0 %) Acute illness (y/n) 24 (82.8 %) 198 (74.4 %) P = 0.324 Antibiotic at ICU admission (y/n) 11/29 (37.9 %) 123/259 (47.5 %) P = 0.328 Age, mean (SD) 60.4 (13.3) 60.1 (15.7) P = 0.921 APACHE II, mean (SD) 26.7 (8.0) 19.6 (8.6) P < 0.0005 ICU-LOS, median (IQR) 28 (21–41.5) 10.5 (6–19) P < 0.0005 (LN) Geometric mean (SD) 27.2 (1.8) 10.6 (2.2) Length of MV, median (IQR) 19 (14.5–28.5) 6 (3–11) P < 0.0005 (LN) Geometric mean (SD) 18.8 (1.8) 6.2 (2.3) SD standard deviation, IQR interquartile range, LOS length of stay, LN log-transformed variable, MV mechanical ventilation, N number of patients Mechanical stress during application of CHX 2 % may have played a role in hospital A, where the solution was initially applied using Kocher’s forceps with gauzes and where the incidence seemed to have reduced after changing to application using a syringe. Hospital B had applied CHX 2 % with a gauze wrapped around a gloved finger.

stress during application of CHX 2 % may have played a role in hospital A, where the solution was initially applied using Kocher’s forceps with gauzes and where the incidence seemed to have reduced after changing to application using a syringe. Hospital B had applied CHX 2 % with a gauze wrapped around a gloved finger. In 12 patients symptoms predominantly consisted of pronounced white plaques at the tongue and other localizations in the mouth, in some resembling Candida infection (Supplementary Picture 3). Yet, the incidence rate ratio between prior respiratory tract colonization with Candida spp. (monitored twice weekly as part of the study protocol and in clinical cultures) and the occurrence of side effects was 0.94 (95 % confidence interval 0.09–1.79, Supplementary Table S3). An association with herpes reactivation could not be determined as reactivation was investigated in five affected patients only (Supplementary Table S1). The study safety committee recommended to replace CHX 2 % mouthwash by a CHX 1 % oral gel in the remaining hospitals. Since then CHX 1 % was withdrawn for reasons of intolerance in 2 of 419 (0.5 %) patients in four hospitals, after 12 and 30 days of use. On the basis of these findings, we recommend against the use of 2 % chlorhexidine digluconate mouthwash in ICU patients. Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary material 1 (DOCX 192 kb) N.L. Plantinga and B.H.J. Wittekamp contributed equally to this work as co-first authors.

On the basis of these findings, we recommend against the use of 2 % chlorhexidine digluconate mouthwash in ICU patients. Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary material 1 (DOCX 192 kb) N.L. Plantinga and B.H.J. Wittekamp contributed equally to this work as co-first authors. Compliance with ethical standards Conflicts of interest On behalf of all authors, the corresponding author states that there is no conflict of interest.

Introduction Sepsis in resource-limited settings will often have different aetiologies to those in western settings, including severe malaria, severe dengue, viral haemorrhagic fevers, mellioidosis, typhus, and leptospirosis. The Surviving Sepsis Campaign (SSC) guidelines [1] are mainly based on evidence from studies on bacterial sepsis. These guidelines are widely applicable, but there are also exceptions. We focus here on disease-specific recommendations for the management of severe falciparum malaria and severe dengue. An international team with extensive practical experience in resource-limited intensive care units (ICUs) identified key questions concerning the SSC’s management recommendations on these diseases. Pertinent evidence from resource-limited settings was evaluated using the grading of recommendations assessment, development and evaluation (GRADE) tools.

sive practical experience in resource-limited intensive care units (ICUs) identified key questions concerning the SSC’s management recommendations on these diseases. Pertinent evidence from resource-limited settings was evaluated using the grading of recommendations assessment, development and evaluation (GRADE) tools. Recommendations for the management of severe malaria and severe dengue in resource-limited settings Severe falciparum malaria Severe falciparum malaria is a multi-organ disease caused by Plasmodium falciparum transmitted by Anopheles mosquitoes. The highest transmission and disease burden is in sub-Saharan Africa, where severe malaria is largely a paediatric disease, as older children and adults become partly immune. In Asia and South America, all age groups may be affected. Independent of age, the presenting symptoms with the strongest prognostic significance are coma (cerebral malaria), metabolic (lactic) acidosis and renal dysfunction. Hypotension occurs infrequently (~12% of cases). One of the main pathophysiologic differences of severe falciparum malaria compared to bacterial sepsis is microcirculatory impairment caused by sequestration of parasite-infected erythrocytes, red cell rigidity and red cell clumping. Management requires rapid parasitological diagnosis by microscopy or rapid diagnostic testing (RDT) and prompt initiation of parenteral artesunate [2]. The SSC recommends that, in patients with sepsis-induced tissue hypoperfusion and suspicion of hypovolemia, with either hypotension or hyperlactatemia, an initial fluid challenge of at least 30 ml/kg of crystalloids be administered, of which a portion may be albumin equivalent [1]. Both paediatric and adult patients with severe malaria and tissue hypoperfusion are volume depleted intravascularly. A large trial on fluid bolus therapy in 3138 African children with severe infections and compensated shock, 57% of whom had falciparum malaria, showed an overall 40% increase in mortality with fluid bolus therapy (20 or 40 ml/kg with either saline or 5% albumin). In the 1793 children with severe P. falciparum malaria, mortality in the bolus groups was 51% higher [RR 1.51 (1.17–1.95)] [3]. In Asian studies of adult severe malaria, rapid fluid resuscitation did not improve metabolic acidosis [4, 5] and transpulmonary thermodilution-guided rapid fluid resuscitation resulted in pulmonary oedema in 8/28 (29%) patients [5].

parum malaria, mortality in the bolus groups was 51% higher [RR 1.51 (1.17–1.95)] [3]. In Asian studies of adult severe malaria, rapid fluid resuscitation did not improve metabolic acidosis [4, 5] and transpulmonary thermodilution-guided rapid fluid resuscitation resulted in pulmonary oedema in 8/28 (29%) patients [5]. One observational study from Myanmar showed no deterioration in renal function or plasma lactate with maintenance fluid therapy between 1.3 and 2.2 ml/kg/h [6]. We recommend against the use of fluid bolus therapy in normotensive patients with severe falciparum malaria (1A). In normotensive patients, we suggest initial (24 h) crystalloid fluid therapy of 2–4 ml/kg/h, which may subsequently be reduced to 1 ml/kg/h in patients receiving additional fluids, e.g. through enteral tube feeding (2D). We suggest fluid bolus therapy (30 ml/kg) with an isotonic crystalloid in patients with hypotensive shock and, if available, early initiation of vasopressor support (ungraded) (Table 1).

kg/h, which may subsequently be reduced to 1 ml/kg/h in patients receiving additional fluids, e.g. through enteral tube feeding (2D). We suggest fluid bolus therapy (30 ml/kg) with an isotonic crystalloid in patients with hypotensive shock and, if available, early initiation of vasopressor support (ungraded) (Table 1). The SSC suggests administering enteral feeding within the first 48 h after diagnosing severe sepsis. In resource-limited settings, endotracheal intubation of comatosed patients is often not practised. In a randomised trial in non-intubated predominantly adult Bangladeshi patients with cerebral malaria, early enteral feeding (<36 h), was associated with aspiration pneumonia in 9/27 (33%) as compared to 0/29 when feeding was commenced after 60 h [7]. No difference in hypoglycaemia incidence was observed. We suggest initiating enteral feeding in non-intubated adult patients with cerebral malaria after 60 h (2B). There are insufficient data on paediatric patients with cerebral malaria from African settings. For patients with sepsis-induced respiratory failure, low tidal volume (6 mL/kg) mechanical ventilation is recommended conform the SSC. In patients with cerebral malaria, we suggest against the use of permissive hypercapnia to achieve this goal, since this may exacerbate brain swelling (ungraded).

rom African settings. For patients with sepsis-induced respiratory failure, low tidal volume (6 mL/kg) mechanical ventilation is recommended conform the SSC. In patients with cerebral malaria, we suggest against the use of permissive hypercapnia to achieve this goal, since this may exacerbate brain swelling (ungraded). Severe dengue Severe dengue is caused by dengue virus transmitted by Aedes mosquitoes. Approximately 1–5% of patients will develop severe manifestations. The defining feature is a vasculopathy with increased capillary permeability, causing plasma leakage, reduced intravascular volume and, if severe, life-threatening hypovolemic shock [8]. This ‘critical phase’ typically starts during the period of defervescense, and lasts for approximately 48 h. Bleeding complications and organ involvement of the brain, liver, kidney and heart may be additional features, and occur more frequently in adult cases [9]. Diagnosis is commonly with combined dengue antigen (NS1) and antibody RDT [8]. No antiviral treatment is currently available (Table 1).

sts for approximately 48 h. Bleeding complications and organ involvement of the brain, liver, kidney and heart may be additional features, and occur more frequently in adult cases [9]. Diagnosis is commonly with combined dengue antigen (NS1) and antibody RDT [8]. No antiviral treatment is currently available (Table 1). Unlike in bacterial sepsis, capillary leak in patients with severe dengue results in haemoconcentration. Haemorrhage, in particular from the gut, can contribute to hypovolemic shock [9]. Myocarditis is rare, but some depression of myocardial contractility is common. The World Health Organisation (WHO) guidelines on fluid resuscitation recommend restoration of the circulation guided by pulse pressure, capillary refill time, haematocrit and urine output [8]. Cautious but prompt fluid administration is essential and should be restricted as soon as the critical phase is over to avoid pulmonary oedema [10]. We recommend fluid administration according to the WHO guidelines (1C). We recommend that rapid (<30 min) administration of large (>15 ml/kg) fluid boluses are avoided, unless the patient is hypotensive (1D). In patients with compensated shock, colloids are not superior to normal saline or Ringer’s lactate for shock reversal, or for the prevention of recurrent shock, and we recommend that colloids are not used (1A) [11–13]. There is insufficient evidence to recommend fluid choice in dengue hypotensive shock. The use of corticosteroids is not recommended (1B) [14]. Although thrombocytopenia is an inherent feature of severe dengue, the cause of bleeding is multifactorial, including a prominent vasculopathy. Bleeding is not prevented by platelet transfusion [15]; we do not recommend platelet transfusion for thrombocytopenia in the absence of active bleeding complications or other risk factors (1C). In cases of bleeding complications, we suggest transfusion of fresh-frozen plasma (or cryoprecipitate) and platelet concentrate (ungraded).

ot prevented by platelet transfusion [15]; we do not recommend platelet transfusion for thrombocytopenia in the absence of active bleeding complications or other risk factors (1C). In cases of bleeding complications, we suggest transfusion of fresh-frozen plasma (or cryoprecipitate) and platelet concentrate (ungraded). Conclusions The management of severe malaria and severe dengue differs in some important aspects from the treatment of bacterial sepsis, in particular, regarding fluid management. Table 1 Recommendations and suggestions for the management of patients with severe malaria and severe dengue in resource-limited settings (with grading)

The management of severe malaria and severe dengue differs in some important aspects from the treatment of bacterial sepsis, in particular, regarding fluid management. Table 1 Recommendations and suggestions for the management of patients with severe malaria and severe dengue in resource-limited settings (with grading) Fluid management of severe malaria We recommend not to use fluid bolus therapy in normotensive patients with severe falciparum malaria (1A). We suggest that patients receive maintenance isotonic crystalloid fluid therapy (2–4 ml/kg/h), which may subsequently be reduced to 1 ml/kg/h in patients receiving additional fluids, e.g. through enteral tube feeding (2D). We suggest that, in patients with hypotensive shock, fluid bolus therapy (30 ml/kg) with isotonic crystalloids be commenced (ungraded) and, if available, early initiation of vasopressor medication (ungraded) Timing of enteral feeding in cerebral malaria We suggest initiating enteral feeing in non-intubated adult patients with cerebral malaria after 60 h, in order to limit the possibility of aspiration pneumonia (2B). There are insufficient data to make this recommendation for children with cerebral malaria Permissive hypercapnia in ventilated cerebral malaria We suggest not to use a strategy of permissive hypercapnia to achieve ventilation with low tidal volumes in patients with cerebral malaria, because of the high incidence of brain swelling in these patients (ungraded) Fluid management in severe dengue We recommend that fluid resuscitation in severe dengue is executed promptly and guided by pulse pressure, capillary refill time, haematocrit and urine output according to WHO guidelines, and that fluid therapy should be restricted as soon as the critical phase of the disease is over to avoid pulmonary oedema (1C). We recommend that rapid administration of large fluid boluses should be avoided, unless the patient is hypotensive (1D). We recommend that, in dengue patients with compensated shock, colloid fluids are not used (1A) Use of corticosteroids in severe dengue We recommend not to use corticosteroids in the treatment of severe dengue (1B) Use of prophylactic platelet transfusion in severe dengue We recommend not to use prophylactic platelet transfusion for thrombocytopenia in the absence of active bleeding complications, or other risk factors (uncontrolled arterial hypertension, recent stroke, head trauma or surgery, continuation of an anticoagulant treatment, or existing haemorrhagic diathesis) (1B)

re dengue We recommend not to use prophylactic platelet transfusion for thrombocytopenia in the absence of active bleeding complications, or other risk factors (uncontrolled arterial hypertension, recent stroke, head trauma or surgery, continuation of an anticoagulant treatment, or existing haemorrhagic diathesis) (1B) Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary material 1 (DOCX 88 kb) Electronic supplementary material The online version of this article (doi:10.1007/s00134-016-4602-2) contains supplementary material, which is available to authorized users. Acknowledgements Dengue and malaria subgroup members: Sanjib Mohanty (Ispat General Hospital, Rourkela, Sundargarh, Odisha, India), Marcus Schultz (Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands and Faculty of Tropical Medicine, Mahidol University, Bangkok, Thailand), Louise Thwaites (Centre for Tropical Medicine and Global Health, Nuffield Department of Medicine, University of Oxford, Oxford, UK), Martin Dünser (Department of Critical Care, University College of London Hospital, London, UK), Jane Nakibuuka (Mulago National Referral and University Teaching Hospital, Kampala, Uganda). We thank Prof. Bridget Wills for her input on the recommendations for severe dengue management.

Correction to: Intensive Care Med 10.1007/s00134-018-5148-2 This article was originally published under a CC BY-NC 4.0 license, but has now been made available under a CC BY 4.0 license. The PDF and HTML versions of the paper have been modified accordingly. Open access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary material 1 (DOCX 28 kb) The original article can be found online at 10.1007/s00134-018-5148-2. Electronic supplementary material The online version of this article (10.1007/s00134-018-5254-1) contains supplementary material, which is available to authorized users.

Take-home message In NHS hospitals, deteriorating ward patients referred to ICU are vulnerable: one in eight will die within a week of the referral, and half of those deaths will occur without ICU admission. While ICU admission delays are common, this study shows that prompt admission reduces mortality after allowing for both observed and unobserved differences in prognosis. Introduction Recent policy stresses the importance of identifying and responding to the deteriorating ward patient [1]. Current guidelines recommend that critical care admission should be delivered within 4 h [2]. However, supporting evidence is limited because randomised evaluation of prompt admission to critical care is deemed unethical. Yet without quantification of the benefits, it is difficult to assess the magnitude and importance of this problem. Non-randomised evaluations are primarily confounded by treatment allocation bias [3]. Patients are prioritised on the basis of clinical severity so prompt admissions tend to have poorer prognoses. Risk adjustment will help remove this bias, but depends heavily on adequate measurement of all factors driving the decision about how to treat. In general, measured severity is an incomplete description, and often there are other end of the bed factors prompting clinicians to recommend prompt admission to critical care.

isk adjustment will help remove this bias, but depends heavily on adequate measurement of all factors driving the decision about how to treat. In general, measured severity is an incomplete description, and often there are other end of the bed factors prompting clinicians to recommend prompt admission to critical care. An alternative to experimental randomisation is to seek an instrument that naturally randomises patients to prompt admission or not. This natural randomisation is known as instrumental variable (IV) analysis, and has been similarly used to remove unmeasured confounding in the assessment of influenza vaccine efficacy and cardiac catheterisation [4, 5]. Here, we used critical care unit strain, measuring bed occupancy rates at the specific time of the patient’s assessment, for this purpose. This approach circumvents selection bias found in previous observational studies comparing prompt to delayed admission [6–13]. Delay can only be defined with hindsight. The bedside choice is to ‘admit now’ or to ‘watch and wait’, not to admit now or deliberately delay. A clinician neither admits the watched patient who improved nor can admit the watched patient who unexpectedly died. The delayed admissions are the subgroup of watchful-waiting controls who survive but continue to deteriorate.

side choice is to ‘admit now’ or to ‘watch and wait’, not to admit now or deliberately delay. A clinician neither admits the watched patient who improved nor can admit the watched patient who unexpectedly died. The delayed admissions are the subgroup of watchful-waiting controls who survive but continue to deteriorate. A fair comparison requires prospective follow-up of all watchful-waiting controls. Our prospective study does this and also exploits the learning opportunity created by a constrained supply of critical care beds. Critical care bed provision in the UK is lower than in the majority of European (6.6 adult critical care beds versus a European average of 11.5 beds per 100,000 population) and North American health care systems [14, 15]. In settings where critical care capacity is less constrained, using strain to evaluate the effect of delay would be difficult. We first describe the effect of critical care strain on decision-making, and the delivery of critical care for all deteriorating ward patients referred to critical care. We explore whether delays to admission engendered by high strain allow us to estimate the effect of effect of delay on patient outcome. Finally, we focus on the subgroup recommended for critical care by the bedside clinician.

d the delivery of critical care for all deteriorating ward patients referred to critical care. We explore whether delays to admission engendered by high strain allow us to estimate the effect of effect of delay on patient outcome. Finally, we focus on the subgroup recommended for critical care by the bedside clinician. Methods Study design, participants and procedures The full study protocol is available on the Intensive National Audit and Research Centre’s (ICNARC) website. In brief, the (SPOT)light study was a prospective cohort study of the deteriorating ward patient referred for assessment by critical care. The assessment had to be conducted on an inpatient ward by either a member of the critical care medical staff or the critical care outreach team (CCOT). Repeat visits and re-admissions were excluded as were patients where intensive care units (ICU) admission was either a priori refused (treatment limitation orders) or inevitable (cardiac arrests, admissions temporarily housed in theatre recovery etc.). Admissions following surgery, where delay was due to the process of care, were also excluded. The study was registered on the National Institute of Health Research (NIHR) research portfolio (No. 9139). Hospitals were eligible for inclusion if they participated in the national clinical audit for critical care—the ICNARC case mix programme (CMP). Research teams at each hospital attended a data set familiarisation course and were given a manual of data definitions. The ICNARC clinical trials unit provided support for the study.

s were eligible for inclusion if they participated in the national clinical audit for critical care—the ICNARC case mix programme (CMP). Research teams at each hospital attended a data set familiarisation course and were given a manual of data definitions. The ICNARC clinical trials unit provided support for the study. Reporting was via a secure online web portal that performed real-time field and record level validation. Hospitals were asked to report all consecutive ward referrals to critical care. Contemporaneous data collection was recommended, but missed referrals were sought and accepted retrospectively. We used the proportion of unplanned ward admissions to critical care in the CMP that were successfully linked to the (SPOT)light database to monitor data capture each month. Data from specific months in which data linkage rates fell below 80% were excluded from the primary analysis (but explored in sensitivity analyses). Further online validation reports were completed by all hospitals before the database was locked in September 2012. Fact and date of death were then requested from the NHS Information Service. CCOT provision was reported by participating hospitals. CMP and hospital episode statistics (HES) data were used to define critical care provision and hospital characteristics.

by all hospitals before the database was locked in September 2012. Fact and date of death were then requested from the NHS Information Service. CCOT provision was reported by participating hospitals. CMP and hospital episode statistics (HES) data were used to define critical care provision and hospital characteristics. Definitions Physiology measurements at the time of the ward assessment were abstracted. From these, the ICNARC physiology score, the NHS National Early Warning Score (NEWS) and the Sequential Organ Failure Assessment (SOFA) score were calculated with missing values given zero weights as recommended [16–18]. The patient’s existing dependency at assessment was defined using the UK critical care minimum data set (CCMDS) levels of care: levels 0 and 1 are most commonly provided on normal wards while levels 2 and 3 are within high dependency (HDU) and ICU respectively [19]. The assessor was asked to recommend a future level of care, and recommendations for levels 2 or 3 were considered as recommendations for critical care admission. Prompt admission was defined as one within 4 h of ward assessment, in line with recently published UK guidelines [19].

ndency (HDU) and ICU respectively [19]. The assessor was asked to recommend a future level of care, and recommendations for levels 2 or 3 were considered as recommendations for critical care admission. Prompt admission was defined as one within 4 h of ward assessment, in line with recently published UK guidelines [19]. The indicator of critical care unit strain was the difference between the maximum number of beds reported to ICNARC and the number of actively treated patients (not medically fit for discharge) occupying those beds at the time the ward patient was assessed. Units were defined as being under low, medium or high strain corresponding to having two or more, one, or zero or fewer empty beds (the last of these where strain exceeded reported capacity) respectively. Statistical analysis The aim of the primary analysis was to estimate the effect of prompt critical care admission versus watchful waiting on 90-day mortality. We repeated the analysis in the subgroup recommended for critical care at the bedside assessment.

The indicator of critical care unit strain was the difference between the maximum number of beds reported to ICNARC and the number of actively treated patients (not medically fit for discharge) occupying those beds at the time the ward patient was assessed. Units were defined as being under low, medium or high strain corresponding to having two or more, one, or zero or fewer empty beds (the last of these where strain exceeded reported capacity) respectively. Statistical analysis The aim of the primary analysis was to estimate the effect of prompt critical care admission versus watchful waiting on 90-day mortality. We repeated the analysis in the subgroup recommended for critical care at the bedside assessment. We first built orthodox proportional hazards models with risk adjustment to handle the anticipated treatment allocation bias according to those risk factors that were observed. We then built IV models in two stages. In the first stage, a selection model was constructed to predict prompt admission including the effect of strain. In the second stage, an outcome model was built replacing strain with the fitted prediction of prompt admission from the selection model. All models were adjusted for patient-level confounders including age, sex, reported referral delay, sepsis diagnosis, peri-arrest status, existing CCMDS level of care, and severity of illness using the ICNARC, SOFA and NEWS scores.

eplacing strain with the fitted prediction of prompt admission from the selection model. All models were adjusted for patient-level confounders including age, sex, reported referral delay, sepsis diagnosis, peri-arrest status, existing CCMDS level of care, and severity of illness using the ICNARC, SOFA and NEWS scores. We tested for weak instruments using the Kleibergen–Paap F test, and used Huber–White (robust) standard errors to allow for hospital-level clustering and potential heteroscedasticity [20]. Bivariate probit IV models were used to ensure that model predictions were correctly constrained [21]. To aid interpretation, we also calculated the marginalised average treatment effect (ATE) for the population, and converted coefficients to approximate odds ratios (OR) by scaling by 1.6 [22]. Finally, we examined the sensitivity of the results to changes in the data linkage quality threshold (between 70% and 90%). Implementation The study was registered with ClinicalTrials.gov (NCT01099813). The sample size was calculated to evaluate mortality increases from delay to admission using estimates from 2007 ICNARC CMP data. The target sample size was 12,075–20,125 patients referred to critical care, allowing for delays to occur in 10–40% of admissions and mortality effect sizes of 5–10%.

ials.gov (NCT01099813). The sample size was calculated to evaluate mortality increases from delay to admission using estimates from 2007 ICNARC CMP data. The target sample size was 12,075–20,125 patients referred to critical care, allowing for delays to occur in 10–40% of admissions and mortality effect sizes of 5–10%. Categorical data were reported as counts and percentages, and continuous data as mean (SD) or median (IQR) values. Effect measures are reported with their 95% confidence intervals. Analyses were performed in R (version 3.03) except for the IV analysis which used the ivregress and biprobit commands provided in Stata (version 12.1). Results Between September 2010 and December 2011, 49 hospitals (10 university-affiliated) submitted records for 435 study months. After cross-checking against national audit records, reporting was potentially incomplete for 66 (15%) months which were excluded. The primary analysis therefore included 369 study months, equivalent to a median eight study months per site (IQR 5–9) with a mean data linkage rate of 95%.

ed records for 435 study months. After cross-checking against national audit records, reporting was potentially incomplete for 66 (15%) months which were excluded. The primary analysis therefore included 369 study months, equivalent to a median eight study months per site (IQR 5–9) with a mean data linkage rate of 95%. The 369 study months captured 18,122 consecutive ward assessments. We excluded a further 2555 (14%) patients with treatment limitation orders and 1632 (9%) post-critical care follow-up visits. Timing data were unavailable for a further 129 patients. This left 12,380 patients in 48 hospitals available for analysis [a median of 222 patients (IQR 142–304) per hospital] (Fig. 1).Fig. 1 Flow diagram of patients screened: ward referrals assessed for eligibility at participating hospitals, reasons for exclusion, and admission timing following bedside assessment for all patients assessed, and for the subgroup recommended for critical care at assessment Participating hospitals There was a median of 12 critical care beds per hospital (IQR 9–18, mixed level 2 and level 3 beds) most often in a single physical location (45 hospitals). Each unit admitted a median 20 unplanned admissions (IQR 14–26) from the ward per month, representing 36% of all ICU admissions (IQR 31–43%). Critical care outreach was available 24/7 in 14 hospitals, daily but not overnight in 19 hospitals, weekday daytime only in 13 hospitals and was not offered in two hospitals.

s). Each unit admitted a median 20 unplanned admissions (IQR 14–26) from the ward per month, representing 36% of all ICU admissions (IQR 31–43%). Critical care outreach was available 24/7 in 14 hospitals, daily but not overnight in 19 hospitals, weekday daytime only in 13 hospitals and was not offered in two hospitals. Patient characteristics Table 1 shows the baseline data for all ward patients assessed. Sepsis was reported in 7586 (61%) patients; of these, the respiratory system was considered to be the source in 3851 (51%). Organ failure, defined as a SOFA score greater than one, was present in 4227 (34%) of patients. A total of 1173 patients (9%) were in respiratory failure, 2403 (19%) were in renal failure and 3629 (29%) were shocked. There was a clear correlation between physiological severity and short-term (7-day) outcome (Supplemental Fig. 1), but organ support at the time of assessment was uncommon (694 patients, 5%).Table 1 Study patients and those admitted promptly

iratory failure, 2403 (19%) were in renal failure and 3629 (29%) were shocked. There was a clear correlation between physiological severity and short-term (7-day) outcome (Supplemental Fig. 1), but organ support at the time of assessment was uncommon (694 patients, 5%).Table 1 Study patients and those admitted promptly All patients Prompt admission Odds ratio (95% CI) p value (n = 12,380) (n = 2411) Age (years) 18–39 1371 (11.1%) 258 (10.7%) 40–59 2616 (21.1%) 567 (23.5%) 1.19 (1.01–1.41) 0.0346 60–79 5454 (44.1%) 1144 (47.4%) 1.15 (0.99–1.33) 0.0773 80– 2939 (23.7%) 442 (18.3%) 0.76 (0.64–0.90) 0.0018 Sex Female 5863 (47.4%) 1056 (43.8%) Male 6517 (52.6%) 1355 (56.2%) 1.19 (1.09–1.31) 0.0001 Reported sepsis diagnosis Not reported septic 4794 (38.7%) 776 (32.2%) Other/unspecified 1672 (13.5%) 317 (13.1%) 1.21 (1.05–1.40) 0.0093 Genitourinary 882 (7.1%) 175 (7.3%) 1.28 (1.07–1.54) 0.0077 Gastrointestinal 1181 (9.5%) 236 (9.8%) 1.29 (1.10–1.52) 0.0019 Respiratory 3851 (31.1%) 907 (37.6%) 1.60 (1.43–1.78) < 0.0001 Referral timing Timely 10,814 (87.4%) 2079 (86.2%) Delayed 1566 (12.6%) 332 (13.8%) 1.13 (0.99–1.29) 0.0652 CCMDS level of care at visit Level 0 1666 (13.5%) 225 (9.3%) Level 1 8490 (68.6%) 1386 (57.5%) 1.25 (1.07–1.45) 0.0040 Level 2 2147 (17.3%) 779 (32.3%) 3.65 (3.09–4.30) < 0.0001 Acute physiology scores ICNARC 14.0 (10.0–20.0) 18.0 (13.0–24.0) 1.09 (1.08–1.09) < 0.0001 SOFA 3.0 (2.0–4.0) 4.0 (2.0–6.0) 1.29 (1.26–1.31) < 0.0001 NEWS 6.0 (4.0–8.0) 8.0 (5.0–10.0) 1.19 (1.18–1.21) < 0.0001 NEWS risk class None 336 (2.7%) 44 (1.8%) Low 3224 (26.0%) 399 (16.5%) 0.94 (0.67–1.31) 0.7039 Medium 3570 (28.8%) 529 (21.9%) 1.15 (0.83–1.61) 0.3939 High 5250 (42.4%) 1439 (59.7%) 2.51 (1.81–3.46) < 0.0001 Reported to be peri-arrest No 11,815 (95.4%) 2103 (87.2%) Yes 565 (4.6%) 308 (12.8%) 5.53 (4.66–6.57) < 0.0001 Visit recommendation Not for critical care 7820 (63.2%) 161 (6.7%) For critical care 4560 (36.8%) 2250 (93.3%) 46.34 (39.23–54.73) < 0.0001 Critical care admission During 7-day follow-up 4401 (35.5%) 2411 (100.0%) Mortality 7-day 1717 (13.9%) 500 (20.7%) 1.88 (1.68–2.11) < 0.0001 90-day 3736 (30.2%) 885 (36.7%) 1.45 (1.32–1.59) < 0.0001 Data are presented as mean (SD), median (IQR) or number (%). ICNARC, SOFA and NEWS refer to severity of illness scores derived from vital signs and laboratory tests. Odds ratios are calculated from univariate logistic regression for prompt admission to critical care

ay 3736 (30.2%) 885 (36.7%) 1.45 (1.32–1.59) < 0.0001 Data are presented as mean (SD), median (IQR) or number (%). ICNARC, SOFA and NEWS refer to severity of illness scores derived from vital signs and laboratory tests. Odds ratios are calculated from univariate logistic regression for prompt admission to critical care Recommendation for critical care at bedside assessment A total of 4560 (37%) patients were recommended for critical care at the bedside assessment. These patients were younger (by 1.3 years, 95% CI 0.7–1.9) and more acutely unwell (by 4.1 ICNARC physiology points, 95% CI 3.8–4.4) than those not recommended. Patients older than 80 years were less likely to be recommended for critical care (OR 0.61, 95% CI 0.53–0.71) even after risk adjustment (Supplemental Table 1). Prompt admission to critical care Overall, 36% (4401 patients) were admitted to critical care in the following week rising to 72% (3296) of those recommended on assessment. The median time from assessment to admission was 3 h (IQR 1–9) overall and 2 h (IQR 1–5) for those recommended. Admissions were prompt for 19% (2411 patients) rising to 49% (2250 patients) for those with a bedside recommendation. A total of 1636 (13%) patients died in the week following the assessment with 856 deaths (52%) following critical care admission and 780 deaths (48%) on the ward without admission. For those recommended for critical care at assessment, there were 862 deaths (19%) of which 663 (77%) received critical care before death and 199 (23%) occurred on the ward without ICU admission.

g the assessment with 856 deaths (52%) following critical care admission and 780 deaths (48%) on the ward without admission. For those recommended for critical care at assessment, there were 862 deaths (19%) of which 663 (77%) received critical care before death and 199 (23%) occurred on the ward without ICU admission. Prompt admission and risk-adjusted mortality For those admitted promptly, 90-day mortality was 36.7% (885 deaths) compared to 28.6% (2851) for the watchful-waiting control group. Patients who were admitted promptly had higher physiological severity scores; for example, the ICNARC physiology score was 4.4 ICNARC physiology points [95% CI 4.0–4.7] higher on average. Without risk adjustment, the proportion of patients who died prior to 90 days was higher for prompt critical care admissions with a hazard ratio (HR) of 1.42 (95% CI 1.32–1.54). With risk adjustment (Supplemental Table 3), survival was equivalent [HR 0.98 (95% CI 0.88–1.09), p = 0.702]. For the 4560 patients recommended for critical care, 90-day mortality was 37.2% (837 deaths) for prompt admissions versus 34.4% (794 deaths) for controls. Prompt admissions again had higher physiological severity [2.0 ICNARC physiology points (95% CI 1.6–2.5)]. Before risk adjustment, survival was worse (HR 1.12, 95% CI 1.02–1.24), but again was again equivalent after risk adjustment [HR 0.99 (95% CI 0.85–1.10), p = 0.852, Supplemental Table 4].

794 deaths) for controls. Prompt admissions again had higher physiological severity [2.0 ICNARC physiology points (95% CI 1.6–2.5)]. Before risk adjustment, survival was worse (HR 1.12, 95% CI 1.02–1.24), but again was again equivalent after risk adjustment [HR 0.99 (95% CI 0.85–1.10), p = 0.852, Supplemental Table 4]. Critical care strain There were 10,039 (81%) bedside assessments when there were two or more empty beds on the critical care unit, 1353 (11%) when there was just one empty bed and 988 (8%) when the unit was already fully occupied (Table 2). As strain increased, the proportion of prompt admissions fell (21%, 15% and 9%, p < 0.0001), corresponding upward trend in the median time to admission: 3 h (IQR 1–8), 4 h (IQR 1–12) and 5 h (IQR 2–16) (p = 0.0009, Supplementary Fig. 2).Table 2 Effects of strain on the admission pathway: recommendation for, and prompt admission to critical care, severity of illness, and outcomes stratified by critical care unit occupancy at the time of the bedside assessment

admission: 3 h (IQR 1–8), 4 h (IQR 1–12) and 5 h (IQR 2–16) (p = 0.0009, Supplementary Fig. 2).Table 2 Effects of strain on the admission pathway: recommendation for, and prompt admission to critical care, severity of illness, and outcomes stratified by critical care unit occupancy at the time of the bedside assessment Critical care beds Test for trend ≤ 0 1 ≥ 2 p value Patients referred 988 (8.0%) 1353 (10.9%) 10,039 (81.1%) Critical care Recommended 354 (35.8%) 471 (34.8%) 3735 (37.2%) 0.1407 Admitted 247 (25.0%) 425 (31.4%) 3775 (37.6%) < 0.0001 Prompt admission 87 (8.8%) 196 (14.5%) 2128 (21.2%) < 0.0001 Death without critical care 92 (9.3%) 79 (5.8%) 614 (6.1%) 0.0002 Time to critical care, hours 5.0 (2.2–15.8) 4.0 (1.0–12.0) 3.0 (1.0–8.0) 0.0009 ICNARC physiology score At referral 15.2 (7.1) 15.1 (7.2) 15.2 (7.2) 0.8266 Change between referral and admission 4.5 (9.2) 3.3 (9.1) 3.1 (9.2) 0.0301 Mortality 7-day 147 (14.9%) 179 (13.2%) 1391 (13.9%) 0.6226 90-day 312 (31.6%) 417 (30.8%) 3007 (30.0%) 0.2326 Trends are tested using the Cochrane–Armitage test for categorical outcomes, and by evaluating continuous variables in a linear regression model Strain varied by time of day, day of the week, and season; however, the relationship between prompt admission and occupancy remained even after adjustment (Supplemental Table 2), and there was a strong negative correlation between strain and prompt critical care admission (Kleibergen–Paap F statistic 83, p < 0.0001).

Critical care beds Test for trend ≤ 0 1 ≥ 2 p value Patients referred 988 (8.0%) 1353 (10.9%) 10,039 (81.1%) Critical care Recommended 354 (35.8%) 471 (34.8%) 3735 (37.2%) 0.1407 Admitted 247 (25.0%) 425 (31.4%) 3775 (37.6%) < 0.0001 Prompt admission 87 (8.8%) 196 (14.5%) 2128 (21.2%) < 0.0001 Death without critical care 92 (9.3%) 79 (5.8%) 614 (6.1%) 0.0002 Time to critical care, hours 5.0 (2.2–15.8) 4.0 (1.0–12.0) 3.0 (1.0–8.0) 0.0009 ICNARC physiology score At referral 15.2 (7.1) 15.1 (7.2) 15.2 (7.2) 0.8266 Change between referral and admission 4.5 (9.2) 3.3 (9.1) 3.1 (9.2) 0.0301 Mortality 7-day 147 (14.9%) 179 (13.2%) 1391 (13.9%) 0.6226 90-day 312 (31.6%) 417 (30.8%) 3007 (30.0%) 0.2326 Trends are tested using the Cochrane–Armitage test for categorical outcomes, and by evaluating continuous variables in a linear regression model Strain varied by time of day, day of the week, and season; however, the relationship between prompt admission and occupancy remained even after adjustment (Supplemental Table 2), and there was a strong negative correlation between strain and prompt critical care admission (Kleibergen–Paap F statistic 83, p < 0.0001). There was evidence that the delay in admission translated into further ongoing physiological deterioration (ICNARC physiology scores increased by 3.1, 3.3 and 4.5 points respectively, p = 0.03). Regardless of timing, the overall probability of receiving critical care also fell (38%, 32% and 25% for assessments during times of low, medium and high strain, p < 0.0001).

anslated into further ongoing physiological deterioration (ICNARC physiology scores increased by 3.1, 3.3 and 4.5 points respectively, p = 0.03). Regardless of timing, the overall probability of receiving critical care also fell (38%, 32% and 25% for assessments during times of low, medium and high strain, p < 0.0001). Within the subgroup recommended, the proportion of prompt admissions also fell as strain increased (53%, 38% and 23%, p < 0.0001) and median time to admission increased: 2 h (IQR 1–4), 3 h (IQR 1–6) and 4 h (IQR 2–9) (p < 0.0001). The proportion of patients managed without critical care increased from 25%, to 33%, to 50% during low, medium and high strain periods (Supplemental Table 5). Critical care strain and mortality The 90-day mortality was 30.0% (3007 deaths), 30.8% (417 deaths) and 31.6% (312 deaths) for deteriorating ward patients assessed at times of low, medium and high critical care strain respectively (Fig. 2). Using the change in mortality driven by critical care strain, the instrumental variable model estimated a reduction in 90-day mortality of 7.4% (95% CI 1.7–18.5%, p = 0.117) which was equivalent to an odds ratio of 0.68 (95% CI 0.42–1.10, Table 3 and Supplemental Table 6).Fig. 2 Patient disposition over time following bedside assessment: proportion of patients who are alive in critical care, who died within or following a critical care admission, or who died without admission to critical care by critical care unit strain at the time of the bedside assessment

3 and Supplemental Table 6).Fig. 2 Patient disposition over time following bedside assessment: proportion of patients who are alive in critical care, who died within or following a critical care admission, or who died without admission to critical care by critical care unit strain at the time of the bedside assessment Table 3 Instrumental variable model for the effect of prompt admission on 90-day mortality: for all patients, and for the subgroup with recommended to critical care at the bedside assessment

3 and Supplemental Table 6).Fig. 2 Patient disposition over time following bedside assessment: proportion of patients who are alive in critical care, who died within or following a critical care admission, or who died without admission to critical care by critical care unit strain at the time of the bedside assessment Table 3 Instrumental variable model for the effect of prompt admission on 90-day mortality: for all patients, and for the subgroup with recommended to critical care at the bedside assessment All patients Recommended for critical care Odds ratio p value Odds ratio p value Visiting timing Winter 1.02 (0.93–1.11) 0.734 0.98 (0.85–1.14) 0.816 Weekend (Saturday–Sunday) 1.02 (0.94–1.12) 0.624 1.13 (0.98–1.30) 0.083 Out-of-hours (7 p.m.–7 a.m.) 0.95 (0.87–1.04) 0.285 1.05 (0.90–1.23) 0.553 Age (per year) < 80 years 1.03 (1.03–1.03) < 0.001 1.02 (1.02–1.03) < 0.001 ≥ 80 years 1.02 (1.01–1.04) 0.004 1.03 (1.00–1.06) 0.087 Male sex 1.09 (1.00–1.17) 0.043 1.09 (0.96–1.23) 0.183 Reported sepsis diagnosis Not septic Unspecified sepsis 1.06 (0.93–1.20) 0.375 1.02 (0.83–1.26) 0.834 Genitourinary sepsis 0.58 (0.48–0.69) < 0.001 0.51 (0.38–0.68) < 0.001 Abdominal sepsis 0.84 (0.73–0.97) 0.020 0.85 (0.67–1.06) 0.147 Chest sepsis 1.24 (1.13–1.37) < 0.001 1.31 (1.12–1.53) 0.001 Level of care at time of visit Level 0 1.04 (0.90–1.20) 0.629 1.09 (0.84–1.41) 0.525 Level 1 Level 2 0.95 (0.84–1.08) 0.469 0.90 (0.76–1.06) 0.220 Delayed referral to critical care 1.00 (0.89–1.13) 0.950 0.94 (0.79–1.13) 0.517 Reported to be peri-arrest 0.95 (0.78–1.15) 0.589 1.08 (0.86–1.34) 0.517 Acute physiology score NEWS 1.06 (1.05–1.08) < 0.001 1.06 (1.04–1.09) < 0.001 ICNARC 1.03 (1.02–1.03) < 0.001 1.02 (1.01–1.03) < 0.001 SOFA 1.14 (1.12–1.17) < 0.001 1.13 (1.09–1.17) < 0.001 Level of care recommended Level 0 1.01 (0.83–1.23) 0.923 Level 1 Reference Level 2 1.29 (1.02–1.62) 0.034 Level 3 1.76 (1.18–2.62) 0.006 1.53 (1.17–2.02) 0.002 Prompt admission (within 4 h) 0.68 (0.42–1.10) 0.118 0.46 (0.22–0.96) 0.036 The coefficients from the underlying bivariate probit model have been scaled by 1.6 to give OR. Age was entered into the model using a linear spline with a knot at 80 years to account for the age bias in the selection model

53 (1.17–2.02) 0.002 Prompt admission (within 4 h) 0.68 (0.42–1.10) 0.118 0.46 (0.22–0.96) 0.036 The coefficients from the underlying bivariate probit model have been scaled by 1.6 to give OR. Age was entered into the model using a linear spline with a knot at 80 years to account for the age bias in the selection model Among those recommended for critical care, unadjusted 90-day mortality was 35.0% (1307 deaths), 41.8% (197 deaths) and 35.9% (127 deaths) for ward assessments at times of low, medium and high strain. For this subgroup, the average reduction in 90-day mortality for prompt admission was 16.2% (95% CI 1.1–31.3%, p = 0.036) equivalent to an odds ratio of 0.46 (95% CI 0.22–0.96, p = 0.036). Discussion This prospective cohort study describes the outcomes of more than 12,000 ward patients assessed by critical care teams in 48 acute NHS hospitals. Delay to admission is common. Those patients admitted promptly to critical care are manifestly more unwell, with a higher unadjusted mortality (37% vs 29%). Risk adjustment for observed severity is unable to show a benefit for a prompt admission strategy. However, we argue that unobserved differences in baseline risk are part of a clinician’s bedside assessment, and our risk adjustment is likely to be incomplete.

ore unwell, with a higher unadjusted mortality (37% vs 29%). Risk adjustment for observed severity is unable to show a benefit for a prompt admission strategy. However, we argue that unobserved differences in baseline risk are part of a clinician’s bedside assessment, and our risk adjustment is likely to be incomplete. We exploit the random variation in critical care strain to additionally allow for unobserved differences in baseline severity. We show that prompt critical care admission is less likely for patients referred during times of strain. Using this natural experimental set-up, we found that prompt admission reduces 90-day mortality without reaching statistical significance. For the subgroup recommended for critical care, the estimated reduction in 90-day mortality is larger and statistically significant. There is a large amount of uncertainty in the precise estimate of these effects, but the weight of evidence favours a prompt admission strategy, especially where the bedside assessor recommends critical care. The prompt admission strategy might confer a survival advantage in two distinct ways: explicitly be delivering critical care to a particular patient more promptly; or implicitly, by increasing the opportunity for critical care in the population. Of note, half of the deaths in the first week occur without admission to critical care, and death without critical care is more common when strain is greater.

: explicitly be delivering critical care to a particular patient more promptly; or implicitly, by increasing the opportunity for critical care in the population. Of note, half of the deaths in the first week occur without admission to critical care, and death without critical care is more common when strain is greater. It is not easy to separate out these two effects. Although previous studies compare the prompt to the ‘delayed’ admission, the delay is defined with hindsight which creates exclusion and survivor bias [6–13]. The real bedside decision is to ‘admit now’ or to ‘watch and wait’. Direct patient-level randomisation to evaluate a ‘prompt’ versus a ‘watchful waiting’ strategy is ethically challenging. Recent attempts have resorted to cluster randomisation of a strategy to increase referrals to critical care (specifically in the elderly and focussing on access not timing) [23]. This increased access but also increased severity of illness leading to the same issues with adjusting for unobserved differences in baseline severity [24].

have resorted to cluster randomisation of a strategy to increase referrals to critical care (specifically in the elderly and focussing on access not timing) [23]. This increased access but also increased severity of illness leading to the same issues with adjusting for unobserved differences in baseline severity [24]. Our attempt to minimise treatment selection bias (confounding by indication) relies on the assumption that critical strain acts as a natural randomisation event. We defend this assumption in three ways. Firstly, critical care occupancy at a specific time is unpredictable. It is the net result of all the decisions that lead to admission, discharge and death in a complex, maybe chaotic, system. Secondly, we have included controls where occupancy might be associated with baseline severity (i.e. increased strain during the winter). Thirdly, although strain might translate as a crowding effect thereby impairing the quality of care within the ICU, the evidence for this is conflicting [25, 26]. Moreover, crowding could not possibly affect the outcomes of the two thirds of patients never admitted to critical care. We additionally observed that patients assessed when there were no available beds deteriorated more prior to admission. This suggests a causal pathway that is independent of any effect of crowding after admission.

ding could not possibly affect the outcomes of the two thirds of patients never admitted to critical care. We additionally observed that patients assessed when there were no available beds deteriorated more prior to admission. This suggests a causal pathway that is independent of any effect of crowding after admission. An important limitation of IV analyses is that they are notoriously weak, and, without very large sample sizes, there is a risk of not detecting a true difference when one exists [27, 28]. This may be understood if we consider IV analysis as a randomised controlled trial but with poor compliance. Critical care strain (at the time of the bedside assessment) is the random coin toss and prompt admission is the treatment randomised. However, clinical teams may find ways to deliver prompt admission even at times of high strain (perhaps by accelerating discharge or flexing staffing). We saw 92 patients admitted promptly even though there were no beds at the instant of bedside assessment. Imperfect compliance requires either large samples or large effect sizes to achieve significance. Of note, our effect size was larger for the most unwell patients, and only then achieved significance at the 5% threshold.

92 patients admitted promptly even though there were no beds at the instant of bedside assessment. Imperfect compliance requires either large samples or large effect sizes to achieve significance. Of note, our effect size was larger for the most unwell patients, and only then achieved significance at the 5% threshold. Our finding of harm from reduced or delayed access to critical care does not stand in isolation. At a patient level, reducing exposure to critical care by premature discharge rather than delayed admission has also increased mortality [29]. At a population level, expanding critical care capacity through an increase in funding during health service reforms is similarly associated with an improvement in outcomes [30]. The one previous study (five hospitals, 749 patients) of prompt admission to critical care that also used the watchful-waiting cohort as controls also found benefit for prompt admission [31]. We also need to acknowledge that in an observational study of this size, there are limitations in the quality of the data recorded. We did not perfectly capture all admissions to critical care in the study database, and we must assume that a proportion of referrals were also missed. However, we tested our findings by raising threshold for judging data capture to 90% so that the median proportion of eligible admissions was 97%. We found no consistent difference in any result other than a fall in precision as the quality threshold increased, and the sample size inevitably fell.

als were also missed. However, we tested our findings by raising threshold for judging data capture to 90% so that the median proportion of eligible admissions was 97%. We found no consistent difference in any result other than a fall in precision as the quality threshold increased, and the sample size inevitably fell. Aspects of the study stand independent of these limitations. Regardless of the effect of prompt admission to critical care, we have identified a cohort of hospital patients at very high risk. This risk is heavily front-loaded, and the window for intervention is short. The bedside assessment is an effective but imperfect triage tool, as the mortality in those initially refused admission is high. Given that we already excluded patients with treatment limitations, it is of concern that nearly half of these early deaths occur without a trial of critical care. A substantial proportion of patients recommended for critical care are not offered a bed, and this proportion increases when capacity is limited. Although expanding critical care bed numbers may help, supply quickly saturates this expensive resource [32]. Given that the benefit of prompt critical care is unlikely to be equal for all referrals, the challenge now is to better prioritise. Interestingly, this same approach has also been highlighted as crucial to both patients and the public [33]. Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary material 1 (DOCX 55 kb) Supplementary material 2 (DOCX 7702 kb)

A substantial proportion of patients recommended for critical care are not offered a bed, and this proportion increases when capacity is limited. Although expanding critical care bed numbers may help, supply quickly saturates this expensive resource [32]. Given that the benefit of prompt critical care is unlikely to be equal for all referrals, the challenge now is to better prioritise. Interestingly, this same approach has also been highlighted as crucial to both patients and the public [33]. Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary material 1 (DOCX 55 kb) Supplementary material 2 (DOCX 7702 kb) This article was originally published under a CC BY-NC 4.0 license, but has now been made available under a CC BY 4.0 license. The PDF and HTML versions of the paper have been modified accordingly. Electronic supplementary material The online version of this article (10.1007/s00134-018-5148-2) contains supplementary material, which is available to authorized users. A correction to this article is available online at https://doi.org/10.1007/s00134-018-5254-1. Change history 6/8/2018 This article was originally published under a CC BY-NC 4.0 license, but has now been made available under a CC BY 4.0 license. The PDF and HTML versions of the paper have been modified accordingly. Acknowledgements Sheila Harvey (Clinical Trials Manager), Rahi Jahan (Research Associate), Sarah Power (Statistician), Emma Walmsley (Research Associate): Intensive Care National Audit and Research Centre, Napier House, 24 High Holborn, London, WC1 V 6AZ.

This article was originally published under a CC BY-NC 4.0 license, but has now been made available under a CC BY 4.0 license. The PDF and HTML versions of the paper have been modified accordingly. Acknowledgements Sheila Harvey (Clinical Trials Manager), Rahi Jahan (Research Associate), Sarah Power (Statistician), Emma Walmsley (Research Associate): Intensive Care National Audit and Research Centre, Napier House, 24 High Holborn, London, WC1 V 6AZ. Funding The study was funded by the Wellcome Trust, via a Clinical Research Training Fellowship (awarded to SH), and ICNARC (KR/DH), the London School of Hygiene and Tropical Medicine (SH/CS), the NIHR Clinical Research Network and the NIHR UCLH/UCL Biomedical Research Centre (MS). The funders of the study had no role in the study design; gathering, analysis and interpretation of these data; writing of the report; and decision to submit the report for publication. The corresponding author had full access to all data (including statistical reports and tables); takes responsibility for the integrity of these data and the accuracy of the analysis; and takes final responsibility for the decision to submit for publication. Compliance with ethical standards Conflicts of interest The authors declare that they have no conflict of interest.

Introduction Following the publication in the early 1980s of the Acute Physiology and Chronic Health Evaluation Score (APACHE [1]), Simplified Acute Physiology Score (SAPS [2]), and-some years later—APACHE II [3] systems, outcome prediction became an important topic among European intensivists. Ten years later, a new generation of these instruments was published: APACHE III [4], SAPS II [5], and Mortality Probability Model (MPM) II [6]. All of these newer systems were developed by using sophisticated statistical techniques in large multinational databases, and were found to perform better than their predecessors [7, 8]. The availability of such sophisticated methods for risk adjustment facilitated outcome research in critically ill patients, which became increasingly important over time. Risk adjustment systems now have a fixed place in critical care research for various purposes. At the patient level, the reporting of severity of illness and the use of risk-adjusted mortality rates to draw inferences from their results are a prerequisite for any study to be published. At the intensive care unit (ICU) level, observed-to-expected mortality ratios (or the use of direct standardisation techniques based on severity scores) have become standard for assessing the impact of ICU-related factors on outcome, such as the effects of organisation and management [9, 10].

any study to be published. At the intensive care unit (ICU) level, observed-to-expected mortality ratios (or the use of direct standardisation techniques based on severity scores) have become standard for assessing the impact of ICU-related factors on outcome, such as the effects of organisation and management [9, 10]. However, a series of studies assessing the performance of risk adjustment systems unveiled a lack of prognostic performance of these systems: In most cases, lack of calibration was evident over several subgroups of patients, often accompanied by an underestimation of mortality in low-risk patients and an overestimation in high-risk patients. This pattern was observed for all published outcome prediction models in several countries [11, 12, 13, 14, 15, 16, 17, 18] and seemed to be worsening over time [19]. For this reason, several researchers tried to improve the prognostic performance of various systems through recalibration, using one of two possible approaches. A level 1 customization requires calculation of a new equation for the prediction of hospital mortality (without changing the weights of the constituent variables). A level 2 customization involves a reweighting of each variable contained in the model. Although recalibration was able to improve prognostic accuracy in some cases [13, 14], it generally did not solve the various problems inherent in the models.

al mortality (without changing the weights of the constituent variables). A level 2 customization involves a reweighting of each variable contained in the model. Although recalibration was able to improve prognostic accuracy in some cases [13, 14], it generally did not solve the various problems inherent in the models. These problems can be classified as either user-, patient-, or model-dependent. User-dependent problems include differences in the definitions and application criteria [20, 21]. Patient-dependent problems are mainly shifts in the baseline characteristics of the populations over time [22]: age distribution, distribution of illnesses, and the development of new treatments, all of which affect prognosis. Model-dependent problems have many different causes, such as the lack of important prognostic variables (e.g., diagnostic information [4, 23]) or the presence, location and aetiology of infection [24, 25, 26]. Confounding variables and statistically wrong assumptions [9, 27] also distort performance results. If recalibration is not sufficient to improve the performance of the prognostic model, the only alternative is to develop a new model that takes into account the results of studies done since the original model was developed. This means incorporating missing variables that have been shown to affect outcome, minimizing problems with the application of the model, and reducing the possibility of other confounders.

y alternative is to develop a new model that takes into account the results of studies done since the original model was developed. This means incorporating missing variables that have been shown to affect outcome, minimizing problems with the application of the model, and reducing the possibility of other confounders. The objective of the SAPS 3 project was to cope with the above-stated problems by developing a new model for improved risk adjustment in critically ill patients. Another important goal was to make the new model available free of charge for use in the scientific community. In the SAPS 3 study (which took place at the end of 2002), data about risk factors and outcomes in an international multicentric cohort of critically ill patients were prospectively collected so that a high-quality database would be available for further analysis of the associations between risks and outcomes in our patients.

SAPS 3 study (which took place at the end of 2002), data about risk factors and outcomes in an international multicentric cohort of critically ill patients were prospectively collected so that a high-quality database would be available for further analysis of the associations between risks and outcomes in our patients. Materials and methods Project Organization The SAPS 3 project was conducted by the SAPS 3 Outcomes Research Group. The project was endorsed by the European Society for Intensive Care Medicine (ESICM, http://www.esicm.org) and conducted in cooperation with the Section on Health Services Research and Outcome of the ESICM. The SAPS 3 Outcomes Research Group consists of a project coordinator and a steering group. The steering group was responsible for the scientific conduct and consistency of the project. An additional advisory board integrated further scientists with special expertise who were asked for comments on the scientific content and for help in conducting the project. The complete board lists can be found in Appendix D of the Electronic Supplementary Material (ESM).

ific conduct and consistency of the project. An additional advisory board integrated further scientists with special expertise who were asked for comments on the scientific content and for help in conducting the project. The complete board lists can be found in Appendix D of the Electronic Supplementary Material (ESM). During the data collection phase, a coordination and communications centre (CCC) was installed. The CCC was responsible for the management and control of the project. This included the administration of all project tasks and implementation of actions and activities as necessary; communication between project partners (e.g., centres, researchers and institutions) through sampling and distribution of necessary information; and pooling and administration of the data provided by project participants. In addition, the CCC provided almost around-the-clock service to answer urgent questions and resolve problems during the phase of data collection. In each country, a country coordinator was responsible for operational management and direct communication with the participating ICUs in that country, including giving specific help when necessary. The country coordinator was responsible for ensuring completion of the various tasks required of the participating ICUs. The list of country coordinators can be found in Appendix E of the ESM.

management and direct communication with the participating ICUs in that country, including giving specific help when necessary. The country coordinator was responsible for ensuring completion of the various tasks required of the participating ICUs. The list of country coordinators can be found in Appendix E of the ESM. At the ICU level, an ICU coordinator was responsible for local activities, such as obtaining approval from the local ethics or data-protection committees where applicable. In addition, the ICU coordinator was responsible for supervising the daily data collection, problem management, controlling the completeness of the data, data quality control, training medical and nonmedical staff for data collection, management of the data, and transmission of the data to the CCC or country coordinator. The list of ICU coordinators can be found in Appendix F of the ESM.

aily data collection, problem management, controlling the completeness of the data, data quality control, training medical and nonmedical staff for data collection, management of the data, and transmission of the data to the CCC or country coordinator. The list of ICU coordinators can be found in Appendix F of the ESM. Data collection Patient data were recorded by using either online data collection software (provided by iMDsoft, Tel Aviv, Israel) or the SAPS 3 stand-alone database system (provided by the CCC). The latter software used a Microsoft Access database (Microsoft Corporation, Redmond, WA, USA) for data storage and needed no Internet connection for data entry. Both systems maintained a variety of plausibility controls to ensure the quality of the recorded data. Each variable was precisely defined before the start of data collection (see Appendix C of the ESM). Detailed definitions of the variables were available to participants in both paper and electronic form. To facilitate plausibility checking, each variable was assigned a probability range, encompassing the range of probable values for that variable. In addition, a range of possible values (storage range) for that variable was defined (e.g., for FiO2, no values <21% or >100% could be accepted). Thus, formal plausibility controls in the software systems were used wherever possible and ensured the maximum of data quality checking during data collection.

hat variable. In addition, a range of possible values (storage range) for that variable was defined (e.g., for FiO2, no values <21% or >100% could be accepted). Thus, formal plausibility controls in the software systems were used wherever possible and ensured the maximum of data quality checking during data collection. Participants who could not use one of the two software options were allowed to record the data on paper forms and submit them to the CCC (n=26 ICUs). Patient data were then entered into the SAPS 3 stand-alone software system and thus checked for plausibility. In cases of uncertainty, ICU coordinators were contacted for clarification. In addition, each ICU received a questionnaire with detailed questions about ICU structures and about resources available in other areas of the hospital. Data were collected at ICU admission, on days 1, 2 and 3, and on the last day of the ICU stay. Data from the day of admission (aside from sociodemographic data such as age and sex) were categorized into different levels: (i) data about the condition of the patient before ICU admission, such as chronic conditions and medical diseases; (ii) data about the patient’s condition at ICU admission, such as the reason for admission, infection at admission, and surgical status; and (iii) data about the patient’s physiologic derangement at ICU admission. These data were collected within an hour before or after ICU admission.

c conditions and medical diseases; (ii) data about the patient’s condition at ICU admission, such as the reason for admission, infection at admission, and surgical status; and (iii) data about the patient’s physiologic derangement at ICU admission. These data were collected within an hour before or after ICU admission. On the following days of the ICU stay, further information was collected: severity of illness, as measured by the SAPS II [5]; number and severity of organ dysfunction, as measured by the Sequential Organ Failure Assessment (SOFA) [28]; length of ICU and hospital stay; and outcome data, including vital status at ICU and hospital discharge. All patients were subjected to mandatory follow-up until hospital discharge, but not longer than 90 days after ICU admission. Patients still remaining in the hospital at 90 days were at that time classified as being “still in the hospital”.

l stay; and outcome data, including vital status at ICU and hospital discharge. All patients were subjected to mandatory follow-up until hospital discharge, but not longer than 90 days after ICU admission. Patients still remaining in the hospital at 90 days were at that time classified as being “still in the hospital”. To record diagnoses, a three-level system was used. (i) An acute medical disease was recorded for all patients, independent of surgical status, i.e., the acute (or acute on chronic) disease that best explained the ICU admission. If the reason for ICU admission was infectious disease, then this was recorded. (ii) Surgical status at admission and the anatomic site of surgery were recorded for all patients undergoing surgery during the hospital stay before ICU admission. (iii) A concrete reason for admission had to be selected. At least one reason for admission was required, but several selections were possible (one within each organ system). If no other reason was present, at least “basic and observational care” had to be selected.

urgery during the hospital stay before ICU admission. (iii) A concrete reason for admission had to be selected. At least one reason for admission was required, but several selections were possible (one within each organ system). If no other reason was present, at least “basic and observational care” had to be selected. All participants received detailed documentation of patient- and ICU-based data items as well as a detailed description of the data collection process. Moreover, specific forms to check the completeness of the patient-based documentation were provided. Additionally, a training session for ICU coordinators was organised at the 15th Annual Congress of the ESICM in Barcelona, Spain, before the start of data collection. Throughout the project, the project website (http://www.saps3.org) provided all necessary information. In addition, the CCC was available to answer questions by email, fax and phone. Data were to be collected from all consecutively admitted patients between 14 October and 15 December 2002. ICUs with a high number of beds (and thus also admissions) could stop patient enrolment after contributing 100 patients. Database Data were collected and pooled by the CCC. The final database file was then imported into the SAS system, Version 8e (SAS Institute Inc., Cary, USA). Data cleaning was accomplished through the application of a variety of additional plausibility controls and cross-checking of information between redundant data fields.

e Data were collected and pooled by the CCC. The final database file was then imported into the SAS system, Version 8e (SAS Institute Inc., Cary, USA). Data cleaning was accomplished through the application of a variety of additional plausibility controls and cross-checking of information between redundant data fields. A total of 22,791 admissions were recorded in the 309 participating ICUs during the study period. For patients who were admitted more than once (n=1,455), only the first admission was included, giving 21,336 admitted patients. Patients who were <16 years of age (n=628), those without ICU admission or discharge data (n=1,074), and those with records that lacked an entry in the field “ICU outcome” (n=57) were excluded. The Basic SAPS 3 Cohort thus comprises 19,577 patients from 307 ICUs. For the development of a predictive model for hospital mortality as outcome, patients with a missing entry in the field of “vital status at hospital discharge” (n=2,540) or an entry of “still in the hospital” at the end of the follow-up period (n=253) were further excluded. The SAPS 3 Hospital Outcome Cohort thus comprises 16,784 patients from 303 ICUs. Because the study was an observational study and no additional interventions were performed, the need for informed consent was waived by the institutional review board. Each ICU, however, was made responsible for obtaining local permissions as necessary.

For the development of a predictive model for hospital mortality as outcome, patients with a missing entry in the field of “vital status at hospital discharge” (n=2,540) or an entry of “still in the hospital” at the end of the follow-up period (n=253) were further excluded. The SAPS 3 Hospital Outcome Cohort thus comprises 16,784 patients from 303 ICUs. Because the study was an observational study and no additional interventions were performed, the need for informed consent was waived by the institutional review board. Each ICU, however, was made responsible for obtaining local permissions as necessary. Data quality Recorded data were evaluated for completeness of the documentation and reliability. Interrater quality control was performed through rescoring of the data from day 0 (the day of ICU admission) for three randomly selected patients in each ICU. From the rescored data, kappa coefficients and intra-class correlation coefficients were calculated, as appropriate. Availability of the variables necessary to calculate the SAPS II was used as an indicator for the completeness of the data.

0 (the day of ICU admission) for three randomly selected patients in each ICU. From the rescored data, kappa coefficients and intra-class correlation coefficients were calculated, as appropriate. Availability of the variables necessary to calculate the SAPS II was used as an indicator for the completeness of the data. Statistical analysis Statistical analysis was performed using the SAS system, version 8e (SAS Institute Inc., Cary, NC, USA). A P value of <0.05 was considered significant. Unless otherwise specified, results are expressed as median and interquartile ranges (quartile). Observed-to-expected (O/E) mortality ratios were calculated by dividing the number of observed deaths per group by the number of expected deaths per group (as predicted by the SAPS II). To test for statistical significance, we calculated 95% confidence intervals (CI) according to the method described by Hosmer and Lemeshow [29]. The Hosmer-Lemeshow goodness-of-fit Ĥ-statistic and Ĉ-statistic [30] were used to evaluate the calibration of the SAPS II. Discrimination was tested by measuring the area under the receiver operating characteristic (aROC) curve, as described by Hanley and McNeil [31]. Reliability of data collection was analysed using K-statistics or intra-class correlation coefficients, as appropriate. Statistical methods used for the development of the predictive model are described in Part 2 of this report.

the receiver operating characteristic (aROC) curve, as described by Hanley and McNeil [31]. Reliability of data collection was analysed using K-statistics or intra-class correlation coefficients, as appropriate. Statistical methods used for the development of the predictive model are described in Part 2 of this report. Results Data quality Four hundred eighty-three rescored patients could be identified and linked to their original counterparts (2.5% of admitted patients). Data quality was found to be excellent, with the majority of coefficients being >0.85. Only two of the more than 50 tested variables had coefficients <0.80 (body weight, 0.79; positive end-expiratory pressure, 0.72), and only one was <0.70 (leukocytes [maximum], 0.57). For a detailed list of coefficients see Table E1 in the ESM. Data completeness was also found to be satisfactory, with 1 [0–3] SAPS II parameter missing per patient. Description of ICUs The Basic SAPS 3 cohort includes 307 ICUs from 35 countries. On average each ICU contributed 50 (27–78) patients to the cohort. To assess heterogeneity of results between different geographic regions, seven regions were defined: Australasia, Central and South America, Central and Western Europe, Eastern Europe, North America, Northern Europe, and Southern Europe and Mediterranean countries. The allocation of countries to these regions can be seen from Table E10 of the ESM.

sults between different geographic regions, seven regions were defined: Australasia, Central and South America, Central and Western Europe, Eastern Europe, North America, Northern Europe, and Southern Europe and Mediterranean countries. The allocation of countries to these regions can be seen from Table E10 of the ESM. Seventy percent of the participating ICUs identified themselves as mixed medical-surgical (Table E2, ESM). Roughly half of the ICUs (46%) were located in university-affiliated or teaching hospitals. Eighty-four percent of ICUs (n=258) reported having a full-time medical director, and 272 (88.6%) reported having a full-time nursing director. On weekdays, 76.6% of ICUs reported having an intensivist available on the ICU 24 hours per day, whereas 6.2% had an intensivist available in the hospital. In 12.1% of ICUs, the intensivist was at home, on-call, during the daytime. During weekends, this proportion did not change much (74.3%, 5.5%, and 15.0% on the ICU, in the hospital, and on-call, respectively). None of the participating ICUs reported having no intensivists available during night or weekend shifts.

ital. In 12.1% of ICUs, the intensivist was at home, on-call, during the daytime. During weekends, this proportion did not change much (74.3%, 5.5%, and 15.0% on the ICU, in the hospital, and on-call, respectively). None of the participating ICUs reported having no intensivists available during night or weekend shifts. Description of patients The Basic SAPS 3 Cohort comprises 19,577 patients admitted to participating ICUs during the study period. More than 70% of patients were admitted from the same hospital as the ICU, with operating rooms, emergency departments and normal wards contributing most of the patients (Table 1). Almost two thirds of the admissions were classified as unplanned. The mean age of patients was 60.0±17.7 years (Fig. 1), and 39.2% were female. Comorbidities were recorded in 65% of admitted patients, with arterial hypertension, chronic obstructive pulmonary disease, and chronic heart failure being the most frequent (Table E3, ESM).Table 1 ICU admission data for the two cohorts (Basic cohort: SAPS 3 basic cohort; HO cohort: SAPS 3 Hospital Outcome Cohort; n: number of patients)

recorded in 65% of admitted patients, with arterial hypertension, chronic obstructive pulmonary disease, and chronic heart failure being the most frequent (Table E3, ESM).Table 1 ICU admission data for the two cohorts (Basic cohort: SAPS 3 basic cohort; HO cohort: SAPS 3 Hospital Outcome Cohort; n: number of patients) Basic cohort HO cohort n % n % Number of patients 19,577 16,784 100.0 Gender Female 7,678 39.2 6,610 39.4 Male 11,881 60.7 10,161 60.5 Missing 18 0.1 13 0.1 Age, years (median, quartiles) 63 49-74 64 49-74 Origin Home 2,810 14.4 2,343 14.0 Same hospital 13,926 71.1 12,063 71.9 Chronic care facility 74 0.4 64 0.4 Public place 519 2.7 432 2.6 Other hospital 2,125 10.9 1,791 10.7 Other 80 0.4 59 0.4 Missing 43 0.2 32 0.2 Intra-hospital location before ICU admission Emergency room 5,419 27.7 4,630 27.6 Intermediate care unit/High dependency unit 562 2.9 475 2.8 Operating room 7,537 38.5 6,449 38.4 Other 552 2.8 413 2.5 Other ICU 698 3.6 611 3.6 Recovery room 482 2.5 400 2.4 Ward 3,411 17.4 3,036 18.1 Missing 916 4.7 770 4.6 ICU admission status Planned admission 6,750 34.5 5,598 33.4 Unplanned admission 12,338 63.0 10,801 64.4 Missing 489 2.5 385 2.3 Acute Infection at ICU admission No infection 15,254 77.9 12,968 77.3 Clinically improbable/colonization 342 1.7 298 1.8 Clinically probable/documented 2,761 14.1 2,422 14.4 Microbiologically documented 1206 6.2 1,083 6.5 Missing 13 0.1 13 0.1 Surgical status No surgical procedure 8,437 43.1 7,305 43.5 Scheduled surgery 6,800 34.7 5,700 34.0 Emergency surgery 3,321 17.0 2,930 17.5 Missing 1,019 5.2 849 5.1 Fig. 1 Age distribution and associated mortality. The figure shows the age distribution of the Basic SAPS 3 Cohort (n=19,577) and the corresponding ICU mortality rates for each stratum. Columns: Number of patients as percentages of the whole cohort; squares: ICU mortality rates for the corresponding stratum

5.2 849 5.1 Fig. 1 Age distribution and associated mortality. The figure shows the age distribution of the Basic SAPS 3 Cohort (n=19,577) and the corresponding ICU mortality rates for each stratum. Columns: Number of patients as percentages of the whole cohort; squares: ICU mortality rates for the corresponding stratum Cardiovascular, respiratory and neurologic diseases were the most frequent organ-specific reasons for admission (Table E4, ESM). The acute medical diseases necessitating ICU admission included a broad spectrum of diagnoses (Table E5, ESM). Approximately one half of the patients underwent surgery before ICU admission, with abdominal, cardiac and vascular surgery being the most frequent procedures (Table E6, ESM). Regarding discharge details (Table 2), it is notable that a high percentage of patients were discharged unplanned (8.15%), i.e., without at least a 12-hour planning window. 15.2% of patients from the SAPS 3 Basic cohort died within the ICU. As can be seen from Table 3, patient cohorts differed significantly between regions. Both, ICU and hospital mortality rates exhibited a broad spectrum between ICUs: hospital mortality was on average 28% (17–42%) in the SAPS 3 Hospital outcome cohort.Table 2 ICU discharge and outcome data for the two cohorts (Basic cohort: SAPS 3 basic cohort; HO cohort: SAPS 3 Hospital Outcome Cohort; n: number of patients; ICU LOS: ICU length of stay; IMCU/HDU: intermediate care unit/high dependency unit; Q1, Q3: lower and upper interquartile range, respectively)

l outcome cohort.Table 2 ICU discharge and outcome data for the two cohorts (Basic cohort: SAPS 3 basic cohort; HO cohort: SAPS 3 Hospital Outcome Cohort; n: number of patients; ICU LOS: ICU length of stay; IMCU/HDU: intermediate care unit/high dependency unit; Q1, Q3: lower and upper interquartile range, respectively) Basic cohort HO cohort n % n % Number of patients 19,577 100.0 16,784 100.0 ICU LOS, days (median, quartiles) 2 1–6 2 1–6 ICU discharge—destination Home 438 2.2 361 2.2 Same hospital 14,946 76.3 12,477 74.3 Other hospital 1,029 5.3 852 5.1 Missing 3,164 16.2 3,094 18.4 Intrahospital discharge Emergency room 58 0.3 50 0.3 IMCU/HDU 2,222 11.4 1,873 11.2 Other 303 1.5 257 1.5 Other ICU 583 3.0 479 2.9 Recovery room 306 1.6 218 1.3 Ward 12,250 62.6 10,291 61.3 Missing 3,855 19.7 3,616 21.5 ICU discharge—status Planned discharge 14,872 76.0 12,262 73.1 Unplanned discharge 1,595 8.1 1,467 8.7 Missing 3,110 15.9 3,055 18.2 Risk adjustment SAPS II score (median, Q1–Q3) 30 20–42 31 21–43 SOFA score (median, Q1–Q3) 9 6–11 9 6–11 Outcome ICU mortality (%) 15.2 17.7 Table 3 ICU admission and discharge data for the seven defined geographic regions (SAPS 3 Basic Cohort; n=19,577)

anned discharge 1,595 8.1 1,467 8.7 Missing 3,110 15.9 3,055 18.2 Risk adjustment SAPS II score (median, Q1–Q3) 30 20–42 31 21–43 SOFA score (median, Q1–Q3) 9 6–11 9 6–11 Outcome ICU mortality (%) 15.2 17.7 Table 3 ICU admission and discharge data for the seven defined geographic regions (SAPS 3 Basic Cohort; n=19,577) Australasia Central & South America Eastern Europe Central and Western Europe Northern Europe Southern Europe and Mediterranean countries North America Number of patients 2,235   2,540   1,084   4,712   355   7,854   797 Females, % 38.0   44.6   42.2   40.1   42.5   36.9   38.0 Age, years (median, quartiles) 59 45–71 31 46–74 62 48–72 65 52–75 66 51–76 64 59–74 64 49–75 SAPS II score (median, Q1-Q3) 28 19–40 30 20–42 27 18–43 29 21–40 35 25–46 32 22–44 29 20–39 SOFA score (median, Q1-Q3) 8 6–10 8 6–11 9 7–11 8 5–11 9 7–11 9 7–11 9 6–11 ICU mortality, % 12.7   17.4   16.9   10.8   20.6   18.1   8.5

years (median, quartiles) 59 45–71 31 46–74 62 48–72 65 52–75 66 51–76 64 59–74 64 49–75 SAPS II score (median, Q1-Q3) 28 19–40 30 20–42 27 18–43 29 21–40 35 25–46 32 22–44 29 20–39 SOFA score (median, Q1-Q3) 8 6–10 8 6–11 9 7–11 8 5–11 9 7–11 9 7–11 9 6–11 ICU mortality, % 12.7   17.4   16.9   10.8   20.6   18.1   8.5 Performance of the SAPS II The performance of the original SAPS II model [5] (using data from the first 24 hours) was tested in the SAPS 3 Hospital Outcome Cohort (n=16,784). Discrimination was good with an aROC of 0.83 (95% CI: 0.824–0.838). SAPS II showed underestimation of hospital mortality: The O/E ratio of the overall cohort was 1.08 (1.06–1.10). O/E ratios significantly differed between regions: from 0.86 (0.81–0.91) for Central and Western Europe to 1.32 (1.25–1.38) for Central and South America, with four out of the seven defined regions exhibited O/E ratios significantly different from 1 (Table E7, ESM). Calibration, as assessed by the Hosmer-Lemeshow Ĥ + Ĉ statistics, was poor for the overall cohort: Ĥ 227.21, Ĉ 184.70; both p<0.0001; This lack of calibration was present for all tested subgroups except for the region of North America (see Table E7, ESM). Discussion To the best of our knowledge, the SAPS 3 study is the largest prospective epidemiologic multicentre, multinational study conducted in health services and outcomes research in intensive care medicine to date.

Performance of the SAPS II The performance of the original SAPS II model [5] (using data from the first 24 hours) was tested in the SAPS 3 Hospital Outcome Cohort (n=16,784). Discrimination was good with an aROC of 0.83 (95% CI: 0.824–0.838). SAPS II showed underestimation of hospital mortality: The O/E ratio of the overall cohort was 1.08 (1.06–1.10). O/E ratios significantly differed between regions: from 0.86 (0.81–0.91) for Central and Western Europe to 1.32 (1.25–1.38) for Central and South America, with four out of the seven defined regions exhibited O/E ratios significantly different from 1 (Table E7, ESM). Calibration, as assessed by the Hosmer-Lemeshow Ĥ + Ĉ statistics, was poor for the overall cohort: Ĥ 227.21, Ĉ 184.70; both p<0.0001; This lack of calibration was present for all tested subgroups except for the region of North America (see Table E7, ESM). Discussion To the best of our knowledge, the SAPS 3 study is the largest prospective epidemiologic multicentre, multinational study conducted in health services and outcomes research in intensive care medicine to date. The project was first intended to focus on Europe because it was believed such a strategy would produce a more homogeneous cohort of patients, which would in turn provide a more stable reference line for further comparisons. This idea was discussed during several investigator meetings and finally abandoned—first, because interest from outside Europe was enormous: 39% of ICUs that registered for the project were located outside Europe. The SAPS 3 board members thus agreed that such a high level of interest should not be ignored. Second, some investigators questioned whether a concentration on European ICUs would be successful in reducing heterogeneity anyway. Provision of intensive care in Europe is extremely variable, with enormous differences in severity of illness, provision of treatments and mortality from north to south and from west to east [32, 33].

investigators questioned whether a concentration on European ICUs would be successful in reducing heterogeneity anyway. Provision of intensive care in Europe is extremely variable, with enormous differences in severity of illness, provision of treatments and mortality from north to south and from west to east [32, 33]. For these reasons ICUs from regions outside Europe were invited to participate. Our results prove that we were right in our assumptions: First, one can easily see that the four European regions (as defined in our study) are hardly comparable: severity of illness as measured by the SAPS II varied from 27 to 35 points, and ICU mortality ranged from 10.8 to 20.6%—almost a doubling of mortality figures (Table 3). Second, almost a third of the patient cohort (28.5%) was contributed from regions outside Europe.

ons (as defined in our study) are hardly comparable: severity of illness as measured by the SAPS II varied from 27 to 35 points, and ICU mortality ranged from 10.8 to 20.6%—almost a doubling of mortality figures (Table 3). Second, almost a third of the patient cohort (28.5%) was contributed from regions outside Europe. Although the decision to accept ICUs worldwide probably increased the heterogeneity of our sample, it also allowed the SAPS 3 database to better reflect important differences in patients’ and health care systems’ baseline characteristics that are known to affect outcome. These include, for example, different genetic makeups, different styles of living or a heterogeneous distribution of major diseases within different regions, as well as issues such as access to the health care system in general and to intensive care in particular, or differences in availability and use of major diagnostic and therapeutic measures within the ICUs [32, 34]. Although the integration of ICUs outside Europe and the U.S. surely increased it’s representativeness, it must be acknowledged, that the extent to which the SAPS 3 database reflects case-mix on ICUs worldwide cannot be determined yet.

bility and use of major diagnostic and therapeutic measures within the ICUs [32, 34]. Although the integration of ICUs outside Europe and the U.S. surely increased it’s representativeness, it must be acknowledged, that the extent to which the SAPS 3 database reflects case-mix on ICUs worldwide cannot be determined yet. It should additionally be noted that allocation of countries to regions does not always follow geographic borders (Table 3; see also Table E10 in the ESM). Partitioning of the sample was done to adjust for some of the above-stated differences between different populations and to develop a system that uses several different reference lines to compare ICUs on a similar level. Thus, patients are not necessarily representative of their respective regions. To minimize possible seasonal influences, we chose late fall in the Northern Hemisphere for data collection. Thus, participants in both late fall/winter (Northern Hemisphere) and spring/summer (Southern Hemisphere) are represented in our cohort. A recent study [35] showed, moreover, that differences in seasonal mortality rates, at least in a sample of ICUs in the United Kingdom, were related to variations in case mix rather than to a specific impact of season on outcome.

orthern Hemisphere) and spring/summer (Southern Hemisphere) are represented in our cohort. A recent study [35] showed, moreover, that differences in seasonal mortality rates, at least in a sample of ICUs in the United Kingdom, were related to variations in case mix rather than to a specific impact of season on outcome. Performance of the SAPS II was, not surprisingly, found to be similar to that in previous studies: acceptable discrimination but lack of calibration. Possible reasons for this have already been alluded to in the Introduction. In contrast to previous studies, however, we found an underestimation of hospital mortality, which contradicts the rationale that the shifting in calibration is due only to the development of new and possibly better therapies and thus to better ICU performance [19]. Analyzing the various geographic regions provides evidence that the underestimation of hospital mortality by the SAPS II might be partially attributable to the composition of the cohort: SAPS 3 is the first large international study on severity of illness systems to include patients from all continents. South America, for example, where provision of intensive care is much more limited than it is in Europe or North America, contributed extensively to the patient cohort. High O/E ratios have already been reported for this continent [36] and are probably linked to the limited availability of resources.

om all continents. South America, for example, where provision of intensive care is much more limited than it is in Europe or North America, contributed extensively to the patient cohort. High O/E ratios have already been reported for this continent [36] and are probably linked to the limited availability of resources. Data quality was one of our major concerns. Completeness of the documentation was found to be satisfactory: The amount of missing data is in fact smaller than reported from previous cohort studies on severity of illness systems [11, 12, 16]. With respect to reliability, intraclass-correlation coefficients and kappa coefficients were generally similar to or even better than those found in previous studies, showing a high degree of interrater agreement (see Table E1 in ESM) [37, 38]. We did, however, experience problems with the cohort of rescored patients: First, we had to exclude all rescored patients for whom the original counterpart was also excluded due to the application of any of the exclusion criteria. Second, in some cases the original patient identification was either missing or documented in such a way that a unique allocation was not possible. Both of these exclusions reduced the number of rescored patients available for analysis.

nterpart was also excluded due to the application of any of the exclusion criteria. Second, in some cases the original patient identification was either missing or documented in such a way that a unique allocation was not possible. Both of these exclusions reduced the number of rescored patients available for analysis. Two strategies to build up a cohort are available: first, to recruit only patients who meet well-documented inclusion criteria (such as documented vital status at hospital discharge) or, second, to document all patients and then exclude patients based on a predefined set of exclusion criteria. For the SAPS 3 study we chose the second option—to form two different cohorts—because we needed to provide a basic cohort for all further analyses of the SAPS 3 database. Since some studies will focus on different outcomes (e.g., ICU outcome rather than hospital outcome), we decided to use missing ICU outcome (and not hospital outcome) as an exclusion criterion for the basic cohort.

rent cohorts—because we needed to provide a basic cohort for all further analyses of the SAPS 3 database. Since some studies will focus on different outcomes (e.g., ICU outcome rather than hospital outcome), we decided to use missing ICU outcome (and not hospital outcome) as an exclusion criterion for the basic cohort. A possible limitation of the SAPS 3 database is that vital status at hospital discharge was not available for all admitted patients. Despite several efforts from the CCC and sufficient time to allow for a close follow-up, we did not succeed to receive all hospital outcomes documented. Recording of hospital outcome (or later outcomes) still poses major problems for ICUs in European and non-European hospitals, either because of technical problems or possibly because of data security algorithms in the hospitals. Exclusion of these patients did, however, not affect major criteria, such as geographic representation, ICU admission or discharge data, co-morbidities, or the distribution of the reasons for admission (Tables 1 and 2). We conclude that the SAPS 3 database is within the above discussed limits of high quality and reflects the heterogeneity of current intensive care provision. As such, it provides an excellent basis for the development of a new risk adjustment system. Electronic Supplementary Material (PDF 794 KB)

A possible limitation of the SAPS 3 database is that vital status at hospital discharge was not available for all admitted patients. Despite several efforts from the CCC and sufficient time to allow for a close follow-up, we did not succeed to receive all hospital outcomes documented. Recording of hospital outcome (or later outcomes) still poses major problems for ICUs in European and non-European hospitals, either because of technical problems or possibly because of data security algorithms in the hospitals. Exclusion of these patients did, however, not affect major criteria, such as geographic representation, ICU admission or discharge data, co-morbidities, or the distribution of the reasons for admission (Tables 1 and 2). We conclude that the SAPS 3 database is within the above discussed limits of high quality and reflects the heterogeneity of current intensive care provision. As such, it provides an excellent basis for the development of a new risk adjustment system. Electronic Supplementary Material (PDF 794 KB) Acknowledgements The SAPS 3 project was endorsed in June 2002 by the European Society of Intensive Care Medicine (ESICM). It received support from the Austrian Centre for Documentation and Quality Assurance in Intensive Care Medicine (ASDI), the Portuguese Society of Intensive Care (SPCI), and the Medical Economics and Research Centre (MERCS) in Sheffield, U.K.. An unrestricted educational grant from Merck Sharp & Dohme Portugal to the SPCI made possible the installation of the CCC in Lisbon. iMDsoft (Tel Aviv, Israel) developed and provided the Internet-based data collection software free of charge.

I), and the Medical Economics and Research Centre (MERCS) in Sheffield, U.K.. An unrestricted educational grant from Merck Sharp & Dohme Portugal to the SPCI made possible the installation of the CCC in Lisbon. iMDsoft (Tel Aviv, Israel) developed and provided the Internet-based data collection software free of charge. Statistical analysis was supported by a grant from the Fund of the Austrian National Bank, Project # 10995 ONB. Statistical analysis was further supported by Lorenz Dolanski and Johanna Einfalt, both from the Department of Medical Statistics, University of Vienna, Vienna, Austria. Our thanks to the participants from all over the world who dedicated a significant amount of their time and effort to this project, proving that it is still possible to conduct a worldwide academic study. The SAPS 3 is primarily their study, and we are deeply indebted to them for the honour of conducting it.

Introduction One of the crucial steps in the evaluation of risk-adjusted outcomes is the choice of the reference database for estimating adequate reference lines for the analyzed variables. For the SAPS 3 to reflect the standard of practices and outcome in intensive care at the beginning of the 21st century, we decided to collect data from a large sample of intensive care units (ICUs) worldwide. Other models have restricted data collection to large ICUs in Europe or North America—SAPS II [1], MPM II [2], APACHE II [3] and APACHE III [4], a strategy that minimizes the heterogeneity of the sample but restricts the generalization of the results.

t data from a large sample of intensive care units (ICUs) worldwide. Other models have restricted data collection to large ICUs in Europe or North America—SAPS II [1], MPM II [2], APACHE II [3] and APACHE III [4], a strategy that minimizes the heterogeneity of the sample but restricts the generalization of the results. At the statistical level, there is also a need for change, in order to take into account the hierarchic nature of our data [5, 6]. Current general outcome prediction models do not consider the existence of clinical and nonclinical factors, aggregated at the ICU level, that can have an important impact on prognosis. Instead, they assume that these factors are either not important or are randomly distributed throughout large samples and that the variation between ICUs is small. This assumption is not likely to be borne out at the ICU level for either nonclinical factors (e.g. organization and management, organizational culture) or clinical factors (e.g. clinical management, diagnostic and therapeutic strategies). If the variation between ICUs is not negligible, it will compromise the stability of the equations used to compute predicted mortality. Furthermore, the published models consider the relation between performance and severity of illness to be constant, and that may not be the case, since performance can vary within ICUs according to the severity of illness of the patients [7, 8]. To overcome this problem, we chose to adopt a new strategy for the development of the SAPS 3 score and to apply statistical modelling techniques that control for the clustering of patients within ICUs instead of assuming the independence of observations. Conceptually, the SAPS 3 admission core comprises the following parts:

ercome this problem, we chose to adopt a new strategy for the development of the SAPS 3 score and to apply statistical modelling techniques that control for the clustering of patients within ICUs instead of assuming the independence of observations. Conceptually, the SAPS 3 admission core comprises the following parts: First, the SAPS 3 ADMISSION SCORE, represented by the arithmetic sum of three subscores, or boxes:Box I: What we know about the patient characteristics before ICU admission: age, previous health status, co-morbidities, location before ICU admission, length of stay in the hospital before ICU admission, and use of major therapeutic options before ICU admission. Box II: What we know about the circumstances of ICU admission: reason(s) for ICU admission, anatomic site of surgery (if applicable), planned or unplanned ICU admission, surgical status and infection at ICU admission. Box III: What we know about the presence and degree of physiologic derangement at ICU admission (within 1 h before or after admission). Second, the SAPS 3 PROBABILITY OF DEATH during a certain period of time (in the case of the main model, the probability of death at hospital discharge). Given our objective of evaluating not only individual patient outcome but also the effectiveness of ICU practices, we focused the model on data available at ICU admission or shortly thereafter. This model will be completely open and available free of any direct or indirect charges to the scientific community.

Second, the SAPS 3 PROBABILITY OF DEATH during a certain period of time (in the case of the main model, the probability of death at hospital discharge). Given our objective of evaluating not only individual patient outcome but also the effectiveness of ICU practices, we focused the model on data available at ICU admission or shortly thereafter. This model will be completely open and available free of any direct or indirect charges to the scientific community. Methods and statistical analysis Primary variable selection Based on the SAPS 3 Hospital Outcome Cohort as described in Part 1 of this report, continuous predictive variables were categorized in mutually exclusive categories based on smoothed curves such as LOWESS [9], showing the univariate dependence of hospital mortality on the predictive variables. Classes of categorical variables were also collapsed according to their univariate hospital mortality levels using multidimensional tables and clinical judgment as appropriate, depending on the nature of the data. Additively, regression trees (MART) [10] were applied to check the cutoffs. Missing values were coded as the reference or “normal” category for each variable. When dual data collection was used—maximum and minimum values recorded during a certain time period—missing maximum values of a variable were replaced by the minimum, if documented, and vice versa. Some regression imputations were performed if noticeable correlations to available values could be exploited. For a detailed description of data collection and handling, see Part 1 of this report.

d first- and second-level customization strategies [23–25]. However, the value of these techniques is for the moment limited, usually because they are based on regional databases [24–26] that prevent extrapolation to other settings; moreover, their superiority in even the regional setting still needs to be established. Finally, the SAPS 3 conceptually dissociates evaluation of the individual patient from evaluation of the ICU. Thus, for individual patient assessment, the system separates the relative contributions to prognosis of (i) chronic health status and previous therapy, (ii) the circumstances related to ICU admission, and (iii) the presence and degree of physiologic dysfunction. It is interesting to note that one half of the predictive power of the model is achieved with Box I, i.e., with the information that is available before ICU admission. The prognostic capabilities of the model can be further improved by 22.5% by using data related to the circumstances of the ICU admission (Box II), and by another 27.5% by the incorporation of physiologic data (Box III). These numbers are different from those published by Knaus et al. [4] but are based on what we have learned in the last years about prognostic determinants in the critically ill patient.

ing a certain time period—missing maximum values of a variable were replaced by the minimum, if documented, and vice versa. Some regression imputations were performed if noticeable correlations to available values could be exploited. For a detailed description of data collection and handling, see Part 1 of this report. Selection of variables was done according to their association with hospital mortality, together with expert knowledge and definitions used in other severity of illness scoring systems. The objective of using this combination of techniques rather than regression-based criteria alone was to reach a compromise between over-sophistication of the model and knowledge from sources beyond the sample with its specific case mix and ICU characteristics. Cross validation For being able to cross-validate the model, we randomly extracted five roughly equal-sized parts based on number of patients from the database, as suggested previously [11]. In a second approach, partitioning was based on ICUs and not on patients. It was thus possible to run the model-building procedure five times in each of the two approaches, each time taking four parts of the sample as a development set and the remaining one as the validation set. This allowed to estimate the variability of prediction resulting from the construction process of the prognostic score. A further check of the stability of the predictions was made by partitioning the sample according to major patient characteristics, such as surgical status and infection status.

as the validation set. This allowed to estimate the variability of prediction resulting from the construction process of the prognostic score. A further check of the stability of the predictions was made by partitioning the sample according to major patient characteristics, such as surgical status and infection status. The quality of predictions in the validation sets was assessed by looking at the goodness-of-fit in terms of the p values for the Hosmer-Lemeshow tests Ĉ and Ĥ [13] and the discriminative capability of the models by the use of the area under the receiver operating characteristic (aROC) curve [14, 15]. Another criterion to judge the appropriateness of the model was the fit in certain subsamples, defined according to major patient typologies [16]. Reducing model complexity To reduce the complexity of the model classes, we concentrated on logistic regression. In the first step a stepwise logistic regression was used to identify the significant predictors in each of the five subsamples. A threshold of 0.01 for the p value was generally applied for inclusion in the model to separate irrelevant predictors [12]. At this stage we also evaluated if interactions among these predictors would influence results. Interactions, however, did not make a valuable contribution for the prediction.

he five subsamples. A threshold of 0.01 for the p value was generally applied for inclusion in the model to separate irrelevant predictors [12]. At this stage we also evaluated if interactions among these predictors would influence results. Interactions, however, did not make a valuable contribution for the prediction. Significant predictors (n=70) were in a second step entered into a logistic regression model. The criterion for a predictor to enter the model was homogeneity across the five model-building processes: in principle, predictors should enter the model in all five development sets, but depending on the frequency of the predictor in the samples, the magnitude of the effect, and medical reasoning, some predictors were included if they appeared in the model in at least three subsamples. An example is the presence of Acquired Immunodeficiency Syndrome (AIDS): it was selected as a comorbidity in only 81 patients (0.48%), but the mortality—without controlling for other variables—in these patients was 42%. By taking all the above steps to identify the set of predictors, although deliberately not using any formal numeric criterion, we reduced the complexity of the model to minimize the amount of overfitting: This process resulted in 61 item classes (representing 20 variables) remaining in the final model.

atients was 42%. By taking all the above steps to identify the set of predictors, although deliberately not using any formal numeric criterion, we reduced the complexity of the model to minimize the amount of overfitting: This process resulted in 61 item classes (representing 20 variables) remaining in the final model. Using the parameter estimates from the logistic regression as starting values, a multilevel model was applied in the next step, using patient characteristics as fixed effects and ICUs as a random effect. Estimates were again calculated for the five development sets (for both, patient and ICU -based development subsamples). At this stage it was checked if rounding of coefficients (which allows for an easier manual computation of the score) would influence results, which was found not to be the case. Consequently, this was the approach chosen for the final construction of the SAPS 3 admission score sheet. The stability of the processes of variable selection and reducing complexity was further checked by bootstraping with replacement the total sample 100 times, both at patient level and at ICU level.

a related to the circumstances of the ICU admission (Box II), and by another 27.5% by the incorporation of physiologic data (Box III). These numbers are different from those published by Knaus et al. [4] but are based on what we have learned in the last years about prognostic determinants in the critically ill patient. For performance evaluation, several reference lines should be used, with risk-adjusted mortality in different patient typologies and not only O/E mortality ratios at hospital discharge in the overall ICU population [27]. The results of the SAPS 3 study showing that different O/E ratios were observed in different regions of the world should be explored further, since, apart from regional differences in case mix (not taken into account by the model), they can also be related to regional variations in structures and organization of acute medical care, to different lifestyles (e.g., prevalence of obesity, or alcohol and tobacco use) and/or—though less likely—to genetic differences among populations.

At this stage it was checked if rounding of coefficients (which allows for an easier manual computation of the score) would influence results, which was found not to be the case. Consequently, this was the approach chosen for the final construction of the SAPS 3 admission score sheet. The stability of the processes of variable selection and reducing complexity was further checked by bootstraping with replacement the total sample 100 times, both at patient level and at ICU level. Predicting hospital mortality After this step was completed, a shrinking power transformation was applied. This procedure uses log-transformation of the score to reduce the influence of extreme score values (outliers) on the mortality prediction. For this purpose, the SAPS 3 score and the transformed log (SAPS 3 + g) scores were used to predict hospital mortality. Conventional logistic regression was used in the evaluation of this step because of convergence problems for the corresponding multilevel model in a few subsamples. The best shrinkage model then was selected (excluding the trivial model with the SAPS 3 score as the single predictor) by checking which of the terms in the model contributed best to the prediction and was moreover stable over the respective validation sets and specific subsamples. This procedure was applied on both, patient and ICU -based subsamples. After finishing these steps of cross-validation, the final estimates for the selected predictors of the SAPS 3 score as well as the selected shrinkage procedure were then calculated from the total sample of patients.

Predicting hospital mortality After this step was completed, a shrinking power transformation was applied. This procedure uses log-transformation of the score to reduce the influence of extreme score values (outliers) on the mortality prediction. For this purpose, the SAPS 3 score and the transformed log (SAPS 3 + g) scores were used to predict hospital mortality. Conventional logistic regression was used in the evaluation of this step because of convergence problems for the corresponding multilevel model in a few subsamples. The best shrinkage model then was selected (excluding the trivial model with the SAPS 3 score as the single predictor) by checking which of the terms in the model contributed best to the prediction and was moreover stable over the respective validation sets and specific subsamples. This procedure was applied on both, patient and ICU -based subsamples. After finishing these steps of cross-validation, the final estimates for the selected predictors of the SAPS 3 score as well as the selected shrinkage procedure were then calculated from the total sample of patients. To arrive at the customised models for each major geographic region, specific customised equations were calculated, relating, by logistic regression, the transformed log (SAPS 3 + g) admission scores computed as described above to the vital status at hospital discharge. This process allows both the intercept and the slope of the curve relating the SAPS 3 admission score to change across different regions. The goodness-of-fit of these equations was evaluated by means of the same methodology used for the global sample.

mputed as described above to the vital status at hospital discharge. This process allows both the intercept and the slope of the curve relating the SAPS 3 admission score to change across different regions. The goodness-of-fit of these equations was evaluated by means of the same methodology used for the global sample. SAS for Windows, version 8.02 (SAS Institute Inc., Cary, NC, USA) and MLwiN version 1.10.0007 (Centre for Multilevel Modelling, Institute of Education, London, UK) and the R Software Package (http://www.r-project.org) were used for the development of the model. Results Based on the methodology described, 20 variables were selected for the SAPS 3 admission score (Tables 1 and 2):Five variables for evaluating Box I: age, co-morbidities, use of vasoactive drugs before ICU admission, intrahospital location before ICU admission, and length of stay in the hospital before ICU admission; Five variables for evaluating Box II: reason(s) for ICU admission, planned/unplanned ICU admission, surgical status at ICU admission, anatomic site of surgery, and presence of infection at ICU admission and place acquired; Ten variables for evaluating Box III: lowest estimated Glasgow coma scale, highest heart rate, lowest systolic blood pressure, highest bilirubine, highest body temperature, highest creatinine, highest leukocytes, lowest platelets, lowest hydrogen ion concentration (pH), and ventilatory support and oxygenation. Table 1 SAPS 3 admission scoresheet—Part 1

Ten variables for evaluating Box III: lowest estimated Glasgow coma scale, highest heart rate, lowest systolic blood pressure, highest bilirubine, highest body temperature, highest creatinine, highest leukocytes, lowest platelets, lowest hydrogen ion concentration (pH), and ventilatory support and oxygenation. Table 1 SAPS 3 admission scoresheet—Part 1 Box I 0 3 5 6 7 8 9 11 13 15 18 Age, years <40 >=40<60 >60<70 >=70<75 >=75<80 >=80 Co-Morbidities Cancer therapy 2) Chron, HF (NYHA IV), Haematological cancer 3),4) Cirrhosis, AIDS 3) Cancer 5) Length of stay before ICU admission, days 1) <14 >=14<28 >=28 Intra-hospital location before ICU admission Emergency room Other ICU Other 6) Use of major therapeutic options before ICU admission Vasoactive drugs Box II 0 3 4 5 6 9 ICU admission: Planned or Unplanned Unplanned Reason(s) for ICU admission please see Part 2 of the scoresheet Surgical status at ICU admission Scheduled surgery No surgery 7) Emergency surgery Anatomical site of surgery please see Part 2 of the scoresheet Acute infection at ICU admission Nosocomial 8) Respiratory 9) Box III 15 13 11 10 8 7 5 3 2 0 2 4 5 7 8 Estimated Glasgow Coma Scale (lowest), points 3–4 5 6 7–12 >=13 Total bilirubine (highest), mg/dL <2 >=2<6 >=6 Total bilirubine (highest), µmol/L <34.2 >=34.2<102.6 >=102.6 Body temperature (highest), Degrees Celsius <35 >=35 Creatinine (highest), mg/dL <1.2 >=1.2<2 >=2<3.5 >=3.5 Creatinine (highest), µmol/L 3–4 5 6 <106.1 >=106.1<176.8 >=176.8<309.4 >=309.4 Heart rate (highest), beats/minute <120 >=120<160 >=160 Leukocytes (highest), G/L <15 >=15 Hydrogen ion concentration (lowest), pH <=7.25 >7.25 Plateletes (lowest), G/L <20 >=20<50 >=50<100 >=100 Systolic blood pressure (lowest), mm Hg <40 >=40<70 >=70<120 >=120 Oxygenation 10), 11) PaO2/FiO2<100 and MV PaO2/FiO2>=100 and MV PaO2<60 and no MV PaO2>=60 and no MV The definition for all variables can be found in detail in Appendix C of the ESM. For names and abbreviations which are differing from those in the ESM, explanations are given below. Generally, it should be noted that no mutually exclusive conditions exist for the following fields: Comorbidities, Reasons for ICU admission, and Acute infection at ICU admission. Thus, if a patient has more than one condition listed for a specific variable, points are assigned for all applicable combinations.

n below. Generally, it should be noted that no mutually exclusive conditions exist for the following fields: Comorbidities, Reasons for ICU admission, and Acute infection at ICU admission. Thus, if a patient has more than one condition listed for a specific variable, points are assigned for all applicable combinations. 1 This variable is calculated from the two data fields: ICU Admission date and time—Hospital admission date and time (see Appendix C of the ESM) 2 Cancer Therapy refers to the data definitions in Appendix C of the ESM: Co-Morbidities: Chemotherapy, Immunosupression other, Radiotherapy, Steroid treatment 3 If a patient has both conditions he/she gets double points. 4 Chronic HF (NYHA IV)/Haematological cancer refer both to the data definitions in Appendix C of the ESM: Co-Morbidities: Chronic heart failure class IV NYHA, Haematological cancer. 5 Cancer refers to the data definitions in Appendix C of the ESM: Co-Morbidities: Metastatic cancer. 6 Other refers to the data definitions in Appendix C of the ESM: Intra-hospital location before ICU admission: Ward, Other. 7 No surgery refers to the data definitions in Appendix C of the ESM: Surgical Status at ICU Admission: Patient not submitted to surgery. 8 Nosocomial refers to the data definitions in Appendix C of the ESM: Acute infection at ICU admission—Acquisition: Hospital-acquired. 9 Respiratory refers to the data definition in Appendix C of the ESM: Acute infection at ICU admission—Site: Lower respiratory tract: Pneumonia, Lung asbcess, other.

7 No surgery refers to the data definitions in Appendix C of the ESM: Surgical Status at ICU Admission: Patient not submitted to surgery. 8 Nosocomial refers to the data definitions in Appendix C of the ESM: Acute infection at ICU admission—Acquisition: Hospital-acquired. 9 Respiratory refers to the data definition in Appendix C of the ESM: Acute infection at ICU admission—Site: Lower respiratory tract: Pneumonia, Lung asbcess, other. 10 PaO2, FIO2 refer to the data definitions in Appendix C of the ESM: Arterial oxygen partial pressure (lowest), Inspiratory oxygen concentration. 11 MV refers to the data definition in Appendix C of the ESM: Ventilatory support and mechanical ventilation. Table 2 SAPS 3 admission scoresheet – Part 2

9 Respiratory refers to the data definition in Appendix C of the ESM: Acute infection at ICU admission—Site: Lower respiratory tract: Pneumonia, Lung asbcess, other. 10 PaO2, FIO2 refer to the data definitions in Appendix C of the ESM: Arterial oxygen partial pressure (lowest), Inspiratory oxygen concentration. 11 MV refers to the data definition in Appendix C of the ESM: Ventilatory support and mechanical ventilation. Table 2 SAPS 3 admission scoresheet – Part 2 Box II – continued ICU admission 12) 16 Reason(s) for ICU admission Cardiovascular: Rhythm disturbances 13) –5 Neurologic: Seizures 13) –4 Cardiovascular: Hypovolemic hemorrhagic shock, Hypovolemic non hemorrhagic shock. / Digestive: Acute abdomen, Other 3) 3 Neurologic: Coma, Stupor, Obtuned patient, Vigilance disturbances, Confusion, Agitation, Delirium 4 Cardiovascular: Septic shock. / Cardiovascular: Anaphylactic shock, mixed and undefined shock 3) 5 Hepatic: Liver failure 6 Neurologic: Focal neurologic deficit 7 Digestive: Severe pancreatitis 9 Neurologic: Intracranial mass effect 10 All others 0 Anatomical site of surgery Transplantation surgery: Liver, Kidney, Pancreas, Kidney and pancreas, Transplantation other –11 Trauma – Other, isolated: (includes Thorax, Abdomen, limb); Trauma – Multiple –8 Cardiac surgery: CABG without valvular repair –6 Neurosurgery: Cerebrovascular accident 5 All others 0 12)Every patient gets an offset of 16 points for being admitted (to avoid negative SAPS 3 Scores). 13) If both reasons for admission are present, only the worse valve (–4) is scored.

Box II – continued ICU admission 12) 16 Reason(s) for ICU admission Cardiovascular: Rhythm disturbances 13) –5 Neurologic: Seizures 13) –4 Cardiovascular: Hypovolemic hemorrhagic shock, Hypovolemic non hemorrhagic shock. / Digestive: Acute abdomen, Other 3) 3 Neurologic: Coma, Stupor, Obtuned patient, Vigilance disturbances, Confusion, Agitation, Delirium 4 Cardiovascular: Septic shock. / Cardiovascular: Anaphylactic shock, mixed and undefined shock 3) 5 Hepatic: Liver failure 6 Neurologic: Focal neurologic deficit 7 Digestive: Severe pancreatitis 9 Neurologic: Intracranial mass effect 10 All others 0 Anatomical site of surgery Transplantation surgery: Liver, Kidney, Pancreas, Kidney and pancreas, Transplantation other –11 Trauma – Other, isolated: (includes Thorax, Abdomen, limb); Trauma – Multiple –8 Cardiac surgery: CABG without valvular repair –6 Neurosurgery: Cerebrovascular accident 5 All others 0 12)Every patient gets an offset of 16 points for being admitted (to avoid negative SAPS 3 Scores). 13) If both reasons for admission are present, only the worse valve (–4) is scored. An estimation of the variability of the coefficients in the overall sample and in the five disjoint subsamples is given in Table E8 of the Electronic Supplementary Material (ESM), together with their respective coefficients (unrounded and rounded) and p values. The SAPS 3 admission score can thus, in theory, vary from a minimum of 0 points to a maximum of 217 points. The distribution of the SAPS 3 admission score in our sample is presented in Fig. 1. The minimum value observed was 5, and the maximum value was 124, with a mean of 49.9±16.6 (mean ± SD) and a median of 48 (38–60). The highest explanatory power came from Box I, with Box II and Box III being less important for the outcome; the three boxes represent 50%, 22.5% and 27.5%, respectively, of the total Nagelkerke’s R-Square. The relationship between the SAPS 3 and vital status at hospital discharge is given by the equation:

48 (38–60). The highest explanatory power came from Box I, with Box II and Box III being less important for the outcome; the three boxes represent 50%, 22.5% and 27.5%, respectively, of the total Nagelkerke’s R-Square. The relationship between the SAPS 3 and vital status at hospital discharge is given by the equation: Logit = −32.6659 +ln(SAPS 3 score +20.5958) ×7.3068 and the probability of mortality by the equation: Probability of death = elogit/(1+elogit).Fig. 1 Distribution of the SAPS 3 admission score in the SAPS 3 database

48 (38–60). The highest explanatory power came from Box I, with Box II and Box III being less important for the outcome; the three boxes represent 50%, 22.5% and 27.5%, respectively, of the total Nagelkerke’s R-Square. The relationship between the SAPS 3 and vital status at hospital discharge is given by the equation: Logit = −32.6659 +ln(SAPS 3 score +20.5958) ×7.3068 and the probability of mortality by the equation: Probability of death = elogit/(1+elogit).Fig. 1 Distribution of the SAPS 3 admission score in the SAPS 3 database The relationship between the SAPS 3 admission score and the respective probability of death in the hospital is described in Fig. 2. Overall, no combined discrepancy between observed and expected outcomes across all of the strata was outside sampling variability as demonstrated a Hosmer-Lemeshow goodness-of-fit test Ĥ of 10.56 (p=0.39) and a Hosmer-Lemeshow goodness-of-fit test Ĉ of 14.29 (p=0.16) (Figs. 3, 4 and Table E9, ESM). The overall discriminatory capability of the model, as measured by aROC curve, was 0.848. The goodness-of-fit according to major patient typologies (surgical status, trauma, and infection) can be found in Table 3. Calibration and discrimination presented differences across different geographic areas: the best predictive results were achieved in patients from Northern Europe (observed-to-expected [O/E] mortality ratio 0.96 [0.83–1.09]) and the worst predictive results were obtained in patients from Central and South America (O/E mortality ratio, 1.30 [1.23–1.37]); see also Table 4 and Fig. 5 and Appendix B in the ESM.Fig. 2 Relationship between the SAPS 3 admission score and the respective probabilities of hospital mortality

y ratio 0.96 [0.83–1.09]) and the worst predictive results were obtained in patients from Central and South America (O/E mortality ratio, 1.30 [1.23–1.37]); see also Table 4 and Fig. 5 and Appendix B in the ESM.Fig. 2 Relationship between the SAPS 3 admission score and the respective probabilities of hospital mortality Fig. 3 Hosmer-Lemeshow goodness-of-fit test Ĉ in the overall sample. Predicted risk of hospital death, observed hospital mortality rate, and the corresponding number of patients per decile are shown. Columns: Number of patients; squares: mean SAPS 3-predicted mortality per decile; circles: mean observed mortality per decile Fig. 4 Hosmer-Lemeshow goodness-of-fit test Ĥ in the overall sample. Predicted risk of hospital death, observed hospital mortaliy rate, and the corresponding number of patients per decile are shown. Columns: Number of patients; squares: mean SAPS 3-predicted mortality per decile; circles: mean observed mortality per decile Table 3 Performance of the model across major patient typologies

Fig. 4 Hosmer-Lemeshow goodness-of-fit test Ĥ in the overall sample. Predicted risk of hospital death, observed hospital mortaliy rate, and the corresponding number of patients per decile are shown. Columns: Number of patients; squares: mean SAPS 3-predicted mortality per decile; circles: mean observed mortality per decile Table 3 Performance of the model across major patient typologies Patient characteristics GOF test Ĥ p GOF test Ĉ p O/E ratio 95% CI aROC Trauma patients 19.92 0.03 9.03 0.53 1.03 0.93–1.12 0.854 Non-operative admissionsa 14.86 0.14 17.8 0.06 1.01 0.98–1.04 0.825 Scheduled surgerya 11.5 0.32 27.39 <0.01 0.97 0.90–1.03 0.825 Emergency surgerya 4.97 0.89 12.88 0.23 1.00 0.95–1.05 0.809 No infectionb 8.57 0.57 14.77 0.14 1.00 0.97–1.02 0.846 Community-acquired infectionc 8.4 0.59 11.76 0.3 1.00 0.96–1.05 0.786 Hospital-acquired infectiond 15.21 0.12 7.11 0.72 1.02 0.97–1.07 0.77 GOF: Hosmer-Lemeshow goodness-of-fit; O/E: observed-to-expected mortality; CI: 95% confidence interval; aROC: area under receiver operating characteristic (curve) aNon-operative admissions, scheduled surgery emergency surgery: see data definitions appendix C, ESM bNo infection: Patients not infected at ICU admission cCommunity-acquired infection: Patients with community-acquired infection at ICU admission dHospital-acquired infection: Patients with hospital-acquired infection at ICU admission Table 4 Performance of the model in the global sample and in different geographic areas

aNon-operative admissions, scheduled surgery emergency surgery: see data definitions appendix C, ESM bNo infection: Patients not infected at ICU admission cCommunity-acquired infection: Patients with community-acquired infection at ICU admission dHospital-acquired infection: Patients with hospital-acquired infection at ICU admission Table 4 Performance of the model in the global sample and in different geographic areas Regions GOF test Ĥ p GOF test Ĉ p O/E ratio 95% CI aROC Australasia 15.25 0.12 8.09 0.62 0.92 0.85–0.99 0.839 Central, South America 78.01 <0.01 80.82 <0.01 1.30 1.23–1.37 0.855 Central, Western Europe 56.45 <0.01 47.89 <0.01 0.84 0.79–0.90 0.861 Eastern Europe 19.45 0.03 18.69 0.04 1.09 1.00–1.19 0.903 North Europe 2.44 0.99 2.34 0.99 0.96 0.83–1.09 0.814 Southern Europe, Mediterranean countries 14.18 0.16 20.78 0.02 1.02 0.98–1.05 0.834 North America 10.57 0.39 9.63 0.47 0.91 0.78–1.04 0.812 Global database 10.56 0.39 14.29 0.16 1 0.98–1.02 0.848 GOF: Hosmer-Lemeshow goodness-of-fit; O/E: observed-to-expected mortality; CI: 95% confidence interval; aROC: area under the receiver operating characteristic (curve). Fig. 5 Observed-to-expected (O/E) mortality ratios by region. Observed-to-expected (O/E) mortality ratios are shown by region. Bars indicate 95% confidence intervals

Regions GOF test Ĥ p GOF test Ĉ p O/E ratio 95% CI aROC Australasia 15.25 0.12 8.09 0.62 0.92 0.85–0.99 0.839 Central, South America 78.01 <0.01 80.82 <0.01 1.30 1.23–1.37 0.855 Central, Western Europe 56.45 <0.01 47.89 <0.01 0.84 0.79–0.90 0.861 Eastern Europe 19.45 0.03 18.69 0.04 1.09 1.00–1.19 0.903 North Europe 2.44 0.99 2.34 0.99 0.96 0.83–1.09 0.814 Southern Europe, Mediterranean countries 14.18 0.16 20.78 0.02 1.02 0.98–1.05 0.834 North America 10.57 0.39 9.63 0.47 0.91 0.78–1.04 0.812 Global database 10.56 0.39 14.29 0.16 1 0.98–1.02 0.848 GOF: Hosmer-Lemeshow goodness-of-fit; O/E: observed-to-expected mortality; CI: 95% confidence interval; aROC: area under the receiver operating characteristic (curve). Fig. 5 Observed-to-expected (O/E) mortality ratios by region. Observed-to-expected (O/E) mortality ratios are shown by region. Bars indicate 95% confidence intervals For a more precise estimation of the probability of death in the hospital across the different geographic regions, specific customised equations were calculated (Table 5). This customised approach allows each ICU to choose its own reference line for the prediction of hospital mortality: either the overall SAPS 3 hospital mortality sample or its own regional subsample. This approach can be supplemented in the future by customised equations at the country level if data are available and if a more precise estimation of outcome in a specific setting is needed. The overall goodness-of-fit of these customised equations for each region is presented in Table 5. A complete list of the number of patients and the respective O/E mortality ratios by country, according to the global equation and the regional equations, are presented in Tables E10 and E11 of the ESM, with point estimates varying at the global level from 0.68 (0.56–0.80) to 2.05 (1.27–2.82). Most O/E ratios are close to the identity line, as expected for a stable model.Table 5 Customized SAPS 3 admission equations for the different geographic areas

regional equations, are presented in Tables E10 and E11 of the ESM, with point estimates varying at the global level from 0.68 (0.56–0.80) to 2.05 (1.27–2.82). Most O/E ratios are close to the identity line, as expected for a stable model.Table 5 Customized SAPS 3 admission equations for the different geographic areas Area Equation GOF Ĥ p GOF Ĉ p O/E CI Australasia Logit=−22.5717 + ln (SAPS 3 score + 1) ×5.3163 10.43 0.40 2.20 0.99 1.00 0.93–1.07 Central, South America Logit=−64.5990 + ln (SAPS 3 score + 71.0599) ×13.2322 8.94 0.54 7.03 0.72 1.00 0.94–1.06 Central, Western Europe Logit=−36.0877 + ln (SAPS 3 score + 22.2655) ×7.9867 15.13 0.13 12.15 0.27 1.00 0.94–1.06 Eastern Europe Logit=−60.1771 + ln (SAPS 3 score + 51.4043) ×12.6847 10.13 0.43 7.12 0.71 1.00 0.92–1.08 North Europe Logit=−26.9065 + ln (SAPS 3 score + 5.5077) ×6.2746 3.45 0.97 2.22 0.99 1.00 0.86–1.14 Southern Europe, Mediterranean countries Logit=−23.8501 + ln (SAPS 3 score + 5.5708) ×5.5709 5.28 0.87 13.12 0.22 1.00 0.97–1.03 North America Logit=−18.8839 + ln (SAPS 3 score + 1) ×4.3979 4.22 0.93 4.47 0.92 1.00 0.86–1.14 GOF Ĥ: Hosmer-Lemeshow goodness-of-fit Ĥ test; GOF Ĉ: Hosmer-Lemeshow goodness-of-fit Ĉ test; p: respective p-values; O/E: observed-to-expected mortality ratio; CI: 95% confidence interval

5.28 0.87 13.12 0.22 1.00 0.97–1.03 North America Logit=−18.8839 + ln (SAPS 3 score + 1) ×4.3979 4.22 0.93 4.47 0.92 1.00 0.86–1.14 GOF Ĥ: Hosmer-Lemeshow goodness-of-fit Ĥ test; GOF Ĉ: Hosmer-Lemeshow goodness-of-fit Ĉ test; p: respective p-values; O/E: observed-to-expected mortality ratio; CI: 95% confidence interval Discussion We have presented the results of a large multicentric, multinational study aimed at updating the SAPS II model. This study was necessary for several reasons. First, the reference line used by SAPS II was derived from a database collected in the early 1990s; since that time, there have been changes in the prevalence of major diseases and in the availability and use of major diagnostic and therapeutic methods that are associated with a shift toward poor calibration of older models such as SAPS II and APACHE III [17, 18]. Second, SAPS II was developed from a database built exclusively from patients in Europe and North America. This sample may not be representative of the case mix and medical practices that constitute the reality of intensive care medicine in the rest of the world (e.g. Australasia or South America), where variability in structures and organization is probably related to outcome [19].

rom patients in Europe and North America. This sample may not be representative of the case mix and medical practices that constitute the reality of intensive care medicine in the rest of the world (e.g. Australasia or South America), where variability in structures and organization is probably related to outcome [19]. Third, since computation of predicted mortality is based on a reference database, the user should be able to choose between them, i.e., a global database, which provides a broader comparison at the potential cost of less relevance to local conditions, and a regional database, which provides a better comparison with ICUs in geographic proximity but at the cost of losing comparability with ICUs in other parts of the world. A third possibility could be added—a country-representative database—but such a database would raise the problem of whether the ICUs selected were representative of a certain country.

des a better comparison with ICUs in geographic proximity but at the cost of losing comparability with ICUs in other parts of the world. A third possibility could be added—a country-representative database—but such a database would raise the problem of whether the ICUs selected were representative of a certain country. Fourth, the development of computers in recent years has created easy access to strong computational power. One of the implications of this is that it is now possible to develop a new outcome prediction model, based on digital data acquisition and analysis, with minimal differences in definitions and application criteria. These advances were coupled with extensive automatic logical and error-checking capabilities and the availability of data collection manuals online. Moreover, developers of the SAPS 3 model could take advantage of computer-intensive methods of data selection and analysis, such as the use of additive partition trees and logistic regression with random effects. Several new statistical techniques have been used in recent years to allow a more stable prediction of outcome, such as genetic algorithms and artificial neural networks [20, 21], dynamic microsimulation techniques [22], and first- and second-level customization strategies [23–25]. However, the value of these techniques is for the moment limited, usually because they are based on regional databases [24–26] that prevent extrapolation to other settings; moreover, their superiority in even the regional setting still needs to be established.

ional differences in case mix (not taken into account by the model), they can also be related to regional variations in structures and organization of acute medical care, to different lifestyles (e.g., prevalence of obesity, or alcohol and tobacco use) and/or—though less likely—to genetic differences among populations. We would like to re-emphasize that the model presented here is based exclusively on data (including physiologic data) available within 1 h of ICU admission and calibrated for manual data acquisition; consequently, it should be expected to overestimate mortality when an automatic patient data management system with a high sampling rate is used [28, 29]. Limiting acquisition of physiologic data to the hour of ICU admission should minimise the impact of this factor when compared with models based on the most deranged data from the first 24 h after ICU admission, probably at the expense of a small decrease in the ROC curve, a greater sensitivity to the exact time point at which admission to ICU occurs, and therefore more reliant on the assumption that measured physiology alone (as opposed to changes in physiology) predict outcome. It also allows the prediction of mortality to be done before ICU interventions take place. This gives the SAPS 3 admission model a major advantage over existing systems, such as the SAPS II or the APACHE II and III, since all these systems can be affected by the so-called Boyd and Grounds effect: the occurrence of more abnormal physiologic values during the first 24 h in the ICU, leading to an increase in computed severity of illness and a corresponding increase in predicted mortality. These increases may, however, be due not to a greater intrinsic severity of illness of the patient but to the provision of suboptimal care in the first 24 h of ICU admission, when a stable patient may be allowed to deteriorate [30].

se in computed severity of illness and a corresponding increase in predicted mortality. These increases may, however, be due not to a greater intrinsic severity of illness of the patient but to the provision of suboptimal care in the first 24 h of ICU admission, when a stable patient may be allowed to deteriorate [30]. Further studies should be done of factors occurring after ICU admission that influence risk-adjusted mortality. We should keep however in mind that this approach comes with one potential pitfall: a possible decrease in the amount of data available for the computation of the model; also, the shorter time period for data collection can eventually increase the likelihood of missing physiological data and the reliance on the assumption that missing physiological data are normal. This effect should be small, considering the widespread availability of monitoring and point-of-case analysers. Having demonstrated the internal validity of the SAPS 3 admission model by the extensive use of cross-validation techniques, we should stress that external validation is also necessary. The fact that the overall database was not collected to be representative of the global case-mix (and especially the case-mix of specific regional areas or patient typologies such as specific diseases) should be empirically tested. Furthermore, the rate of deterioration of our estimates over time should be followed by the appropriate use of temporal validation, especially to avoid what Popovich called grade inflation [18].

cially the case-mix of specific regional areas or patient typologies such as specific diseases) should be empirically tested. Furthermore, the rate of deterioration of our estimates over time should be followed by the appropriate use of temporal validation, especially to avoid what Popovich called grade inflation [18]. The SAPS 3 system was developed to be used free of charge by the scientific community; no proprietary information regarding the scientific content is retained. All the coefficients needed for the computation of outcome probabilities are available in the published material. The SAPS 3 can even be computed manually, using a simple scoresheet, although it was designed to be integrated into computerised data acquisition and storage systems that allow the automatic check of the quality of the registered data.

or the computation of outcome probabilities are available in the published material. The SAPS 3 can even be computed manually, using a simple scoresheet, although it was designed to be integrated into computerised data acquisition and storage systems that allow the automatic check of the quality of the registered data. In conclusion, we can say that at the end of this stage of the project, we have been able to overcome some of the problems inherent in current risk-adjustment systems. We have minimized user-dependent problems through the publication of careful, detailed definitions and criteria for data collection [31]. We have also addressed the patient-dependent problems by expanding the reference database and making it more representative of reality, in order to include the maximum possible range of variations for patient-centred variables and resulting patient-centred outcomes. This approach was complemented by the development of specific customised equations for major areas of the world, allowing ICUs to choose a reference line for outcome prediction—the global database or the regional database for their own area.

riations for patient-centred variables and resulting patient-centred outcomes. This approach was complemented by the development of specific customised equations for major areas of the world, allowing ICUs to choose a reference line for outcome prediction—the global database or the regional database for their own area. Users of these models should keep in mind that benchmarking is a process of comparing an ICU with a reference population. The appropriate choice of reference population is difficult, and we cannot simply change it because the observed-to-predicted mortality rate is not the one we want. For this reason, the choice should depend on the objective of the benchmark: more precise estimation will need local or regional equations, developed from a more homogeneous case mix. A generalisable estimation will, on the other hand, need more global equations developed from a more representative case mix. Last but not least, we have successfully addressed some of the problems of prognostic model development, especially those related to the underlying statistical assumptions for the use of specific methods for selection and weighting of variables and the conceptual development of outcome prediction models. In the future, multi-level modelling with varying slopes (and not just random intercepts) might be able to give a better answer to researchers but for the moment they would make the models to complex to be managed outside a research environment. Electronic Supplementary Material (PDF 794 KB)

Last but not least, we have successfully addressed some of the problems of prognostic model development, especially those related to the underlying statistical assumptions for the use of specific methods for selection and weighting of variables and the conceptual development of outcome prediction models. In the future, multi-level modelling with varying slopes (and not just random intercepts) might be able to give a better answer to researchers but for the moment they would make the models to complex to be managed outside a research environment. Electronic Supplementary Material (PDF 794 KB) Acknowledgements The SAPS 3 project was endorsed in June 2002 by the European Society of Intensive Care Medicine (ESICM). It received support from the Austrian Centre for Documentation and Quality Assurance in Intensive Care Medicine (ASDI), the Portuguese Society of Intensive Care (SPCI), and the Medical Economics and Research Centre in Sheffield (UK). An unrestricted educational grant from Merck Sharp & Dohme Portugal to the SPCI allowed for the installation of the Coordination and Communication Centre in Lisbon. iMDsoft (Tel Aviv, Israel) developed and provided free of charge the Internet-based data collection software. Statistical analysis was supported by a grant from the Fund of the Austrian National Bank, Project # 10995 ONB. Statistical analysis was further supported by Lorenz Dolanski and Johanna Einfalt, both: Dept. of Medical Statistics, University of Vienna, Vienna, Austria.

Acknowledgements The SAPS 3 project was endorsed in June 2002 by the European Society of Intensive Care Medicine (ESICM). It received support from the Austrian Centre for Documentation and Quality Assurance in Intensive Care Medicine (ASDI), the Portuguese Society of Intensive Care (SPCI), and the Medical Economics and Research Centre in Sheffield (UK). An unrestricted educational grant from Merck Sharp & Dohme Portugal to the SPCI allowed for the installation of the Coordination and Communication Centre in Lisbon. iMDsoft (Tel Aviv, Israel) developed and provided free of charge the Internet-based data collection software. Statistical analysis was supported by a grant from the Fund of the Austrian National Bank, Project # 10995 ONB. Statistical analysis was further supported by Lorenz Dolanski and Johanna Einfalt, both: Dept. of Medical Statistics, University of Vienna, Vienna, Austria. Our thanks to the participants from all over the world who dedicated a significant amount of their time and effort to this project, proving that it is still possible to conduct a worldwide academic study. The SAPS 3 is primarily their study, and we are deeply indebted to them for the honour of conducting it.

Intensive Care Med (2005) 31:1345–1355 Table 1: Box III, the entry “Leukocytes (highest)” should have read “Leukocytes (lowest)”. Explanation: Patients with high “lowest” leukocytes values higher than 15,000 G/L presented with an excess risk which turned out to be significant. Although leukocytes values collected within one hour prior to one hour after ICU admission will not change dramatically it is necessary to clarify this issue. The same mistake occurred on p. 65 of the electronic supplementary material accompanying the article. The online version of the original article can be found at http://dx.doi.org/10.1007/s00134-005-2763-5.

Introduction High-frequency ventilation (HFV) has been compared with conventional mechanical ventilation (CMV) since the 1980s. In HFV, patients are ventilated with small tidal volumes, even smaller than the dead space of their airways, at high frequencies, normally between 5 and 10 Hz. Because HFV combines high mean airway pressures with small tidal volumes, this technique of ventilation has been regarded by some to be the most optimal form in patients with infant respiratory distress syndrome (IRDS), adult respiratory distress syndrome (ARDS), and other forms of severe lung disease [1].

5 and 10 Hz. Because HFV combines high mean airway pressures with small tidal volumes, this technique of ventilation has been regarded by some to be the most optimal form in patients with infant respiratory distress syndrome (IRDS), adult respiratory distress syndrome (ARDS), and other forms of severe lung disease [1]. The HFV has been extensively investigated in premature neonates with IRDS, a population specifically at risk for chronic lung disease (CLD). Unfortunately, the results of these studies were equivocal [2, 3]; thus, the question remains whether or not HFV better prevents CLD than conventional mechanical ventilation (CMV) in patients with severe lung disease. A significant number of meta-analyses have been performed to answer this question [4, 5, 6, 7, 8]. Pooled estimates of pulmonary outcomes failed to show clinically relevant differences among HFV and CMV [7]; however, significant heterogeneity existed between studies included in these meta-analyses. In a recent cumulative meta-analysis we identified improvements of the conventional treatment of IRDS and ventilation strategies applied in both HFV and CMV as important sources of heterogeneity [4]. These associations could be confounded by other explanatory variables. Although a meta-analysis may pool results from randomized trials, differences among trials will not be randomly or independently distributed. A meta-analysis constitutes an observational study of trials, subjected to bias inherent to observational research. In a meta-regression analysis it is possible to adjust for confounding covariates. A number of alternative hypotheses have been formulated to explain heterogeneity between trials: The observed regression of the cumulative relative risks to the level of unity was due to publication bias.

e on demand CXR was performed. Similar to the ICU physicians, the radiologist structurally interpreted these on demand CXRs for each patient (i.e., the radiologist ticked whether radiological abnormalities were absent or present and, if an abnormality was present, whether it was judged to be an “old” or “new” finding). If a predefined finding was unexpectedly found, then we determined whether any action was taken because of the new unexpected finding. To do this, two of us (M.G. and M.J.S.) and two independent observers carefully read the medical records, checked the patient data management system (Metavision, iMDsoft, Sassenheim, The Netherlands) and searched the hospital information system for the following: orders for sputum cultures or performance of a bronchoalveolar lavage for culture, or start of, or a change in, antimicrobial therapy in case of unexpected infiltrates on the CXR; repositioning of tubes in case of malposition of orotracheal tubes (ignoring planned extubations); ultrasound of the thorax in case of pleural effusion on the CXR, start or change in medication (diuretics); insertion of a pleural drain; and repositioning of devices in the case of malposition of medical devices other than orotracheal tubes (ignoring planned changes such as removal of intravenous lines). The observers were not involved in the daily care of the patients, and ICU physicians were not aware of this part of the observation. As a consequence, the clinical relevance of the predefined abnormalities could not be evaluated in some cases, specifically in cases of large atelectasis and severe pulmonary congestion, since start of physiotherapy, changes in levels of positive end-expiratory pressure, and the use of diuretics might have been triggered by other (clinical) findings.

erent to observational research. In a meta-regression analysis it is possible to adjust for confounding covariates. A number of alternative hypotheses have been formulated to explain heterogeneity between trials: The observed regression of the cumulative relative risks to the level of unity was due to publication bias. Use of the Sensormedics ventilator resulted in better results in HFV treated patients. A prolonged ventilation on CMV before initiating HFV treatment could reduce the benefits of HFV. In subgroups of more premature neonates with lower birth weight with a higher susceptibility for CLD, HFV could result in better pulmonary outcome. With outcome rates increasingly representing more severe disease, HFV could have an increasing advantage over CMV [9, 10]; therefore, we used meta-regression analysis to better estimate relative treatment effects through adjustments for factors that could explain trial heterogeneity. Methods Trials were included based on a previous meta-analysis that we had conducted [4]. The same search strategy, as well as the same inclusion and exclusion criteria as in our previous meta-analysis, were used for an update, yielding two more studies that could be included for this meta-regression analysis. Validity of studies was assessed by criteria published by Jadad et al. [11]. The validity was generally deemed as high with adequate allocation concealment in all trials. Blinding of treatment was not possible due to the nature of the interventions.

more studies that could be included for this meta-regression analysis. Validity of studies was assessed by criteria published by Jadad et al. [11]. The validity was generally deemed as high with adequate allocation concealment in all trials. Blinding of treatment was not possible due to the nature of the interventions. Data extraction was performed as has been reported in our previous meta-analysis. The following outcome measures were used: mortality, chronic lung disease (CLD) as defined by supplemental oxygen need or ventilator dependency at the age of 30–36 weeks post-menstrual. A number of explanatory variables were extracted as well: year of publication; type of ventilator used for HFV (Sensormedics 3100A ventilator versus other); ventilation strategies applied in the HFV and CMV treatment groups were obtained as previously described [4]; time on CMV before study initiation; gestational age and birth weight; and outcome rates in the control population were taken as proxy for baseline disease severity in the source population. The Sensormedics ventilator was singled out because previous research suggested better performance compared with other oscillator ventilators [2, 4].

tudy initiation; gestational age and birth weight; and outcome rates in the control population were taken as proxy for baseline disease severity in the source population. The Sensormedics ventilator was singled out because previous research suggested better performance compared with other oscillator ventilators [2, 4]. Statistical analysis All data were extracted according to the intention-to-treat principle. The number of patients surviving without chronic lung disease was subtracted from the total number of randomized patients in each treatment arm to calculate the composite outcome of death or CLD. To calculate the risk of CLD, the number of surviving patients was put in the denominator. Publication bias was assessed by visual appraisal of symmetry of funnel plots and performing rank tests. Smaller studies could show different results than larger studies which could suggest publication bias, but in fact was caused by systematic differences among studies; therefore, an analysis of publication bias stratified for ventilation strategies was performed to determine whether the observed association between the inverse of the standard error with the risk ratio was confounded by ventilation strategies. Meta-regression analysis was used to evaluate other hypotheses. The dependent variables, RR of CLD and RR of CLD or death, were natural log transformed to linearize the regression models. Individual studies were weighted by inverse variances of relative risks of outcomes of interest so that the more precise studies had more influence in the analysis. Firstly, linear regression analyses were applied to explanatory variables. Secondly, linear regression analyses with continuous covariates were conducted stratified by HLVS, LPVS, and use of surfactant. Finally, multivariable linear regression analyses were performed to calculate adjusted contributions of different explanatory variables of rivalling hypotheses to changes in RR. The relative effects of covariates were evaluated by relative risk ratios (RRR). A relative risk ratio quantifies the relative change in RR that is associated with a specified change of a covariate. For continuous variables the RRR was calculated for the ranges of minimum and maximum values of covariates that were reported in trials. For example, the RRR for year of publication was calculated by using the range between the publication year of the first year and the publication year of the last included trial.

ate. For continuous variables the RRR was calculated for the ranges of minimum and maximum values of covariates that were reported in trials. For example, the RRR for year of publication was calculated by using the range between the publication year of the first year and the publication year of the last included trial. The RRR for year of publication thus estimates the relative change in RR due to the difference in years of publication between the first and last trials. All analyses were conducted using SPSS 12.0.1 for Windows software (SPSS, Chicago, Ill.) and Excel (Microsoft, Redmond, Wash.). Results For the analyses 15 studies were available that specified either CLD in survivors or death or CLD as outcome measures [2, 3, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24]. In 11 trials a high frequency oscillatory ventilator was used [2, 3, 12, 13, 16, 17, 19, 20, 22, 23, 24], in 7 of these trials this was the SensorMedics ventilator [2, 12, 13, 17, 22, 23]. Two studies used a high-frequency jet ventilator [14, 15] and in two studies a high-frequency flow interrupter ventilator was used [18, 21]. In the HFV group a total of 1141 patients were included for the outcome of CLD with 373 events and a total of 1457 patients with 671 events for the outcome death or CLD. In the CMV group a total 1159 patients were reported for the outcome of CLD with 428 events and a total of 1473 patients with 730 events for the outcome death or CLD. A forest plot of these trials can be found in the Electronic Supplement.

vents and a total of 1457 patients with 671 events for the outcome death or CLD. In the CMV group a total 1159 patients were reported for the outcome of CLD with 428 events and a total of 1473 patients with 730 events for the outcome death or CLD. A forest plot of these trials can be found in the Electronic Supplement. Table 1 summarizes the main characteristics of the studies. The outcome of CLD was available in all studies but one [22]. Time on CMV to start of the study was not reported by Plavka et al. [17] and Craft et al. [21]. In only one study was surfactant not used as concomitant treatment [12]. A high lung volume strategy (HLVS) was used in all but two studies [14, 16]. A ventilation strategy in the CMV-treated patients that could qualify as lung protective (LPVS) was reported in the most recent 9 studies [2, 3, 18, 19, 20, 21, 22, 23, 24]. Studies were published over a range of 13 years. Other reported ranges of covariates were 8.7 h average time on CMV before start of study, 5 weeks average gestational age, and 0.65 kg average birth weight. These ranges were used to calculate RRRs. Two studies dominated the analyses by virtue of the weight they received in the analyses: Johnson et al. [3] and Courtney et al. [2] (together 69% for CLD and 73% for death or CLD as outcome). Table 1 Study characteristics

rage gestational age, and 0.65 kg average birth weight. These ranges were used to calculate RRRs. Two studies dominated the analyses by virtue of the weight they received in the analyses: Johnson et al. [3] and Courtney et al. [2] (together 69% for CLD and 73% for death or CLD as outcome). Table 1 Study characteristics Reference Year Time on CMV Age Birth weight SensorM HLVS LPVS Surf CLD lnRR Weight Death or CLD lnRR Weight [12] 1992 9.0 28 1.100 Y Y N N -1.29 0.01 -0.58 0.01 [13] 1996 3.0 31 1.500 Y Y N Y -0.67 0.04 -0.55 0.02 [14] 1996 7.2 27 0.950 N N N Y 0.02 0.01 -0.23 0.10 [15] 1997 8.0 27 1.020 N Y N Y -0.70 0.03 0.48 0.03 [16] 1998 1.0 28 1.100 N N N Y 0.00 0.00 0.31 0.00 [17] 1999 26 0.850 Y Y N Y -1.03 0.01 -0.74 0.01 [18] 1999 0.5 27 0.870 N Y Y Y 0.09 0.06 0.01 0.04 [19] 2001 2.6 26 0.840 Y Y Y Y -0.98 0.02 -0.59 0.02 [20] 2001 0.3 28 0.990 N Y Y Y -0.20 0.05 -0.06 0.05 [2] 2002 2.7 26 0.850 Y Y Y Y -0.06 0.16 -0.22 0.13 [3] 2002 1.0 26 0.850 N Y Y Y -0.01 0.54 -0.02 0.60 [23] 2003 1.0 29 1.200 Y Y Y Y 0.32 0.03 0.27 0.04 [22] 2003 14.0 27 0.980 Y Y Y Y -0.04 0.05 [21] 2003 26 0.726 N Y Y Y 0.10 0.05 0.09 0.03 [24] 2005 0.3 27 0.880 N Y Y Y -1.44 0.01 -1.20 0.00 Year: year of publication. Time CMV: Mean time on CMV before start of the study in hours. Age: Mean gestational age (weeks). Birth weight: mean birth weight (kg). HLVS: high lung volume strategy in the HFV group, LPVS: lung protective ventilation strategy in the CMV group, Surf: use of surfactant in the study. CLD: chronic lung disease, defined as on oxygen at 30–36 weeks postmenstrual age, LnRR: natural log of the relative risk

age (weeks). Birth weight: mean birth weight (kg). HLVS: high lung volume strategy in the HFV group, LPVS: lung protective ventilation strategy in the CMV group, Surf: use of surfactant in the study. CLD: chronic lung disease, defined as on oxygen at 30–36 weeks postmenstrual age, LnRR: natural log of the relative risk A funnel plot of the inverse of the standard error against the natural logarithm of the RR for CLD was indicative of publication bias because of asymmetry round the line of the pooled effect (Fig. 1). A rank test showed a p-value of 0.112. A stratified analysis of publication bias is indicated by different colors in Fig. 1. To visually evaluate publication bias within subgroups of ventilation strategy, the distribution of trials round the corresponding colored lines (mean effect size within subgroup) was assessed. Stratification by ventilation strategy (HLVS and LPVS vs either no HLVS and/or no LPVS) showed p-values of 0.456 and 0.851, respectively, indicating less evidence of publication bias. The distribution of stratified studies round the lines of pooled estimates showed less asymmetry (Fig. 1). Publication bias for the composite outcome of death or CLD was less likely with a p-value of 0.329. Stratified analysis showed p-values of 0.677 and 1.000. Fig. 1 Funnel plot. Selection bias in reporting RR of chronic lung disease (CLD) as suggested by asymmetry of the distribution of studies. x-axis: inverse of the standard error of the RR; y-axis: natural logarithm of the RR. Blue diamonds: studies with either no high lung volume strategy (HLVS) or no lung protective volume strategy (LPVS); pink diamonds: studies with both HLVS and LPVS; dotted line: estimated pooled RR including all studies; dashed colored lines: pooled RR of subgroups of studies. Publication bias was visually appraised by assessing symmetry of distribution of studies around the lines of pooled estimates. CMV conventional mechanical ventilation

amonds: studies with both HLVS and LPVS; dotted line: estimated pooled RR including all studies; dashed colored lines: pooled RR of subgroups of studies. Publication bias was visually appraised by assessing symmetry of distribution of studies around the lines of pooled estimates. CMV conventional mechanical ventilation Figures 2, 3, 4 show the results of the linear meta-regression analyses for continuous explanatory variables with relative risk of CLD as dependent variable. Two studies dominate these figures, designated by the weight they received in the analyses [2, 3]. Over the years the reported benefit of HFV over CMV seemed to diminish (Fig. 2). A longer time on CMV prior to study initiation and a higher gestational age and increase of birth weight (data not shown) seemed to be positively associated with a relatively better outcome in HFV (Figs. 3, 4). Table 2 shows the results of linear meta-regression analyses, showing significant associations with year of publication (3.1 times higher RR with change of publication year from 1992 to 2005) and whether or not a protective ventilation strategy was applied (1.9 times higher RR with change of protective ventilation from no to yes; Table 2). In the linear regression analyses with death or CLD as composite outcome no significant associations were detected. Whether or not a Sensormedics high-frequency oscillatory ventilator was used and baseline incidence in CMV (0.75 vs 0.08) treated patients displayed the smallest effects on trial outcome (RRR = 0.84 and 0.90 for CLD and RRR = 0.85 and 0.99 for death or CLD, respectively). Fig. 2 Linear regression analyses. Crude and subgroup linear regression analyses of the effect of year of publication, prior time on CMV and gestational age with natural logarithm of RR of CLD as dependent variable. y-axis: natural logarithm of the RR; x-axis: explanatory variables. Blue diamonds: studies with either no HLVS or no LPVS; pink diamonds: studies with both HLVS and LPVS. The size of the diamonds reflects the weights the individual trials contribute to the analyses. Thin blue line: regression line including all studies; thick pink line: regression line including only studies with both HLVS and LPVS. CMV conventional mechanical ventilation

pink diamonds: studies with both HLVS and LPVS. The size of the diamonds reflects the weights the individual trials contribute to the analyses. Thin blue line: regression line including all studies; thick pink line: regression line including only studies with both HLVS and LPVS. CMV conventional mechanical ventilation Fig. 3 Same as Fig. 2 Fig. 4 Same as Fig. 2 Table 2 Univariable linear regression analysis

pink diamonds: studies with both HLVS and LPVS. The size of the diamonds reflects the weights the individual trials contribute to the analyses. Thin blue line: regression line including all studies; thick pink line: regression line including only studies with both HLVS and LPVS. CMV conventional mechanical ventilation Fig. 3 Same as Fig. 2 Fig. 4 Same as Fig. 2 Table 2 Univariable linear regression analysis 95% confidence interval 95% confidence interval Crude B Sig. Lower Upper RRR Lower Upper boundary boundary boundary boundary All studies CLD Year 0.09 0.025 0.01 0.16 3.13 1.18 8.27 SensorM -0.17 0.351 -0.55 0.21 0.84 0.58 1.24 (no to yes) TimeCMV -0.09 0.055 -0.19 0.00 0.44 0.19 1.02 Age -0.08 0.237 -0.23 0.06 0.66 0.32 1.36 Weight -0.76 0.163 -1.87 0.35 0.54 0.22 1.33 HLVS -0.11 0.883 -1.74 1.52 0.89 0.17 4.57 LPVS 0.64 0.009 0.19 1.10 1.91 1.21 3.00 Surf 1.21 0.168 -0.59 3.00 3.34 0.56 20.03 CMV -0.18 0.774 -1.53 1.17 0.90 0.42 1.92 Death or CLD Year 0.05 0.096 -0.01 0.12 2.01 0.86 4.65 SensorM -0.17 0.132 -0.39 0.06 0.85 0.67 1.06 TimeCMV -0.01 0.590 -0.05 0.03 0.92 0.65 1.29 Age -0.02 0.733 -0.13 0.10 0.91 0.52 1.61 Weight -0.22 0.611 -1.16 0.71 0.84 0.40 1.77 HLVS -0.37 0.698 -2.44 1.69 0.69 0.09 5.45 LPVS 0.19 0.275 -0.18 0.56 1.21 0.84 1.76 Surf 0.52 0.289 -0.51 1.56 1.69 0.60 4.75 CMV -0.02 0.963 -0.91 0.87 0.99 0.60 1.63 Studies with surfactant, HLVS, and LPVS CLD Year 0.00 0.971 -0.23 0.22 0.96 0.05 17.34 TimeCMV -0.05 0.698 -0.34 0.25 0.66 0.05 8.75 Age 0.04 0.727 -0.22 0.30 1.22 0.33 4.49 Weight 0.41 0.693 -1.99 2.81 1.38 0.20 9.44 Death or CLD Year 0.01 0.846 -0.15 0.17 1.20 0.15 9.72 TimeCMV 0.00 0.819 -0.05 0.04 0.96 0.65 1.43 Age 0.06 0.406 -0.10 0.21 1.34 0.61 2.92 Weight 0.55 0.396 -0.89 1.99 1.55 0.49 4.90 Simple linear regression analyses were calculated for chronic lung disease (CLD), defined as on oxygen at 30–36 weeks postgestational age, and death or CLD.

1 0.846 -0.15 0.17 1.20 0.15 9.72 TimeCMV 0.00 0.819 -0.05 0.04 0.96 0.65 1.43 Age 0.06 0.406 -0.10 0.21 1.34 0.61 2.92 Weight 0.55 0.396 -0.89 1.99 1.55 0.49 4.90 Simple linear regression analyses were calculated for chronic lung disease (CLD), defined as on oxygen at 30–36 weeks postgestational age, and death or CLD. The following co-variates were evaluated: Year: Number of years after the first included study; SensorM: Whether or not a Sensormedics type of HFV was used; TimeCMV: Mean time on CMV before start of the study in hours; Age: Mean gestational age (weeks); Weight: mean birth weight (kg); HLVS: high lung volume strategy in the HFV group; LPVS: lung protective ventilation strategy in the CMV group; Surf: use of surfactant in the study; B was the estimated crude coefficient; RRR: relative risk ratio = RRcovariate=1/RRcovariate=0, for binary variables (SensorM, HLVS and LPVS), for continues variables the extreme values reported in the studies were used to calculate the ranges, 13 for years, 8.7 for Time on CMV, 5 for Age, 0.8 for Weight and 0.65 for CMV (RRRyears = RRyear=2005/RRyear=1992, RRRtime on CMV = RRtime=9 h/RRtime=0.3 h, RRRage = RRage=31 weeks/RRyear=26 weeks, RRRweight = RRweight=1.5 kg/RRyear=0.7 kg, RRRincidence of CLD in CMV=0.75/RRincidence=0.08)

used to calculate the ranges, 13 for years, 8.7 for Time on CMV, 5 for Age, 0.8 for Weight and 0.65 for CMV (RRRyears = RRyear=2005/RRyear=1992, RRRtime on CMV = RRtime=9 h/RRtime=0.3 h, RRRage = RRage=31 weeks/RRyear=26 weeks, RRRweight = RRweight=1.5 kg/RRyear=0.7 kg, RRRincidence of CLD in CMV=0.75/RRincidence=0.08) Figure 5 shows how the incidence of CLD in the CMV treated patients was related to the incidence in HFV treated patients for each of the studies. The diagonal line represents the line of no effect in this figure. A trend line was fitted by weighted linear regression, showing a small effect of change in incidence in CMV on incidence in HFV-treated patients. Fig. 5 Linear regression analysis of incidence of CLD in CMV on incidence of CLD in HFV. y-axis: incidence of CLD in HFV; x-axis: incidence of CLD in CMV. Thin pink line: regression line including all studies

d linear regression, showing a small effect of change in incidence in CMV on incidence in HFV-treated patients. Fig. 5 Linear regression analysis of incidence of CLD in CMV on incidence of CLD in HFV. y-axis: incidence of CLD in HFV; x-axis: incidence of CLD in CMV. Thin pink line: regression line including all studies Year of publication was not related to change in relative risk of CLD in the subgroup of studies with HLVS, LPVS and concomitant use of surfactant (RRR = 0.96). There was only a small increase in relative risk for death or CLD (RRR = 1.20; Fig. 2; Table 2). Opposite effects of gestational age (RRR = 1.22 for CLD and 1.38 for death or CLD vs RRR = 0.66 for CLD and 0.91 for death or CLD, respectively) and birth weight were detected in the subgroup analysis (Fig. 4; Table 2). Prior time on CMV exerted less effect on outcome compared with the crude analysis, RRR = 0.66 for CLD and 0.96 for death or CLD and RRR = 0.44 for CLD and 0.92 for death or CLD in the adjusted and crude analyses, respectively (Fig. 3; Table 2).

and birth weight were detected in the subgroup analysis (Fig. 4; Table 2). Prior time on CMV exerted less effect on outcome compared with the crude analysis, RRR = 0.66 for CLD and 0.96 for death or CLD and RRR = 0.44 for CLD and 0.92 for death or CLD in the adjusted and crude analyses, respectively (Fig. 3; Table 2). Multivariable linear regression analyses were conducted to assess the independent contributions to change in RR by explanatory variables (Table 3). The RRRs in Table 3 have the same meaning as in Table 2, only they represented adjusted RRRs. Year of publication was not considered as an independent explanatory variable but rather as proxy for changes in treatment and patient population. Gestational age and birth weight were collinearly related by nature; only gestational age was fitted in the model. One study contributed to the fact that surfactant was not used; therefore, surfactant was not used in the multiple linear regression analyses. Two models were fitted. Model A used Sensormedics, time on CMV, gestational age, HLVS, and LPVS as covariates. The largest estimated effects were caused by ventilation strategies, HLVS, and LPVS, adjusted for use of Sensormedics ventilator, prior time on CMV, and gestational age. These estimations were consistent for the outcomes CLD (RRR = 0.42 and RRR = 2.02 for HLVS and LPVS, respectively) and death or CLD (RRR = 0.42 and RRR = 1.98 fro HLVS and LPVS, respectively). Use of a Sensormedics ventilator seemed to have a much smaller effect on RR for outcome. The RRR of gestational age, comparing 26 weeks with 31 weeks, for CLD and death or CLD were larger (RRR = 1.17 and RRR = 1.47). The effect of a difference in prior time on CMV of 8.7 h on CLD vs death or CLD was not consistent (RRR = 0.85 and RRR = 1.07, respectively). Table 3 Multivariable linear regression analysis

. The RRR of gestational age, comparing 26 weeks with 31 weeks, for CLD and death or CLD were larger (RRR = 1.17 and RRR = 1.47). The effect of a difference in prior time on CMV of 8.7 h on CLD vs death or CLD was not consistent (RRR = 0.85 and RRR = 1.07, respectively). Table 3 Multivariable linear regression analysis Adjusted 95% confidence interval 95% confidence interval B Sig. Lower boundary Upper boundary RRR Lower boundary Upper boundary Model A CLD (Constant) -0.66 0.900 -13.03 11.70 SensorM -0.04 0.884 -0.75 0.66 0.96 0.47 1.94 TimeCMV -0.02 0.903 -0.38 0.34 0.85 0.04 19.22 Age 0.03 0.850 -0.36 0.42 1.17 0.16 8.32 HLVS -0.88 0.306 -2.80 1.04 0.42 0.06 2.84 LPVS 0.70 0.506 -1.73 3.14 2.02 0.18 23.12 Death or CLD (Constant) -1.86 0.412 -7.22 3.49 SensorM -0.17 0.309 -0.55 0.21 0.85 0.58 1.24 TimeCMV 0.01 0.722 -0.05 0.06 1.07 0.68 1.69 Age 0.08 0.299 -0.09 0.25 1.47 0.62 3.47 HLVS -0.88 0.407 -3.38 1.62 0.42 0.03 5.06 LPVS 0.68 0.127 -0.28 1.65 1.98 0.76 5.19 Model B CLD (Constant) 0.07 0.904 -1.21 1.35 SensorM -0.06 0.698 -0.38 0.26 0.94 0.69 1.30 HLVS -0.81 0.203 -2.14 0.52 0.44 0.12 1.68 LPVS 0.72 0.011 0.21 1.23 2.06 1.23 3.43 Death or CLD (Constant) SensorM -0.11 0.318 -0.33 0.12 0.90 0.72 1.13 HLVS -0.79 0.363 -2.66 1.08 0.45 0.07 2.93 LPVS 0.46 0.089 -0.09 1.01 1.59 0.92 2.74 Multiple linear regression analyses were calculated for chronic lung disease (CLD), defined as on oxygen at 30-36 weeks postgestational age, and death or CLD. The following co-variates were evaluated: SensorM whether or not a Sensormedics type of HFV was used; TimeCMV mean time on CMV before start of the study (in hours). HLVS high lung volume strategy in the HFV group, LPVS lung protective ventilation strategy in the CMV group. B was the adjusted estimated coefficient. RRR relative risk ratio=RRcovariate=1/RRcovariate=0, for binary variables (SensorM, HLVS and LPVS), for continues variables the extreme values reported in the studies were used, 8.7 for Time on CMV (RRRage = RRage=31 weeks/RRyear=26 weeks)

ventilation strategy in the CMV group. B was the adjusted estimated coefficient. RRR relative risk ratio=RRcovariate=1/RRcovariate=0, for binary variables (SensorM, HLVS and LPVS), for continues variables the extreme values reported in the studies were used, 8.7 for Time on CMV (RRRage = RRage=31 weeks/RRyear=26 weeks) A sensitivity analysis was conducted by fitting a second model (model B) with the most important variables, HLVS and LPVS, combined with whether or not a Sensormedics ventilator was used. The reported RRRs were comparable to those in the first model. Type of ventilator did not have a large effect compared with ventilation strategies (RRR = 0.94 and RRR = 0.90). The HLVS was associated with a decrease of the RRs comparing HFV with CMV (RRR = 0.44 and RRR = 0.45), while LPVS increased the RRs to the line of no effect (RRR = 2.06 and RRR = 1.59).

e in the first model. Type of ventilator did not have a large effect compared with ventilation strategies (RRR = 0.94 and RRR = 0.90). The HLVS was associated with a decrease of the RRs comparing HFV with CMV (RRR = 0.44 and RRR = 0.45), while LPVS increased the RRs to the line of no effect (RRR = 2.06 and RRR = 1.59). Discussion Our meta-regression analysis showed a clear trend of decreasing differences in pulmonary outcome between HFV and CMV in randomized trials conducted in premature neonates with IRDS over the years. The most likely hypothesis for this trend was the application of a LPVS in the most recent studies. Use of surfactant could also have a significant contribution, but only one study did not use surfactant [12]. In previous meta-analyses, subgroup analyses or cumulative methods were used to explore heterogeneity [4, 5, 7]. Subgroup analysis is equivalent to meta-regression with a categorical trial-level covariate. Considering subgroup analysis formally as a meta-regression has advantages, since it focuses on differences between subgroups as is appropriate, rather than the effects in each subgroup separately. Furthermore, it is appropriate to use meta-regression to explore sources of heterogeneity, even if an initial overall test for heterogeneity is non-significant. This test often has low power and therefore a non-significant result does not reliably identify lack of heterogeneity [25].

ffects in each subgroup separately. Furthermore, it is appropriate to use meta-regression to explore sources of heterogeneity, even if an initial overall test for heterogeneity is non-significant. This test often has low power and therefore a non-significant result does not reliably identify lack of heterogeneity [25]. In this meta-regression analysis we evaluated in a quantitative way a number of hypotheses that were raised to account for different results between randomized trials. A relatively large proportion of well-conducted trials were available for the analyses. For most explanatory variables there were important differences among trials. The effects of the two most important covariates, HLVS and LPVS, were consistent in the different models and were even increased in effect size by adjusting for other covariates. None of the competing hypotheses were more likely to influence results as shown by calculating the RRRs. Common pitfalls in meta-regression analysis can occur, such as multiple or post-hoc analyses, and lead to data dredging and a high probability of false-positive conclusions [25]. We, therefore, restricted our analyses to a limited number of pre-specified explanatory covariates.

results as shown by calculating the RRRs. Common pitfalls in meta-regression analysis can occur, such as multiple or post-hoc analyses, and lead to data dredging and a high probability of false-positive conclusions [25]. We, therefore, restricted our analyses to a limited number of pre-specified explanatory covariates. Publication bias was considered unlikely as an explanation of the apparent diminishing relative effect of HFV. Publication bias is selection bias. If trials are selectively published either because of their size or because of significant results, this would result in an association between trial size and/or precision and the trial outcome. Strictly speaking, funnel plots probe whether studies with little precision (small studies) give different results from studies with greater precision (larger studies). Asymmetry in the funnel plot may therefore result not from a systematic under-reporting of negative trials but from an essential difference between smaller and larger studies that arises from inherent between-study heterogeneity [26]; thus, if larger studies were also associated with changes in ventilation strategies and these strategies resulted in changes in reported RRs, the assumed publication bias would be, in fact, a real association between ventilation strategy and study outcome; therefore, we conditioned the association between precision and effect size, presumably caused by publication bias, on ventilation strategies. This resulted in a lower p-value for publication bias and more symmetrical distribution of studies in subgroups in the funnel plots; therefore, what appeared to be publication bias could also be explained by differences in ventilation strategies related to both study size and observed relative risks. However, it should be pointed out that the strength of this evidence is difficult to assess because fewer studies in the subgroups automatically resulted in less power to detect publication bias.

ias could also be explained by differences in ventilation strategies related to both study size and observed relative risks. However, it should be pointed out that the strength of this evidence is difficult to assess because fewer studies in the subgroups automatically resulted in less power to detect publication bias. Other alternative hypotheses that have been formulated to explain differences among studies were also less compatible with the evidence [9]. The type of ventilator, Sensormedics vs other types of high-frequency ventilators, displayed RRR close to one. In the crude analyses, prior time on CMV before study initiation showed contradictory effects to what was hypothesized [10]. The adjusted analyses showed conflicting results depending on the outcome. Gestational age and birth weight could also influence the magnitude of the effect of HFV compared with CMV. In the adjusted analysis gestational age did not change the RR for CLD but showed an increase of the RR for less premature neonates. Finally, an increased risk of CLD was not accompanied by a greater relative benefit of HFV as compared with CMV.

could also influence the magnitude of the effect of HFV compared with CMV. In the adjusted analysis gestational age did not change the RR for CLD but showed an increase of the RR for less premature neonates. Finally, an increased risk of CLD was not accompanied by a greater relative benefit of HFV as compared with CMV. The observed effects of continuous variables, such as time on CMV or gestational age, could be exaggerated by small studies with outlying results. For the covariate, time on CMV, the two largest studies showed results that were compatible with the hypothesis that this had no important impact on the results of these trials [2, 3]. The same fact applied to the effect of baseline incidence of CLD or death or CLD. Gestational age and weight were comparable between the two largest trials, which made it more difficult to ascertain the relevance of the hypothesis that in smaller and more premature infants HFV performed better than CMV treatment. The observed direction of the effect of gestational age and birth weight, however, was opposite to what the hypothesis predicted. If gestational age was to be interpreted as a higher risk of acquiring CLD, one would expect that an increase in the incidence of CLD was associated with a relatively lower incidence of CLD in HFV treated patients; however, linear regression analysis showed perfectly equal increase in both treatment groups. Still, the possibility remains that the relationship with patient averages, such as gestational age and birth weight, across trials was not the same as the relationship for patients within trials, and therefore an effect of these patient characteristics cannot be excluded but only considered in relation to other covariates [25].

possibility remains that the relationship with patient averages, such as gestational age and birth weight, across trials was not the same as the relationship for patients within trials, and therefore an effect of these patient characteristics cannot be excluded but only considered in relation to other covariates [25]. Similar findings of the effects of ventilation strategies have been reported by us and other authors as well [4, 5]; however, meta-analyses are subject to bias when differences among trials are used to explain differences in reported RRs. In this meta-regression analysis we were able to estimate adjusted association measures, thereby diminishing the effects of possible confounders/effect-modifiers. By calculating less biased estimates of the effects of ventilation strategies and the effect of using a Sensormedics ventilator instead of other ventilators on the outcome in the different HFV trials we were able to reinforce the hypothesis that ventilation strategies are more important than type of ventilator to prevent CLD.

ulating less biased estimates of the effects of ventilation strategies and the effect of using a Sensormedics ventilator instead of other ventilators on the outcome in the different HFV trials we were able to reinforce the hypothesis that ventilation strategies are more important than type of ventilator to prevent CLD. The results of this meta-analysis stresses the importance of using appropriate ventilation strategies to prevent ventilator-induced lung damage in a highly vulnerable group of patients; therefore, in clinical practice the question of how to use the ventilator is more important than the question of which ventilator should be used. The major theoretical advantage of HFV to CMV is delivery of smaller tidal volumes to an optimally recruited lung. As this meta-regression analysis did not confirm that subgroups of more premature neonates, avoidance of CMV prior to initiating HFV, or neonates with higher risk of CLD were more likely to benefit form elective HFV in IRDS, future research should be directed at identifying patients in whom HFV does have a benefit over CMV. To improve the robustness of these conclusions and to avoid the limitations of meta-analysis of trials, an individual-patient-data-based meta-regression analysis should be conducted.

o benefit form elective HFV in IRDS, future research should be directed at identifying patients in whom HFV does have a benefit over CMV. To improve the robustness of these conclusions and to avoid the limitations of meta-analysis of trials, an individual-patient-data-based meta-regression analysis should be conducted. Conclusion In conclusion, confining randomized trails to smaller or more premature children with IRDS did not seem to result in better pulmonary outcomes of HFV compared with CMV. A generally held opinion that a prolonged ventilation time on CMV prior to initiating HFV diminished the benefits of HFV was not in agreement with the current evidence. The most important effects resulting in differences among trials were probably caused by ventilation strategies applied in HFV- and CMV-treated patients. Electronic supplementary material Electronic Supplementary Material (DOC 514K) Electronic supplementary material The online version of this article (doi:10.1007/s00134-007-0545-y) contains supplementary material, which is available to authorized users

Introduction Chest radiographs (CXRs) are frequently obtained as a complement to physical examination of critically ill patients [1, 2]. There are two different schools of thought regarding the utility of CXRs in the intensive care unit (ICU): The CXRs should be made on indication only, specifically when there is a sound reason to obtain a film (so-called on demand CXRs); or CXRs should be obtained routinely every day, that is, without any specific reason (so-called daily routine CXRs). Argument for the latter strategy is the high prevalence of findings on CXRs of ICU patients [3]; however, interpretation of studies on the usefulness of daily routine CXRs is hampered because of major differences in methodology [4]. Importantly, most studies did not attempt to discriminate between clinically relevant and irrelevant findings. We recently demonstrated that daily routine CXRs hardly ever reveal potentially important abnormalities and seldom result in a change in therapy [5]. While it can be recommended to discontinue a daily routine CXR practice in ICU patients, elimination of these CXRs may have several disadvantages. Firstly, eliminating daily routine CXRs bears the risk that the number of on demand CXRs increases. In addition, elimination of daily routine CXRs might result in on demand CXRs being obtained more frequently during off-time hours, which may cause an inverse rise of costs. Secondly, length of stay (LOS) in ICU, readmission rate and mortality rate might be negatively influenced by this change in CXR practice.

creases. In addition, elimination of daily routine CXRs might result in on demand CXRs being obtained more frequently during off-time hours, which may cause an inverse rise of costs. Secondly, length of stay (LOS) in ICU, readmission rate and mortality rate might be negatively influenced by this change in CXR practice. To evaluate the impact of elimination of daily routine CXRs we determined the change in on demand CXR practice in our multidisciplinary ICU, where a daily routine CXR strategy was applied until performance of this study. In addition, we evaluated the diagnostic and therapeutic value of on demand CXRs before and after this intervention. Finally, LOS in ICU, readmission rate, and mortality rate during a daily routine CXR strategy were compared with those during an on demand CXR strategy.

ategy was applied until performance of this study. In addition, we evaluated the diagnostic and therapeutic value of on demand CXRs before and after this intervention. Finally, LOS in ICU, readmission rate, and mortality rate during a daily routine CXR strategy were compared with those during an on demand CXR strategy. Materials and Methods Subjects A prospective, nonrandomized, controlled design with an intervention was used for this study. Of all patients, all CXRs taken in the adult ICU department of the Academic Medical Center in Amsterdam, Netherlands, from 1 March 2004 to 31 July 2004 and from 1 September 2004 to 31 January 2005 were studied. This department is a closed-format tertiary care, referral, 28-bed multidisciplinary ICU. The patient population consists of cardiothoracic surgery patients, medical patients (including cardiology patients and pulmonary disease patients), and surgery patients (including trauma patients and neurosurgery patients). Patients who were admitted during the period in between phases 1 and 2, as well as patients that were readmitted, were not analyzed. The study protocol was approved by the local ethics committee.

iology patients and pulmonary disease patients), and surgery patients (including trauma patients and neurosurgery patients). Patients who were admitted during the period in between phases 1 and 2, as well as patients that were readmitted, were not analyzed. The study protocol was approved by the local ethics committee. Protocol The study period was divided into two parts: phase 1, a 5-month phase before the intervention during which the daily routine CXR strategy was practiced; and phase 2, a 5-month phase which began 1 month after the intervention. The intervention consisted of a change in the ordering practice of CXRs: no standing orders for daily routine CXRs; each (on demand) CXR required a clinical indication, such as admittance to the ICU, insertion of central venous lines, intra-aortic balloon pump or tracheal and chest tubes, an increase in oxygen requirement, or a change in pulmonary secretions with or without fever (see Table E1). For phases 1 and 2, CXR volume data were collected prospectively. Type of, and reason for, admission was registered for all patients. Severity of illness was scored by means of acute physiology and chronic health evaluation (APACHE) II for all patients. Data on LOS in ICU, readmission to ICU as well as ICU, and hospital mortality rate, were collected from the National Intensive Care Evaluation (NICE) database [6]. The LOS was calculated from day and time of arrival at, and discharge from, ICU. The total number of hours in ICU were divided by 24 to determine the exact LOS in ICU in days.

readmission to ICU as well as ICU, and hospital mortality rate, were collected from the National Intensive Care Evaluation (NICE) database [6]. The LOS was calculated from day and time of arrival at, and discharge from, ICU. The total number of hours in ICU were divided by 24 to determine the exact LOS in ICU in days. Diagnostic and therapeutic value of on demand CXR Diagnostic and therapeutic value of on demand CXRs was determined as described previously for daily routine CXRs [5]. In short, the attending physician completed a specially developed data sheet on radiological abnormalities which was printed on the back of the normal CXR request form. It was to be ticked whether a certain finding was expected, and whether it was “old” (i.e., already present on preceding CXR) or “new” (i.e., not present on preceding CXR). All CXRs were interpreted by an independent radiologist on the day the on demand CXR was performed. Similar to the ICU physicians, the radiologist structurally interpreted these on demand CXRs for each patient (i.e., the radiologist ticked whether radiological abnormalities were absent or present and, if an abnormality was present, whether it was judged to be an “old” or “new” finding).

elevance of the predefined abnormalities could not be evaluated in some cases, specifically in cases of large atelectasis and severe pulmonary congestion, since start of physiotherapy, changes in levels of positive end-expiratory pressure, and the use of diuretics might have been triggered by other (clinical) findings. Statistical analysis All data are expressed as means (± SD), or medians (interquartile ranges). A Mann-Whitney U-test was used for analyzing continuous variables. A chi-square test was used to compare the groups in phase 1 and phase 2. The incidences of expected and unexpected findings, and clinically important abnormalities, were compared by chi-square test. A p-value < 0.05 was considered to be statistically significant. All calculations were performed using SPSS version 12.0.1 software (SPSS, Chicago, Ill.). Results Study population We evaluated 1376 patients over the two periods. Patient profiles on entering this study are summarized in Table 1. A total of 3894 CXRs were obtained from 754 patients in phase 1; these included 2457 daily routine CXRs and 1437 on demand CXRs. A total of 1267 CXRs were obtained from 622 patients in phase 2. These CXRs were, by definition, all on demand CXRs. Table 1 Demographic data. APACHE-II Acute Physiology and Chronic Health Evaluation II, CXRs chest radiographs, CI confidence interval

se 1; these included 2457 daily routine CXRs and 1437 on demand CXRs. A total of 1267 CXRs were obtained from 622 patients in phase 2. These CXRs were, by definition, all on demand CXRs. Table 1 Demographic data. APACHE-II Acute Physiology and Chronic Health Evaluation II, CXRs chest radiographs, CI confidence interval Phase 1 Phase 2 Significance (p) No. of patients 754 622 Age (years; mean, SD) 60 (16) 62 (16) 0.02 Gender (male; n) 475 (63%) 398 (64%) 0.70 CXRs while patients being mechanically ventilated (n)a 3194 (82%) 1115 (88%) < 0.001 APACHE-II score 16.4 ± 6.9 16.4 ± 7.2 1.00 Patient subgroups Cardiac surgery (n) 317 (42%) 306 (49%) 0.01 Medical (n) 197 (26%) 119 (19%) Surgical (n) 144 (19%) 131 (21%) Neurosurgical/neurology (n) 69 (9%) 46 (7%) Other (n) 27 (4%) 20 (3%) Length of stay in ICU (days; median IQR) 1.9 (1.0–4.6) 1.9 (0.9–4.6) 0.95 Mortality ICU (n) 94 (12%) 62 (10%) 0.49 Hospital (n) 132 (18%) 104 (17%) 0.70 Predicted hospital mortality (%) 181 (24%) 155 (25%) 0.69 Observed/predicted ratio (95% CI) 0.73 (0.59–0.90) 0.67 (0.53–0.83) aAll patients were mechanically ventilated at any time during stay in ICU. Expressed is the percentage of CXRs during which patients were on the ventilator while the CXR was performed

(17%) 0.70 Predicted hospital mortality (%) 181 (24%) 155 (25%) 0.69 Observed/predicted ratio (95% CI) 0.73 (0.59–0.90) 0.67 (0.53–0.83) aAll patients were mechanically ventilated at any time during stay in ICU. Expressed is the percentage of CXRs during which patients were on the ventilator while the CXR was performed Utility of CXRs The number of CXRs per day for the whole ICU declined from 22.6 ± 4.9 to 8.2 ± 3.2 (p < 0.05; Fig. 1). Adjusting for patient volume, the ratio of CXRs per patient day decreased from 1.1 ± 0.3 to 0.6 ± 0.4 after the intervention (p < 0.05). The median number of CXRs per patient for the complete stay in ICU declined from 3 (range 2–5) during phase 1, to 1 (range 1–2) after the intervention. The number of on demand CXRs increased minimally after the intervention, and the distribution over 24 h, did not change (see ESM, Fig. E1). Fig. 1 Number of CXRs/day during the study. Phase 1: daily routine CXR strategy, i.e., a daily routine CXR was made every morning, from March to July; phase 2: on demand CXR strategy, i.e., each CXR required a clinical indication, from September to January. Open bars: mean number (± SD) of on demand CXRs/day; closed symbols: mean number (± SD) of all CXRs/day

y. Phase 1: daily routine CXR strategy, i.e., a daily routine CXR was made every morning, from March to July; phase 2: on demand CXR strategy, i.e., each CXR required a clinical indication, from September to January. Open bars: mean number (± SD) of on demand CXRs/day; closed symbols: mean number (± SD) of all CXRs/day Diagnostic and therapeutic value of on demand CXRs The diagnostic and therapeutic value of on demand CXRs increased with elimination of daily routine CXRs (Tables 2, 3). Before intervention, 38 expected predefined abnormalities were found (2.6% of all on demand CXRs in 4.9% of all patients), and after the intervention 64 expected predefined abnormalities were found (5.0%; p < 0.05) in 9.5% of cases (p < 0.05). All these findings led to a change in therapy. Before intervention, 147 unexpected predefined abnormalities were found (10.2% of all on demand CXRs in 15.9% of all patients), of which 57 (4.0 in 6.4%) led to a change in therapy. After intervention 156 unexpected predefined abnormalities were found (11.6% of all on demand CXRs in 19.1% of all patients), of which 64 (4.8 in 9.5%; p < 0.05) led to a change in therapy. Subgroup analysis revealed no differences between phases 1 and 2, except for medical patients, in which there was a significant rise in the number of on demand CXRs that showed an unexpected predefined major abnormality (p < 0.05 vs phase 1; see ESM, Table E2). Table 2 Expected and unexpected findings on on demand chest radiographs

ysis revealed no differences between phases 1 and 2, except for medical patients, in which there was a significant rise in the number of on demand CXRs that showed an unexpected predefined major abnormality (p < 0.05 vs phase 1; see ESM, Table E2). Table 2 Expected and unexpected findings on on demand chest radiographs Phase 1 (n = 1437) Phase 2 (n = 1267) Abnormalities Expected Expected+found Unexpected+found Expected Expected+found Unexpected+found Large atelectasis 37 (2.6) 2 (0.1) 13 (0.9) 49 (3.9) 3 (0.2) 15 (1.2) Large infiltrates 57 (4.0) 3 (0.2) 21 (1.5) 69 (5.4) 5 (0.4) 27 (2.1) Pulmonary congestion 98 (6.8) 8 (0.6) 25 (1.7) 104 (8.2) 14 (1.1) 22 (1.7) Pleural effusion 41 (2.9) 3 (0.2) 17 (1.2) 43 (3.4) 4 (0.3) 27 (2.1) Pneumothorax or pneumomediastinum 68 (4.7) 4 (0.3) 17 (1.2) 39 (3.1)c 3 (0.2) 12 (0.9) Malposition of invasive devices 350 (24.4) 18 (1.3) 54 (3.8) 392 (30.9)c 35 (2.7)c 52 (4.1) Total no. of abnormalities 651 38 147 696 64c 155 Total no. of CXRs with abnormalitiesa 641 (44.6) 38 (2.6) 133 (9.2) 384 (30.3)c 63 (5.0)c 147 (11.6)c Total no. of patients with CXRs with abnormalitiesb 580 (76.9) 37 (4.9) 120 (15.9) 223 (35.9)c 58 (9.5)c 119 (19.1) Numbers in parentheses are percentages a Absolute number of chest radiographs (CXRs; percentage of all daily routine CXRs) b Absolute number of patients (percentage of all patients with on demand CXRs) c p < 0.05 vs phase 1 Table 3 Unexpected findings on on demand chest radiographs resulting in a chance in therapy. ND not defined

Phase 1 (n = 1437) Phase 2 (n = 1267) Abnormalities Expected Expected+found Unexpected+found Expected Expected+found Unexpected+found Large atelectasis 37 (2.6) 2 (0.1) 13 (0.9) 49 (3.9) 3 (0.2) 15 (1.2) Large infiltrates 57 (4.0) 3 (0.2) 21 (1.5) 69 (5.4) 5 (0.4) 27 (2.1) Pulmonary congestion 98 (6.8) 8 (0.6) 25 (1.7) 104 (8.2) 14 (1.1) 22 (1.7) Pleural effusion 41 (2.9) 3 (0.2) 17 (1.2) 43 (3.4) 4 (0.3) 27 (2.1) Pneumothorax or pneumomediastinum 68 (4.7) 4 (0.3) 17 (1.2) 39 (3.1)c 3 (0.2) 12 (0.9) Malposition of invasive devices 350 (24.4) 18 (1.3) 54 (3.8) 392 (30.9)c 35 (2.7)c 52 (4.1) Total no. of abnormalities 651 38 147 696 64c 155 Total no. of CXRs with abnormalitiesa 641 (44.6) 38 (2.6) 133 (9.2) 384 (30.3)c 63 (5.0)c 147 (11.6)c Total no. of patients with CXRs with abnormalitiesb 580 (76.9) 37 (4.9) 120 (15.9) 223 (35.9)c 58 (9.5)c 119 (19.1) Numbers in parentheses are percentages a Absolute number of chest radiographs (CXRs; percentage of all daily routine CXRs) b Absolute number of patients (percentage of all patients with on demand CXRs) c p < 0.05 vs phase 1 Table 3 Unexpected findings on on demand chest radiographs resulting in a chance in therapy. ND not defined Phase 1 (n = 1437) Phase 2 (n = 1267) Abnormalities Resulting in a change in therapy Resulting in a change in therapy Large atelectasis ND ND Large infiltrates 10 (0.7%) 14 (1.1%) Pulmonary congestion ND ND Pleural effusion 11 (0.8%) 12 (0.9%) Pneumothorax or pneumomediastinum 11 (0.8%) 9 (0.7%) Malposition of invasive devices 25 (1.7%) 29 (2.3%) Total no. of abnormalities 57 64 Total no. of CXRs with abnormalitiesa 56 (3.9%) 61 (4.8%) Total no. of patients with CXRs with abnormalitiesb 48 (6.4%) 59 (9.5%)c a Absolute number of chest radiographs (CXRs; percentage of all daily routine CXRs)

m 11 (0.8%) 9 (0.7%) Malposition of invasive devices 25 (1.7%) 29 (2.3%) Total no. of abnormalities 57 64 Total no. of CXRs with abnormalitiesa 56 (3.9%) 61 (4.8%) Total no. of patients with CXRs with abnormalitiesb 48 (6.4%) 59 (9.5%)c a Absolute number of chest radiographs (CXRs; percentage of all daily routine CXRs) b Absolute number of patients percentage of all patients with on demand CXRs) c p < 0.05 vs phase 1 LOS in ICU, readmission rate and mortality rate The LOS in ICU was not different in phase 1 as compared with phase 2 (Table 1). Total readmission rate was similar (8.4% in phase 1 vs 7.6% and phase 2, risk difference 0.8% (95% CI: 2.1–3.7%, P = 0.6), and did not change with the intervention for the different subgroups. There were no statistically significant differences in ICU and hospital mortality rates before and after the intervention (Table 1). Discussion The present study demonstrates the impact of elimination of daily routine CXRs in a mixed medical–surgical ICU. We found a sharp decline in the total number of CXRs, while only a minimal increase in the number of on demand CXRs was observed. In addition, the number of CXRs in off-hours was similar between the two periods. Elimination of daily routine CXRs did neither affect LOS in ICU and readmission rate nor ICU and hospital mortality rate.

arp decline in the total number of CXRs, while only a minimal increase in the number of on demand CXRs was observed. In addition, the number of CXRs in off-hours was similar between the two periods. Elimination of daily routine CXRs did neither affect LOS in ICU and readmission rate nor ICU and hospital mortality rate. Although the diagnostic and therapeutic value of on demand CXRs was significantly higher after the intervention, we considered this difference clinically irrelevant. When one considers the increase in diagnostic and therapeutic value of on demand CXRs after elimination of daily routine CXRs indirect proof of the “value” of daily routine CXRs, one must also recognize its futility regarding the therapeutic value. Indeed, the percentage of CXRs with unexpected findings that truly led to a change in therapy was similar in the two study phases. Since readmission rate and mortality rate remained unchanged after the intervention, we conclude that the true value of daily routine CXRs in our multidisciplinary ICU is very low. Interestingly, only in medical patients did the number of CXRs that showed an unexpected predefined major abnormality increase after elimination of daily routine CXRs. The reason for this finding remains unexplained. The distribution of abnormalities encountered on CXRs of these patients was similar in the two study phases; however, neither readmission rate nor differences in raw or risk-adjusted ICU and hospital mortality rates of medical patients was affected by the change in CXR practice.

on for this finding remains unexplained. The distribution of abnormalities encountered on CXRs of these patients was similar in the two study phases; however, neither readmission rate nor differences in raw or risk-adjusted ICU and hospital mortality rates of medical patients was affected by the change in CXR practice. One interesting finding was the decrease in abnormalities presumed to be present on CXRs. Indeed, a 30% reduction in expected predefined findings was observed in phase 2. This finding remains unexplained and we can only speculate on its cause. Firstly, it may be that physicians learned from experience that many of their expectations proved to be untrue during the actual carrying out of the study. This may have caused them to be more reluctant in scoring for expected findings. Alternatively, physicians may have become less enthusiastic about the study, which might have resulted in failure to comply with study rules at some moments (i.e., they did not fill in the back of the formal CXR request form); however, there was no change in expectations of physicians regarding abnormalities that truly led to a change in therapy. More importantly, if the backside of the formal forms were not filled out, as a rule the CXR was simply not obtained. Indeed, collection of data was complete regarding this issue, there were no on demand CXRs without a completed form.

ctations of physicians regarding abnormalities that truly led to a change in therapy. More importantly, if the backside of the formal forms were not filled out, as a rule the CXR was simply not obtained. Indeed, collection of data was complete regarding this issue, there were no on demand CXRs without a completed form. Our study has, at least partially, overlap with two other studies [7, 8]. Price et al. performed a nonrandomized controlled study on the financial impact of elimination of daily routine CXRs [7]. They showed that elimination of daily routine CXRs in a pediatric ICU resulted in decreased variability in ordering practice, fewer CXRs per patient, and an accompanying cost savings, while not influencing LOS. In addition, cost reduction with the change in radiology policy was significant in their study. This is in line with our results, since we found a substantial decline in radiology costs (see ESM). Besides the fact that this study was performed in a pediatric ICU, making generalization of study results difficult, their study did not include all patient categories. Indeed, postoperative cardiovascular surgery patients continued to receive daily routine CXR. We specifically included this patient group in our study because cardiovascular patients form one of the largest categories in many adult ICU. Krivopal et al. performed a randomized controlled trial to determine whether there is any difference in diagnostic, therapeutic, and outcome efficacy between protocols utilizing daily routine CXRs and those utilizing on demand CXRs in mechanically ventilated patients [8]. In their study a daily routine CXR strategy compared with an on demand CXR strategy was not associated with a negative effect on LOS or mortality; however, this study was small, including not more than 94 patients.

utilizing daily routine CXRs and those utilizing on demand CXRs in mechanically ventilated patients [8]. In their study a daily routine CXR strategy compared with an on demand CXR strategy was not associated with a negative effect on LOS or mortality; however, this study was small, including not more than 94 patients. We did not collect information on less evident findings on CXRs. Less evident findings (such as atelectasis less than two lobes, infiltrates less than one lobe, or small pleural effusions [5]), however, might still influence daily management of ICU patients. Since LOS in ICU was not altered for the whole group, readmission rate and mortality rate remained unchanged after the intervention, we suggest that changes of less evident CXR findings are not at all important, at least in our ICU. In other ICUs, such as open-format ICUs, less evident findings might be of more clinical importance, however; therefore, our results must be interpreted with caution, it might be that our results are not easily translated to other types of ICU.

of less evident CXR findings are not at all important, at least in our ICU. In other ICUs, such as open-format ICUs, less evident findings might be of more clinical importance, however; therefore, our results must be interpreted with caution, it might be that our results are not easily translated to other types of ICU. Several important drawbacks of our study must be mentioned. Firstly, our study did not include a strict method for tracking complications as a result of elimination of daily routine CXRs. Indeed, several abnormalities might have been missed (or discovered too late) which might (or do) have impact on clinical outcome. Examples of these types of abnormalities include pneumothorax causing weaning problems, the malposition of devices such as central venous lines, causing extravasation of fluid, or orotracheal tubes, potentially causing injury to the vocal cords. Considering these examples, such a strict method may mandate a daily check of all invasive devices. Although possible complications of elimination of daily routine CXRs could be discussed in daily bedside rounds, daily radiology conferences and daily multidisciplinary meetings during the performance of our study, no clinically important complications were reported as the result of elimination of daily routine CXRs; thus, although we assume that the elimination of daily routine CXRs does not cause any complications, we cannot be certain that this was truly the case. Secondly, as mentioned previously, it is of importance to realize that results that come from one center may simply not be similar for other centers: differences in staffing; especially during off-hours, and differences in case mix may be of great influence on outcome when abandoning daily routine CXRs. Thirdly, as mentioned previously, we found a reduction in expected predefined finding in phase 2. We assumed that the cause of this reduction might be that the physicians became less enthusiastic about the study, which might be seen as a limitation of the study.

influence on outcome when abandoning daily routine CXRs. Thirdly, as mentioned previously, we found a reduction in expected predefined finding in phase 2. We assumed that the cause of this reduction might be that the physicians became less enthusiastic about the study, which might be seen as a limitation of the study. Conclusion In conclusion, in our mixed medical–surgical ICU elimination of daily routine CXRs leads to a sharp decline in the total number of CXRs, while only minimally increasing the number of on demand CXRs. Although we cannot be certain whether we missed important findings by abandoning daily routine CXRs, its elimination did neither affect LOS in ICU, nor readmission rate and ICU and hospital mortality rates. Electronic supplementary material Electronic Supplementary Material (DOC 165K) Electronic supplementary material The online version of this article (doi:10.1007/s00134-007-0542-1) contains supplementary material, which is available to authorized users

Sir, We read with interest the paper by Sprung et al. on attitudes of European physicians, nurses, patients, and families regarding end-of-life decisions [1]. The study compares preferences for end-of-life decisions between the above-mentioned groups from six European countries with different societal backgrounds. We agree with the authors that it is remarkable that one-third to half of the respondent groups wanted active euthanasia for pain, both when they would be diagnosed with a terminal illness and when they would be permanently unconscious; however, we question the interpretation of the answers for the second scenario, as in the questionnaire, active euthanasia was defined as the hastening of death at the patient's explicit request. In our opinion, this scenario is theoretical, since an unconscious patient cannot explicitly request euthanasia. In such a situation, alleviation of pain, without the intention of hastening death, is a much more plausible scenario. In all participating countries, the law does not allow hastening of death without the patient's explicit request. In addition, the numbers probably represent an underestimation of the respondents' desire for euthanasia, as studies have shown that pain is not the only reason for a euthanasia request [2, 3]. In these studies, euthanasia requests were typically related to patients' sense of suffering without improvement and loss of dignity. Rietjens et al. showed that in 36% of the patients that received euthanasia, pain was the main reason for their request [2].

pain is not the only reason for a euthanasia request [2, 3]. In these studies, euthanasia requests were typically related to patients' sense of suffering without improvement and loss of dignity. Rietjens et al. showed that in 36% of the patients that received euthanasia, pain was the main reason for their request [2]. Regrettably, a multivariate logistic regression analysis was only presented for value of life vs quality of life, and the desire to go into the ICU. Especially for the euthanasia scenario, it would be interesting to compare countries, as it is known that marked differences between countries exist [4, 5]. An author's reply to this comment is available at: http://dx.doi.org/10.1007/s00134-007-0571-9

Sir: It is hypothesized that increased intestinal permeability can induce or enhance septic complications in intensive care patients by facilitating bacterial translocation. A reliable and safe detection method would aid in identifying patients with increased intestinal permeability. Intestinal permeability has frequently been measured by tests based on the differential sugar absorption principle. In these tests the ratio of urinary recovery after orally administration of a small permeant sugar probe and large sugar probe, impermeant in the uncompromised intestine, is used as an indication of intestinal permeability [1]. The principle of these tests is that premucosal factors (i.e., gastric retention) and postmucosal factors (i.e., metabolism and renal function) are excluded because these should affect both probes similarly.Therefore only mucosal factors (i.e., intestinal permeability) is indicated. The most commonly used test is the lactulose mannitol test (LMT). Oudemans-van Straaten et al. [2] identified confounding factors when performing the LMT in ICU patients. Their study was conducted in severely ill patients with multiple organ failure. The LMT could still be applicable in patients with milder disease.

d. The most commonly used test is the lactulose mannitol test (LMT). Oudemans-van Straaten et al. [2] identified confounding factors when performing the LMT in ICU patients. Their study was conducted in severely ill patients with multiple organ failure. The LMT could still be applicable in patients with milder disease. We performed the LMT on trauma patients admitted to an ICU, including all patients, with a variety of injury severity. Thirteen trauma patients were included who underwent three tests each. The trauma patients' median Injury Severity Score was 24 (range 16–38). Renal function was within normal range in all patients (creatinine < 120 μmol/l; ureum < 7.5 μmol/l). In 19 tests (61%) confounding factors were identified. Of the confounding factors 53% were therapy related (i.e., mannitol use as therapeutic agent or in saline adenosine glucose mannitol). In the remaining 47% of biased tests, the factors consisted of administering problems (i.e., stomach retention). Thus two groups of confounding factors cause a problem: (a) the use of test substances for clinical applications and (b) administering problems with the test solution. The first problem affects the probes differently, thus causing invalid test results. This can be overcomeby discarding mannitol as probe. The administering problem affects the two probes similarly, and thus the ratio remains constant. In the case of gastric retention probes do not reach the intestine and cannot be measured in urine. A solution would be to administer the test fluid in the small intestine by a nasoduodenal tube.

scarding mannitol as probe. The administering problem affects the two probes similarly, and thus the ratio remains constant. In the case of gastric retention probes do not reach the intestine and cannot be measured in urine. A solution would be to administer the test fluid in the small intestine by a nasoduodenal tube. We conclude, and stress this as a warning to future researchers, that the LMT is inappropriate not only for patients with multiple organ failure but also for trauma patients in an ICU setting (without organ failure).

Sir: Recently Ciroldi et al. reported on the ability of family members to predict patient's consent to critical care research [1]. They showed that family members designated to serve as surrogate decision makers fail to accurately consent to research in one-third to one-half of cases.

Sir: Recently Ciroldi et al. reported on the ability of family members to predict patient's consent to critical care research [1]. They showed that family members designated to serve as surrogate decision makers fail to accurately consent to research in one-third to one-half of cases. We would like to recall a previous report on the same topic. In 2005 we asked representatives of 714 patients to grade their competence in giving surrogate informed consent for the temporarily incapable patient, from 1 (“very incompetent”) to 10 (“very competent”) [2]. This was done for two hypothetical situations: (a) in the case that a medical decision had to be made (i.e., the performance of a surgical procedure) and (b) in the event that the patient was considered to be a potential candidate to participate in critical care research (i.e., participation in a randomized controlled trial comparing a new therapy with standard care). If patients regained competence before discharge from the intensive care unit (ICU), they were asked to grade their representative for the task of giving informed consent in the same way. The most important finding of this study was that patients felt their surrogate decision makers were very capable of representing them in respect of medical decisions as well as participation in research [patients graded their surrogate decision makers 8.0 (interquartile range 7.0–9.0) for both decisions]; there was no significant difference in rating of the patients for the different representatives (spouses, parents, children, siblings, and others). We also found that representatives felt very confident in their ability to represent the patients [they graded themselves 8.0 (7.0–9.0)]; while there was a significant difference in confidence for the task of representing the patient in participation in research between spouses and parents at 8.0 (7.0–9.0), children at 7.5 (6.0–9.0), siblings at 7.0 (5.5–7.0), and others at 6.0 (5.5–7.5) (p = 0.02), confidence level concerning medical decisions were similar for the different representatives.

ce in confidence for the task of representing the patient in participation in research between spouses and parents at 8.0 (7.0–9.0), children at 7.5 (6.0–9.0), siblings at 7.0 (5.5–7.0), and others at 6.0 (5.5–7.5) (p = 0.02), confidence level concerning medical decisions were similar for the different representatives. The differences between the two studies are remarkable. One vital dissimilarity between our study and the one by Ciroldi et al. was timing of consent. Indeed, while Ciroldi et al. obtained their data at the time of discharge from the ICU, we acquired data early after admission to the ICU. It remains uncertain, however, whether this really explains the differences between the two studies. Ciroldi et al. hypothesize that patients and surrogates may be less inclined to give consent in a hypothetical situation than in real life. Although this may be the case for France, our data at least suggest this not to be the case for the Netherlands. Another difference between France and the Netherlands appears to speak against the suggestion that formal designation of surrogates improves the representation of patient's wishes: in contrast to France, where 10% of representatives are formal designated surrogates, in the Netherlands only 0.5% of the patients in our study had a legal representative. The different results from these two studies underline the importance of studying this topic in a broader context: differences between countries should be recognized when changing legislation.

Introduction Delirium is a neuropsychiatric disorder secondary to a general medical condition, and must be considered a serious complication of the underlying disease or its treatment. In the revised fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV-TR®), delirium is defined by four concurrent diagnostic criteria: (1) acute onset and fluctuation of (2) a disturbance of consciousness with reduced ability to focus, shift or maintain attention and (3) a change of cognition with memory deficit, disorientation, language disturbance, perceptual disturbances or hallucinations, (4) caused by the direct physiological consequences of a general medical condition [1]. It is frequently seen in critically ill adult and geriatric patients [2–5] and is associated with a poor prognosis, reflected by longer hospital stay, worse functional and cognitive outcome, and a higher mortality rate after discharge from hospital [3]. In mechanically ventilated critically ill adults, delirium is an independent predictor of elevated 6-month mortality and a longer hospital stay [6]. If appropriate diagnostic tools validated for bedside use by non-psychiatrists [e.g. Delirium Rating Scale (DRS), Confusion Assessment Instrument for the Intensive Care Unit (CAM-ICU)] are used, delirium is diagnosed in over 80% of critically ill adult patients [7]. Thus, systematic monitoring for delirium and appropriate treatment with haloperidol in critically ill adult patients were included in the recently published clinical practice guidelines for sedatives and analgesia of the Society of Critical Care Medicine [8]. However, the optimal management of patients with delirium and the effects of the pharmacological treatment on the outcome are still key concerns for today [9]. Given lack of age-appropriate diagnostic criteria and assessment tools in children, even less is known about the incidence, clinical presentation, response to treatment and consequences of childhood delirium in general, and in critically ill children in particular [10–12].

outcome are still key concerns for today [9]. Given lack of age-appropriate diagnostic criteria and assessment tools in children, even less is known about the incidence, clinical presentation, response to treatment and consequences of childhood delirium in general, and in critically ill children in particular [10–12]. The few available published data on childhood delirium suggest that morbidity and mortality are higher in children with than in children without delirium [13]. Therefore, delirium in children should be considered a serious complication and be treated accordingly. Unfortunately, while there are comprehensive guidelines on the diagnosis and treatment of delirium in adults, clinical guidelines for delirium in children are nonexistent. The aim of this study was to investigate the incidence, patient and population characteristics, clinical presentation and response to treatment of delirium in a cohort of critically ill children admitted to a tertiary pediatric intensive care unit (PICU). Given the necessarily multidisciplinary approach to assessment and treatment of these children, the input of four disciplines – child psychiatry, pediatric intensive care medicine, child neurology and adult neuropsychiatry – was used.

critically ill children admitted to a tertiary pediatric intensive care unit (PICU). Given the necessarily multidisciplinary approach to assessment and treatment of these children, the input of four disciplines – child psychiatry, pediatric intensive care medicine, child neurology and adult neuropsychiatry – was used. Methods Design, setting and patients A descriptive study was carried out over a 4-year period (January 2002 to December 2005) in an eight-bed tertiary PICU. This PICU is a tertiary referral center for both general and surgically critically ill children in the southeastern region of the Netherlands (population 1.4 million, 350 annual admissions). Critically ill children, acutely, non-electively and consecutively admitted, were prospectively sampled. Both mechanically ventilated and non-ventilated patients were included.

h general and surgically critically ill children in the southeastern region of the Netherlands (population 1.4 million, 350 annual admissions). Critically ill children, acutely, non-electively and consecutively admitted, were prospectively sampled. Both mechanically ventilated and non-ventilated patients were included. Diagnostic approach All children with (1) confusion, agitation, anxiety, moaning, discomfort, or behavioral disturbances with no acceptable medical explanation or (2) failure of standard analgosedative treatment were systematically assessed for the presence of delirium in a two-step diagnostic approach. The standard analgosedative treatment can be summarized as follows: children who required analgosedation because of obvious or expected pain or because of stress related to their underlying disease or treatment received adequate doses of opioids and/or benzodiazepines according to internationally published guidelines for analgesia and sedation in critically ill children [14]. Drug doses were individually tailored to achieve optimal patient comfort and were slowly reduced in order to avoid a withdrawal syndrome. In the event that a withdrawal syndrome was suspected, based on clinical observation or the revised Finnegan score, specific treatment with long-acting benzodiazepines (e.g. lorazepam) or opioids (e.g. methadone) was started according to internationally published guidelines [15, 16].

der to avoid a withdrawal syndrome. In the event that a withdrawal syndrome was suspected, based on clinical observation or the revised Finnegan score, specific treatment with long-acting benzodiazepines (e.g. lorazepam) or opioids (e.g. methadone) was started according to internationally published guidelines [15, 16]. At the time the assessment for delirium was initiated, none of the patients had signs of imminent life-threatening respiratory, circulatory or neurological failure, while ongoing asphyxia, respiratory acidosis, metabolic disturbances, fighting the ventilator due to inappropriate ventilator settings and withdrawal syndrome all had been excluded systematically as an explanation for the observed behavior.

of imminent life-threatening respiratory, circulatory or neurological failure, while ongoing asphyxia, respiratory acidosis, metabolic disturbances, fighting the ventilator due to inappropriate ventilator settings and withdrawal syndrome all had been excluded systematically as an explanation for the observed behavior. The first step of the diagnostic approach was a systematic assessment by a child neuropsychiatrist (J.S.) using DSM-IV criteria for delirium. Criteria were evaluated on the basis of (1) hetero-anamnestic information from parents, nurses, intensivists, and child neurologists about behavior and behavioral changes and (2) child psychiatric examination. Based on the findings, patients were categorized as having a (probable) delirium or not. In a second step, the provisional diagnosis of delirium was further tested in a daily multidisciplinary consensus meeting. The team consisted of the child neuropsychiatrist, the attending pediatric intensivist, and occasionally a geriatric neuropsychiatrist specialized in delirium in geriatric patients and/or a child neurologist. If this team agreed that alternative explanations for a child's behavior were unlikely, the consensus diagnosis was delirium.

sted of the child neuropsychiatrist, the attending pediatric intensivist, and occasionally a geriatric neuropsychiatrist specialized in delirium in geriatric patients and/or a child neurologist. If this team agreed that alternative explanations for a child's behavior were unlikely, the consensus diagnosis was delirium. Based on the dominant clinical presentation, cases of delirium were classified as “hyperactive” when psychomotor agitation was present, and as “hypoactive” when retardation and/or inhibition was present. A number of children presented with cognitive and/or attentional disturbances in the context of severe anxiety states, often accompanied with moaning and restlessness, but without clear agitation or retardation. This latter group was classified as “emerging” or “veiled” delirium [17], in reference to the way delirious syndromes have been described as “partial delirium” in adult ICU patients or as “subsyndromal delirium” in elderly medical patients [7, 18, 19]. The different presentation forms were not always clear-cut, and some fluctuated dramatically over time. The severity of illness was scored according to the Pediatric Index of Mortality (PIM) and Pediatric Risk of Mortality (PRISM) [20].

ICU patients or as “subsyndromal delirium” in elderly medical patients [7, 18, 19]. The different presentation forms were not always clear-cut, and some fluctuated dramatically over time. The severity of illness was scored according to the Pediatric Index of Mortality (PIM) and Pediatric Risk of Mortality (PRISM) [20]. Therapeutic approach Whenever delirium was identified or suspected, a two-track treatment approach consisting of both psychosocial and pharmacological interventions was implemented. Psychosocial interventions – the parents' presence and comforting throughout the day (and night), familiar music, favorite toys, pictures of home and pets, friends, school, lighting schedules, sometimes even fragrances – are standard in the PICU. The parents also received an information leaflet on childhood delirium [21]. All patients were also treated with antipsychotic medication after the referring pediatric intensivist had agreed and the parents, because of the off-label use, had given informed consent, which was never refused. In children with psychomotor agitation that was acutely threatening to their health status, haloperidol at a loading dosage of 0.15–0.25 mg i.v. was used, given slowly over a period of 30–45 min, followed by a maintenance dose of 0.05–0.5 mg/kg/24 h i.v. [11, 22, 23]. In less acute situations, and when oral medication was possible, risperidone at a loading dose of 0.1–0.2 mg p.o. was used, followed by a maintenance dose of 0.2–2.0 mg/24 h p.o. as the treatment of choice. Clinical response and side effects were recorded by the child neuropsychiatrist and the pediatric intensivists. In order to tailor the treatment for delirium, daily discussions were held with the multidisciplinary team. Adjustment of treatment was based on the clinical observations and judgements of the parents, nurses, intensivists and child psychiatrist.

e recorded by the child neuropsychiatrist and the pediatric intensivists. In order to tailor the treatment for delirium, daily discussions were held with the multidisciplinary team. Adjustment of treatment was based on the clinical observations and judgements of the parents, nurses, intensivists and child psychiatrist. Children were followed up for 6 weeks after discharge from the hospital either at the outpatient clinic or by contacting the parents by telephone. As the study solely involved the structured recording of routine clinical practice, under Dutch law no institutional review board approval was required. Results From January 2002 to December 2005, there were 877 acute, non-elective admissions to the PICU. Distribution of age and gender are shown in Table 1. In 61 cases (7%), a systematic assessment by a child neuropsychiatrist was requested, usually for agitation, anxiety, moaning, discomfort, behavioral disturbance or problematic analgosedation. Table 1 Number and incidence of delirium in the total sample* by age and gender Age Total sample* Patients with delirium Incidence (%) 0–2.99 years 513 14 2.7 Male 310 9 2.9 Female 203 5 2.5 3–5.99 years 106 4 3.8 Male 56 3 5.4 Female 50 1 2.0 6–8.99 years 80 6 7.5 Male 46 1 2.2 Female 34 5 14.7 9–11.99 years 77 3 3.9 Male 61 3 4.9 Female 16 0 0 12–14.99 years 70 7 10 Male 35 5 14.3 Female 35 2 5.7 15–18 years 31 6 19.4 Male 13 4 30.8 Female 18 2 11.1 Total 877 40 4.6 Male 521 25 4.8 Female 356 15 4.2 *Critically ill children, acutely, non-electively and consecutively admitted during a 4-year period

34 5 14.7 9–11.99 years 77 3 3.9 Male 61 3 4.9 Female 16 0 0 12–14.99 years 70 7 10 Male 35 5 14.3 Female 35 2 5.7 15–18 years 31 6 19.4 Male 13 4 30.8 Female 18 2 11.1 Total 877 40 4.6 Male 521 25 4.8 Female 356 15 4.2 *Critically ill children, acutely, non-electively and consecutively admitted during a 4-year period Of these 61 patients, 40 (61%) were diagnosed with delirium, yielding a cumulative incidence of 5% (boys 5%; girls 4%). Age-specific incidences increased from 3% in the 0–3 years age group (boys 3%; girls 3%), to 19% in the 16–18 years age group (boys 31%; girls 11%) (Table 1, Fig. 1). Fig. 1 Incidence of delirium in the sub-groups by age and gender The child psychiatric disorders diagnosed in the 61 referrals are summarized in Table 2. Table 2 Child psychiatric diagnosis at the first consultation (n = 61) n Diagnosis 40 delirium 5 adjustment disorders with anxiety and depressed mood, post operative 4 psychological–psychiatric factors affecting the medical condition 3 anxiety disorder 3 emotional and behavioral problems during chronic ventilation 2 adjustment disorders with depressed mood 1 mood disorder 1 adjustment disorder with anxiety 1 sleeping problem 1 feeding problem Table 3 summarizes the population characteristics of the sample diagnosed with delirium, while Table 4 lists the underlying somatic disease or causative pharmacological treatment. Table 3 Population characteristics of the 40 PICU cases with delirium, 2002–2005

n Diagnosis 40 delirium 5 adjustment disorders with anxiety and depressed mood, post operative 4 psychological–psychiatric factors affecting the medical condition 3 anxiety disorder 3 emotional and behavioral problems during chronic ventilation 2 adjustment disorders with depressed mood 1 mood disorder 1 adjustment disorder with anxiety 1 sleeping problem 1 feeding problem Table 3 summarizes the population characteristics of the sample diagnosed with delirium, while Table 4 lists the underlying somatic disease or causative pharmacological treatment. Table 3 Population characteristics of the 40 PICU cases with delirium, 2002–2005 Characteristics Frequency (total n = 40) Age (mean ± SD) 7.6 ± 5.9 Male n 25 age (mean ± SD) 7.6 ± 6.1 Female n 15 age (mean ± SD) 7.6 ± 5.8 Ethnicity White 36 (90%) African 3 (7.5%) Asian 1 (2.5%) Mechanical ventilation 34 (85%) PIM score (mean ± SD) 9.96 ± 16.20 PRISM score (mean ± SD) 23.54 ± 24.94 Major somatic pharmacological features Recent increase or decrease of analgosedative medication 22 (55%) Neurological disorders 21 (52%) Infectious disorders 20 (50%) Respiratory disorders 12 (30%) PIM, Pediatric Index of Mortality; PRISM, Pediatric Risk of Mortality Table 4 Patient characteristics of the 40 PICU cases with delirium 2002–2005 No.

Characteristics Frequency (total n = 40) Age (mean ± SD) 7.6 ± 5.9 Male n 25 age (mean ± SD) 7.6 ± 6.1 Female n 15 age (mean ± SD) 7.6 ± 5.8 Ethnicity White 36 (90%) African 3 (7.5%) Asian 1 (2.5%) Mechanical ventilation 34 (85%) PIM score (mean ± SD) 9.96 ± 16.20 PRISM score (mean ± SD) 23.54 ± 24.94 Major somatic pharmacological features Recent increase or decrease of analgosedative medication 22 (55%) Neurological disorders 21 (52%) Infectious disorders 20 (50%) Respiratory disorders 12 (30%) PIM, Pediatric Index of Mortality; PRISM, Pediatric Risk of Mortality Table 4 Patient characteristics of the 40 PICU cases with delirium 2002–2005 No. Sex Age Primary diagnosis on admission PICU Mechanical ventilation Delirium type Treatment 1 M 3 months Multiple congenital malformations + Emerging Haloperidol 2 F 4 months Meningococcal septic shock + Emerging Risperidone 3 M 4.5 months Severe CLD + Emerging Risperidone 4 F 10 months Near drowning + Emerging Haloperidol 5 F 1 year Pneumonia + Emerging Haloperidol 6 M 1 year Sepsis due to perforated appendicitis + Emerging Haloperidol 7 M 1 year Subarachnoid bleeding + Emerging Haloperidol 8 M 1 year Pericardial effusion with pretamponade + Hyperactive Haloperidol 9 M 1 year Multiple dysmorphia, upper airway obstruction – Emerging Haloperidol 10 F 2 years Meningococcal septic shock + Hyperactive Haloperidol 11 M 2 years ADEM + Hyperactive Risperidone 12 M 2 years Cervical mass, upper airway obstruction + Emerging Haloperidol 13 M 2 years Aspiration and pneumothorax + Hypoactive – 14 F 2 years Meningococcal meningitis with sepsis and DIC + Hyperactive Haloperidol 15 F 3 years Cystic fibrosis and pneumonia + Hypoactive Haloperidol 16 M 4 years Intracerebral hemorrhage, Marfan syndrome + Hypoactive Haloperidol 17 M 5 years Medulloblastoma post surgery + Emerging Risperidone 18 M 5 years Upper respiratory tract infection + Hyperactive Haloperidol 19 M 6 years Multiple trauma + Emerging Risperidone 20 F 8 years Meningo-encephalitis + Hyperactive Risperidone 21 F 8 years Viral encephalitis + Hyperactive – 22 F 9 years Status asthmaticus + Hyperactive Risperidone 23 F 9 years TBI, gunshot wound + Hypoactive Risperidone 24 M 9 years Status asthmaticus + Hyperactive Haloperidol 25 M 9 years Neural tube defect and drain dysfunction – Emerging first Haloperidol, then Risperidone 26 M 11 years Hypovolemic shock, typhus abdominalis – Hyperactive Haloperidol 27 F 12 years TBI + Hypoactive Haloperidol 28 M 12 years TBI and fracture of lower leg + Emerging Haloperidol 29 M 13 years Sepsis, paronychia – Hyperactive Haloperidol 30 M 13 years Status epilepticus + Emerging Haloperidol 31 F 14 years TBI + Hyperactive Haloperidol 32 F 15 years Post TBI + Hypoactive Haloperidol 33 M 15 years Postoperative state + Hypoactive Risperidone 34 M 15 years Acute lymphoblastic leukemia – Hyp

ridol 29 M 13 years Sepsis, paronychia – Hyperactive Haloperidol 30 M 13 years Status epilepticus + Emerging Haloperidol 31 F 14 years TBI + Hyperactive Haloperidol 32 F 15 years Post TBI + Hypoactive Haloperidol 33 M 15 years Postoperative state + Hypoactive Risperidone 34 M 15 years Acute lymphoblastic leukemia – Hyp oactive Haloperidol 35 M 15 years TBI + Emerging Haloperidol 36 F 15 years Status asthmaticus + Hyperactive Haloperidol 37 M 16 years Multiple trauma – Hypoactive Haloperidol 38 F 16 years Bacterial meningitis + Hyperactive Risperidone 39 M 16 years Respiratory failure, Duchenne disease + Emerging Haloperidol 40 M 17 years Septic shock + Emerging Haloperidol CLD, chronic lung disease; ADEM, acute disseminated encephalomyelitis; DIC, diffuse intravascular coagulation; TBI, traumatic brain injury The underlying features were: recent increase or decrease in analgosedative medication (n = 22), neurological disorders (n = 21), infections (n = 20) and respiratory disorders (n = 12). Usually, a combination of these existed.

oactive Haloperidol 35 M 15 years TBI + Emerging Haloperidol 36 F 15 years Status asthmaticus + Hyperactive Haloperidol 37 M 16 years Multiple trauma – Hypoactive Haloperidol 38 F 16 years Bacterial meningitis + Hyperactive Risperidone 39 M 16 years Respiratory failure, Duchenne disease + Emerging Haloperidol 40 M 17 years Septic shock + Emerging Haloperidol CLD, chronic lung disease; ADEM, acute disseminated encephalomyelitis; DIC, diffuse intravascular coagulation; TBI, traumatic brain injury The underlying features were: recent increase or decrease in analgosedative medication (n = 22), neurological disorders (n = 21), infections (n = 20) and respiratory disorders (n = 12). Usually, a combination of these existed. All but two patients were treated with an antipsychotic drug. Twenty-seven children were given haloperidol, 10 risperidone, and 1 child received both drugs in succession. In most cases, the beneficial results were observed rapidly, especially in the hyperactive forms, sometimes even after a single dose [11]. Sometimes it took a few hours or days. Two patients experienced acute dystonia as a likely side effect of the haloperidol, responding well to biperidene. Two children received no medication: one because of lack of consensus in our expert team, and one because of an endotracheal intubation at the time that medication was indicated. In most cases, the medication was stopped or tapered off successfully during hospitalization or afterwards in an outpatient setting. Five children (12.5%) with delirium died of their underlying disease; the mean PIM was 10% and the mean PRISM 24% (Table 3).

tracheal intubation at the time that medication was indicated. In most cases, the medication was stopped or tapered off successfully during hospitalization or afterwards in an outpatient setting. Five children (12.5%) with delirium died of their underlying disease; the mean PIM was 10% and the mean PRISM 24% (Table 3). Discussion This is the first systematic multidisciplinary study of the phenomenology and treatment of delirium in 40 critically ill children in a PICU context. The low cumulative incidence of 5% is mainly the result of the low incidence in the younger age groups (< 9 years old), this segment constituting the majority of the total sample (80%). A clearly higher incidence is seen in the older age groups. However, in critically ill adult patients substantially higher incidences have been reported, ranging from 10–30% in general hospital settings to 50% in postoperative patients and up to 80% in the terminally ill [2].

ing the majority of the total sample (80%). A clearly higher incidence is seen in the older age groups. However, in critically ill adult patients substantially higher incidences have been reported, ranging from 10–30% in general hospital settings to 50% in postoperative patients and up to 80% in the terminally ill [2]. There are several possible explanations for this difference. First, the incidence of delirium in young critically ill children may be truly low: differences in age-related resilience and underlying conditions may contribute to true differences in the incidence between the very young and the very old. However this explanation seems unlikely, given the tendency in the very young to develop delirium under even much less severe circumstances [1, 11, 24]. A second explanation may relate to the fact that extensive psychosocial interventions are provided as a routine in Dutch PICU settings, with a possibly preventive effect on delirium in much the same way as observed in geriatric patients [25]. A third factor may be an anti-delirium effect of the routinely used analgosedative medication, although especially benzodiazepines may have excitatory side effects in children. A fourth factor may relate to reluctance on the part of the intensivists and/or child neurologists to request psychiatric evaluation in the case of behavioral changes for fear of adding stigmatization to an already burdened system. Perhaps the fifth, most likely and important explanation is that parents, nurses, pediatric intensivists and child neurologists do not easily recognize delirium, because the medical condition of these critically ill PICU children is so complex and changeable. It is possible that a psychiatric consultation was readily requested only in cases of anxiety and/or hyperactive delirium, not in the hypoactive and or veiled ones.

and child neurologists do not easily recognize delirium, because the medical condition of these critically ill PICU children is so complex and changeable. It is possible that a psychiatric consultation was readily requested only in cases of anxiety and/or hyperactive delirium, not in the hypoactive and or veiled ones. The differential diagnosis of pediatric delirium consists of acute stress reactions, acute anxiety states, adjustment disorders with mixed emotions, dissociative and/or regressive states and childhood-onset psychosis (see also Table 2). However, differentiating delirium from extreme stress and agitation due to acute and life-threatening conditions is not only impossible, it is also unwanted, because it is usually irrelevant at that point of time in the process of medical care. Causative treatment, if possible, is always the first step to be taken. Furthermore, the diagnosis of delirium in children is complicated by the fact that the criteria for adult delirium are not always easily applicable to children because of important differences in age and developmental levels. The DSM-IV criteria for delirium are not always useful in the case of pediatric delirium, especially in a PICU context. This is a source of concern. Moreover, delirium is not mentioned in the DSM-IV section on child psychiatry. DSM-IV describes as an essential feature of delirium the “disturbance of consciousness” leading to “impairment of the ability to focus, shift and sustain attention”. This, however, is of little relevance in the critically ill in a PICU context, where disturbance of attention is routinely present due to the disorder(s) itself. In fact, attention is often the first “to go” [17, 26]. In addition, patients almost always require treatment with opioids and benzodiazepines, which also have a strong impact on attention. Furthermore, it has been hypothesized that a disturbance of consciousness is not a discriminating characteristic of delirium in an ICU setting [27].

is often the first “to go” [17, 26]. In addition, patients almost always require treatment with opioids and benzodiazepines, which also have a strong impact on attention. Furthermore, it has been hypothesized that a disturbance of consciousness is not a discriminating characteristic of delirium in an ICU setting [27]. Our case series suggests that in addition to the hyper- and hypoactive forms of delirium, a third form may be characterized by anxiety, moaning, and/or restlessness. This was referred to as an “emerging” or veiled delirium, as described previously [17]. In the PICU population this state did not develop into a frank hyper- or hypoactive form, but appeared to exist in its own right, accompanied by disturbances of consciousness and cognition. Although atypical presentations of disorders are often not captured in classification systems, the high prevalence of “emerging” delirium in our sample (17/40) stresses the importance of further phenomenological study. Adhering too strongly to DSM-IV criteria for adult delirium, for clinical use in a PICU context, may result in persistent under-diagnosis. On the other hand, the incidence of delirium may be overestimated by using the CAM-ICU, so most important is the issue of what constitutes delirium in critical illness [28].

study. Adhering too strongly to DSM-IV criteria for adult delirium, for clinical use in a PICU context, may result in persistent under-diagnosis. On the other hand, the incidence of delirium may be overestimated by using the CAM-ICU, so most important is the issue of what constitutes delirium in critical illness [28]. There are accumulating indications that pediatric delirium can be subtle and accompanied or even dominated by other neuropsychiatric signs such as: reduction of awareness of the caregiver and/or the surrounding environment, purposeless actions, restlessness, inconsolability, signs of autonomic dysregulation and other subtle higher cortical dysfunctions [29–32]. Parents and nurses have a way of discriminating patterns in their children that may be diagnostically important. Thus, parents sometimes state: “This is not my child anymore.” Pediatric delirium therefore may have various subtle presentations and can be considered a spectrum disorder, which makes it sometimes difficult to diagnose [33]. Neither haloperidol nor risperidone is registered for the treatment of childhood delirium, although both are mentioned as the treatment of choice for adults [1]. Moreover, haloperidol is not registered for i.v. administration, even though it is used in this way in many places. We prefer risperidone in non-acute situations because of the theoretically lower risk of extrapyramidal side effects. Haloperidol and risperidone have been used for other indications in young children as well, such as childhood psychosis [34] or aggression in autism [35]. There are two limitations regarding our observations of treatment response. First, no severity scale for pediatric delirium was used, because none exists for this PICU population. Second, because no studies on the natural course of childhood delirium exist that have established the rate of spontaneous remission, spontaneous remissions may have been misclassified as response to treatment. In our opinion, however, the time frame of response points towards a medication effect.

this PICU population. Second, because no studies on the natural course of childhood delirium exist that have established the rate of spontaneous remission, spontaneous remissions may have been misclassified as response to treatment. In our opinion, however, the time frame of response points towards a medication effect. Most medications employed in pediatrics and child psychiatry are used off-label [36], which means that special attention should be paid to information and informed consent procedures. Given the relatively high incidence of extrapyramidal symptoms with haloperidol, the “off-label” use needs further study [37]. This study has several limitations. First, it was a study of routine clinical practice. Observations were based on referrals emerging from care as usual. Although the focus on delirium may have altered referral paths and rates, we did not actively advocate any change, nor did we screen all admissions for delirium. In the absence of a hard clinical indication, no routine blood level measurements were performed to rule out persistent high levels of sedatives as a possible explanation for any neuropsychiatric symptom. Next, the lack of DSM-IV criteria for pediatric delirium and the disputed relevance of its main criterion in a (P)ICU setting make a standardized diagnosis difficult. Finally, treatment was provided in an open setting and based on consensus, especially among child psychiatrist, pediatric intensivist and child neurologist.

t, the lack of DSM-IV criteria for pediatric delirium and the disputed relevance of its main criterion in a (P)ICU setting make a standardized diagnosis difficult. Finally, treatment was provided in an open setting and based on consensus, especially among child psychiatrist, pediatric intensivist and child neurologist. In conclusion, critically ill children in a PICU can develop delirium, with a hyperactive, hypoactive or veiled presentation, despite adequate analgosedation and intensive psychosocial interventions. The approach using an algorithmic structuring and an intensifying of daily clinical care, including the use of multidisciplinary daily consensus meetings, appeared effective in assessing, diagnosing and treating childhood delirium at the PICU. The findings suggest that the incidence is much lower than in adults, but a likely explanation is that delirious states requiring child psychiatric referral are still frequently under-diagnosed. There is also still a great need for developing delirium criteria in critically ill patients, children and adults alike, in a (P)ICU setting. Treatment with haloperidol or risperidone was successful in all patients. Future research is necessary to identify the risk factors for pediatric delirium in a multivariate prospective approach, to develop “easy to use” bedside tools for non-psychiatric trained team members for the early detection of delirium in all pediatric PICU patients, and to study the effects of interventions in a double-blind and ideally placebo-controlled fashion.

k factors for pediatric delirium in a multivariate prospective approach, to develop “easy to use” bedside tools for non-psychiatric trained team members for the early detection of delirium in all pediatric PICU patients, and to study the effects of interventions in a double-blind and ideally placebo-controlled fashion. Disclosure Jim van Os is a speaker or member of the advisory board for Lilly, BMS, Janssen-Cilag and AstraZeneca. He received grant or research support from Lilly, GSK, BMS and AstraZeneca. Albert Leentjens participated in research for Boehringer Ingelheim and is on the advisory board of a study related to Parkinson's disease by the same company. The other authors have no financial relationships to disclose. Acknowledgements We gratefully acknowledge the cooperation with, and help from, our colleagues in the study group on pediatric delirium, especially Kirsten Venrooij and Richel Lousberg, M.Sc., Ph.D, and our colleagues in the PICU: the pediatric intensivists, child neurologists and nurses. J. N. M. Schieveld and P. L. J. M. Leroy contributed equally to this paper.

Case A 51-year-old man was admitted to the emergency room because of progressive lethargy, slurred speech, weakness of the right arm and low-grade fever. Physical examination showed no additional neurological deficits. He had a history of a tick bite 6 months previously which had been treated with antibiotics. Apart from leukocytosis (10.5 × 109/l) and raised C-reactive protein (62 mg/l), additional blood examination was normal. Cerebral contrast-enhanced computed tomography (CT) revealed no abnormalities. Cerebrospinal fluid (CSF) examination showed an elevated cell count of 66 × 106 cells/l (of which 97% were lymphocytes), slightly elevated protein concentration of 1.03 g/l, a normal glucose level of 3.7 mmol/l and an elevated pressure of 29 cmH2O. Case discussion The onset of progressive lethargy over a period of days, slurred speech and the weakness indicates a lesion of the central nervous system. Given the progression of lethargy, the fever and the absence of acute onset, an infectious origin seemed more likely. This was suggested by the CSF examination, showing an elevated cell count, especially lymphocytes. The differential diagnosis consisted of (viral) meningo-encephalitis – most likely caused by enteroviruses, arboviruses or herpesviruses – neuroborelliosis and vasculitis. Antibiotics and antiviral drugs were started.

ly. This was suggested by the CSF examination, showing an elevated cell count, especially lymphocytes. The differential diagnosis consisted of (viral) meningo-encephalitis – most likely caused by enteroviruses, arboviruses or herpesviruses – neuroborelliosis and vasculitis. Antibiotics and antiviral drugs were started. Case Within 2 days, the patient's condition deteriorated into a deep coma requiring mechanical ventilation. Initial treatment consisted of acyclovir and ceftriaxone, anticipating a possible neuroborelliosis and herpes simplex encephalitis. However, repeated blood and CSF cultures were normal, making an infective origin less likely. Serological tests for Lyme disease and repeated polymerase chain reaction (PCR) testing for herpes simplex virus (HSV) were negative. Systemic vasculitis was ruled out. The patient remained in an areactive coma without focal signs and abnormal brainstem reflexes. An electro-encephalogram (EEG) after 1 week showed non-specific diffuse slow activity. Three weeks after admission cerebral gadolinium-enhanced magnetic resonance imaging (MRI) was performed, showing numerous hyperintense lesions, on T2-weighted images, of the brainstem, the left cerebellar hemisphere, the basal ganglia on both sides and in periventricular locations (Fig. 1a–c). Fig. 1 a–c T2-weighted MRI of the brain, demonstrating hyperintense lesions of a brainstem and left cerebellar peduncle, b basal ganglia and c left periventricular region. c–e T2-weighted MRI of the brain 2 months after therapy, demonstrating a decrease in white matter lesions

in periventricular locations (Fig. 1a–c). Fig. 1 a–c T2-weighted MRI of the brain, demonstrating hyperintense lesions of a brainstem and left cerebellar peduncle, b basal ganglia and c left periventricular region. c–e T2-weighted MRI of the brain 2 months after therapy, demonstrating a decrease in white matter lesions Case discussion On ICU admission the main medical problems were (1) respiratory insufficiency requiring mechanical ventilation, (2) progressive loss of consciousness and (3) the unknown origin of the coma. Maintenance of an adequate airway and concomitant aspiration pneumonia necessitated mechanical ventilation. There are numerous possible causes of progressive loss of consciousness, including cerebrovascular accidents, cerebral infections (viral, bacterial, parasites, tuberculosis, Lyme disease), neoplastic and auto-immune diseases (sarcoidosis, vasculitis, SLE). Despite the elevated white cell count and the presence of protein in the CSF, no infectious origin was determined. Screening for HIV and coagulation disorders was negative. Repeated blood and CSF cultures were negative, and there was no response after 14 days of ceftriaxone and acyclovir. Nevertheless CSF analysis after treatment was interpreted cautiously. Many viruses responsible for meningo-encephalitis cannot be identified using common tests. MRI showed multiple hyperintense brain lesions, especially of the brainstem. Because of the massive involvement of the brainstem and persistence of coma for 2 weeks, the prognosis was considered poor.

was interpreted cautiously. Many viruses responsible for meningo-encephalitis cannot be identified using common tests. MRI showed multiple hyperintense brain lesions, especially of the brainstem. Because of the massive involvement of the brainstem and persistence of coma for 2 weeks, the prognosis was considered poor. Case Acute disseminated encephalomyelitis (ADEM) could not be ruled out, so treatment was continued. Additionally, treatment with intravenous dexamethasone (4 mg four times a day) was started. Upon treatment with corticosteroids for 4 weeks the patient made a slow but full recovery within 2 months. After 2 months cerebral gadolinium-enhanced MRI was repeated, showing substantial amelioration of the intracerebral lesions (Fig. 1d–f). One year after treatment the patient still functions well, without relapse. Case discussion Elevated pressure, white cell count and protein level in the CSF in combination with the extensive white matter lesions confirmed the diagnosis of ADEM. ADEM is a rare cause of prolonged coma with complete recovery under specific treatment, and intensivists should be aware of this disease. Withdrawal or withholding of care in patients with prolonged coma can hardly be discussed as long as the cause of the coma remains unknown and the prognosis cannot be accurately established. Treatment with intravenous corticosteroids resulted in a full recovery. The absence of a previous infection is a peculiar aspect of our case.

r withholding of care in patients with prolonged coma can hardly be discussed as long as the cause of the coma remains unknown and the prognosis cannot be accurately established. Treatment with intravenous corticosteroids resulted in a full recovery. The absence of a previous infection is a peculiar aspect of our case. Comments This case illustrates the remarkable outcome after a long period of coma. Good recovery from a coma is observed in only 10% of reported cases and depends on the aetiology, the accompanying clinical signs, and the depth and duration of the coma [1]. ADEM is a rare monophasic illness that is thought to develop from antigenic mimicry, with antibodies having cross-reactivity to host epitopes in the nervous system. The disorder typically occurs following vaccination or a viral prodrome and is predominantly seen in children and in Japan, where viral encephalitis is more common [2, 3, 4, 5]. A preceding infection is reported in 50–75% of cases [3, 4, 5]. The estimated incidence is 0.8 per 100,000 population per year [3]. Some cases of ADEM among adults and the elderly have been reported, but the incidence is estimated to be considerably lower. The diagnosis of ADEM is reached on clinical grounds, evidence of white matter lesions on MRI and exclusion of other causes. These classic features were present in our case. However, diagnosis of ADEM may be difficult. Diagnostic overlap with multiple sclerosis (MS) may lead to underestimation of the prevalence [4, 5]. Some 0–33% of children and 35% of adults initially diagnosed with ADEM will eventually develop MS [4, 5, 6]. Therefore this diagnosis should be considered, especially in adults. Follow-up by repeated MRI, at intervals not shorter than 6 months, may anticipate this development [7]. The differential diagnosis consists of meningo-encephalitis, meningitis and cerebral involvement in auto-immune diseases [8]. In severe cases, ADEM may lead to coma. There is no standard treatment for ADEM. It is a rare disease, and no formal clinical trials of any therapeutic agent have been published. Thus the management of the disease rests on strategies that have appropriate effects on the plausible disease mechanisms. Present treatments rely on immunosuppression and immunomodulation.

s no standard treatment for ADEM. It is a rare disease, and no formal clinical trials of any therapeutic agent have been published. Thus the management of the disease rests on strategies that have appropriate effects on the plausible disease mechanisms. Present treatments rely on immunosuppression and immunomodulation. Intravenous corticosteroids (preferably methylprednisolone) is the first choice of treatment and usually leads to full recovery. Plasma exchange and intravenous immunoglobulin should be considered by deterioration on corticosteroids [9, 10]. In conclusion, treatment of coma should be continued when the cause of the coma is unclear, even if there are numerous lesions in the brain. ADEM should be considered in comatose patients with raised CSF pressure, protein and lymphocytes, negative CSF cultures, and multiple white matter lesions on cerebral MRI. ADEM is predominantly seen in children, but also occurs in adults. Recovery may last weeks. The Case Discussion format is intended to describe instructive cases in the field of intensive care medicine in a stepwise manner, allowing each step to be commented or discussed. This format might help to delineate the different steps of clinical reasoning in intensive care and highlight the important messages conveyed by the case

Sir: With interest we read Mackenzie and Woodhouse's paper about serum C-reactive protein (CRP) levels in patients with and without liver dysfunction during bacteraemia [1]. The authors concluded that patients with biochemical evidence of liver disease show significantly lower serum CRP levels during bacteraemia than patients without liver dysfunction. This remarkable result was only achieved when patients with biopsy-proven liver cirrhosis were included in the analysis. When cirrhotic patients were excluded, however, and patients with liver dysfunction, defined as an elevated plasma bilirubin level and increased prothrombin time, were compared with controls, between-group differences did not reach statistical significance. Therefore, we feel that from the available data it cannot be concluded that patients with liver diseases, in general, generate lower serum CRP levels during bacteraemia.

plasma bilirubin level and increased prothrombin time, were compared with controls, between-group differences did not reach statistical significance. Therefore, we feel that from the available data it cannot be concluded that patients with liver diseases, in general, generate lower serum CRP levels during bacteraemia. As multiple studies have shown greater CRP production in cirrhotic non-infected patients than in healthy controls [2, 3], it could be argued that cirrhotic patients produce less CRP during periods of infection. However, a recent large study in critically ill subjects reported that CRP levels during infection in cirrhotic patients did not significantly differ from CRP levels in non-cirrhotic patients [4]. Also, no differences were found in relation to the severity of the cirrhosis. Because interleukin-6 stimulates hepatocytes to produce CRP, an interesting study to aid the understanding of these apparently conflicting results would be to relate the interleukin-6 to CRP levels in cirrhotic patients. However, as far as we know, this has not been done. Finally, we believe that the use of CRP as a valuable screening test for infection in the ICU has certain limitations. The decision on starting antibiotics must be individualized and not be based on a certain CRP level. It should be based on suspicion of the presence of infection, the stability of the patient's clinical condition, the presence of comorbidity and the risk of death or complications. An author's reply to this comment is available at: http://dx.doi.org/10.1007/s00134-006-0525-7

Introduction Lactic acid was first found and described in sour milk by the Swedish chemist Karl Wilhelm Scheele (1742–1786) in 1780 [1]. The Swedish chemist Jöns Jakob Berzelius (1779–1848) found lactic acid in fluid extracted from meat in 1808 [2, 3], and the German chemist Justus von Liebig (1803–1873), who established the world's first school of chemistry at Giessen, proved that lactic acid was always present in muscular tissue of dead organisms [4]. In 1859, Emil Heinrich du Bois-Reymond (1818–1896) published several articles on the influence of lactic acid on muscle contraction [5–9]. Araki and Zillessen found that if they interrupted oxygen supply to muscles in mammals and birds, lactic acid was formed and increased [10–14]. This was the first demonstration of the relationship between tissue hypoxia and the formation of lactate. The occurrence of increased lactic acid in blood (hyperlactataemia) nowadays reflects severe illness, in which the increased blood lactate levels may result from both anaerobic and aerobic production or from a decreased clearance. It was the German physician–chemist Johann Joseph Scherer who first demonstrated the occurrence of lactic acid in human blood under pathological conditions after death in 1843 and 1851 [15, 17], and Carl Folwarczny in 1858 who first demonstrated lactic acid in blood of a living patient. In this article we wish to honour Scherer's forgotten observations and describe the influence of his finding on further research on lactic acid at the end of the 19th century.

l conditions after death in 1843 and 1851 [15, 17], and Carl Folwarczny in 1858 who first demonstrated lactic acid in blood of a living patient. In this article we wish to honour Scherer's forgotten observations and describe the influence of his finding on further research on lactic acid at the end of the 19th century. Biography of Johann Joseph Scherer Born on 18 March 1814 in Aschaffenburg, Germany, Scherer studied medicine, chemistry, geology and mineralogy at the university of Würzburg between 1833 and 1836. He obtained his PhD in medicine and surgery in 1838 with a thesis entitled “Versuche über die Wirkung einiger Gifte auf verscheidene Thierclassen” (Experiments on the action of some poisons on several classes of animals). He practised medicine in Wipfeld, but inspired by the chemist Ernst von Bibra (1806–1878) he completed his studies in chemistry at the University of Munich between 1838–1840 [18]. In 1840 he was employed at the laboratory of Justus Liebig at Giessen, and became professor at the medical faculty in 1842, professor of organic chemistry in 1847, and later professor of general, anorganic and pharmaceutical chemistry. His work especially concerned quantitative research on blood and urine in pathological conditions. In 1843 he published his book ‘Chemische und Mikroskopische Untersuchungen zur Pathologie angestellt an den Kliniken des Julius-Hospitales zu Würzburg’ (Chemical and microscopic investigations of pathology carried out at the Julius Clinic at Würzburg) [15] (Fig. 1), in which he described 72 case reports, giving details on clinical course, diagnosis, and results obtained during autopsy and analysis of body fluids. Scherer died on 17 February 1869 [18]. Fig. 1 Title page of Scherer's 1843 book

c investigations of pathology carried out at the Julius Clinic at Würzburg) [15] (Fig. 1), in which he described 72 case reports, giving details on clinical course, diagnosis, and results obtained during autopsy and analysis of body fluids. Scherer died on 17 February 1869 [18]. Fig. 1 Title page of Scherer's 1843 book The 1843 cases In one chapter in his 1843 book entitled 'Untersuchungen von krankhaften Stoffen bei der im Winter 1842–1843 in Würzburg und der Umgegend herrschenden Puerperal-Fieber-Epidemie' (Investigations of pathological substances obtained during the epidemic of puerperal fever which occurred in the winter of 1842–1843 in and around Würzburg) Scherer described the cases of seven young women who all died peripartum.

im Winter 1842–1843 in Würzburg und der Umgegend herrschenden Puerperal-Fieber-Epidemie' (Investigations of pathological substances obtained during the epidemic of puerperal fever which occurred in the winter of 1842–1843 in and around Würzburg) Scherer described the cases of seven young women who all died peripartum. One of the women, the 23-year-old primipara Eva Rumpel, gave birth to a healthy child on 9 January 1843. The same night she developed a painfully swollen abdomen and became ill, feverish, and sweaty, with rapid pulse and severe thirst. The initiated treatment was bloodletting and clystering. The next evening she deteriorated, became delirious, with anxious breathing, a tense abdomen, cold extremities and rapid pulse, finally losing consciousness. Again, bloodletting followed. At 4:30 a.m., 36 h after the onset of the first symptoms, she died. During autopsy, severe purulent endometritis, vaginal pus, pulmonary oedema, and shock liver and shock spleen were found. The blood that was obtained directly from the heart was chemically analysed, in which lactic acid was found. Most likely this unfortunate woman had died from a fulminant septic shock caused by group A haemolytic streptococci (Streptococcus pyogenes). Scherer diagnosed this case as perimetritis with secondary peritonitis.

e blood that was obtained directly from the heart was chemically analysed, in which lactic acid was found. Most likely this unfortunate woman had died from a fulminant septic shock caused by group A haemolytic streptococci (Streptococcus pyogenes). Scherer diagnosed this case as perimetritis with secondary peritonitis. Another patient, the 28-year-old, 7 months pregnant (second pregnancy) Margaretha Glück, was, after being icteric, nauseous, vomiting and complaining about epigastric pain for 8 days, admitted to the lying-in birth clinic on 6 February 1843. Four days later she was transferred to the hospital with severe nosebleeds and generalised exanthema or purpura. In the evening she suffered from severe gastric bleeding and epistaxis, showing rapid pulse, cold extremities and dizziness. The next morning, she was transferred back to the birth clinic, where she gave birth to a premature child (30 weeks) and suffered from a severe post-partum fluxus. She was again transferred to the hospital with the following symptoms: cold clammy skin, tachycardia, severe lochia and persistent exanthema or purpura, but without signs of an acute abdomen. During the night of February 11, she became aphasic and restless, followed by chills and profound sweating. On the morning of February 13, she further deteriorated and bilirubinuria was detected. The next day she was comatose, finally developed rattling breathing and convulsions. Death occurred during the following night. Autopsy revealed a small intracerebral haematoma, normal lungs without pulmonary oedema, ascites and an anaemic, foul smelling uterus filled with purulent and decayed tissue and pus. Blood was also obtained directly from the heart during autopsy and lactic acid was found.

ons. Death occurred during the following night. Autopsy revealed a small intracerebral haematoma, normal lungs without pulmonary oedema, ascites and an anaemic, foul smelling uterus filled with purulent and decayed tissue and pus. Blood was also obtained directly from the heart during autopsy and lactic acid was found. In this case we could think of a haemorrhagic shock and cerebral haemorrhage due to clotting disorders possibly resulting from either acute fatty liver of pregnancy/HELLP syndrome, idiopathic thrombocytopenic purpura, thrombotic microangiopathy (TTP/HUS) or DIC. The case was most likely complicated by a sepsis (endometritis). Scherer himself diagnosed this case as septic endometritis. In the conclusions of his 1843 book, Scherer attached high importance to the fact that he found lactic acid in cases of puerperal fever, which he had not found before in healthy persons. He held the opinion that lactic acid was formed in blood during bodily deterioration in severe diseases like puerperal fever. Lactic acid was thus described for the first time in human blood and was demonstrated for the first time as a symptom of septic and haemorrhagic shock.

e had not found before in healthy persons. He held the opinion that lactic acid was formed in blood during bodily deterioration in severe diseases like puerperal fever. Lactic acid was thus described for the first time in human blood and was demonstrated for the first time as a symptom of septic and haemorrhagic shock. In the same period a junior obstetrician in Vienna, Ignaz Philipp Semmelweis (1818–1865), discovered in 1847 that physicians carried infectious particles on their hands from the mortuary to the obstetrical clinic, causing puerperal fever and puerperal sepsis, and he introduced a successful method for its prevention. Louis Pasteur (1822–1894) found in 1879 that infection with streptococci was the most important cause of puerperal fever [16].

carried infectious particles on their hands from the mortuary to the obstetrical clinic, causing puerperal fever and puerperal sepsis, and he introduced a successful method for its prevention. Louis Pasteur (1822–1894) found in 1879 that infection with streptococci was the most important cause of puerperal fever [16]. The 1851 article Scherer worked closely with the famous pathologist Rudolf Virchow (1821–1902) on several projects (Fig. 2). In 1851 Virchow performed an autopsy on a patient who had died from leukaemia and offered Scherer blood from this patient for analysis. The results of this analysis were published the same year in the ‘Verhandlungen der Physikalisch-Medicinischen Gesellschaft in Würzburg’  [17]. Virchow and Scherer had previously studied the spleens of patients who died from leukaemia, and were curious if they could find the same results in the blood. Scherer reached the conclusion that: the blood of this patient contains: “Ameisensäure, Essigsäure und Milchsäure, die gleichfalls von mir schon früher als in der Milzflüssigkeit vorkommend bezeichnet wurden” (Formic acid, acetic acid, and lactic acid, as also found by me previously in fluids from the spleen). Fig. 2 Johann Joseph Scherer (left) and Rudolf Virchow (right) in 1849

contains: “Ameisensäure, Essigsäure und Milchsäure, die gleichfalls von mir schon früher als in der Milzflüssigkeit vorkommend bezeichnet wurden” (Formic acid, acetic acid, and lactic acid, as also found by me previously in fluids from the spleen). Fig. 2 Johann Joseph Scherer (left) and Rudolf Virchow (right) in 1849 Further research Scherer's observations inspired others to conduct further research, primarily in patients with leukaemia [19–22], but also in patients with other conditions and diseases and in animal experiments with dogs and rabbits [23]. While Scherer found lactic acid in blood obtained after death during autopsy, Mosler and Körner [19] mention an observation made by Carl Folwarczny, published in the Allgemeinen Wiener Medicinischen Zeitung in 1858, where blood was withdrawn from a leukaemia patient during life, analysed according to Scherer's method, and found positive for lactic acid. In addition, Carl Folwarczny described in 1863 in his 'Handbuch der Physiologischen Chemie' [24] that lactic acid can be found in the blood of patients with leukaemia, septicaemia (pyaemia) and in conditions leading to septicaemia like puerperal fever, the latter probably after Scherer's observations.

for lactic acid. In addition, Carl Folwarczny described in 1863 in his 'Handbuch der Physiologischen Chemie' [24] that lactic acid can be found in the blood of patients with leukaemia, septicaemia (pyaemia) and in conditions leading to septicaemia like puerperal fever, the latter probably after Scherer's observations. In an extensive article, the Berliner physician Georg Salomon [25], who had serious doubts that the occurrence of lactic acid in blood was mostly related to leukaemia, proved in 1878 that lactic acid was also present in the blood of patients who were suffering and died from other diseases. He studied blood obtained during autopsy from cadavers, but also blood from patients obtained by bloodletting or cupping, and in some cases he compared the blood before and after death. He was able to demonstrate lactic acid in the blood of patients suffering from leukaemia, (pernicious) anaemia, congestive heart failure, chronic obstructive pulmonary disease, pleuritis, pericarditis, pneumonia and several solid malignant tumours.

upping, and in some cases he compared the blood before and after death. He was able to demonstrate lactic acid in the blood of patients suffering from leukaemia, (pernicious) anaemia, congestive heart failure, chronic obstructive pulmonary disease, pleuritis, pericarditis, pneumonia and several solid malignant tumours. Gaglio [26] is often erroneously mentioned as the first author to find lactic acid in blood [27–29]. He was able to demonstrate lactic acid in fresh arterial blood withdrawn from dogs and rabbits after bloodletting. Berlinerblau [30] confirmed these observations in mammalian and venous human blood. Both Gaglio and Berlinerblau, however, neglected previous research, as indignantly described by Salomon in 1888 [“Ich erlaube mir, den Inhalt meiner Arbeiten, die von Gaglio nur ganz flüchtig, von Berlinerblau gar nicht berührt sind, in Kürze zu reproduciren” (I take the liberty of summarizing the contents of my work, which was mentioned only briefly by Gaglio and not at all by Berlinerblau)] [31].

y described by Salomon in 1888 [“Ich erlaube mir, den Inhalt meiner Arbeiten, die von Gaglio nur ganz flüchtig, von Berlinerblau gar nicht berührt sind, in Kürze zu reproduciren” (I take the liberty of summarizing the contents of my work, which was mentioned only briefly by Gaglio and not at all by Berlinerblau)] [31]. The Japanese chemist Trasaburo Araki showed that the amount of lactic acid in exhausted muscle results from muscle activation [11]. Irisawa [32], inspired by the results obtained by Salomon and Gaglio, obtained fresh blood of 11 dying patients with serious conditions. In six cases he found hyperlactataemia, in four cases normal values. He speculated on the aetiology of hyperlactataemia, the most plausible cause being the severe hypoxia during the dying process. In an experiment in which he made a dog anaemic for several days, he found a rise in lactic acid levels during the time leading up to death. In Cambridge (UK), Walter Morley Fletcher (1873–1933) and Frederick Gowland Hopkins (1861–1947) worked together on the metabolic changes occurring in muscular contractions and rigor mortis under anaerobic conditions, and found that lactate was the product of carbohydrate metabolism [33]. Their classic 1907 paper demonstrated rigorously that muscle contraction is accompanied by the anaerobic formation of lactic acid, which is removed aerobically, at a rate depending on the level of exposure to oxygen [34].

nder anaerobic conditions, and found that lactate was the product of carbohydrate metabolism [33]. Their classic 1907 paper demonstrated rigorously that muscle contraction is accompanied by the anaerobic formation of lactic acid, which is removed aerobically, at a rate depending on the level of exposure to oxygen [34]. Poul Astrup and John Severingshaus mentioned Scherer's 1851 article as first demonstration of lactic acid in blood, but overlooked the 1843 cases and Folwarczny's work [35]. In conclusion, Scherer's 1843 case reports [15] should be cited as the first description of lactic acid in human blood and also as the first demonstration of lactic acid as a pathological finding in septic and haemorrhagic shock. Folwarczny, in 1858, was the first to demonstrate lactic acid in blood in a living patient. Acknowledgements We would like to thank Mrs Helga Seifert, librarian, Pathologisches Institut der Universität Würzburg Bibliothek, Würzburg, Germany, for providing copies of Scherer's 1851 paper and Gaglio's 1888 article.

Introduction Posttraumatic stress disorder (PTSD) is the development of psychological and physical symptoms following exposure to one or more traumatic events [1, 2]. PTSD symptoms include intrusive recollections (re-experiencing the trauma in flashbacks, memories or nightmares); avoidant and numbing symptoms (including diminished emotions and avoidance of situations that are reminders of the traumatic event); and hyperarousal (including increased irritability, exaggerated startle reactions or difficulty sleeping or concentrating) [3]. PTSD symptoms have a major impact on life, illustrated by the fact that the patients have a reduced quality of life [4] and frequently suffer from depression [5].

ders of the traumatic event); and hyperarousal (including increased irritability, exaggerated startle reactions or difficulty sleeping or concentrating) [3]. PTSD symptoms have a major impact on life, illustrated by the fact that the patients have a reduced quality of life [4] and frequently suffer from depression [5]. Events that typically trigger the development of PTSD include exposure to violent events such as rape, domestic violence, child abuse, war, accidents, natural disasters and political torture, all of which include a threat to life [6–8]. Increasingly PTSD has also been found in patients who have survived a major, life-threatening disease and patients who have spent a significant amount of time in an intensive care unit (ICU) [9–13]. Severe peritonitis (or abdominal sepsis) is such a disease where typically an episode of acute and severe illness [14, 15] is followed by a lengthy ICU stay and a long recovery period that often includes multiple surgical and non-surgical interventions [16–22]. This combination of factors could make this patient group particularly vulnerable for developing PTSD symptoms. To date, little is known about the presence and severity of PTSD and possible risk factors in patients recovering from severe peritonitis [15].

cludes multiple surgical and non-surgical interventions [16–22]. This combination of factors could make this patient group particularly vulnerable for developing PTSD symptoms. To date, little is known about the presence and severity of PTSD and possible risk factors in patients recovering from severe peritonitis [15]. Therefore, our aims were to determine the presence and level of symptoms of PTSD in patients surviving abdominal sepsis. In addition, we searched for demographic and disease-related factors associated with higher levels of PTSD symptoms. Identification of such factors may be important to determine possible targets of intervention and to select patients for psychological assessment interviews. Methods Study design Our study was embedded in an ongoing randomized clinical trial (the RELAP Trial) evaluating two surgical treatment strategies for patients with secondary peritonitis after the initial emergency laparotomy. Patients were enrolled between December 2001 and February 2005 in two academic medical centers and seven regional teaching hospitals in The Netherlands. All patients were followed up for 12 months after initial (index) laparotomy. The study was approved by the medical ethics committee of the Academic Medical Center, Amsterdam. All patients gave informed consent to participate in this study.

Methods Study design Our study was embedded in an ongoing randomized clinical trial (the RELAP Trial) evaluating two surgical treatment strategies for patients with secondary peritonitis after the initial emergency laparotomy. Patients were enrolled between December 2001 and February 2005 in two academic medical centers and seven regional teaching hospitals in The Netherlands. All patients were followed up for 12 months after initial (index) laparotomy. The study was approved by the medical ethics committee of the Academic Medical Center, Amsterdam. All patients gave informed consent to participate in this study. Study population Patients were eligible for the RELAP trial if they had a clinical diagnosis of secondary peritonitis requiring emergency laparotomy and an Acute Physiology and Chronic Health Evaluation II (APACHE-II) score above 10. Further details of the study population can be found elsewhere [23]. Data collection All self-administered PTSD questionnaires were distributed by mail to patients who survived at least 12 months following initial emergency laparotomy, with a reminder by phone within 2 weeks in the case of no response. After 1 month without response a new questionnaire including a reminder letter was sent. Instruments assessing the level of PTSD symptoms We used two instruments with good psychometric characteristics [24, 25] for measuring PTSD symptoms in research settings: the Post-Traumatic Stress Scale 10 [26] and the Impact of Event Scale–Revised [27, 28].

Data collection All self-administered PTSD questionnaires were distributed by mail to patients who survived at least 12 months following initial emergency laparotomy, with a reminder by phone within 2 weeks in the case of no response. After 1 month without response a new questionnaire including a reminder letter was sent. Instruments assessing the level of PTSD symptoms We used two instruments with good psychometric characteristics [24, 25] for measuring PTSD symptoms in research settings: the Post-Traumatic Stress Scale 10 [26] and the Impact of Event Scale–Revised [27, 28]. The Post-Traumatic Stress Syndrome Scale 10 (PTSS-10) was originally designed to diagnose PTSD according to the Diagnostic and Statistical Manual of Mental Disorders III (DSM-III) criteria in victims of natural disasters [14]. The PTSS-10 is now a widely used self-report questionnaire assessing symptoms related to PTSD, particularly in critically ill and ICU patients [4, 11, 12]. The PTSS-10 consists of 10 items, each of which ranges from 1 point (none) to 7 points (always). The total score ranges from 10 to 70, with higher scores indicating more symptoms; scores of 35 or above are considered indicative of PTSD [11, 29].

ated to PTSD, particularly in critically ill and ICU patients [4, 11, 12]. The PTSS-10 consists of 10 items, each of which ranges from 1 point (none) to 7 points (always). The total score ranges from 10 to 70, with higher scores indicating more symptoms; scores of 35 or above are considered indicative of PTSD [11, 29]. The Impact of Events Scale–Revised (IES-R) is one of the most commonly used self-report questionnaires for determining PTSD symptomatology following a trauma [27]. The IES-R consists of 22 items, each ranging from 0 (no problems) to 4 (frequent problems), with the total score ranging from 0 to 66. Scores above 24 points are generally considered indicative of PTSD, with higher scores indicating more symptoms [28]. The IES-R has been developed based on DSM-IV criteria and therefore has three distinct subscales, the avoidance subscale (eight questions), the intrusion subscale (eight questions) and the hyperarousal subscale (six questions) [28, 30], and is one of the most frequently used self-report questionnaires in both the clinic and in PTSD research [27]. Potential risk factors Potential risk factors were selected from previous studies [31] examining factors for increased mortality and morbidity [17–22, 32, 33] in secondary peritonitis supplemented with specific factors mentioned in the PTSD literature [6, 9–11, 14, 34, 35]. We divided these factors into three distinct categories. General patient characteristics included age, gender and the presence of major comorbidity (cardiovascular disease; COPD; renal failure; diabetes; malignancy).

Potential risk factors Potential risk factors were selected from previous studies [31] examining factors for increased mortality and morbidity [17–22, 32, 33] in secondary peritonitis supplemented with specific factors mentioned in the PTSD literature [6, 9–11, 14, 34, 35]. We divided these factors into three distinct categories. General patient characteristics included age, gender and the presence of major comorbidity (cardiovascular disease; COPD; renal failure; diabetes; malignancy). Disease characteristics and postoperative course included severity of disease measured at the time of initial laparotomy using the APACHE-II score. As several components of the APACHE-II score are already considered in a univariate analysis (namely age and comorbidity); we chose to replace the APACHE-II score with the APS score as a potential predictor of PTSD [36]. The APS comprises only the acute components of the APACHE-II score, without age and comorbidity. Postoperative characteristics included administration of hydrocortisone during ICU stay [13, 37], development of acute respiratory distress syndrome (ARDS), [4, 12, 20], number of relaparotomies, duration of ICU and hospital stay, the development of a disease-related major morbidity during 6 months' follow-up [23] and an enterostomy present after 6 months' follow-up.

tion of hydrocortisone during ICU stay [13, 37], development of acute respiratory distress syndrome (ARDS), [4, 12, 20], number of relaparotomies, duration of ICU and hospital stay, the development of a disease-related major morbidity during 6 months' follow-up [23] and an enterostomy present after 6 months' follow-up. Traumatic memories of ICU/hospital stay were assessed using the four-item adverse experiences questionnaire, which captures four types of traumatic memories of the stay in the ICU or hospital ward: nightmares, fear and panic, pain, and difficulty in breathing [13]. Patients scored the frequency of traumatic memories of their stay in the ICU or hospital ward using a four-point scale of 0 (never), 1 (sometimes), 2 (regularly) or 3 (often), administered concurrently with the PTSS-10 and IES-R questionnaires after at least 12 months' follow-up. These were subsequently summed and classified into three graded categories of traumatic memories: 0 (no traumatic memories), 1–4 (some traumatic memories), more than 4 (many traumatic memories).

larly) or 3 (often), administered concurrently with the PTSS-10 and IES-R questionnaires after at least 12 months' follow-up. These were subsequently summed and classified into three graded categories of traumatic memories: 0 (no traumatic memories), 1–4 (some traumatic memories), more than 4 (many traumatic memories). We also collected data on whether patients had experienced other traumas or whether a close family member or friend had experienced a trauma within the previous 3 years. We used questions 29 and 30 from the Life Stressor Checklist–Revised [38], administered at the same time as the PTSS-10, IES-R and Beck Depression Inventory II (BDI-II) questionnaires. Responses were given dichotomously as yes or no, and patients were subsequently asked to specify the event type [38]. These questions were asked to determine to what extent the PTSD symptomatology found in this patient group was due to their peritonitis or to other traumatic events.

nventory II (BDI-II) questionnaires. Responses were given dichotomously as yes or no, and patients were subsequently asked to specify the event type [38]. These questions were asked to determine to what extent the PTSD symptomatology found in this patient group was due to their peritonitis or to other traumatic events. Data analysis We used two instruments aimed at measuring the presence and severity of PTSD symptoms in our population, each with their own cut-off value. Combining data from two instruments measuring the same construct (PTSD symptoms) may lead to a more robust classification of patients. To preserve the natural ordering of patients who scored below the cut-off value on both questionnaires (‘low-scoring patients’), patients scoring above the cut-off on only one of these questionnaires (‘moderate scoring patients’) and patients scoring above the cut-off on both questionnaires (‘high-scoring patients’), we applied ordinal regression modeling. The proportion of patients in each of these three categories is presented with 95% confidence intervals (95% CI) using the method of Wilson [39]. Potential predictors for PTSD symptoms were analyzed using an ordinal logistic regression model. This ordinal regression model is an extension of the binary logistic model and is appropriate when a continuous trait is grouped into several categories by using cut-offs [40].

Data analysis We used two instruments aimed at measuring the presence and severity of PTSD symptoms in our population, each with their own cut-off value. Combining data from two instruments measuring the same construct (PTSD symptoms) may lead to a more robust classification of patients. To preserve the natural ordering of patients who scored below the cut-off value on both questionnaires (‘low-scoring patients’), patients scoring above the cut-off on only one of these questionnaires (‘moderate scoring patients’) and patients scoring above the cut-off on both questionnaires (‘high-scoring patients’), we applied ordinal regression modeling. The proportion of patients in each of these three categories is presented with 95% confidence intervals (95% CI) using the method of Wilson [39]. Potential predictors for PTSD symptoms were analyzed using an ordinal logistic regression model. This ordinal regression model is an extension of the binary logistic model and is appropriate when a continuous trait is grouped into several categories by using cut-offs [40]. All potential predictors for PTSD symptoms were first examined in univariate ordinal regression models. Factors with a p-value of less than 0.1 were entered in a multivariate ordinal logistic regression model. If variables within a group of predictors were strongly correlated, only the factor with the strongest univariate relationship and/or most relevant clinical interpretation was added to the model. Because the literature on PTSD and ICU studies shows them to be clinically relevant, age and gender were always included in the multivariate model regardless of the strength of their associations [34].

factor with the strongest univariate relationship and/or most relevant clinical interpretation was added to the model. Because the literature on PTSD and ICU studies shows them to be clinically relevant, age and gender were always included in the multivariate model regardless of the strength of their associations [34]. In addition, a factor comprised of other non-related traumas that the patient had experienced within the previous 3 years was included in the final model to assess its potential confounding role. The fit and validity of the model was evaluated by checking the discriminatory properties (overlap in risk scores of patients with different outcomes), the proportional odds assumption (test for parallel lines) and calibration (closeness in expected and observed numbers of patients evaluated by an extension of the Hosmer–Lemeshow goodness-of-fit statistic). Calibration was checked by comparing expected and observed number of patients in each of the three outcome categories across deciles of expected risk and tested for significance by using an extension of the Hosmer–Lemeshow goodness-of-fit statistic [41].

The fit and validity of the model was evaluated by checking the discriminatory properties (overlap in risk scores of patients with different outcomes), the proportional odds assumption (test for parallel lines) and calibration (closeness in expected and observed numbers of patients evaluated by an extension of the Hosmer–Lemeshow goodness-of-fit statistic). Calibration was checked by comparing expected and observed number of patients in each of the three outcome categories across deciles of expected risk and tested for significance by using an extension of the Hosmer–Lemeshow goodness-of-fit statistic [41]. Nomogram: A nomogram was developed to visualize the prognostic strength of the different factors from the multivariate model in a single diagram. A nomogram allows readers to calculate an expected distribution of PTSD symptomatology (‘low-scoring’, ‘moderate-scoring’ and ‘high-scoring’ patients) based on a specific profile of a patient. The number of points for each predictor was based on the original coefficient from the multivariate ordinal model. The total number of points derived by specifying values for all predictors was used to calculate the expected probabilities that a patient would be a ‘low-scoring patient’, a ‘moderate-scoring patient’ or a ‘high-scoring patient’. Analyses were performed using SAS software version 9.1 (SAS Institute Inc., Cary, NC, USA).

Nomogram: A nomogram was developed to visualize the prognostic strength of the different factors from the multivariate model in a single diagram. A nomogram allows readers to calculate an expected distribution of PTSD symptomatology (‘low-scoring’, ‘moderate-scoring’ and ‘high-scoring’ patients) based on a specific profile of a patient. The number of points for each predictor was based on the original coefficient from the multivariate ordinal model. The total number of points derived by specifying values for all predictors was used to calculate the expected probabilities that a patient would be a ‘low-scoring patient’, a ‘moderate-scoring patient’ or a ‘high-scoring patient’. Analyses were performed using SAS software version 9.1 (SAS Institute Inc., Cary, NC, USA). Results Of the total of 132 patients eligible for this study, 108 (80%) patients responded to the questionnaire (Fig. 1). On average the responses were provided approximately 12.5 months following initial emergency laparotomy. There were no significant differences in any of the patient or disease characteristics between respondents and non-respondents. Fig. 1 Flowchart summarizing inclusion and response

nded to the questionnaire (Fig. 1). On average the responses were provided approximately 12.5 months following initial emergency laparotomy. There were no significant differences in any of the patient or disease characteristics between respondents and non-respondents. Fig. 1 Flowchart summarizing inclusion and response The median age of patients was 66.8 years and 54% were male. Patients were severely ill, with a median APACHE-II score of 14 and a median APS score of 6, and 5% had a major comorbidity (Table 1). Ninety-six patients (89%) were admitted to ICU: their median ICU-stay was 7 days, and these patients were mechanically ventilated for a median of 5 days. Patients were hospitalized for a median period of 28 days (IQR 19–55 days). Fifty-one percent of patients also underwent another trauma in the 3 years prior to filling in the PTSD questionnaires. Table 1 Association between severity of PTSD symptoms (three categories) and patient, disease operative and postoperative characteristics: results from univariate ordinal regression models

9–55 days). Fifty-one percent of patients also underwent another trauma in the 3 years prior to filling in the PTSD questionnaires. Table 1 Association between severity of PTSD symptoms (three categories) and patient, disease operative and postoperative characteristics: results from univariate ordinal regression models Overall PTSD symptoms a(n = 107) Univariate ordinal regression b None to mild (n = 66) Moderate (n = 30) High (n = 11) p-value General patient characteristics Median age (IQR) 66.8 (57–73) 70.2 (60–74) 58.7 (47–72) 57.8 (49–65) 0.004 Male gender (%) 54% 53% 53% 64% 0.847 Major comorbidity present (%) c 53% 55% 50% 55% 0.670 Peritonitis and postoperative characteristics Initial Median APS score (IQR) 6 (4–8) 6 (4–8) 7 (5–9) 8 (3–8.5) 0.271 Hydrocortisone in first 14 days in ICU (median days) 2 (0–7) 1.5 (0–8) 1 (0–8) 5 (1–7) 0.749 ARDS 6% 3% 10% 9% 0.192 One or more relaparotomies 67% 70% 63% 64% 0.515 Admitted to ICU 89% 85% 93% 100% 0.110 Median length of ICU stay (IQR) 7 (4–15) 7 (4–12) 7 (4–19) 9 (6–16) 0.042 Median ventilation time (IQR) 5 (1–8) 4 (1–7) 5 (1–10) 7 (4–13) 0.073 Median length of hospital stay (IQR) 28 (19–55) 26 (18–47) 31 (23–60) 56 (19–72) 0.102 Follow-up Disease-related major morbidity at 6-month follow-up 15% 9% 27% 18% 0.068 Enterostomy at 6-month follow-up 51% 47% 55% 70% 0.183 IQR, Interquartile range; a Three graded outcomes: none to mild, moderate and high; two patients' data are based on only one completed questionnaire; b All models were checked for parallel lines to see if an ordinal test for significance was appropriate; c Major comorbidity included cardiovascular disease, COPD, renal failure, diabetes and malignancy

e graded outcomes: none to mild, moderate and high; two patients' data are based on only one completed questionnaire; b All models were checked for parallel lines to see if an ordinal test for significance was appropriate; c Major comorbidity included cardiovascular disease, COPD, renal failure, diabetes and malignancy Prevalence of PTSD symptoms The proportion of ‘moderate-scoring’ PTSD patients was 28% (95% CI 20–37%), whilst 10% (95% CI 6–17%) of patients were ‘high-scoring’ patients (Table 1). Detailed information on depression and PTSD symptoms is presented in the electronic supplementary material (ESM). Predictive factors Results from the univariate analysis are presented in Table 1 and Table 2, and descriptive details can be found in the ESM. Table 2 Association between severity of PTSD symptoms (three categories) and other traumatic experiences following peritonitis PTSD symptoms (n = 105) Univariate ordinal regression None to mild (n = 64) a Moderate (n = 30) a High (n = 11) p-value Traumatic memories of ICU or hospital stay Nightmares 39% 61% 82% 0.002 Fear and panic 24% 61% 100% < 0.001 Pain 67% 70% 82% 0.002 Difficulty breathing 33% 76% 100% < 0.001 Traumatic memories None (0) 41% 50% 9% < 0.001 Moderate (1–4) 7% 47% 47% Severe (> 4) 0% 18% 82% a Two patients not included in final analysis due to missing data on traumatic memories during ICU or hospital stay

0.002 Fear and panic 24% 61% 100% < 0.001 Pain 67% 70% 82% 0.002 Difficulty breathing 33% 76% 100% < 0.001 Traumatic memories None (0) 41% 50% 9% < 0.001 Moderate (1–4) 7% 47% 47% Severe (> 4) 0% 18% 82% a Two patients not included in final analysis due to missing data on traumatic memories during ICU or hospital stay The final multivariate model included age, gender, length of ICU stay, disease-related morbidity during the 6-month follow-up, traumatic memories of the ICU or hospital stay and other traumatic factors within the previous 3 years (Table 3). Table 3 Association between severity of PTSD symptoms and patient, disease operative and postoperative characteristics and other traumatic experiences following peritonitis in a multivariate analysis Final model (n = 105) a OR 95% CI p-value Lower Upper Ten years increase in age 0.74 0.53 1.04 0.084 Female 0.9 0.94 2.3 0.822 Length of ICU stay (log2 transformed) 1.4 1.1 1.7 < 0.003 Major disease-related morbidity during 6-month follow-up (including index hospital admittance) 2.1 0.61 7.11 0.238 Traumatic memories of ICU or hospital stay Moderate (1–4) 4.9 0.95 24.9 0.058 Severe (> 4) 55.5 9.4 328.0 < 0.001 Other trauma within previous 3 years 2.4 0.94 6.3 0.085 a This multivariate ordinal analysis included a test for parallel lines (p = 0.694)

ollow-up (including index hospital admittance) 2.1 0.61 7.11 0.238 Traumatic memories of ICU or hospital stay Moderate (1–4) 4.9 0.95 24.9 0.058 Severe (> 4) 55.5 9.4 328.0 < 0.001 Other trauma within previous 3 years 2.4 0.94 6.3 0.085 a This multivariate ordinal analysis included a test for parallel lines (p = 0.694) In our final model, increasing age was associated with a lower likelihood of PTSD symptomatology (OR = 0.74 per 10 years increase in age, p = 0.084). Gender was not predictive of PTSD symptoms (OR = 0.90, p = 0.82). Disease-related morbidity at the 6-month follow-up (OR = 2.1, p = 0.24) was no longer independently predictive of PTSD symptoms. Memories of the ICU/hospital stay (patients that reported some memories: OR = 4.9, p < 0.057; patients that reported many memories: OR = 55.5, p < 0.001) were the most prominent independent risk factor for increased PTSD symptomatology. Length of ICU stay was also significantly predictive in the development of PTSD symptomatology in the multivariate model (OR = 1.4, p = 0.004). The relative strengths of these relationships are visualized in the nomogram (Fig. 2). In this nomogram, one can calculate for the individual patient given his/her risk profile the probability that he/she will score either no to mild, moderate or high PTSD symptoms according to the PTSS-10 and IES-R. Fig. 2 Nomogram for prediction of severity of PTSD symptoms in patients with secondary peritonitis. Graded outcome categories are: none to mild (negative on both instruments), moderate (positive on one instrument), and severe (positive on both instruments)

erate or high PTSD symptoms according to the PTSS-10 and IES-R. Fig. 2 Nomogram for prediction of severity of PTSD symptoms in patients with secondary peritonitis. Graded outcome categories are: none to mild (negative on both instruments), moderate (positive on one instrument), and severe (positive on both instruments) The proportional odds assumption was not violated as indicated by a p-value of 0.694 for the test of parallel lines. Calibration of the model (closeness between predicted and observed probabilities) was good, with a p-value of 0.987 for the goodness-of-fit test for ordinal models. A graphical impression of the model's discriminative ability is shown in Fig. 3. This figure shows that the mean risk score is significantly different between all three PTSD symptom severity categories (p < 0.001), although there is some substantial overlap in the risk scores between patients from different categories of PTSD symptom severity. Fig. 3 Distribution of total points from nomogram (risk score) for the prediction of the severity of PTSD symptoms with use of the risk factors taken from the multivariate ordinal model. PTSD categories are graded according to severity: none to mild (negative on both instruments), moderate (positive on one instrument), severe (positive on both instruments)

om nomogram (risk score) for the prediction of the severity of PTSD symptoms with use of the risk factors taken from the multivariate ordinal model. PTSD categories are graded according to severity: none to mild (negative on both instruments), moderate (positive on one instrument), severe (positive on both instruments) Discussion The proportion of patients with ‘high-scoring’ PTSD symptomatology 12 months after peritonitis was 10%, and the number of ‘medium-scoring’ patients (28%) was in line with earlier studies measuring PTSD symptoms in critically ill patients who had been admitted to ICU [9–13, 15, 42]. The prevalence of PTSD recorded in the general population varies between 0.9% and 2.9% (the ESEMeD study) [7, 43, 44]. From our study, the following observations can be made. Firstly, the development of PTSD symptoms is not directly related to the severity of the disease at presentation. The APS score at baseline was not predictive for the development of PTSD symptoms. The APS measures the severity of disease score solely on the weight of the acute clinical features and does not incorporate age and comorbidity [36].

f PTSD symptoms is not directly related to the severity of the disease at presentation. The APS score at baseline was not predictive for the development of PTSD symptoms. The APS measures the severity of disease score solely on the weight of the acute clinical features and does not incorporate age and comorbidity [36]. The development of PTSD symptoms was, however, predominantly related to a more complicated course of secondary peritonitis. Longer ICU and hospital stays and major disease-related morbidity during the 12-month follow-up were associated with more PTSD symptoms. In concordance with earlier studies [15, 34], the strongest predictor of having PTSD symptoms following abdominal sepsis was having traumatic memories and experiences during their initial hospital or ICU stay. These results suggest that the presence of traumatic memories is one of the most relevant aspects for the development of PTSD-related symptoms. Earlier studies also found that subjective interpretation of the intensive care experience emerged as a consistent predictor of adverse emotional outcome in both the short and the long term [13, 34].

hat the presence of traumatic memories is one of the most relevant aspects for the development of PTSD-related symptoms. Earlier studies also found that subjective interpretation of the intensive care experience emerged as a consistent predictor of adverse emotional outcome in both the short and the long term [13, 34]. Age plays a critical role in the development of PTSD symptoms. Younger patients are much more likely to develop and report such symptoms. This finding confirms the results of an earlier, retrospective study of a different cohort of patients 4–10 years after hospital admission for severe peritonitis [15]. These findings suggest that older patients are more able to adapt to the limitations that are associated with experiencing such a major disease, most likely because they have already experienced a co-morbid illness and health-related problems. In contrast to some other studies, gender did not play a role in the development of PTSD symptoms [34].

older patients are more able to adapt to the limitations that are associated with experiencing such a major disease, most likely because they have already experienced a co-morbid illness and health-related problems. In contrast to some other studies, gender did not play a role in the development of PTSD symptoms [34]. Patients with abdominal sepsis suffering from ARDS did not report more PTSD symptomatology than those without ARDS. In earlier ICU studies, ARDS patients reported considerable PTSD symptoms [12, 45]. In our cohort of abdominal sepsis patients we found different predictive factors for PTSD than those found in the ARDS patients [10, 12, 15]. Secondary peritonitis in itself may have been severe enough, with ICU admission and extended mechanical ventilation days, to cause PTSD symptoms; therefore, the added risk by ARDS may be moot. Lack of power may also be a factor, because the proportion of patients developing ARDS in this study was modest.

[10, 12, 15]. Secondary peritonitis in itself may have been severe enough, with ICU admission and extended mechanical ventilation days, to cause PTSD symptoms; therefore, the added risk by ARDS may be moot. Lack of power may also be a factor, because the proportion of patients developing ARDS in this study was modest. We did not find an association between hydrocortisone administration during ICU stay and PTSD symptoms within this peritonitis cohort, as has been demonstrated for other critically ill patient groups [13, 37, 46, 47]. Hydrocortisone did not protect against developing PTSD symptoms, whereas other studies have found that administration of hydrocortisone during ICU can lead to a reduction in PTSD symptoms after discharge. In this study corticosteroid use during only the first 14 days of ICU was included in our analyses. The effect of prolonged use of hydrocortisone or late-stage use during conditional adrenal insufficiency cannot be excluded.

stration of hydrocortisone during ICU can lead to a reduction in PTSD symptoms after discharge. In this study corticosteroid use during only the first 14 days of ICU was included in our analyses. The effect of prolonged use of hydrocortisone or late-stage use during conditional adrenal insufficiency cannot be excluded. Unfortunately, due to the acute and life-threatening nature of secondary peritonitis, it was not possible to collect baseline information on PTSD or data on earlier psychological disorders. However, as recommended in a recent review by Griffiths and colleagues [48], to account for possible earlier traumas we considered information pertaining to comorbid diseases. Furthermore, we collected data on other, non-disease-related traumatic events that had occurred within the previous 3 years. These non-disease-related events were indeed associated with having more PTSD symptoms, and altered the initial ORs of the other factors to the extent that we considered it a moderate confounder. Timing plays an important role in collecting data on PTSD symptoms in critically ill and ICU patients [48]. We set the period for the recording of PTSD symptoms at 12 months for a very specific reason; in this severely ill patient group, we did not want to record patient recovery. Past studies have shown that critically ill patients develop PTSD symptoms only after their physical recovery period has passed, hence with a delayed onset [9].

iod for the recording of PTSD symptoms at 12 months for a very specific reason; in this severely ill patient group, we did not want to record patient recovery. Past studies have shown that critically ill patients develop PTSD symptoms only after their physical recovery period has passed, hence with a delayed onset [9]. Although these self-report questionnaires are frequently used, the diagnostic value of such instruments in relation to a DSM-IV diagnosis obtained by a structured interview is still being researched and discussed [49]. Some studies have reviewed the diagnostic value of the questionnaires, but in general these studies were methodologically limited [11, 29, 50]. In this study we did not include a structured clinical interview for establishing a definite DSM-IV criteria diagnosis of PTSD, although this is highly recommended in clinical psychology. However, we feel that the use of questionnaires is more feasible in the ICU [48], and patients who report many symptoms on these self-report questionnaires [51] can subsequently be referred to an appropriate mental healthcare provider [50].

diagnosis of PTSD, although this is highly recommended in clinical psychology. However, we feel that the use of questionnaires is more feasible in the ICU [48], and patients who report many symptoms on these self-report questionnaires [51] can subsequently be referred to an appropriate mental healthcare provider [50]. In this study we have tried to learn from two questionnaires, one commonly used and validated in particular for critically ill patients, together with one of the most frequently used screening instruments for PTSD, the IES-R. The prevalence of PTSD symptomatology in the present study was based on whether or not a patient scored above the cut-offs of the IES-R and the PTSS-10. We employed the two questionnaires as complementary for the detection of PTSD symptoms and not to compare results deduced from both questionnaires separately [49]. Combining the results of both questionnaires in our analysis was anticipated to lead to a more robust assessment of the factors associated with more PTSD symptoms. Although these two instruments aim to measure the presence of PTSD symptoms, their concordance in classification of patients was not perfect, with 30 patients (28%) being positive on one questionnaire but not on the other. This demonstrates the difficulty in measuring PTSD symptoms by questionnaire, but also means that both questionnaires are informative in their own right. Combining the two instruments may therefore lead to a more robust classification of patients based on their level of PTSD symptoms and may be a more useful tool in screening patients following ICU stay, while potentially reducing biases due to instrument variation [52].

uestionnaires are informative in their own right. Combining the two instruments may therefore lead to a more robust classification of patients based on their level of PTSD symptoms and may be a more useful tool in screening patients following ICU stay, while potentially reducing biases due to instrument variation [52]. We assessed traumatic experiences during ICU/hospital stay based on the patients' recollections (after 1 year). The patients' perceived traumatic experience may well have contributed to the development of PTSD symptoms, but it is also possible that having PTSD symptoms influenced their perceptions. Future studies should aim to prospectively quantify traumatic experiences during or shortly after ICU stay to draw more causal conclusions, even though this might be difficult in patients with such a lengthy recovery period [9, 48, 53, 54].

but it is also possible that having PTSD symptoms influenced their perceptions. Future studies should aim to prospectively quantify traumatic experiences during or shortly after ICU stay to draw more causal conclusions, even though this might be difficult in patients with such a lengthy recovery period [9, 48, 53, 54]. In the clinical setting, there is a continuing debate on whether to intervene in the more acute peritraumatic psychological processes or in a later phase, when symptoms or prodromes of PTSD are observed. By improving our understanding of which factors play an important role in the development of PTSD, we can better prevent PTSD symptoms in high-risk patients and decide when best to intervene. The aim of our predictive model is for it to be used by treating physicians, following the acute episode and phase of secondary peritonitis in which survival and physical recovery are the main concerns, to recognize high-risk PTSD patients. This relatively simple model can aid the surgeon, for instance, during the first outpatient visit in determining which patients are at higher risk for the development of PTSD symptoms. However, before this nomogram can be used to actually predict PTSD symptomatology in clinical practice, it must be externally validated in another cohort of patients with secondary peritonitis.

ance, during the first outpatient visit in determining which patients are at higher risk for the development of PTSD symptoms. However, before this nomogram can be used to actually predict PTSD symptomatology in clinical practice, it must be externally validated in another cohort of patients with secondary peritonitis. In conclusion, 10% of peritonitis patients report ‘high’ PTSD symptomatology and another 28% ‘moderate’ PTSD symptoms. Factors that were related to more PTSD symptoms included younger age, traumatic memories of the period of hospitalization and length of ICU stay. Knowledge of these predictive factors is required to increase awareness, and to develop tailored early treatment options for these high-risk patients our nomogram may assist in identifying patients with PTSD symptoms. Electronic supplementary material Electronic Supplementary Material (DOC 21K) Electronic Supplementary Material (DOC 52K) Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited. Appendix: RELAP trial clinical centers and investigators of the Dutch Peritonitis Study Group All investigators are from departments of surgery unless specified as clinical epidemiology and biostatistics (E), intensive care (I) or medical psychology (MP).

Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited. Appendix: RELAP trial clinical centers and investigators of the Dutch Peritonitis Study Group All investigators are from departments of surgery unless specified as clinical epidemiology and biostatistics (E), intensive care (I) or medical psychology (MP). O. van Ruler, K. R. Boer (E), J. B. Reitsma (E), C. W. Mahler, E. A. Reuland, J. W. O. van Till, B. C. Opmeer (E), P. M. M. Bossuyt (E), M. J. Schultz (I), M. A. Sprangers (MP), H. Obertop, D. J. Gouma, C. A. J. M. de Borgie (E), M. A. Boermeester: Academic Medical Center, Amsterdam; E. P. Steller, P. Tanis, H. Hart (I): Sint Lucas Andreas Hospital, Amsterdam; M. F. Gerhards, M. Guijt, H. M. Oudemans (I): Onze Lieve Vrouwe Gasthuis, Amsterdam; K. Bosscha, E. Ritchie, M. Vermeer: Bosch Medical Centre, Den Bosch; P. W. de Graaf, B. van Etten, C. Haazer, E. Salm (I): Reinier de Graaf Hospital, Delft; B. Lamme, E. J. Hesselink, H. Rommes (I): Gelre Hospitals, Lukas Hospital, Apeldoorn; R. J. Oostenbroek, L. te Velde, G. Govaert, H. H. Ponssen (I): Albert Schweitzer Hospital, Dordrecht; H. G. Gooszen, M. K. Dinkelman, L. P. H. Leenen (I): University Medical Centre Utrecht; E. G. J. M. Pierik, K. W. W. Lansink, J. Bakker (I): Isala Clinics, Zwolle.

ommes (I): Gelre Hospitals, Lukas Hospital, Apeldoorn; R. J. Oostenbroek, L. te Velde, G. Govaert, H. H. Ponssen (I): Albert Schweitzer Hospital, Dordrecht; H. G. Gooszen, M. K. Dinkelman, L. P. H. Leenen (I): University Medical Centre Utrecht; E. G. J. M. Pierik, K. W. W. Lansink, J. Bakker (I): Isala Clinics, Zwolle. Key staff and steering committee at coordinating center of RELAP trial: O. van Ruler (investigator), E. A. Reuland (data management), C. W. Mahler (investigator), J. B. Reitsma (epidemiologist), C. A. J. M. de Borgie (epidemiologist), K. R. Boer (quality of life investigator), B. C. Opmeer (economist), M. A. Boermeester (principal investigator, project supervisor) from the Department of Surgery, Academic Medical Center Amsterdam, The Netherlands. Abbreviations APACHE-IIAcute Physiology and Chronic Health Evaluation II ARDSAcute respiratory distress syndrome APSAcute Physiology Score DSM-IVDiagnostic and Statistical Manual of Mental Disorders–Fourth Edition ICUIntensive care unit IQRInterquartile range IES-RImpact of Events–Revised Questionnaire NNumber of patients n. a.Not applicable OROdds ratio PTSDPosttraumatic stress disorder PTSS-10Post-Traumatic Stress Syndrome 10-Questions Inventory SCIDStandardized Clinical Interview Diagnosis SDStandard deviation 95% CI95% Confidence intervals Financial support: Supported by the Dutch Organization for Health Research and Development (ZonMW), The Hague, The Netherlands. Health Care Efficiency Program, grant number 945-02-028. The RELAP trial clinical centers and the investigators of the Dutch Peritonitis Study Group are listed in the Appendix.

95% CI95% Confidence intervals Financial support: Supported by the Dutch Organization for Health Research and Development (ZonMW), The Hague, The Netherlands. Health Care Efficiency Program, grant number 945-02-028. The RELAP trial clinical centers and the investigators of the Dutch Peritonitis Study Group are listed in the Appendix. Competing interests: There are no competing interests for any of the authors for the content of this paper. Authors' contributions: M. A. B., K. R. B., C. A. J. M. dB. and M. A. S. designed the study and advised on PTSD and depression. All information pertaining to surgical procedures and ICU stay for the final manuscript were considered by M. A. B. and S. E. dR. K. R. B. and O. vR. were responsible for the coordination of the study, including contacting patients, collecting data and entering data. K. R. B., J. B. R. and M. A. B. analyzed data, and K. R. B. was responsible for the final manuscript. K. R. B., O. vR., C. A. J. M. dB., J. B. R., M. A. S., S. E. dR., A. A. P. vE. and M. A. B. interpreted and discussed all data. All authors read and approved the final manuscript. Electronic supplementary material The online version of this article (doi:10.1007/s00134-007-0941-3) contains supplementary material, which is available to authorized users.

Sir: Dellinger et al. are to be congratulated on their new Surviving Sepsis Campaign recommendations as published recently in this journal [1]. The efforts of Dellinger and his group are of great value for all of us who take care of patients with sepsis. Indeed, guidelines like those from the Surviving Sepsis Campaign, an initiative of the European Society of Intensive Care Medicine, the International Sepsis Forum, and the Society of Critical Care Medicine, are very much welcomed and appreciated by the intensive care community. They have been and continue to be an important tool in improving care of septic patients globally. However, in our opinion, the new recommendations are flawed with respect to glucose control. The new recommendations for septic patients advise maintaining the blood glucose concentration (BGC) below 150 mg/dl (8.3 mmol/l) instead of adhering to the more strict thresholds as in the two randomized controlled trials by van den Berghe et al. (BGC 80–110 mg/dl, 4.0–6.1 mmol/l) [2, 3]. The Surviving Sepsis Campaign justifies this departure on several grounds: new studies show conflicting results incidence of severe hypoglycemia with lower BGC thresholds, with a presumed association between severe hypoglycemia and death. We disagree with the advised thresholds, and argue against the reasoning for not retaining the original stricter BGC thresholds.

parture on several grounds: new studies show conflicting results incidence of severe hypoglycemia with lower BGC thresholds, with a presumed association between severe hypoglycemia and death. We disagree with the advised thresholds, and argue against the reasoning for not retaining the original stricter BGC thresholds. First, the evidence. Simply, there are no studies that provide evidence for glucose control in sepsis using the now recommended threshold of 150 mg/dl. Using the 150 mg/dl upper limit, more septic patients will be in the higher BGC range. However, with more successful glucose control (i. e., BGC closer to 80 mg/dl) more benefit is achieved [4, 5]. Indeed, the lowered BGC rather than the insulin dose is related to reduced mortality, critical illness polyneuropathy, bacteremia, and inflammation [4].

t, more septic patients will be in the higher BGC range. However, with more successful glucose control (i. e., BGC closer to 80 mg/dl) more benefit is achieved [4, 5]. Indeed, the lowered BGC rather than the insulin dose is related to reduced mortality, critical illness polyneuropathy, bacteremia, and inflammation [4]. Second, the conflicting information that comes from new (yet unpublished) studies [6, 7]. Unfortunately, the multicenter VISEP trial on glucose control in Germany by the SepNet group was discontinued prematurely because of identical mortality rates in the treatment groups but a higher incidence of hypoglycemia in the glucose control group (12.1% vs. 2.1%) [6]. Another study, the European GLUControl trial, was also stopped before inclusion was completed [7]. As in the VISEP study, the primary reason for stopping inclusion of patients was the relatively high incidence of hypoglycemia in the glucose control group. Due to the early termination of these studies we are left with two underpowered randomized controlled trials which can by no means be used as evidence in the discussion on potential benefit of glucose control.

stopping inclusion of patients was the relatively high incidence of hypoglycemia in the glucose control group. Due to the early termination of these studies we are left with two underpowered randomized controlled trials which can by no means be used as evidence in the discussion on potential benefit of glucose control. Third, blood glucose control with insulin carries the risk of hypoglycemia. Indeed, the incidence of severe hypoglycemia (defined as a blood glucose concentration < 40 mg/dl) is 5 to 10 times higher than with a conventional blood glucose strategy [2, 3]. Thus, the rise in incidence of hypoglycemia in the glucose control groups in the two newer studies as mentioned above was not surprising. Similar findings came from numerous other studies reporting on some form of glucose control (merely studies with a before–after design) [8]. Fear of severe hypoglycemia has, at least in part, hampered broad implementation of glucose control [8]. Although many ICUs have adopted some form of glucose control, frequently the applied glucose control regimens have higher thresholds than those used in the original studies. Of note, with higher BGC thresholds, the incidence of severe hypoglycemia does not differ greatly from that with more strict BGC thresholds [8].

. Although many ICUs have adopted some form of glucose control, frequently the applied glucose control regimens have higher thresholds than those used in the original studies. Of note, with higher BGC thresholds, the incidence of severe hypoglycemia does not differ greatly from that with more strict BGC thresholds [8]. Fourth, the assumed association between hypoglycemia and poor outcome. We recently analyzed the short-term consequences (seizures, coma, and death) of hypoglycemia [9]. The hazard ratio for in-hospital death was 1.03 (95% confidence interval 0.68–1.56; P = 0.88) in patients with a first occurrence of hypoglycemia relative to the controls without hypoglycemia. Results were corrected for duration of intensive care unit admittance before hypoglycemia, age, sex, and Acute Physiology, Age and Chronic Health Evaluation II score at admission. In addition, no cases of hypoglycemia-associated death were reported. Although the evidence for glucose control with strict thresholds does not yet support a grade A recommendation (based on the highest level of evidence), it does appear to be stronger than the evidence in support of a strategy of tolerating higher BGC thresholds. Most importantly, however, the only evidence we have supports a BGC threshold of 110 mg/dl, not 150 mg/dl. Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.

Although the evidence for glucose control with strict thresholds does not yet support a grade A recommendation (based on the highest level of evidence), it does appear to be stronger than the evidence in support of a strategy of tolerating higher BGC thresholds. Most importantly, however, the only evidence we have supports a BGC threshold of 110 mg/dl, not 150 mg/dl. Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited. An author's reply to this comment is available at: http://dx.doi.org/10.1007/s00134-008-1028-5.

Introduction In the past few decades, Pseudomonas aeruginosa has become a major hospital pathogen, due to both the number and severity of the infections it causes. A national prevalence survey indicated that P. aeruginosa was responsible for 10% of all nosocomial infections in France, only slightly fewer than Escherichia coli and Staphylococcus aureus  [1]. In the hospital setting, intensive care units (ICUs) have a high endemic potential for this bacterium, which causes 18% of nosocomial infections in such units, versus only 6% in surgical and non-surgical units [2]. In many departments, P. aeruginosa is frequently involved in broncho-pulmonary infections and, to a lesser extent, urinary infections, infections of surgical sites, and bacteremia. The severity and excess mortality observed in cases of pulmonary infection and bacteremia are due to a combination of the intrinsic properties of the bacteria – virulence factors and the natural resistance to antibiotics – and the immunocompromized status of the infected patients [3].

of surgical sites, and bacteremia. The severity and excess mortality observed in cases of pulmonary infection and bacteremia are due to a combination of the intrinsic properties of the bacteria – virulence factors and the natural resistance to antibiotics – and the immunocompromized status of the infected patients [3]. P. aeruginosa is an environmental bacterium. It exists as a saprophyte on damp soil, plant material, and in freshwater, wastewater, and seawater [4]. It is independent of humans but may be found as a commensal organism in the digestive tract. P. aeruginosa is rarely carried by subjects in good health, being found in only 2–10% of individuals, whereas it may be found in 50–60% of hospitalized patients, particularly on burns and scabs. The survival of P. aeruginosa seems to be particularly favored by damp environments, such as sinks, taps, shower heads, and other water fittings [5, 6]. This enables this bacterium to contaminate medical and surgical equipment, hospital fittings, and other material [7]. Numerous hospital outbreaks have been blamed on the colonization of diverse items of equipment and/or damp materials [8, 9]. As a result, many hospital hygiene teams place great emphasis on the role of water in all infections with P. aeruginosa, particularly those occurring in ICUs in the absence of an outbreak. The aim of this study was to evaluate the role of the water environment on the colonization of patients hospitalized in ICUs in the absence of a recognized epidemic.

P. aeruginosa is an environmental bacterium. It exists as a saprophyte on damp soil, plant material, and in freshwater, wastewater, and seawater [4]. It is independent of humans but may be found as a commensal organism in the digestive tract. P. aeruginosa is rarely carried by subjects in good health, being found in only 2–10% of individuals, whereas it may be found in 50–60% of hospitalized patients, particularly on burns and scabs. The survival of P. aeruginosa seems to be particularly favored by damp environments, such as sinks, taps, shower heads, and other water fittings [5, 6]. This enables this bacterium to contaminate medical and surgical equipment, hospital fittings, and other material [7]. Numerous hospital outbreaks have been blamed on the colonization of diverse items of equipment and/or damp materials [8, 9]. As a result, many hospital hygiene teams place great emphasis on the role of water in all infections with P. aeruginosa, particularly those occurring in ICUs in the absence of an outbreak. The aim of this study was to evaluate the role of the water environment on the colonization of patients hospitalized in ICUs in the absence of a recognized epidemic. Materials and methods Background This study was carried out in the two adult ICUs (one surgical and one medical) at Besançon University Hospital. Each of these two units has 15 beds, all of which are in individual rooms for the surgical ICU, with two double rooms for the medical ICU. This study was carried out (collection of data and strains) over 8 weeks, from 20 February to 10 April 2006. It was approved by the hospital's review board.

çon University Hospital. Each of these two units has 15 beds, all of which are in individual rooms for the surgical ICU, with two double rooms for the medical ICU. This study was carried out (collection of data and strains) over 8 weeks, from 20 February to 10 April 2006. It was approved by the hospital's review board. Design of the study Environmental samples were taken weekly from the water fittings of the rooms, regardless of the P. aeruginosa infection status of the patients in the rooms. P. aeruginosa colonization status was monitored by testing samples taken for diagnostic purposes and samples taken weekly for screening purposes. We tested for isogenicity between clinical and environmental samples by determining macrorestriction profiles of total DNA, by pulsed-field gel electrophoresis (PFGE). Clinical samples The clinical samples tested were taken for diagnostic and epidemiological (screening) purposes. Diagnostic samples were not taken systematically: they were taken only if there were clinical reasons to suspect infection, and the sampling site was chosen according to the site of suspected infection. Screening samples were taken systematically on admission of the patient to the ICU and once per week thereafter, throughout the patient's stay. They were taken from the nose, the rectum, and from tracheal aspirates. Colonization was defined as a positive result for at least one sample.

the site of suspected infection. Screening samples were taken systematically on admission of the patient to the ICU and once per week thereafter, throughout the patient's stay. They were taken from the nose, the rectum, and from tracheal aspirates. Colonization was defined as a positive result for at least one sample. Environmental samples Two types of environmental samples were taken once per week from the water fittings in each intensive care room: 10 ml of water from the U-bend under the sink and 150 ml of cold water taken directly from the tap immediately after activating. The various volumes of water sampled were collected into sterile flasks containing sodium thiosulphate to inhibit the effects of water chlorination. Culture Diagnostic samples were plated on agar Columbia broth supplemented with 5% horse blood. Screening samples were plated on agar Muëller–Hinton broth. Environmental samples were plated on cetrimide agar after filtration and 24 h enrichment in trypticase-soy (TS) broth. Cetrimide agar plates were incubated for 72 h at 41°C. Identification Cetrimide agar is selective for P. aeruginosa, but other bacterial species may nevertheless develop. Identification of P. aeruginosa was based on positive oxidase activity, resistance to kanamycin, growth at 41°C and not at 4°C, and confirmed by biochemical tests (ID 32 GN, Biomérieux, Marcy l'étoile, France).

fication Cetrimide agar is selective for P. aeruginosa, but other bacterial species may nevertheless develop. Identification of P. aeruginosa was based on positive oxidase activity, resistance to kanamycin, growth at 41°C and not at 4°C, and confirmed by biochemical tests (ID 32 GN, Biomérieux, Marcy l'étoile, France). Genotyping Isolates were typed by determination of their total DNA macrorestriction profile (pulsotype) following digestion with DraI (Boehringer, Mannheim, Germany) as assessed by PFGE (CHEF DRIII, Bio-Rad Ivry sur Seine, France), according to a technique previously developed by our laboratory [10]. We used Gel-Compar (Applied Math, Kortrijk, Belgium) to establish a similarity matrix for the DNA based on calculation of the Dice coefficient (pairwise comparison of strains). A dendrogram was generated with the UPGMA (unweighted pair group using arithmetic means) hierarchical algorithm. We compared gels using S. aureus NCTC 8325 as a reference strain. Typing results were interpreted according to international recommendations [11]. Definitions Clones including clinical isolates were defined as sporadic if they were isolated from only a single patient, microepidemic if they were found in two or three patients, and epidemic if from more than three patients. Clones including environmental isolates were defined as unique if the isolates were found in a single room and multiple if the isolates were present in several rooms.

they were isolated from only a single patient, microepidemic if they were found in two or three patients, and epidemic if from more than three patients. Clones including environmental isolates were defined as unique if the isolates were found in a single room and multiple if the isolates were present in several rooms. Statistical analysis The data were analyzed with EpiInfo version 6.04 (EpiInfo, CDC, Atlanta, GA, USA). Confidence intervals were calculated by the quadratic method of Fleiss. Results Incidence of colonization/infection In total, 123 patients were admitted to the two ICUs (69 to the medical unit and 54 to the surgical unit) for a total of 1416 days of hospitalization (720 in the medical unit and 696 in the surgical unit). Seventeen patients (8 in the medical unit and 9 in the surgical unit) presented at least one sample positive for P. aeruginosa. The overall incidence of colonization was 13.8 (range 8.5–21.5) per 100 patients admitted: 11.6 (range 5.5–22.1) for the medical unit and 16.66 (range 8.4–29.8) for the surgical unit. The overall incidence density was 12.0 (range 7.2–19.6) per 1000 days of hospitalization: 11.1 (range 5.9–22.7) for the medical unit and 12.93 (range 8.3–25.3) for the surgical unit. In total, 63 samples tested positive for P. aeruginosa: 46 screening samples from 16 patients and 17 diagnostic samples from 7 patients.

The overall incidence density was 12.0 (range 7.2–19.6) per 1000 days of hospitalization: 11.1 (range 5.9–22.7) for the medical unit and 12.93 (range 8.3–25.3) for the surgical unit. In total, 63 samples tested positive for P. aeruginosa: 46 screening samples from 16 patients and 17 diagnostic samples from 7 patients. Positivity of the water environment In total, 448 samples were taken from both ICUs. P. aeruginosa was detected in 193 (86.2%) of the 224 U-bend samples and 10 of the 224 samples taken from the tap (4.5%; Table 1). More than half the samples taken from U-bends contained more than one strain of P. aeruginosa, with some samples containing up to four strains. Permanent colonization of the U-bend was observed in five of the 28 rooms. Two of these rooms were in the surgical ICU and the other three were in the medical ICU. Table 1 Positive results (and numbers of clones) for samples taken from U-bends

ne strain of P. aeruginosa, with some samples containing up to four strains. Permanent colonization of the U-bend was observed in five of the 28 rooms. Two of these rooms were in the surgical ICU and the other three were in the medical ICU. Table 1 Positive results (and numbers of clones) for samples taken from U-bends Room Date 21 Feb 27 Feb 6 Mar 13 Mar 20 Mar 27 Mar 3 Apr 10 Apr Medical intensive care unit A1 1a 2 2a 3 2a 0 1 3 A2 1 3 2 2 1 2 2 0 A3 1 1 1 2 2 2 2 2 A4 1 1 3 3 3 1 0 2 A5 1 1 2 3 3 1 1 2 B1 1 3 2 1 2 2 1 2 B2 1a 2a 2a 2a 2a 1 3 0 B3 1 3 1 3 2 2 2 2 B4 1 1 3 0 4 2 2 0 B5 1 1 2 2 1 1 2 0 C1 1 1 4 2 2 0 3 2 C2-3 1 3 3 3 3 3 2 0 C4-5 1 2 4 3 2 0 1 2 Surgical intensive care unit A1 1 2 1 3 2 0 1 0 A2 1 2 3 2 2 4 0 2 A3 1 3 1 2 1 0 1 2 A4 1 2 2 1 2 3 2 2 A5 1 2 1 3 2 2 1 3 A6 1 2 3 2 2 0 2 0 A7 1 2 2 3 2 0 1 0 B1 1 2 2 2 3 1 2 0 B2 1 1 2 2 2 2 0 0 B3 1 1a 0 2 1 1 2 0 B4 1 3 1 2 3 2 1 2 C1 1 0 4 2 2 1 0 1 C2 0 0 2 2 2 1 3 2 C3 1 3 1 2 3 1 2 0 C4 1 0 3 2 1a 0 1 0 a One positive sample from taps

t A1 1 2 1 3 2 0 1 0 A2 1 2 3 2 2 4 0 2 A3 1 3 1 2 1 0 1 2 A4 1 2 2 1 2 3 2 2 A5 1 2 1 3 2 2 1 3 A6 1 2 3 2 2 0 2 0 A7 1 2 2 3 2 0 1 0 B1 1 2 2 2 3 1 2 0 B2 1 1 2 2 2 2 0 0 B3 1 1a 0 2 1 1 2 0 B4 1 3 1 2 3 2 1 2 C1 1 0 4 2 2 1 0 1 C2 0 0 2 2 2 1 3 2 C3 1 3 1 2 3 1 2 0 C4 1 0 3 2 1a 0 1 0 a One positive sample from taps Molecular typing The clinical isolates belonged to 17 clones. Sixteen of these 17 clones were sporadic (isolated from a single patient), and one clone was microepidemic including isolates from two patients. We typed 203 environmental isolates in all. These 203 environmental isolates belonged to 82 pulsotypes – 37 present in the medical ICU, 33 in the surgical ICU and 12 in both units; 54 of these 82 pulsotypes were isolated only once – 29 in the medical unit and 25 in the surgical unit – and the other 28 pulsotypes were multiple. Eight of these strains were present in the medical unit, 8 in the surgical unit, and 12 were present in both units (Table 1). Only two multiple clones included both clinical and environmental isolates: one was isolated from 7 environmental samples and 1 patient, and the other was isolated from 8 environmental samples and 2 patients. The time course of colonization was investigated by considering the timing of the clinical and environmental samples taken for the 17 patients colonized with P. aeruginosa (Table 2). Only 1 patient was colonized with a clone present in the water environment of his room before he gave his first positive sample. In the 3 cases without environmental sample E1, the environment perhaps acted as a reservoir, too. The environment was therefore a possible reservoir for the colonization of patients in 1 of 14 cases. In half the cases, the water environment (U-bends and taps) of the room contained several clones of P. aeruginosa. For a given water fitting, in 9 cases of 10, the tap and the U-bend were colonized with identical clones. Table 2 Timing of the colonization/infection of patients with respect to the timing of positive results for environmental samples from the water fittings

taps) of the room contained several clones of P. aeruginosa. For a given water fitting, in 9 cases of 10, the tap and the U-bend were colonized with identical clones. Table 2 Timing of the colonization/infection of patients with respect to the timing of positive results for environmental samples from the water fittings Patient no. Clone no. First positive environmental sample First clinical sample Second environmental sample 1 Absa 43 61 2 9, 48 23 5 3 43 43 43 4 43 34 31, 43 5 66 42 –b 6 42 40 –b 7 Absa 7 41, 104 8 71 37 42, 64 9 Absa 2 40, 42, 76 10 29 18 20, 38 11 42, 90 36 42 12 2 13 2, 10, 38 13 10, 27 31 10, 27 14 43 22 –b 15 42, 101 65 42 16 2, 66 12, 103 67 17 31, 32, 74 35 –b a Absence of environmental sample b Negative environmental sample Discussion In this study, the incidence of colonization by P. aeruginosa was high, but molecular typing showed that there was no clonal outbreak. This high incidence was revealed by identification of simple colonization through testing screening samples. If diagnostic samples only had been used, the measured incidence would have been about half that reported. The water environment played only a minor role because only one of the 14 cases that could be evaluated was consistent with a colonization from the water sources tested (7.2% of cases).

testing screening samples. If diagnostic samples only had been used, the measured incidence would have been about half that reported. The water environment played only a minor role because only one of the 14 cases that could be evaluated was consistent with a colonization from the water sources tested (7.2% of cases). Testing samples taken from water fittings in surgical and medical intensive care rooms in this study showed, nevertheless, that U-bends were frequently contaminated with P. aeruginosa, consistent with published findings [5, 6, 12, 13]. In half the cases, the water environment (U-bends and taps) of the room contained several clones of P. aeruginosa. This, together with the clonal diversity of the strains isolated from U-bends (differences between and within U-bends), provides evidence for colonization of the U-bend from the exterior rather than originating from the water supply. In addition, the presence of identical strains in the U-bends of sinks in both these physically independent units not shared by the same patients is suggestive of retro-colonization of the U-bend by the microflora present in wastewater pipes, via the biofilm, as previously proposed [14].

ginating from the water supply. In addition, the presence of identical strains in the U-bends of sinks in both these physically independent units not shared by the same patients is suggestive of retro-colonization of the U-bend by the microflora present in wastewater pipes, via the biofilm, as previously proposed [14]. The role of the water environment in the P. aeruginosa colonization of patients was the key issue in this study. Many studies have attributed a major role to water fittings in the incidence of patient colonization with P. aeruginosa in ICUs [5, 12, 14]. Other studies have reported only a weak epidemiological link between environmental colonization and the occurrence of infections in patients [4, 15, 16]. All these previous studies were carried out during outbreaks. Some prospective studies have been published and they report frequencies of patient colonization by the environment of 14.2–50% [3, 6, 12–14, 17, 18]. In our study, in the absence of clonal epidemic, the frequency of colonization of patients via the water environment of their rooms seemed to be lower. Regardless of the epidemiological situation, which may differ between departments, measurement of the frequency with which this event arises requires three methodological conditions which have never before been fulfilled: The identification of all patients colonized by epidemiological sampling (screening) of patients. We found that the use of diagnostic samples alone would have identified only about half the colonized/infected patients.

ch this event arises requires three methodological conditions which have never before been fulfilled: The identification of all patients colonized by epidemiological sampling (screening) of patients. We found that the use of diagnostic samples alone would have identified only about half the colonized/infected patients. The collection of samples from water fittings should be generalized and carried out regularly to make it possible to determine when the patients became infected or colonized. This is rarely the case in surveys motivated by the suspicion of an epidemic. Strain comparisons should be based on a highly reproducible and discriminant typing method. The determination of the total DNA macrorestriction profile by PFGE can be considered to be the gold standard method for typing [10, 19]. The frequency of strains widely present in the environment (multiple clones) but never isolated from patients was high. Observations of this type led Valles et al. to suggest that there may be two different genetic groups in this species: one group of strains that are mostly environmental and not very pathogenic in humans; and one group of strains better adapted to humans with a much higher pathogenic potential [13]. This view is supported by virulence surveys, analyses of bacterial populations by multi-locus sequence typing, and the application of several typing methods to strains of two different origins (clinical and environmental) [4, 20, 21].

oup of strains better adapted to humans with a much higher pathogenic potential [13]. This view is supported by virulence surveys, analyses of bacterial populations by multi-locus sequence typing, and the application of several typing methods to strains of two different origins (clinical and environmental) [4, 20, 21]. Conclusion In conclusion, although water fittings clearly play a role in the acquisition of P. aeruginosa by patients hospitalized in ICUs, the contribution of this phenomenon in non-epidemic situations appears to be much smaller than generally believed by many operational hospital infection control teams. Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.

Sir: Several recent studies clearly identify the development of hyperglycaemia as an important risk factor in terms of mortality and morbidity of critically ill patients. In patients undergoing cardiac surgery, hyperglycaemia has been associated with a substantial mortality risk and delayed extubation [1, 2]. Strong evidence favouring a strategy targeting strict glucose control (4.5–6.0 mmol/l) came from the landmark prospective, randomized, controlled study on intensive insulin therapy in adult surgical critically ill patients [3]. Recently, the clinical benefits of this therapy in adults were largely confirmed in a large, randomized, controlled trial in a strictly medical adult ICU patient population [4]; however, whether the maintenance of normoglycaemia in critically ill children is also beneficial on patient outcome is undetermined. In an effort to analyse the presently available evidence related to paediatric critically ill patients, we performed a structured critical appraisal of the literature.

ent population [4]; however, whether the maintenance of normoglycaemia in critically ill children is also beneficial on patient outcome is undetermined. In an effort to analyse the presently available evidence related to paediatric critically ill patients, we performed a structured critical appraisal of the literature. In critically ill children (patient) does insulin therapy in order to maintain strict normoglycaemia (intervention) improve survival, or reduce morbidity (outcome)? We have performed the following search strategy: Secondary (Cochrane library) and primary (PubMed) sources were included in the search. These databases were searched for: (“Insulin”[Mesh] or “Hyperglycemia”[Mesh] or “Glucose Metabolism Disorders”[Mesh] or “Glucose”[Mesh]) and (“Critical Illness”[Mesh] or (“Intensive Care”[Mesh] or “Intensive Care Units, Pediatric”[Mesh])) and systematic [sb] not (“Diabetic Ketoacidosis”[Mesh] or “Diabetes Mellitus”[Mesh]). Our search yielded 37 results, five relevant to the question, two related to children. One of the publications used a pooled database [4] partly described in a previous study. [3] One study was non-observational only. The three remaining studies are given in Table 1. Table 1 Overview of relevant papers

ellitus”[Mesh]). Our search yielded 37 results, five relevant to the question, two related to children. One of the publications used a pooled database [4] partly described in a previous study. [3] One study was non-observational only. The three remaining studies are given in Table 1. Table 1 Overview of relevant papers Reference Study group Level of evidence Outcome Key results (95% CI) Comments [4] Adult ICU patients, n = 2,748; conventional insulin treatment group n = 1,388; intensive insulin treatment (IIT) group n = 1,360 Meta-analysis (level 1a) Mortality, critical illness polyneuropathy, renal failure, hypoglycaemia Mortality: relative risk reduction (RRR) 19% (95% CI: 2–35%); absolute risk reduction (ARR): 0.03 (0.004–0.056); number needed to treat (NNT): 33 (18–281) Study limited to adults; insulin therapy in the conventional treatment group was initiated when blood glucose levels exceeded 215 mg/dl (12 mmol/l), and adjusted to keep blood glucose between 180 and 200 mg/dl (10–11 mmol/l); in the intensive insulin treatment group, therapy was initiated when blood glucose was above 110 mg/dl (6 mmol/l) and adjusted to maintain blood glucose between 80 and 110 mg/dl (4.5–6 mmol/l) New renal failure: RRR 42 (18–65), ARR 0.032 (0.014–0.050), NNT 31 (20–71) Critical illness polyneuropathy: RRR 40 (25–56), ARR 0.063 (0.038–0.088), NNT 16 (11–26) Hypoglycemia: RRR −528% (629 to −427%), ARR −0.095 (−0.013 to −0.077), NNH 11 (9–13) Strict glucose control reduced in-hospital mortality for patients with > 3 ICU days but did not lead to a160;difference in overall in-hospital mortality Mortality or neurological sequelae in hypoglycaemic patients: RRR 50% (−237 to 100%), ARR −0.001 (−0.005 to 0.003), NNH 1,000 (211 to infinite) [6] Pediatric septic shock patients (n = 57) Single-center, prospective observational cohort study (level 2c) Mortality Peak glucose level in non-survivors 262 ± 110 mg/dl (14.5 ± 6 mmol/l) was higher than in survivors: 167.8 ± 55 mg/dl (9.5 ± 3 mmol/l; p < 0.01) Observational pediatric study; univariate analysis identified three possible factors that could be associated with increased mortality (higher glucose level, male gender, pediatric risk of mortality score II above 10); multivariate analysis demonstrated that peak glucose level was the only independent risk factor associated with mortality Best peak glucose level predicting death was 178 mg/dl (10 mmol/l), sensitivity 0.71, specificity 0.72, relative risk of death in hyperglycaemic patients 2.59 (1.37–4

f mortality score II above 10); multivariate analysis demonstrated that peak glucose level was the only independent risk factor associated with mortality Best peak glucose level predicting death was 178 mg/dl (10 mmol/l), sensitivity 0.71, specificity 0.72, relative risk of death in hyperglycaemic patients 2.59 (1.37–4 .88) Severity of illness was not reported; 941 of 1,927 patients excluded because of absence of glucose measurement, leading to a possible selection bias [5] Pediatric critically ill, non-diabetic patients (n = 942) Single-center, retrospective observational cohort study (level 2c) Mortality and length of stay Peak glucose level above 150 mg/dl (8.3 mmol/l) within 10 days of initial glucose measurement predicted death with a sensitivity of 81% (68–93%), specificity 51% (48–54%), relative risk of death 4.13 (1.83–9.32); length of stay in hyperglycaemic group was higher 6.1 ± 9.6 vs. 4.0 ± 6.0 days (p = 0.001)

.88) Severity of illness was not reported; 941 of 1,927 patients excluded because of absence of glucose measurement, leading to a possible selection bias [5] Pediatric critically ill, non-diabetic patients (n = 942) Single-center, retrospective observational cohort study (level 2c) Mortality and length of stay Peak glucose level above 150 mg/dl (8.3 mmol/l) within 10 days of initial glucose measurement predicted death with a sensitivity of 81% (68–93%), specificity 51% (48–54%), relative risk of death 4.13 (1.83–9.32); length of stay in hyperglycaemic group was higher 6.1 ± 9.6 vs. 4.0 ± 6.0 days (p = 0.001) In critically ill adults, it has been shown that maintenance of strict normoglycaemia (4.5–6.0 mmol/l) with intensive insulin therapy substantially prevents morbidity and reduces mortality [4]. The risk of hypoglycaemia increases with this therapy, but it is unclear whether this is truly harmful in the setting of adult intensive care. In paediatric critically ill patients, hyperglycaemia is prevalent and is associated with a worse outcome [5, 6]. This association in itself, however, does not imply a causal relation. Whereas in adult studies maintenance of blood glucose levels between 80 and 110 mg/dl (4.5–6.0 mmol/l) has shown to reduce both mortality and morbidity [4], in the described paediatric studies higher glucose levels above 150 mg/dl (8.3 mmol/l) or above 178 mg/dl (10 mmol/l) have been found to be predict mortality [5, 6]. Uniform criteria to define hyperglycaemia in critically ill children have yet to be established and may differ between age groups. Moreover, the acute phase of sepsis    in children may differ significantly from the hyperinsulinaemic hyperglycaemia associated with insulin resistance in adult sepsis. In a recent study it has been shown that children in shock due to meningococcal sepsis showed signs of insufficient insulin response to hyperglycaemia, whereas patients without signs of shock were insulin resistant [7]. In paediatric critically ill patients, hypoglycaemia is also prevalent and is related to increased mortality [8]. Hypoglycaemia was significantly more common in patients receiving intensive insulin therapy [4]. Insulin therapy in sick children with high blood glucose levels, exceeding 178 mg/dl (10 mmol/l), can be advocated, but based on current evidence, there is insufficient data to extrapolate to critically ill children that strict glucose control is beneficial.

e common in patients receiving intensive insulin therapy [4]. Insulin therapy in sick children with high blood glucose levels, exceeding 178 mg/dl (10 mmol/l), can be advocated, but based on current evidence, there is insufficient data to extrapolate to critically ill children that strict glucose control is beneficial. Currently, two large registered and ongoing European randomised controlled trials on the subject of tight glucose control in critically ill paediatric patients are being performed: Control of Hyperglycaemia in Paediatric Intensive Care (ISRCTN 61735247, London, http://www.chip-trial.org.uk) and Tight Glycemic Control With Intensive Insulin Treatment in PICU (NCT00214916, Leuven). In conclusion, no randomised controlled studies focusing on strict glucose control in paediatric patients were found. In critically ill adults maintenance of strict normoglycaemia with intensive insulin therapy reduces morbidity and mortality (evidence grade A). In paediatric ill patients, hyperglycaemia is prevalent and levels above 178 mg/dl (10 mmol/l) are associated with increased mortality (grade B).Ill children, however, are also susceptible to hypoglycaemia. Based on current evidence, insulin therapy aimed at strict glucose control [blood glucose levels between 80 and 110 mg/dl (4.5–6 mmol/l)] cannot be recommended in paediatric critically ill patients. Any future change in practice should be based on evidence from current randomised clinical trials.

ible to hypoglycaemia. Based on current evidence, insulin therapy aimed at strict glucose control [blood glucose levels between 80 and 110 mg/dl (4.5–6 mmol/l)] cannot be recommended in paediatric critically ill patients. Any future change in practice should be based on evidence from current randomised clinical trials. Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.

Introduction Glucose control with intensive insulin therapy in critically ill medical and surgical patients reduces morbidity and mortality [1, 2]. However, this therapy carries the risk of hypoglycemia. Although little is known about the impact of hypoglycemia, cases with severe sequelae have been reported [3–5]. Recently a retrospective review found that the occurrence of a single episode of hypoglycemia is related to mortality [6]. Due to the high rate of insulin-induced hypoglycemia a multicenter trial has been stopped, and negative results have led to debate about the overall benefit of glucose control [7, 8]. Therefore a meticulous scheme of glucose measurements and subsequent insulin-dosing adjustments must be implemented. Such a scheme should lead to adequate control with as few glucose measurements as possible with minimal risk of hypoglycemia. Many insulin infusion protocols have been proposed, but performance is difficult to compare [9], and the size of reported patient cohorts is often small. As hypoglycemia is and should be a rare event, no firm conclusions can be drawn from small populations. Furthermore, a research setting in which a protocol is tested on selected patients may not reflect the performance of a protocol when used in routine practice. We implemented a computer-assisted glucose control system in routine practice and evaluated its safety and efficiency in a large cohort of unselected consecutive patients in three ICUs.

etting in which a protocol is tested on selected patients may not reflect the performance of a protocol when used in routine practice. We implemented a computer-assisted glucose control system in routine practice and evaluated its safety and efficiency in a large cohort of unselected consecutive patients in three ICUs. Materials and methods GRIP computer program The Glucose Regulation for Intensive Care Patients (GRIP) computer program gives recommendations on exact insulin pump rates and glucose measurement frequency for all patients. It also acts as watchdog against overdue measurements. Initial results and a description of the GRIP computer program have previously been reported [10]. In short, the GRIP program is installed on a computer directly next to the point of care blood gas analyzer. By default the computer shows an overview of the ICU. The colors of the beds make clear whether a specific action (such as taking a new blood glucose or modifying an insulin pump) should be performed at a certain bed. New glucose values are automatically queried from the hospital information system. These values are combined with clinical data obtained from the nurse, and recommendations on both the insulin pump rate and the measurement interval can be calculated at any moment. The algorithm employed by GRIP is based on both the level and the change in recent glucose values. A more detailed description of the algorithm is given in the accompanying Electronic Supplementary Material. GRIP was designed with practical applicability in mind; where possible, we aimed at limiting the number of blood glucose measurements needed and thus the time spent on glucose control. The total turn-around time for a nurse, from glucose measurement to final pump change was measured to be 4 min.

Supplementary Material. GRIP was designed with practical applicability in mind; where possible, we aimed at limiting the number of blood glucose measurements needed and thus the time spent on glucose control. The total turn-around time for a nurse, from glucose measurement to final pump change was measured to be 4 min. In 2005 we implemented an improved version that incorporates more safety features, such as label printing, rendering the entire procedure paperless. The complete source code of the GRIP computer program is freely available from the website http://grip-glucose.sf.net/. The insulin dosing of GRIP is flexible with regard to its target value. In view of the debated value of aiming for true normoglycemia (4.4–6.1 mmol/l), the medical directors (M.N., B.L., J.R.) preferred a more lenient glucose range of 4.0–7.5 mmol/l, and therefore GRIP was configured to aim for a glucose level of 6.5 mmol/l. It should be noted that changing the algorithm's target value is trivial.

of the debated value of aiming for true normoglycemia (4.4–6.1 mmol/l), the medical directors (M.N., B.L., J.R.) preferred a more lenient glucose range of 4.0–7.5 mmol/l, and therefore GRIP was configured to aim for a glucose level of 6.5 mmol/l. It should be noted that changing the algorithm's target value is trivial. Implementation of GRIP All ICUs that have implemented GRIP are part of the University Medical Center of Groningen, a 1,300-bed tertiary university hospital. All three ICUs are closed-format ICUs with a nursing to patient ratio of one-to-one. The three units are teaching general surgery, anesthesiology, and internal medicine residents, and fellows in intensive care medicine. GRIP was first implemented at the surgical ICU [10], followed by the cardiothoracic and neurosurgical ICU. In dialog with staff the initial target level was 7.0 mmol/l for the cardiothoracic ICU and 7.5 mmol/l for the neurosurgical ICU. Nurses were trained by a “teach-the-teacher” principle, where two or three nurses of each ICU were trained at the surgical ICU, after which they taught the colleagues at their own unit. After a run-in period the GRIP target level was lowered to 6.5 mmol/l.

he cardiothoracic ICU and 7.5 mmol/l for the neurosurgical ICU. Nurses were trained by a “teach-the-teacher” principle, where two or three nurses of each ICU were trained at the surgical ICU, after which they taught the colleagues at their own unit. After a run-in period the GRIP target level was lowered to 6.5 mmol/l. Patients We included 2,800 admissions with 73,188 glucose measurements for analysis; Table 1 presents patients' characteristics. We recorded baseline demographics, reason for admission, and severity of illness as measured by Acute Physiology and Chronic Health Evaluation (APACHE) II scores [median 14, interquartile range (IQR) 10–19]. We excluded the patients who were treated by GRIP in any of the run-in periods. Figure 1 shows a patient selection diagram. All patients were managed by GRIP. Orders of GRIP could be overruled at the discretion of the physician at any time, or a patient could be taken out of GRIP altogether. The latter happened structurally with patients who had recovered enough to be taking meals. During intermittent invasive interventions insulin was stopped. The local institutional review board approved this study. Table 1 Patient characteristics

physician at any time, or a patient could be taken out of GRIP altogether. The latter happened structurally with patients who had recovered enough to be taking meals. During intermittent invasive interventions insulin was stopped. The local institutional review board approved this study. Table 1 Patient characteristics Surgical ICU (n = 1,073, 38%) Thoracic ICU (n = 1,597, 57%) Neurosurgical ICU (n = 130, 5%) Treatment period Feb. 2005–June 2007 Jan. 2006–June 2007 Mar. 2007–June 2007 Age (years) 62 (50–72) 66 (57–74) 55 (46–65) Male sex 663 (62%) 1,096 (69%) 76 (62%) APACHE II 14 (10–19) 14 (10–18) 12 (7–17) Reason for admission Cardiothoracic surgery 13 (1.2%) 1,303 (82%) 1 (0.8%) Abdominal surgery 512 (48%) 17 (1.1%) 9 (6.9%) Neurological 67 (6.2%) 24 (1.5%) 103 (79%) Cardiac arrest 10 (0.9%) 137 (8.6%) 1 (0.8%) Trauma 150 (14%) 12 (0.8%) 5 (3.8%) Vascular surgery 106 (10%) 18 (1.1%) 0 (0%) Medical 57 (5.3%) 30 (1.9%) 8 (6.2%) Posttransplant 45 (4.2%) 10 (0.6%) 0 (0%) Miscellaneous 113 (11%) 46 (2.9%) 3 (2.3%) In-hospital mortality 147 (14%) 117 (7.3%) 19 (15%) Total number of patient-days 8,359 5,137 752 Admission glucose (mmol/l) 7.8 (6.5–9.5) 8.6 (7.3–10) 7.5 (6.5–8.9) Time to reach target range (h) 4 (0–9) 9 (2–13) 2 (0–6) Glucose after 24 h (mmol/l) 6.7 (6.0–7.4) 6.5 (5.9–7.2) 6.8 (6.1–7.7) Glucose SD (mmol/l) 1.1 (0.7–1.7) 1.3 (0.9–1.8) 1.0 (0.7–1.4) Fig. 1 Time chart of patient inclusion per ICU. The dark colored squares indicate a ‘run-in’ period. The patients treated by GRIP in these periods were not analyzed in this study

) 2 (0–6) Glucose after 24 h (mmol/l) 6.7 (6.0–7.4) 6.5 (5.9–7.2) 6.8 (6.1–7.7) Glucose SD (mmol/l) 1.1 (0.7–1.7) 1.3 (0.9–1.8) 1.0 (0.7–1.4) Fig. 1 Time chart of patient inclusion per ICU. The dark colored squares indicate a ‘run-in’ period. The patients treated by GRIP in these periods were not analyzed in this study Glucose parameters Glucose measurements were performed in arterial blood samples when an arterial line was present and in venous blood otherwise. All ICUs had a point-of-care blood gas analyzer on which glucose measurements were performed (ABL 700 or 800 series, Radiometer, Copenhagen, Denmark). For each glucose measurement the recommended measurement time was compared with the actual measurement time. We classified these times in lowest, middle two, and highest quartiles. For measurements in each group we determined whether the glucose level was in the desired range, and what the change in insulin pump rate was after that measurement. Actual pump rates were compared to the recommended pump rate at each time point. We determined the linearly interpolated glucose value of each patient at 3-h intervals after admission for the first 48 h to visualize the general decrease in glucose achieved by GRIP directly after admission to the unit. Low glucose values were scored as mild hypoglycemia (< 3.5 mmol/l) or severe hypoglycemia (< 2.2 mmol/l) and expressed as the proportion of all patients, and as the number of episodes per 1,000 patient-days.

e first 48 h to visualize the general decrease in glucose achieved by GRIP directly after admission to the unit. Low glucose values were scored as mild hypoglycemia (< 3.5 mmol/l) or severe hypoglycemia (< 2.2 mmol/l) and expressed as the proportion of all patients, and as the number of episodes per 1,000 patient-days. To assess the ability of the program to reduce hyperglycemia we determined the time from admission until the first measurement that was in our target range (4.0–7.5 mmol/l). We also determined the amount of time spent within the target range, both for the full length of stay and for the period after first reaching the target range. We also calculated the hyperglycemic index, a measure of overall hyperglycemia [11]. Recently glucose variability has been proposed as control target of glucose management [12]. We therefore determined blood glucose variability by taking the standard deviation of all measurements of a patient. Statistics All statistics were calculated with R version 2.2.4 (http://www.r-project.org). Variables are presented as medians and IQR. Comparisons between groups were made with the Mann–Whitney U-test, Fisher's exact test, or χ2 test. A p-value less than 0.05 was considered statistically significant.

To assess the ability of the program to reduce hyperglycemia we determined the time from admission until the first measurement that was in our target range (4.0–7.5 mmol/l). We also determined the amount of time spent within the target range, both for the full length of stay and for the period after first reaching the target range. We also calculated the hyperglycemic index, a measure of overall hyperglycemia [11]. Recently glucose variability has been proposed as control target of glucose management [12]. We therefore determined blood glucose variability by taking the standard deviation of all measurements of a patient. Statistics All statistics were calculated with R version 2.2.4 (http://www.r-project.org). Variables are presented as medians and IQR. Comparisons between groups were made with the Mann–Whitney U-test, Fisher's exact test, or χ2 test. A p-value less than 0.05 was considered statistically significant. Results Glucose control Median time between patient admission and the first glucose entered into GRIP generating a recommendation was 21 min (6–65). Compliance with GRIP-recommended pump rates was 97%. Compared with the recommended measurement time, the median measurement was 5 min late (IQR 20 min early to 34 min late). Table 2 shows characteristics of glucose measurements and insulin pump interventions for early, on-time, and late measurements. Incidence of measurements less than 4.0 mmol/l was slightly lower with early than on-time measurements (p = 0.04), but late measurements were more often low (p = 0.03). Hypoglycemia was uncommon, with 0.86% of patients experiencing a glucose level lower than 2.2 mmol/l during any moment of their stay and 7% for a level of 3.5 mmol/l. Per 1,000 patient-days of ICU stay there were 1.6 measurements lower than 2.2 mmol/l and 14.4 measurements lower than 3.5 mmol/l. For comparison with other studies, which often report the proportion of hypoglycemic measurements instead of patients, the percentage of measurements lower than 2.2 mmol/l was 0.04%. Table 2 Compliance with measurement time: the lowest (early), middle two (on time), and highest quartile (late). For each category the number of measurements in range is shown and how much the pump rate was adjusted after that measurement. Overall distribution of glucose levels and insulin changes was different for each group (χ2 test, p < 0.001)

urement time: the lowest (early), middle two (on time), and highest quartile (late). For each category the number of measurements in range is shown and how much the pump rate was adjusted after that measurement. Overall distribution of glucose levels and insulin changes was different for each group (χ2 test, p < 0.001) Early 18,357 (25%) On time 36,805 (50%) Late 18,026 (25%) Time > 20 min early –20 min to +34 min > 34 min late Glucose level (percentage of time group) Hypo (< 2.2 mmol/l) 0.03% 0.03% 0.07% Low (2.2–4 mmol/l) 0.8% 1.0% 1.3% In range (4–7.5 mmol/l) 74% 66% 70% High (> 7.5 mmol/l) 25% 33% 29% Insulin change (percentage of time group) Decrease > 1 IU/h 6.3% 11.9% 10.1% Decrease 0.5–1 IU/h 7.3% 9.9% 9.1% Small change (–0.5 to +0.5 IU/h) 68% 52% 58% Increase 0.5–1 IU/h 10.2% 12.5% 10.7% Increase > 1 IU/h 8.1% 13.9% 11.7%

mmol/l) 0.8% 1.0% 1.3% In range (4–7.5 mmol/l) 74% 66% 70% High (> 7.5 mmol/l) 25% 33% 29% Insulin change (percentage of time group) Decrease > 1 IU/h 6.3% 11.9% 10.1% Decrease 0.5–1 IU/h 7.3% 9.9% 9.1% Small change (–0.5 to +0.5 IU/h) 68% 52% 58% Increase 0.5–1 IU/h 10.2% 12.5% 10.7% Increase > 1 IU/h 8.1% 13.9% 11.7% Figure 2 graphically shows glucose control directly after admission to the ICU. Admission glucose decreased from 8.3 mmol/l (6.9–9.8) to 6.6 (6.0–7.4) at 24 h (p < 0.001). Time from admission to a value within the desired range was 5.6 h (0.2–11.8). After that time patients stayed within range for 89% (70–100%) of their remaining stay. Due to the inclusion of a significant number of short stay patients, for which the time to establish initial control represents a relatively large part of their total stay, the overall time in range was 67%. The hyperglycemic index was 1.24 mmol/l (0.8–1.76). The patients in the lowest quartile of time in range had higher APACHE II scores, higher insulin doses, less compliance with the recommended insulin rate, more frequent measurements, and a higher incidence of hypoglycemia. Median number of glucose measurements per day of stay was 5.9 (4.7–7.3), i.e., once every 245 min. The glucose variability was 1.22 mmol/l (0.8–1.75). Fig. 2 Median and interquartile range of glucose levels and insulin pump rates during the first 48 h of ICU stay

nt measurements, and a higher incidence of hypoglycemia. Median number of glucose measurements per day of stay was 5.9 (4.7–7.3), i.e., once every 245 min. The glucose variability was 1.22 mmol/l (0.8–1.75). Fig. 2 Median and interquartile range of glucose levels and insulin pump rates during the first 48 h of ICU stay Discussion This study evaluated computer-assisted glucose control in routine practice over a 3-year period at three ICUs with different patient populations. With a reasonable glucose target level, we found that glucose control was adequate and safe, with low rates of hypoglycemia. Importantly, the protocol was feasible with six measurements per patient per day, and compliance to the protocol was high. Safety is a major concern in contemporary health care. While the efficacy of glucose control is debated, implementation should be particularly safe [8]. Reducing the possibility of human cognitive failure is a generic way to improve safety in health care [13]. Prior to our study it was shown that standardization by paper protocol improves glucose control, and that computer-assisted protocols are feasible and lead to better compliance than protocols on paper [14, 15]. Recommendations of pump rates and measurement intervals and the notification of expiration of the interval, combined with the possibility of overruling, provide a flexible standardization of glucose control in the ICU.

puter-assisted protocols are feasible and lead to better compliance than protocols on paper [14, 15]. Recommendations of pump rates and measurement intervals and the notification of expiration of the interval, combined with the possibility of overruling, provide a flexible standardization of glucose control in the ICU. Comparison with protocols from the literature is difficult because of wide variations in target glucose levels, patient populations, and metrics of control reported [9]. We performed a literature search to find all computerized glucose protocols designed for ICU patients and tested in at least 20 patients. We summarize these findings in Table 3. Similar to what is reported for paper protocols, reported target ranges, patient populations, and measurement means vary greatly. However, we can conclude that compared with the studies published thus far, the present study is one of the largest to date and compares favorably in terms of occurrence of hypoglycemia and achievement of the target range. Interestingly, only one other published computer protocol [16] aiming at achieving control with a low measurement frequency. This is remarkable, as glucose control with a high measurement frequency is associated with significant costs in terms of nursing effort and supplies [17]. The glucose variability, which has recently been proposed as an important determinant of control [12], was 1.22 mmol/l in our large cohort. This is lower than the figures published previously, which were 1.7 ± 1.3 in survivors and 2.3 ± 1.6 in nonsurvivors [12]. Table 3 Comparison with other computer-assisted glucose control protocols and intervention groups from clinical trials (BG, blood glucose)

nant of control [12], was 1.22 mmol/l in our large cohort. This is lower than the figures published previously, which were 1.7 ± 1.3 in survivors and 2.3 ± 1.6 in nonsurvivors [12]. Table 3 Comparison with other computer-assisted glucose control protocols and intervention groups from clinical trials (BG, blood glucose) Reference n Patient type APACHE Target range Performance Hypoglycemia a Measurements per patient per day Hovorka et al. [22] 30 Cardiac surgery ? 4.4–6.1 60% of time in range 0 < 2.9 mmol/l 16 Shulman et al. [23] 50 Mixed ICU 23 4.4–6.1 23 % of time in range 0.2% < 2.2 mmol/l 12.7 Hermayer et al. [24] 66 CABG ? 4.4–6.7 Mean BG 6.4 mmol/l 0.11% < 2.2 mmol/l 16.2 b Rood et al. [15] 66 Mixed ICU 19.5 4.0–7.0 54% of time in range 0.09% of time < 2.5 mmol/l 12.4 b Toschlog et al. [25] 128 Trauma ISS 24.5 4.4–7.2 Mean BG 6.4 mmol/l 32% of patients < 2.8 mmol/l ? Meynaar et al. [16] 179 Mixed ICU 13 4.5–7.5 53% of time in range 0.05% < 2.2 mmol/l 3.4 Boord et al. [26] 204 Surgical ICU ? 4.4–6.1 49% of time in range 0.2% < 2.2 mmol/l Approx. 18 (12–24) c Thomas et al. [27] 603 Mixed ICU 14.4 5.4–7.1 85% of measurements < 8 mmol/l 19 episodes ? Juneja et al. [28] 2,398 Mixed ICU ? 4.4–6.1 61% of measurements in range 0.4% < 2.8 mmol/l Approx. 18 (12–24) c This study 2,800 Mixed surgical 14 4.0–7.5 67% of time in range 0.04% < 2.2 mmol/l 5.9 Davidson et al. [29] 5,808 General medical and surgical wards ? Variable “Stable glucose” 0.6% < 1.8 mmol/l Approx. 18 (12–24) c Van den Berghe et al. [1] 764 Surgical 9 4.4–6.1 Mean BG 5.7 mmol/l 5% of patients < 2.2 mmol/l Approx. 18 (12–24) c Krinsley [30] 800 Mixed 15 < 7.8 Median BG 6.6 mmol/l 0.34% < 2.2 mmol/l Approx. 16 (8–24) d Van den Berghe et al. [2] 595 Medical 23 4.4–6.1 Mean BG 6.2 mmol/l 19% of patients < 2.2 mmol/l Approx. 18 (12–24) a Brunkhorst et al. [7] 247 Severe sepsis 20 4.4–6.1 Mean BG 6.2 mmol/l 17% of patients < 2.2 mmol/l Approx. 12 (6–24) e a Proportion of measurements, not patients, unless otherwise specified; b calculated from number of measurements and length of stay; c one- to two-hourly measurements; d one- to three-hourly measurements; e one- to four-hourly measurements

s 20 4.4–6.1 Mean BG 6.2 mmol/l 17% of patients < 2.2 mmol/l Approx. 12 (6–24) e a Proportion of measurements, not patients, unless otherwise specified; b calculated from number of measurements and length of stay; c one- to two-hourly measurements; d one- to three-hourly measurements; e one- to four-hourly measurements The protocol GRIP used is published as open source, and the program can therefore be used freely by any hospital able to implement the bridge between GRIP and the hospital information system. We recognize that few hospitals have such resources, and we therefore provided a more detailed description of the algorithm in an online supplement, including typical recommendations for two example patients. Computer-based protocols published to date broadly fall into three categories: flow-chart based, empirical formula based, and model based. Most paper protocols that are directly computerized fall into the first category. GRIP uses a formula empirically derived from many protocols and expert knowledge. Model-based approaches model glucose dynamics in different compartments and try to predict future glucose levels and insulin need from that. There have been no direct comparisons of different computer algorithms to date, and more research is needed to identify the best approach.

rom many protocols and expert knowledge. Model-based approaches model glucose dynamics in different compartments and try to predict future glucose levels and insulin need from that. There have been no direct comparisons of different computer algorithms to date, and more research is needed to identify the best approach. Our study has a number of strengths. We studied a large number of patients treated in routine practice. This provides a less biased view on true performance and rate of hypoglycemia than studies of glucose protocols in a research setting. The low number of measurements required by GRIP makes the protocol less time consuming than protocols with hourly measurements. The low average number of measurements GRIP requires is mainly caused by the flexibility of the interval algorithm. GRIP performs frequent measurements when needed, but this is counterbalanced by identification of stable periods during a patient's stay (e.g., with steady glucose levels, full caloric intake and low insulin dose) during which much longer intervals can safely be recommended. Another strength is our analysis of actual compliance to the protocol. Prior publications of protocols rarely mention compliance metrics. Good performance attributed to a protocol that is violated often by nurses (by for instance measuring glucose earlier than recommended) can lead to disappointing results when strictly followed at another unit. In our analysis we find no structural violation of the protocol in terms of early and late measurements, and only late measurements were associated with more glucose values that were out of range.

tance measuring glucose earlier than recommended) can lead to disappointing results when strictly followed at another unit. In our analysis we find no structural violation of the protocol in terms of early and late measurements, and only late measurements were associated with more glucose values that were out of range. Our study also has a number of limitations. Our choice of target glucose level (6.5 mmol/l) may be regarded as either low or high, as the optimal target glucose level is currently an open question [18–20]. True adherence to normoglycemia would require a lower target range than we used. The number of measurements asked for by GRIP depends on the expected risk of hypoglycemia, which becomes higher when aiming for a lower target. We therefore expect that a lower target level would increase GRIP's recommended measurement frequency without a significant increase in hypoglycemic episodes, but we have not yet tested this. Clinical decision support tools may be less likely to be successful in a setting in which they were not developed [21]. As the ICUs in our center function independently, and the development process has been limited solely to the surgical ICU, we expect that limitations of generalizability would have been uncovered in the implementation process, or would have been shown by inferior performance on the two other ICUs.

e not developed [21]. As the ICUs in our center function independently, and the development process has been limited solely to the surgical ICU, we expect that limitations of generalizability would have been uncovered in the implementation process, or would have been shown by inferior performance on the two other ICUs. A computer program also brings benefits that we did not study, such as straightforward auditing of performance and easy incorporation of future changes in practice. For instance, when newly published evidence would favor a different target range, a quick adoption of the new preferred practice is possible without going through a learning period for nurses and physicians. In conclusion, our study shows that a computer-assisted protocol can safely and efficiently guide glucose control at intensive care units in routine practice. Electronic supplementary material Electronic Supplementary Material (DOC 225K) Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited. Electronic supplementary material The online version of this article (doi:10.1007/s00134-008-1091-y) contains supplementary material, which is available to authorized users.

e association between global CBF and different CPP, within the 50–90 mmHg range. No correlation was evident between CPP and global cerebral blood flow in this selected cohort, probably because most of the patients were studied in a normal CPP range, and pressure autoregulation (not tested by the authors) was preserved. Unfortunately testing autoregulation clinically is a complex task. Dynamic methods of measuring cerebral autoregulation have become an accepted alternative to static evaluation, but a reference reliable method is not available in clinical practice. Christ et al. [33] and co-workers tried to improve the cross-correlation method for non-invasive, continuous monitoring of cerebral autoregulation, in comparison to the cuff deflation test. Inter-method agreement in diagnosing an intact or impaired cerebral autoregulation was obtained in 73.5% of examinations. Cross-correlation analysis might serve as a simple, non-invasive, and continuous measure of cerebral autoregulation. Nevertheless, as suggested by the authors, short-term autoregulation tests and monitoring techniques based on slow spontaneous oscillations should not be used interchangeably. Therefore the best method for continuous assessment of autoregulation state remains undefined and requires additional investigations.

In the article by Dellinger et al., published in the January 2008 issue of Intensive Care Medicine, the addition of two tables, labeled Scheme 1 and Scheme 2, and subsequent text changes should appear as follows. On page 19, the first sentence in the Methods section should read as follows. Sepsis is defined as infection plus systemic manifestations of infection (Scheme 1) (12). On page 19, the first full sentence in the second column should read as follows. An example of typical thresholds for identification of severe sepsis is shown in Scheme 2 (12, 13). Scheme 1 and Scheme 2, which were not included in the article, appear as follows.Scheme 1 Diagnostic criteria for sepsis. WBC, white blood cell; SBP, systolic blood pressure; MAP, mean arterial blood pressure; INR, international normalized ration; a PTT, activated partial thromboplastin time

egulation. Nevertheless, as suggested by the authors, short-term autoregulation tests and monitoring techniques based on slow spontaneous oscillations should not be used interchangeably. Therefore the best method for continuous assessment of autoregulation state remains undefined and requires additional investigations. Two studies of a very early post-TBI phase explored the possibility of non-invasive monitoring of intracranial pressure (ICP) and of CPP target treatment, before its invasive monitoring. Geeraerts et al. [34], in order to evaluate a non-invasive method of ICP determination, assessed at ICU admission the relationship between optic nerve sheath diameter (ONSD) and ICP and whether greater ONSD at patient admission was associated with higher ICP in the first 48 h after TBI. A significant relationship between the greatest ONSD and admission ICP was documented and it was a suitable predictor of high ICP. In the early posttraumatic period, ocular ultrasound scans may be attractive for detecting high ICP even if additional demonstrations are required.

An example of typical thresholds for identification of severe sepsis is shown in Scheme 2 (12, 13). Scheme 1 and Scheme 2, which were not included in the article, appear as follows.Scheme 1 Diagnostic criteria for sepsis. WBC, white blood cell; SBP, systolic blood pressure; MAP, mean arterial blood pressure; INR, international normalized ration; a PTT, activated partial thromboplastin time Infection, documented or suspected, and some of the following: General variables Fever (> 38.3°C) Hypothermia (core temperature < 36°C) Heart rate > 90 min−1 or > 2 SD above the normal value for age Tachypnea Altered mental status Significant edema or positive fluid balance (> 20 mL/kg over 24 hrs) Hyperglycemia (plasma glucose > 140 mg/dL or 7.7 mmol/L) in the absence of diabetes Inflammatory variables Leukocytosis (WBC count > 12,000 μL−1) Leukopenia (WBC count < 4000 μL−1) Normal WBC count with > 10% immature forms Plasma C-reactive protein > 2 SD above the normal value Plasma procalcitonin > 2 SD above the normal value Hemodynamic variables Arterial hypotension (SBP < 90 mmHg, MAP < 70 mmHg, or an SBP decrease > 40 mmHg in adults or < 2 SD below normal for age) Organ dysfunction variables Arterial hypoxemia (PaO2/FIO2 < 300) Acute oliguria (urine output < 0.5 ml/kg hr for at least 2 hrs despite adequate fluid resuscitation) Creatinine increase > 0.5 mg/dL or 44.2 micromol/L Coagulation abnormalities (INR > 1.5 or a PTT > 60 secs) Ileus (absent bowel sounds) Thrombocytopenia (platelet count < 100,000 μL−1) Hyperbilirubinemia (plasma total bilirubin > 4 mg/dL or 70 micromol/L) Tissue perfusion variables Hyperlactatemia (> upper limit of lab normal) Decreased capillary refill or mottling Diagnostic criteria for sepsis in the pediatric population are signs and symptoms of inflammation plus infection with hyper- or hypothermia (rectal temperature > 38.5 or < 35°C), tachycardia (may be absent in hypothermic patients), and at least one of the following indications of altered organ function: altered mental status, hypoxemia, increased serum lactate level or bounding pulses. Adapted from Levy MM, Fink MP, Marshall JC, et al: 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis Definitions Conference. Crit Care Med 2003; 31:1250-1256

nts), and at least one of the following indications of altered organ function: altered mental status, hypoxemia, increased serum lactate level or bounding pulses. Adapted from Levy MM, Fink MP, Marshall JC, et al: 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis Definitions Conference. Crit Care Med 2003; 31:1250-1256 Scheme 2 Severe Sepsis = Sepsis-Induced Tissue Hypoperfusion or Organ Dysfunction (any of the following thought to be due to the infection)Sepsis induced hypotension Lactate > upper limits lab normal Urine output < 0.5 ml/kg hr for > 2 hrs despite adequate fluid resuscitation ALI with PaO2/FIO2 < 250 in the absence of pneumonia as infection source ALI with PaO2/FIO2 < 200 in the presence of pneumonia as infection source Creatinine > 2.0 mg/dl (176.8 micromol/L) Bilirubin > 2 mg/dl (34.2 micromol/L) Platelet count < 100,000 Coagulopathy (INR > 1.5) Adapted from Levy, MM, Fink MP, Marshall JC, et al: 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis Definitions Conference. Intensive Care Medicine 2003; 29:530–538. ACCP/SCCM Consensus Conference Committee: American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference: Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. Crit Care Med 1992; 20:864–874 On page 21, the sentence in the second full paragraph should read as follows. The committee assessed whether the desirable effects of adherence will outweigh the undesirable effects, and the strength of a recommendation reflects the group's degree of confidence in that assessment (Table 2).

Adapted from Levy, MM, Fink MP, Marshall JC, et al: 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis Definitions Conference. Intensive Care Medicine 2003; 29:530–538. ACCP/SCCM Consensus Conference Committee: American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference: Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. Crit Care Med 1992; 20:864–874 On page 21, the sentence in the second full paragraph should read as follows. The committee assessed whether the desirable effects of adherence will outweigh the undesirable effects, and the strength of a recommendation reflects the group's degree of confidence in that assessment (Table 2). The following errors appeared in Table 5. Under Blood product administration recommendations: Fresh frozen plasma recommendation circle should be open Antithrombin recommendation circle should be closed Platelet recommendation circle should be open Under Glucose control: Aim to keep blood glucose recommendation < 150 mg/dL (8.3 mmol/L) should have an open circle On page 17 the Indian Society of Critical Care Medicine and the World Federation of Critical Care Nurses should be added to the list of sponsoring organizations who did not participate formally in the revision process. The authors regret the errors. The online version of the original can be found at http://dx.doi.org/10.1007/s00134-007-0934-2.

Experimental studies The “Experimental studies”, published either as original papers or as brief communications, focused on four main subjects, i. e. the mechanisms of lung injury and therapies, the treatment of burns and trauma, multiple organ failures and the pathophysiology and treatment of sepsis, severe sepsis and septic shock. Lung injury: mechanical ventilation – from mechanisms to therapies Aspiration of gastric contents frequently results in acute lung injury and acute respiratory distress syndrome (ARDS). Because the acid aspiration model of acute lung injury has considerable clinical implications, Fraisse et al. [1] sought to define the relationship between oxygenation and hemodynamic profile during an acute phase of acid aspiration in rabbits. Inhalation of human gastric juice (1 ml/kg) markedly decreased PaO2/FiO2 ratio and increased airway plateau pressure within 1 h. Lung injury was characterized by pulmonary edema, hemorrhage, necrosis, polymorphonuclear leukocyte infiltration, and hyaline membranes, using a PaO2/FIO2 threshold value of 150. The levels of pH, mean arterial pressure, and cardiac output were significantly lower after acid aspiration. However, neither left nor right ventricular dysfunction occurred during the 4-h experiment, and no animal experienced circulatory failure. This study provided Intensive Care Medicine readers with a PaO2/FiO2 threshold. If the P/F ratio is well above the threshold, it suggests that the lung injury and inflammatory responses may be independent of any compromised hemodynamic changes following acid aspiration.

t, and no animal experienced circulatory failure. This study provided Intensive Care Medicine readers with a PaO2/FiO2 threshold. If the P/F ratio is well above the threshold, it suggests that the lung injury and inflammatory responses may be independent of any compromised hemodynamic changes following acid aspiration. Several studies examined the mechanisms of ventilator-induced lung injury in a variety of animal models, and provided some novel insights for our readers. Eyal et al. [2] tested the hypothesis that alveolar macrophages can initiate ventilator-induced lung injury. The investigators depleted alveolar macrophages by intratracheal administration of liposomal clodronate 48 h prior to mechanical ventilation in rats. The animals were then ventilated with an injurious strategy at a mean tidal volume of 40 ± 0.7 ml/kg and zero positive end-expiratory pressure (PEEP) for 15 min. The animals were randomized to be sacrificed or to continue being ventilated for an additional 2 h at a tidal volume less than 10 ml/kg. The study demonstrated that oxygenation, lung compliance, and alveolar stability were better preserved during the initial 15-min and the subsequent 2-h ventilation in animals with alveolar macrophages depleted. The lung wet/dry ratio was significantly reduced in the animals with alveolar macrophage depletion compared to the control groups. This study suggested that alveolar macrophages play an important role in initiating ventilator-induced lung injury.

subsequent 2-h ventilation in animals with alveolar macrophages depleted. The lung wet/dry ratio was significantly reduced in the animals with alveolar macrophage depletion compared to the control groups. This study suggested that alveolar macrophages play an important role in initiating ventilator-induced lung injury. Many critically ill, ventilated patients suffer from unexplained immunosuppression that is associated with increased risk for infections. Lachmann et al. [3] investigated the effects of ventilatory strategies on bacterial translocation in piglets. Lung injury was induced by surfactant depletion, followed by infection with intratracheal injection of group B streptococci. The animals were then ventilated for 5 h with either a conventional ventilatory strategy (tidal volume of 7 ml/kg and PEEP of 4–5 cmH2O), an open lung strategy (6 ml/kg and 14–15 cmH2O), or an open lung with high-PEEP strategy (6 ml/kg and 20–21 cmH2O). The investigators observed that the open lung ventilation resulted in significantly less bacterial translocation than either conventional or open lung with high-PEEP ventilation. This study suggested that an optimal ventilatory strategy (i. e., open lung) achieved by using individually tailored settings can be useful to minimize bacterial translocation during mechanical ventilation.

d in significantly less bacterial translocation than either conventional or open lung with high-PEEP ventilation. This study suggested that an optimal ventilatory strategy (i. e., open lung) achieved by using individually tailored settings can be useful to minimize bacterial translocation during mechanical ventilation. A recent cohort study reported that non-depolarizing neuromuscular blocking agents are used in 13% of mechanically ventilated patients with a mean duration of 2 days in the ICU. Testelmans et al. [4] studied the effects of 24 h infusion of rocuronium and cisatracurium on diaphragm function in rats during mechanical ventilation using a tidal volume of 6 ml/kg. They observed that the diaphragm tetanic force was decreased by 33% in the rocuronium-treated group compared to a vehicle control group, while the force was well preserved in the cisatracurium-treated animals. The levels of diaphragm muscle RING-finger protein-1 (MURF-1) mRNA and the protease calpain were increased after rocuronium, while unchanged in the cisatracurium-treated groups. Since MURF-1 and calpain are molecules that may contribute to muscle atrophy, these data suggest a role of certain non-depolarizing neuromuscular agents in ventilator-induced lung injury.

otein-1 (MURF-1) mRNA and the protease calpain were increased after rocuronium, while unchanged in the cisatracurium-treated groups. Since MURF-1 and calpain are molecules that may contribute to muscle atrophy, these data suggest a role of certain non-depolarizing neuromuscular agents in ventilator-induced lung injury. Endotracheal suctioning serves to clean the airway in ventilated patients but may cause adverse effects. To assess the effects on lung volume and compliance changes during suctioning, Lindgren et al. [5] monitored lung volume in four ventral-to-dorsal regions by using electric impedance tomography and a nitrogen washout/washin technique in surfactant-depleted, ventilated pigs. At disconnection, before suctioning, the fractional residual capacity (FRC) decreased by 60% of baseline. During open suctioning, FRC decreased further to 20% of baseline. A deterioration of regional compliance was more pronounced in the dorsal parts of the lung. The authors also reported that the restoration of oxygenation was prolonged under pressure-controlled ventilation compared to volume-controlled ventilation during the post-suction period. It appeared that the dorsal regions of a lavaged lung are more susceptible to disconnection and suctioning, resulting in marked decreases in compliance. Under volume-controlled ventilation, the collapsed lung and oxygenation can be better restored from open suctioning than under pressure-controlled ventilation.

period. It appeared that the dorsal regions of a lavaged lung are more susceptible to disconnection and suctioning, resulting in marked decreases in compliance. Under volume-controlled ventilation, the collapsed lung and oxygenation can be better restored from open suctioning than under pressure-controlled ventilation. Hemodynamic alterations during mechanical ventilation may contribute to poor organ perfusion that leads to multiple system organ dysfunction. Duperret et al. [6] investigated the effect of gradually increased intra-abdominal pressure on the systolic and pulse pressure variations during mechanical ventilation in pigs in the presence and absence of hypovolemia by blood withdrawal. For a given intra-abdominal pressure, the systolic pressure variations were greater in hypovolemic animals than in normovolemic animals. Also, blood flow in the inferior vena cava diminished as the intra-abdominal pressure increased. This study suggested that systemic hemodynamics can be greatly altered by intra-abdominal pressure that leads to changes in pleural pressure, and this effect is more pronounced during hypovolemia.

n in normovolemic animals. Also, blood flow in the inferior vena cava diminished as the intra-abdominal pressure increased. This study suggested that systemic hemodynamics can be greatly altered by intra-abdominal pressure that leads to changes in pleural pressure, and this effect is more pronounced during hypovolemia. Enhanced procoagulant and depressed fibrinolytic activities may be implicated in the pathogenesis of ARDS. Plasma protein C levels decreased in ARDS patients compared to normal volunteers. Thrombomodulin levels in pulmonary edema fluid of ARDS patients were higher than in normal plasma and plasma from ARDS patients. Richard et al. [7] tested the hypothesis that if occluded pulmonary vessels reopen with activated protein C (APC), the ventilation-perfusion would be improved in injured lung. Pigs were treated with intravenous infusion of APC starting 30 min prior to oleic acid injection. The animals were monitored for an additional 180 min under mechanical ventilation. PaO2/FIO2 was significantly lower in the APC-treated group before and after induction of lung injury than in a placebo group. Lung perfusion tended to redistribute towards dorsal regions with APC. Total aerated lung volume was identical in the placebo and the APC-treated animals before and after the injection of oleic acid. The plasma concentrations of IL-6 and IL-8 were higher in the APC group 2 h after oleic acid injection than in the placebo group. In contrast to sepsis models, pretreatment with APC resulted in worsening oxygenation. The author speculated that the administration of APC might have led to ventilation-perfusion mismatch, with more perfusion to dependent non-aerated areas in their model. It is noteworthy that APC exerted beneficial effects only in those critically ill patients. The present animal model of oleic acid injection might not have reproduced severe situations seen clinically.

C might have led to ventilation-perfusion mismatch, with more perfusion to dependent non-aerated areas in their model. It is noteworthy that APC exerted beneficial effects only in those critically ill patients. The present animal model of oleic acid injection might not have reproduced severe situations seen clinically. The above studies addressed mechanisms and discussed several practical issues in the context of ventilator-induced lung injury. Theoretically, lung-protective strategies may be the most important primary approach to minimize ventilator-induced lung injury. This concept is supported by the results of the ARDSnet trial showing a significant reduction in the absolute mortality rate when lower tidal volumes were used compared to higher tidal volumes. Muellenbach et al. [8] proposed that high-frequency oscillatory ventilation (HFOV) may be an optimal lung-protective strategy to minimize ventilator-induced lung injury. They investigated the effects of 24-h mechanical ventilation by using either lung-protective pressure-controlled ventilation (tidal volume of 6 ml/kg) or HFOV (6 Hz) in a pig model of ARDS induced by repeated lung lavage. This study showed in a large animal model of ARDS that the lung inflammation score and the expression of IL-1β mRNA in lung tissue were significantly decreased in the HFOV-treated animals compared to the lung-protective pressure-controlled ventilated group.

(6 Hz) in a pig model of ARDS induced by repeated lung lavage. This study showed in a large animal model of ARDS that the lung inflammation score and the expression of IL-1β mRNA in lung tissue were significantly decreased in the HFOV-treated animals compared to the lung-protective pressure-controlled ventilated group. Organ dysfunction Acute kidney injury secondary to sepsis and septic shock affects approximately 6% of critically ill patients. It has been proposed that high resistance in the renal vasculature resulted in renal hypoperfusion in a variety of animal models in which the study designs were heterogeneous. To understand the nature of renal blood flow and its function during recovery from experimental septic acute kidney injury, Langenberg et al. [9] recorded systemic and renal hemodynamics during septic acute kidney injury and subsequent recovery in sheep. Sepsis was induced by the administration of a bolus of live Escherichia coli followed by a continuous infusion for 48 h. Normal saline was administered to prevent marked hypovolemia. A bolus of gentamicin was injected and E. coli infusion was stopped. The animals were then monitored for 48 h during recovery phase with fluid administration continued. Infusion of E. coli induced a hyperdynamic circulatory state with increased cardiac output and decreased blood pressure. Acute kidney injury was evidenced by decreased creatinine clearance. Renal vasodilatation occurred that was accompanied by an increase in renal blood flow. During recovery, renal blood flow returned to control levels associated with relative renal vasoconstriction. Indices of tubular function decreased during sepsis and returned to control values after 18 h of recovery. Although the mechanisms responsible for the development of recovery from acute kidney injury are yet to be elucidated, the results of this study nevertheless challenge the conventionally held paradigm of renal ischemia as the underlying cause for the development of septic acute kidney injury.

es after 18 h of recovery. Although the mechanisms responsible for the development of recovery from acute kidney injury are yet to be elucidated, the results of this study nevertheless challenge the conventionally held paradigm of renal ischemia as the underlying cause for the development of septic acute kidney injury. Ischemia and reperfusion is associated with excessive inflammatory responses, and the mechanisms are very complex. Glycogen synthase kinase-3β (GSK-3β) is a key regulatory enzyme in glucose metabolism that, when activated, phosphorylates/inactivates target enzymes of the insulin signaling pathway. It has been recently reported that GSK-3β is involved in the regulation of many cell functions, but its role in the regulation of the inflammatory response remains unknown. Cuzzocrea et al. [10] investigated the effects of TDZD-8, a potent and selective GSK-3β inhibitor, on intestinal injury following ischemia/reperfusion induced by splanchnic artery occlusion in rats. TDZD-8 was administered 5 min prior to the 6-h reperfusion phase. The investigators observed that the TDZD-8-treated animals had higher blood pressure and a greater survival rate than a group receiving vehicle control. These beneficial effects were associated with a decrease in neutrophil infiltration in the reperfused intestine, attenuation of the production of inflammatory cytokines and the degranulation of IκB-α, reduction of oxidative stress, and prevention of cell apoptosis. This study opens a new avenue to explore the role of GSK-3β as a therapeutic target in ischemia and reperfusion.

n neutrophil infiltration in the reperfused intestine, attenuation of the production of inflammatory cytokines and the degranulation of IκB-α, reduction of oxidative stress, and prevention of cell apoptosis. This study opens a new avenue to explore the role of GSK-3β as a therapeutic target in ischemia and reperfusion. Recent findings suggest that abdominal compartment syndrome, as reflected by intra-abdominal hypertension (IAH), is associated with an increased rate of the occurrence of multiple organ dysfunction syndrome. Prevention of abdominal compartment syndrome has been shown to decrease the incidence of multiple organ dysfunction. The inflammatory cytokine responses are crucial to trigger multiple organ dysfunction in abdominal compartment syndrome, but studies suggested that decompression may be too late to reverse the inflammatory responses. Thus early detection and prevention of abdominal compartment syndrome is of pivotal interest in successful treatment. Meier et al. [11] evaluated the potential role of microdialysis to detect intra-abdominal organ injury during IAH by measuring energy metabolism in rats. IAH was induced for 3 h and followed by decompression and reperfusion for another 3 h. The prolonged IAH induced persistent abdominal organ injury. Microdialysis analysis demonstrated a significant increase in lactate/pyruvate and glycerol in kidney, intestine, and liver, indicating ischemia, energy failure, and cell membrane damage. This was accompanied by a decrease in glucose level in all organs studied. The deterioration of energy metabolism did not completely recover upon decompression of IAH. This study suggested that continuous microdialysis in the rectus abdominis muscle may provide a useful tool for early detection of IAH-induced metabolic derangements.

panied by a decrease in glucose level in all organs studied. The deterioration of energy metabolism did not completely recover upon decompression of IAH. This study suggested that continuous microdialysis in the rectus abdominis muscle may provide a useful tool for early detection of IAH-induced metabolic derangements. The loss of blood volume would be further increased with an elevated hydrostatic capillary pressure, especially when vascular permeability is increased. If the hypothesis is true, this concept may lead to a potential therapeutic strategy to minimize blood volume substitution in critically ill patients by avoiding high blood pressure. To evaluate interactions between the increased vascular permeability and increased hydrostatic pressure on blood volume loss, Dubniks et al. [12] induced vascular permeability as a result of an anaphylactic reaction by injection of dextran 70 in rats. Plasma volume was measured before and after 5% albumin infusion for volume expansion. Blood pressure was elevated by administration of noradrenalin or decreased by metoprolol/clonidine started after the albumin infusion and continued throughout the experiment. The investigators observed that the plasma volume increased after the albumin infusion in the control group, but to a significantly lesser extent—by 24-fold—in the noradrenalin-treated group in the experimental conditions with increased vascular permeability. On the contrary, the plasma volume increased in the noradrenalin-treated group in the normal permeability state, and this increase was similar in extent to that seen in the metoprolol/clonidine-treated animals under increased permeability conditions. The study provided evidence that an increase in arterial pressure increases the loss of plasma volume during a state of increased permeability, perhaps due to increased hydrostatic capillary pressure.

was similar in extent to that seen in the metoprolol/clonidine-treated animals under increased permeability conditions. The study provided evidence that an increase in arterial pressure increases the loss of plasma volume during a state of increased permeability, perhaps due to increased hydrostatic capillary pressure. Head trauma and burn injury are characterized by hypermetabolism and hypercatabolism mediated by altered regulatory mechanisms of hormones, imbalanced pro- and anti-inflammatory cytokine production, and immune depression. It has been shown that decreased production of interleukin-2 (IL-2) and expression of the surface IL-2 receptor alpha chain (CD25) is associated with immune depression in head injury. Arginine is known as a nitric oxide donor and a precursor of polyamines for IL-2 production. Hamani et al. [13] tested the hypothesis that standard enteral nutrition may not be sufficient to meet specific nutrient demand in head trauma. The investigators examined the role of arginine in the immunomodulatory effects in a rat model of head injury. They showed that the standard diet supplemented with arginine reduced thymic atrophy, maintained thymus weight, increased the expression of CD25 and the production of IL-2, and blunted the enterobacterial translocation and dissemination induced by head injury. Although the benefit of immune-enhancing diets in the management of trauma injury remains controversial, this study suggested that arginine supplement appears to be safe, and may help modulate immune response in head injury.

f IL-2, and blunted the enterobacterial translocation and dissemination induced by head injury. Although the benefit of immune-enhancing diets in the management of trauma injury remains controversial, this study suggested that arginine supplement appears to be safe, and may help modulate immune response in head injury. Ornithine α-ketoglutarate (OKG) has been shown to restore glutamine pools in hypercatabolic patients and to improve wound healing in trauma patients, but the mechanisms remain to be elucidated. Since OKG consists of two molecules of ornithine (Orn) and one molecule of α-ketoglutarate (αKG), Cynober et al. [14] tested the hypothesis that Orn and αKG interact metabolically to produce glutamine in muscles in a rat model of 20% surface burn injury. The animals were fed with enteral nutrition supplemented with either OKG, Orn, or αKG. Glycine was used as an isonitrogenous control. Higher plasma content of glutamine and muscle contents of glutamate and glutamine were observed in the animals treated with the OKG-enriched diet than in those that received diets supplemented with αKG or Orn. This study demonstrated that OKG was more efficient than Orn or αKG alone in restoring glutamine pools in plasma and muscle, suggesting that the metabolic interaction between Orn and αKG is required to produce glutamine in muscles.

th the OKG-enriched diet than in those that received diets supplemented with αKG or Orn. This study demonstrated that OKG was more efficient than Orn or αKG alone in restoring glutamine pools in plasma and muscle, suggesting that the metabolic interaction between Orn and αKG is required to produce glutamine in muscles. Sepsis and pneumonia models Animal models of sepsis and acute lung injury were used in several studies published in Intensive Care Medicine during 2007. Although not perfect, these models were long ago shown to be valuable to test pathogenic hypotheses and potential therapies in sepsis and acute lung injury.

th the OKG-enriched diet than in those that received diets supplemented with αKG or Orn. This study demonstrated that OKG was more efficient than Orn or αKG alone in restoring glutamine pools in plasma and muscle, suggesting that the metabolic interaction between Orn and αKG is required to produce glutamine in muscles. Sepsis and pneumonia models Animal models of sepsis and acute lung injury were used in several studies published in Intensive Care Medicine during 2007. Although not perfect, these models were long ago shown to be valuable to test pathogenic hypotheses and potential therapies in sepsis and acute lung injury. Left ventricular contractile dysfunction is observed in a significant proportion of patients with septic shock. This sepsis-related cardiac failure is usually transitory and reversible. Given the high peripheral oxygen demand and the vasodilatory shock, oxygen delivery may be insufficient in the early phase of septic shock in those patients. A pharmacologic increase of cardiac output may therefore be required in such patients. Dobutamine has been the inotrope of choice for several years in this condition. It is also associated with unwanted ß-agonistic effects, which may limit its use. Dubin et al. tested dobutamine and the new inotrope levosimendan head-to-head in an endotoxemic shock in sheep model [15]. Both drugs increased systemic oxygen delivery by increasing cardiac output. However, only levosimendan induced an increase of intestinal blood flow and prevented the decrease in mucosal acidosis that was observed with dobutamine. This study, along with others, serves as a proof-of-concept that levosimendan may become a useful drug to treat patients with septic-associated left ventricular dysfunction. Energy utilization by the heart is also a critical aspect of cardiac function during sepsis. The heart is one of the few organs using lactate as energetic substrate. Levy and collaborators examined whether lactate deprivation induced a cardiac and hemodynamic dysfunction in a rat endotoxic shock model [16]. The combined pharmacological blockade of the ß2 receptor and pyruvate dehydrogenase markedly decreased muscle production of lactate, heart lactate concentration, and tissue ATP content. This was associated with a worsening in cardiovascular performance and a poorer outcome. These effects could be reversed by the infusion of lactate. This study highlights the role of lactate as an important fuel for the heart and suggests that lactate deprivation might be detrimental during septic shock.

e ATP content. This was associated with a worsening in cardiovascular performance and a poorer outcome. These effects could be reversed by the infusion of lactate. This study highlights the role of lactate as an important fuel for the heart and suggests that lactate deprivation might be detrimental during septic shock. Endotoxemia induces a marked increase of both pro-inflammatory and anti-inflammatory cytokines, which are though to be detrimental in this setting. Many efforts have been devoted to decrease cytokine production or increase cytokine elimination in sepsis models. It remains unclear whether putative beneficial effects of activated protein C (aPC) are related to its anti-thrombotic and pro-fibrinolytic properties or to possible anti-inflammatory effects. In an porcine model of endotoxemia, Nielsen et al. confirmed a pro-fibrinolytic effect of aPC by decreasing plasma levels of plasminogen activator inhibitor 1 [17]. However, the effects on circulating cytokines were modest; aPC infusion did not influence peak levels of IL-6, IL-8, TNF-a, and IL-10, the peak occurring simply later for the two latter. This is in accordance with the PROWESS study, which did not show marked plasma cytokine level differences between patients receiving aPC and controls. Endotoxemia also induces lung inflammation, a phenomenon largely mediated by locally produced IL-1ß. The secretion of mature and bioactive IL-1ß and IL-18 depends on the activity of the enzyme caspase-1, a limiting step in the biosynthesis of these pro-inflammatory cytokines. In a series of elegant experiments, Boost and collaborators showed that a caspase-1 inhibitor administered by inhalation in endotoxemic rats decreased IL-1 and IL-18 levels in bronchoalveolar lavage fluid [18]. Importantly, downstream inflammatory products such as COX-2 and nitric oxide were also decreased in animals treated with the caspase inhibitor. In addition to the identification of a potential novel therapeutic target (caspase-1), this study highlighted inhalation as a promising route for the administration of drugs in critically ill patients.

nflammatory products such as COX-2 and nitric oxide were also decreased in animals treated with the caspase inhibitor. In addition to the identification of a potential novel therapeutic target (caspase-1), this study highlighted inhalation as a promising route for the administration of drugs in critically ill patients. Hemoperfusion columns are theoretically interesting to remove noxious bacterial or host substances during septic shock. Taniguchi et al. perfused blood from lipopolysaccharide (LPS)-treated rats onto a cytokine absorbent column, which resulted in a significant reduction in plasma cytokine concentrations, and decreased mortality, compared with controls [19]. Survival rates were dependent on the size of the column. Oxidative stress-induced programmed cell death has been proposed as a mechanism participating in end-organ dysfunction during sepsis. Ozdemir et al. showed that infant rats injected with endotoxin had increased levels of lipid peroxidation markers and intestinal cell apoptosis [20]. These effects were dampened by the treatment with melatonin. This apparently harmless molecule should be further tested for its potential role as an anti-oxidant and for putative cytoprotective effects in the context of sepsis. In a rat model of septic shock induced by the staphylococcal alpha-toxin, Temmesfeld-Wollbrück et al. showed that post-treatment with the vasoregulatory polypeptide adrenomedullin decreased mortality [21]. The main effect of adrenomedullin was to prevent the massive increased vascular permeability observed in this model, as demonstrated by a smaller diffusion of labeled albumin in remote organs of treated animals. Mortality was also markedly reduced in animals receiving adrenomedullin. Staphylococcal alpha-toxin, being a pore-forming toxin, is likely to induce significantly more vascular damage than other toxins such as LPS. Therefore, the effects of adrenomedullin need to be further investigated in endotoxemic or infectious models.

ality was also markedly reduced in animals receiving adrenomedullin. Staphylococcal alpha-toxin, being a pore-forming toxin, is likely to induce significantly more vascular damage than other toxins such as LPS. Therefore, the effects of adrenomedullin need to be further investigated in endotoxemic or infectious models. Cecal ligation and puncture (CLP) is a useful model, mimicking peritonitis. CLP induces lethality rates depending on the size of the hole(s) performed in the cecum of mouse or rat. The advantage of this model over endotoxemia is that live bacteria represent the infectious challenge. Nitric oxide (NO) is produced in excess during septic shock, due to the increased expression of inducible NO synthase (iNOS). In a resuscitated CLP model, Albuszies et al. showed that glucose production by the liver was higher in mice treated with a specific iNOS inhibitor and in mice deficient for the iNOS gene than in littermates [22]. Increased hepatic activity of phosphoenolpyruvate carboxykinase—a key enzyme of gluconeogenesis—paralleled the increased glucose production. These results suggest that iNOS-induced excess production of NO during septic shock may down-regulate the glucose production by the liver. High glucose production by the liver was maintained when iNOS was blocked or genetically absent during CLP. It has also been suggested that NO was an important molecule participating in the innate control of bacterial infections. Cui et al. showed that the pharmacological inhibition and the gene deletion of the neuronal NOS isoform increased plasma pro-inflammatory cytokine concentrations and mortality in a CLP mouse model [23]. These results suggest that this NOS isoform is important for bacterial clearance and cytokine production during severe bacterial infections. In a P. aeruginosa rat model, inhaled NO (iNO) increased pulmonary endothelial but not epithelial permeability [24]. This effect could not be related with modifications in levels of alveolar inflammation. Further studies are needed to unravel mechanisms of iNO-induced increase in endothelial permeability in this model. In an original study, Tuon et al. submitted rats to CLP, followed by fluid resuscitation and antibiotic treatment [25]. They reported that rats exhibited depressive behavior 10 days after sepsis, as assessed by a forced swimming test. Interestingly, these symptoms could be reversed by the treatment with the anti-depressive drug imipramine.

Tuon et al. submitted rats to CLP, followed by fluid resuscitation and antibiotic treatment [25]. They reported that rats exhibited depressive behavior 10 days after sepsis, as assessed by a forced swimming test. Interestingly, these symptoms could be reversed by the treatment with the anti-depressive drug imipramine. Dehydroepiandrosterone (DHEA) has been shown to improve the outcome in some models of systemic inflammation, although the mechanisms remain unclear. Oberbeck et al. showed that parenteral administration of DHEA (modestly) improved survival in a CLP mouse model. Increased heat shock protein-70 levels, decreased splenocyte apoptosis, and TNF release observed in treated animals may participate in the protective mechanism of DHEA [26].

gh the mechanisms remain unclear. Oberbeck et al. showed that parenteral administration of DHEA (modestly) improved survival in a CLP mouse model. Increased heat shock protein-70 levels, decreased splenocyte apoptosis, and TNF release observed in treated animals may participate in the protective mechanism of DHEA [26]. Although the nature of fluid to be used for volume resuscitation remains a question of debate, starches have gained popularity in some European countries during the past decade. However, both the plasma volume expansion capacity and the secondary effects seem to be governed by their molecular weight and degree of hydroxyethyl substitution. In a pig model of hemorrhagic shock, Eisenbach et al. demonstrated that three different hydroxyethyl starch solutions with molecular weight between 100 and 200 kDa restored macro- and microcirculation similarly [27]. Urine production was, however, higher in a starch preparation of lower molecular weight, whereas plasma clearance was less in starches with a molecular weight of 200 kDa. As a note of caution, all starch solutions had significantly accumulated in various organs by 6 h after infusion. Another controversy is the use of normotonic vs. hypertonic saline solution for volume resuscitation, particularly when effects on lung injury are measured. In a similar shock model in pigs, Roch et al. showed that acute hemorrhage induced a significant lung injury assessed histologically with increased alveolar inflammation and edema [28]. However, when a strict goal-directed resuscitation protocol was applied, lung injury parameters did not differ whether animals were resuscitated with normal saline or with small volumes of hypertonic/hyperoncotic saline.

ignificant lung injury assessed histologically with increased alveolar inflammation and edema [28]. However, when a strict goal-directed resuscitation protocol was applied, lung injury parameters did not differ whether animals were resuscitated with normal saline or with small volumes of hypertonic/hyperoncotic saline. It has been suggested by various clinical investigators that treatment with antibiotics before bronchoalveolar lavage sampling will jeopardize results that can be obtained with this diagnostic procedure. Brandão da Silva et al. modeled this situation in pneumonia models in rats [29]. Not surprisingly, it was found that pretreatment with antibiotics markedly decreased the diagnostic yield of BAL cultures in animals with S. pneumoniae or P. aeruginosa pneumonia. The direct examination of BAL smears, however, showed that a significant proportion of cases of pneumonia could still be diagnosed based on the detection of bacteria inside neutrophils (intracellular microorganisms) despite pretreatment with antibiotics. The importance of sampling patients with suspected pneumonia before initiation of a novel antibiotic therapy was highlighted in an accompanying editorial by Timsit [30].

a could still be diagnosed based on the detection of bacteria inside neutrophils (intracellular microorganisms) despite pretreatment with antibiotics. The importance of sampling patients with suspected pneumonia before initiation of a novel antibiotic therapy was highlighted in an accompanying editorial by Timsit [30]. In a well-described ovine model of lung injury (smoke inhalation plus bacterial pneumonia), Maybauer and collaborators reported that ceftazidime showed numerous beneficial effects besides its antibacterial properties [31]. Ceftazidime treatment was associated with hemodynamic stabilization, better oxygenation, and less bronchial obstruction. Interestingly, ceftazidime blunted the increase in 3-nitrotyrosine observed in lungs from control animals. Clinical studies Brain injury Several articles dealt with treatment, pathophysiology and prognosis of traumatic brain injury (TBI).

In a well-described ovine model of lung injury (smoke inhalation plus bacterial pneumonia), Maybauer and collaborators reported that ceftazidime showed numerous beneficial effects besides its antibacterial properties [31]. Ceftazidime treatment was associated with hemodynamic stabilization, better oxygenation, and less bronchial obstruction. Interestingly, ceftazidime blunted the increase in 3-nitrotyrosine observed in lungs from control animals. Clinical studies Brain injury Several articles dealt with treatment, pathophysiology and prognosis of traumatic brain injury (TBI). Following TBI, impaired autoregulation contributes to the increased sensitivity of the brain to secondary ischemic insults, especially those caused by hypotension. The concept of an individualized treatment is emerging, targeting the optimal cerebral perfusion pressure (CPP) strategy on the basis of an impaired or intact autoregulation. Chieregato et al. [32] exploited an imaging technique for evaluating the relationship between CPP and cerebral blood flow (CBF). Using xenon-CT in 162 patients with severe TBI, they evaluated the association between global CBF and different CPP, within the 50–90 mmHg range. No correlation was evident between CPP and global cerebral blood flow in this selected cohort, probably because most of the patients were studied in a normal CPP range, and pressure autoregulation (not tested by the authors) was preserved.

h higher ICP in the first 48 h after TBI. A significant relationship between the greatest ONSD and admission ICP was documented and it was a suitable predictor of high ICP. In the early posttraumatic period, ocular ultrasound scans may be attractive for detecting high ICP even if additional demonstrations are required. Ract et al. [35] evaluated the usefulness of early transcranial Doppler ultrasound (TCD) goal-directed therapy after severe TBI, before the availability of invasive cerebral monitoring. When admission TCD was abnormal, attending physicians increased CPP by infusing mannitol and/or norepinephrine, with normalization of TCD recordings in 9/11 patients. After invasive monitoring, the abnormal TCD group demonstrated higher ICP. The use of TCD at hospital admission could allow the identification of severely brain-injured patients with potential brain hypoperfusion. In such high-risk patients, early TCD goal-directed therapy appears attractive in restoring normal cerebral perfusion and might potentially help in reducing the extent of secondary brain injury.

D at hospital admission could allow the identification of severely brain-injured patients with potential brain hypoperfusion. In such high-risk patients, early TCD goal-directed therapy appears attractive in restoring normal cerebral perfusion and might potentially help in reducing the extent of secondary brain injury. After ICU admission, standard and advanced monitoring are utilized in order to reduce the incidence of secondary insults and to identify the heterogeneity of this pathology. Longhi et al. [36] measured in 32 TBI patients the brain tissue oxygen tension (PtiO2) in the hypodense area and/or near the core of the contusion and in normal-appearing brain parenchyma on computerized tomography. PtiO2 was lower in pericontusional tissue than in normal-appearing tissue. In pericontusional tissue, a progressive increase of PtiO2 from pathologic to normal values was observed over time, suggestive of a microcirculatory improvement. The use of advanced monitoring system helps the clinician to reach a better understanding of the evolving TBI pathophysiology.

han in normal-appearing tissue. In pericontusional tissue, a progressive increase of PtiO2 from pathologic to normal values was observed over time, suggestive of a microcirculatory improvement. The use of advanced monitoring system helps the clinician to reach a better understanding of the evolving TBI pathophysiology. In this complex pathology, early prognosis is sometimes a difficult task. Ballesteros et al. [37] investigated whether serum drained from the jugular vein of patients with traumatic or hemorrhagic brain injury induced apoptosis of neuronal cells in vitro and whether the apoptotic rate correlated with patients' outcome at 6 months. Regional serum drained from the jugular vein induced higher early apoptosis than systemic serum, and only early apoptotic rate was an independent factor associated with mortality at 6 months. This in vitro technique, combined with clinical and radiological measurements, might improve the value of prognostic models to predict acute brain injury patients' outcome. Korfias et al. [38] examined the relationship between serum S-100B concentrations and injury severity, clinical course, survival, and treatment efficacy after severe TBI. Serum S-100B protein reflects injury severity and improves prediction of outcome after severe TBI. S-100B might also have a role in assessing the efficacy of treatment after severe TBI, demonstrating a reduction after surgery.

ions and injury severity, clinical course, survival, and treatment efficacy after severe TBI. Serum S-100B protein reflects injury severity and improves prediction of outcome after severe TBI. S-100B might also have a role in assessing the efficacy of treatment after severe TBI, demonstrating a reduction after surgery. Subarachnoid, intracerebral hemorrhage and brain death Updated information on aneurysmal subarachnoid hemorrhage (SAH) patients is not widely available. Citerio et al. [39] collected information on clinical practice and current management strategies on 350 sequential cases in 22 Italian neurosurgical hospitals. Despite the increasing trend towards interventional neuroradiologic means of securing aneurysms, aneurysms were mainly clipped and an endovascular approach was utilized in one third of cases, with wide variation among centers. SAH was confirmed to be a multiorgan disease, with frequent extracranial and intracranial complications. Nevertheless, only high ICP and deterioration in neurological status were independent factors related to unfavorable outcome. Confirming the multiorgan nature of this pathology, Terao et al. [40] determined the prevalence and the prognostic significance of microalbuminuria in SAH. The prevalence rates of microalbuminuria were higher in SAH than in the control, and the highest urinary microalbumin/creatinine ratio and the lowest GCS score during the first 8 days were the significant predictors of unfavorable neurological outcome.

valence and the prognostic significance of microalbuminuria in SAH. The prevalence rates of microalbuminuria were higher in SAH than in the control, and the highest urinary microalbumin/creatinine ratio and the lowest GCS score during the first 8 days were the significant predictors of unfavorable neurological outcome. The number of patients under oral anticoagulant (OAC) therapy, and consequently the number of subject with related intracerebral hemorrhagic (ICH) complications, is increasing. Appelboam et al. [41] examined current British practice regarding the emergency medical management of ICH patients whilst receiving OAC and to compare this with established guidelines. Prothrombin complex concentrate (PCC), the gold-standard therapy to normalize hemostasis during OAC, remains underused. There is considerable variation in practice amongst clinicians and, in most cases, practice is not in keeping with guidelines. Vigué studied whether ultra-rapid reversal of OAC could reduce the time to biological and surgical hemostasis, and might improve outcome. All patients, including over-anticoagulated individuals, had complete reversal of anticoagulation immediately after the bolus of PCC. No hemorrhagic or thrombotic adverse effect was observed. A bolus infusion of PCC completely reverses anticoagulation within 3 min. Neurosurgery can be performed immediately in OAC-related intracranial hemorrhage. This study shows that OAC-treated patients can be managed as rapidly as non-anticoagulated patients.

PCC. No hemorrhagic or thrombotic adverse effect was observed. A bolus infusion of PCC completely reverses anticoagulation within 3 min. Neurosurgery can be performed immediately in OAC-related intracranial hemorrhage. This study shows that OAC-treated patients can be managed as rapidly as non-anticoagulated patients. Spinal cord injury after aortic surgery still represents a matter of concern. Maier et al. [42] investigated the effect of the PARP-1 inhibitor INO1001 on aortic occlusion-related porcine spinal cord injury. The selective PARP-1 inhibitor INO1001 markedly reduced aortic occlusion-induced spinal cord injury. INO1001 might improve spinal cord recovery after thoracic aortic cross-clamping and, after these encouraging results, further data are required. Two papers tested two diagnostic procedures during brain death, i. e. CT angiography (CT-a) and bispectral index (BIS). Quesnel et al. [43] evaluated the accuracy of cerebral CT-a for the diagnosis of brain death. In clinically brain-dead patients, CT-a documented opacification of the intracerebral vessels in a significant percentage of the cases. Therefore CT-a cannot be recommended as a means of brain death diagnosis. Wennervirta et al. [44] studied the usefulness of entropy and BIS in brain-dead subjects. Both entropy and BIS showed non-zero values due to artifacts after brain death diagnosis. BIS was more liable to artifacts than entropy. Neither of these indices are diagnostic tools, and care should be taken when interpreting EEG-derived indices in the evaluation of brain death.

entropy and BIS in brain-dead subjects. Both entropy and BIS showed non-zero values due to artifacts after brain death diagnosis. BIS was more liable to artifacts than entropy. Neither of these indices are diagnostic tools, and care should be taken when interpreting EEG-derived indices in the evaluation of brain death. Sepsis-induced brain dysfunction and delirium Sharshar et al. [45] tried to advance in our understanding of sepsis-induced brain dysfunction using in vivo magnetic resonance imaging. This preliminary study showed that sepsis-induced brain lesions (such as multiple ischemic strokes, white and gray matter lesions) can be documented by magnetic resonance imaging. These lesions predominated in the white matter, worsened with increasing duration of shock, suggesting increased blood–brain barrier permeability, and were associated with poor outcome. ICU delirium is a common, undiagnosed and adverse event in critically ill patients. Ouimet et al. [46] investigated whether subsyndromal delirium affects outcome. Patients with no delirium were more likely to be discharged home and less likely to need convalescence or long-term care than those with subsyndromal delirium or clinical delirium. A graded diagnostic scale (Intensive Care Delirium Screening Checklist) permitted detection of subsyndromal delirium which occurs in many ICU patients, and which is associated with adverse outcome.

d home and less likely to need convalescence or long-term care than those with subsyndromal delirium or clinical delirium. A graded diagnostic scale (Intensive Care Delirium Screening Checklist) permitted detection of subsyndromal delirium which occurs in many ICU patients, and which is associated with adverse outcome. Pandharipande et al. [47] identified the prevalence of the motoric subtypes of delirium in surgical and trauma ICU patients. Prevalence of delirium was 70% for the entire study population. Hypoactive delirium was significantly more prevalent than either mixed or hyperactive delirium. In the absence of active monitoring, however, this subtype of delirium goes undiagnosed and the prevalence of delirium in surgical and trauma ICU patients remains greatly underestimated. Acute renal failure and replacement techniques Among the controversies in acute renal dysfunction in the critically ill patient are the mechanisms of renal shutdown and reversibility, the early diagnosis and staging of severity, and issues concerning renal replacement therapy.

Pandharipande et al. [47] identified the prevalence of the motoric subtypes of delirium in surgical and trauma ICU patients. Prevalence of delirium was 70% for the entire study population. Hypoactive delirium was significantly more prevalent than either mixed or hyperactive delirium. In the absence of active monitoring, however, this subtype of delirium goes undiagnosed and the prevalence of delirium in surgical and trauma ICU patients remains greatly underestimated. Acute renal failure and replacement techniques Among the controversies in acute renal dysfunction in the critically ill patient are the mechanisms of renal shutdown and reversibility, the early diagnosis and staging of severity, and issues concerning renal replacement therapy. Renal arterial resistance in septic shock and the effect of increasing mean arterial pressure with norepinephrine thereon was studied with the help of Doppler ultrasonography by Deruddre et al. [48] in 11 patients with septic shock who required fluid resuscitation and norepinephrine to maintain mean arterial pressure at or above 65 mmHg. Norepinephrine was dosed for three successive periods of 2 h to achieve a mean arterial pressure of 65, 75, and 85 mmHg, respectively. At the end of each period, hemodynamic and renal function variables were measured, and Doppler ultrasonography was performed on interlobar arteries to assess the renal resistive index. With increasing mean arterial pressure, urinary output increased from 76 ± 64 to 93 ± 68 ml/h and the resistive index decreased from 0.75 ± 0.07 to 0.71 ± 0.06. No difference was found between 75 mmHg and 85 mmHg. Doppler ultrasonography may thus help to determine in each patient the optimal mean arterial pressure for renal blood flow and may be a relevant end-point for treatment of septic shock. This study suggests that decreased renal blood flow during septic shock may be a contributing factor to renal dysfunction, which may contrast with current animal data that argue against hypoperfusion as a pivotal event in acute renal failure (ARF) during endotoxemia.

a relevant end-point for treatment of septic shock. This study suggests that decreased renal blood flow during septic shock may be a contributing factor to renal dysfunction, which may contrast with current animal data that argue against hypoperfusion as a pivotal event in acute renal failure (ARF) during endotoxemia. Herrera-Gutiérrez et al. [49] wondered whether early diagnosis of acute renal failure could be based on 2-h as opposed to 24-h creatinine clearance (CrCl) in the ICU. They studied 359 patients and compared the measures, also with the estimated clearance from the Cockcroft–Gault (Ck-G) equation. The mean Ck-G value was 87.4 ± 3.0, with 2-h CrCl of 109.2 ± 4.5 ml/min/1.73 m2 and 24-h CrCl of 100.9 ± 4.2 ml/min/1.73 m2 (r2 = 0.88 for CrCl-2h and r2 = 0.84 for Ck-G). The differences from 24-h CrCl were 21.8 ± 3.3 for the Ck-G and 8.3 ± 2.6 for 2-h CrCl. Patients with CrCl < 100 ml/min only showed variability in hyperglycemia during the 24-h period. Hence, 2-h CrCl is an adequate substitute for 24-h CrCl, even in patients who are unstable or who have irregular diuresis where a 24-h collection is unpractical. The Ck-G equation seems less useful. These ideas can be utilized in attempts to improve prognosis of patients by earlier diagnosis of acute renal failure and subsequent measures to prevent further deterioration.

Cl, even in patients who are unstable or who have irregular diuresis where a 24-h collection is unpractical. The Ck-G equation seems less useful. These ideas can be utilized in attempts to improve prognosis of patients by earlier diagnosis of acute renal failure and subsequent measures to prevent further deterioration. Jenq et al. [50] studied whether the new RIFLE (risk, injury, failure, loss and end stage) classification can predict short-term prognosis in critically ill cirrhotic patients. This study analyzed the outcomes of critically ill cirrhotic patients and identified the association between prognosis and the RIFLE classification, in comparison with other five scoring systems, in Taiwan. Thirty-two demographic, clinical and laboratory variables were analyzed as predictors of survival in 134 patients. Overall hospital mortality was 66%. There was a progressive increase in mortality based on RIFLE classification severity. Multiple logistic regression analysis indicated that RIFLE classification and Sequential Organ Failure Assessment (SOFA) score on the first day of ICU admission were independent risk factors for hospital mortality. By using the areas under the receiver operating characteristic curve (AUROC), the RIFLE category and SOFA score both indicated good discrimination (AUROC 0.84 ± 0.04 and 0.92 ± 0.02 respectively). Cumulative survival rates at 6-month follow-up differed for non-ARF vs. R, I, and F. This study again shows the prognostic value of staging ARF by the RIFLE criteria.

ting characteristic curve (AUROC), the RIFLE category and SOFA score both indicated good discrimination (AUROC 0.84 ± 0.04 and 0.92 ± 0.02 respectively). Cumulative survival rates at 6-month follow-up differed for non-ARF vs. R, I, and F. This study again shows the prognostic value of staging ARF by the RIFLE criteria. Maccariello et al. [51] also studied the RIFLE classification in patients with ARF in need of renal replacement therapy. They included 214 patients, and continuous renal replacement therapy was used in 179 (84%); patients were classified as risk (25%), injury (27%), or failure (48%). Overall mortality was 76% but there were no differences according to RIFLE classification (Risk 72%, Injury 79%, Failure 76%). Various variables were selected in multivariate analysis, including start of renal replacement therapy after the first day of ICU, but the RIFLE stages did not contribute. However, a subgroup analysis of patients on mechanical ventilation and vasopressors found the F of RIFLE to be associated with increased mortality. Hence, this study does not confirm the independent prognostic significance of the RIFLE criteria and argues in favor of further refinement.

RIFLE stages did not contribute. However, a subgroup analysis of patients on mechanical ventilation and vasopressors found the F of RIFLE to be associated with increased mortality. Hence, this study does not confirm the independent prognostic significance of the RIFLE criteria and argues in favor of further refinement. ARF in critically ill patients still carries high mortality, and a high proportion of these patients require renal replacement therapy (RRT) during their illness. To evaluate the efficiency of RRT, usually urea kinetic modeling is used, based on a sample of dialysate. The accuracy of this estimation in critically ill patients remains, however, questionable, since it assumes stable urea production and constant distribution of urea, which is not the case in critically ill patients. Ionic dialysance, which is a parameter calculated automatically from the dialysate conductivity, has been correlated to the effective urea clearance during hemodialysis in chronic renal failure patients. Ridel et al. [52] evaluated this method in critically ill ARF patients, using 31 sessions of intermittent hemodialysis (IHD) in 31 patients. Comparing the effectiveness of the RRT, they found a strong correlation between the Kt measured with dialysate sampling and with the ionic dialysance method (r = 0.96, p > 0.01). Despite some limitations of their study (small delivered RRT dose, lack of measurement of the removal of intermediate-molecular weight molecules), they conclude that this methodology might be well suited to monitor and adapt RRT in critically ill ARF patients.

and with the ionic dialysance method (r = 0.96, p > 0.01). Despite some limitations of their study (small delivered RRT dose, lack of measurement of the removal of intermediate-molecular weight molecules), they conclude that this methodology might be well suited to monitor and adapt RRT in critically ill ARF patients. Continuous RRT (CRRT) is often the method of choice in critically ill patients, when intermittent methods seem not to be appropriate. CRRT, however, is not a simple single method—it rather consists of a continuum of different treatment modalities which can be distinguished through a variety of factors. Uchino et al. [53] used data available from the B.E.S.T. study to investigate several aspects of CRRT. They included 1006 subjects treated with CRRT for ARF. Interestingly, they reported that approximately a third of the CRRT treatments (33.1%) were carried out without anticoagulation. Among those who received anticoagulation, unfractionated heparin (UFH) was the most common choice, followed by sodium citrate. Only 11.7% received a treatment dose above 35 ml/kg/h, the average being 20.4 ml/kg/h. Hypotension was the most frequently recorded complication during CRRT (observed in 18.8% of the patients), followed by arrhythmias (4.3%) and bleeding episodes (3.3%). With respect to outcome, approximately one third of the patients died on CRRT. Most of the survivors recovered renal function until their discharge from the hospital. The authors concluded that current RRT practice does not follow established guidelines for optimal therapy and thus might be responsible for significant morbidity.

ct to outcome, approximately one third of the patients died on CRRT. Most of the survivors recovered renal function until their discharge from the hospital. The authors concluded that current RRT practice does not follow established guidelines for optimal therapy and thus might be responsible for significant morbidity. CRRT is associated with less chronic renal failure than intermittent hemodialysis after ARF, according to a large, nationwide retrospective study in Sweden (Bell et al. [54]). The objective of the authors was to determine the impact of type of RRT on renal recovery in patients with acute renal failure. A total of 2,642 patients from 32 ICUs were included. Patients with end-stage renal disease and patients lacking a diagnosis in the inpatient register were excluded. Thus, 2,202 patients were studied and follow-up was complete. Renal recovery and mortality were studied. There were no differences between patients on IHD and those on CRRT techniques regarding baseline characteristics. Of the 1,102 patients surviving 90 days after inclusion, 86% were treated by continuous techniques and 14% by IHD. Seventy-eight patients (8%) never recovered renal function in the CRRT group. The proportion was higher among patients treated by intermittent techniques, where 26 subjects or 16% developed need for chronic dialysis. Mortality did not differ. Nevertheless, these data on a large number of patients support the use of continuous techniques.

nts (8%) never recovered renal function in the CRRT group. The proportion was higher among patients treated by intermittent techniques, where 26 subjects or 16% developed need for chronic dialysis. Mortality did not differ. Nevertheless, these data on a large number of patients support the use of continuous techniques. Baldwin et al. [55] performed a randomized controlled comparison of continuous venovenous hemofiltration and extended daily dialysis with filtration, with respect to effect on small solutes and acid–base balance. Recently, extended intermittent dialytic techniques have been proposed for the treatment of acute renal failure. Sixteen critically ill patients were subjected in a randomized controlled trial to 3 days of treatment with either continuous venous hemofiltration (n = 8) or extended daily dialysis with filtration (n = 8). There was no difference between the two therapies for urea or creatinine levels. One patient in the continuous group developed hypophosphatemia (0.54 mmol/l) at 72 h. After 3 days of treatment, there was a mild but persistent metabolic acidosis in the extended daily dialysis with filtration group. Hence, in spite of similar control of urea, creatinine and electrolytes, acidosis was better controlled by continuous venovenous hemofiltration. This study adds to the ongoing debate on the superiority of continuous over intermittent techniques.

t metabolic acidosis in the extended daily dialysis with filtration group. Hence, in spite of similar control of urea, creatinine and electrolytes, acidosis was better controlled by continuous venovenous hemofiltration. This study adds to the ongoing debate on the superiority of continuous over intermittent techniques. Another controversial issue relates to the mode of anticoagulation of the filter. Joannidis et al. [56] compared enoxaparin with unfractionated heparin, in a randomized controlled crossover study. Continuous venovenous hemofiltration was performed (predilution; 2500 ml/h ultrafiltration and 180 ml/min blood flow rate). Heparin-treated patients received an initial bolus of 30 U/kg and a maintenance infusion at 7 units/kg/h, dosed to achieve an aPTT of 40–45 s. Enoxaparin-treated patients received an initial pre-filter bolus of 0.15 mg/kg and a maintenance infusion starting at 0.05 mg/kg/h, adjusted to maintain anti-Xa activity at 0.25–0.30 U/ml. Each patient received both regimens in a crossover design. Maximum treatment duration for each set was 72 h, and 37 patients completed both study arms. Mean filter life span was 22 ± 17 h for heparin and 31 ± 25 h for enoxaparin. One major bleeding episode occurred during heparin and one during enoxaparin treatment. Daily costs averaged €270 and €240 for heparin and enoxaparin, respectively. Enoxaparin can thus be safely and cost-effectively used for anticoagulation during continuous venovenous Hemofiltration, resulting in longer filter life than with unfractionated heparin. A next step in this debate could be a comparison with citrate-based filtration.

70 and €240 for heparin and enoxaparin, respectively. Enoxaparin can thus be safely and cost-effectively used for anticoagulation during continuous venovenous Hemofiltration, resulting in longer filter life than with unfractionated heparin. A next step in this debate could be a comparison with citrate-based filtration. Continuous hemofiltration techniques carry the disadvantage of losses in the ultrafiltrate. Berg et al. [57] evaluated glutamine kinetics during supplementation and CRRT in 12 patients randomized to receive alanyl-l-glutamine i. v. for 20 h before placebo, or placebo before glutamine, on two consecutive days. Plasma and ultrafiltrate glutamine concentrations were measured, and blood flow across the leg was measured to calculate efflux of glutamine. Glutamine supplementation increased plasma concentrations. Losses into the ultrafiltrate were similar during treatment and control days. Net glutamine balance across the leg was also similar on treatment and control days. The loss of glutamine into the ultrafiltrate suggests an augmented need for exogenous glutamine. Although supplementation did not decrease endogenous production, it did not aggravate losses via the ultrafiltrate. Hence, in dosing of glutamine supplementation in patients on continuous venovenous hemofiltration these factors should be taken into account.

rafiltrate suggests an augmented need for exogenous glutamine. Although supplementation did not decrease endogenous production, it did not aggravate losses via the ultrafiltrate. Hence, in dosing of glutamine supplementation in patients on continuous venovenous hemofiltration these factors should be taken into account. Endocrinology Current paradigms in intensive care medicine include the idea, albeit not beyond doubt, that tight glucose control and selected treatment by substitution doses of hydrocortisone for relative adrenal insufficiency (RAI) improve outcome. However, the mechanisms of such benefits and optimal ways to achieve them in practice are still open to discussion. Weber-Carstens et al. [58] studied the effect of low-dose hydrocortisone on glycemic control in septic shock patients (n = 16). At baseline, a continuous 200 mg/daily hydrocortisone infusion was replaced by a single bolus of 50 mg of the drug. Blood glucose was monitored hourly, and insulin infusion was kept constant for 6 h. Thereafter, hydrocortisone and adjustment of blood glucose were resumed according to standard treatment. Mean blood glucose level at baseline was 7.1 mmol/l. After bolus injection of hydrocortisone, it increased within 6 h to peak levels of mean 8.6 mmol/l. Blood glucose returned to baseline with restoration of continuous hydrocortisone infusion. The data thus indicate that for glycemic control strategies, continuous infusions of hydrocortisone seem to be preferable to bolus injections.

njection of hydrocortisone, it increased within 6 h to peak levels of mean 8.6 mmol/l. Blood glucose returned to baseline with restoration of continuous hydrocortisone infusion. The data thus indicate that for glycemic control strategies, continuous infusions of hydrocortisone seem to be preferable to bolus injections. Another potential confounder in glucose control is the route and continuity of nutrition. Nguyen et al. [59] assessed the relationship between blood glucose concentrations and intolerance to gastric feeding in critically ill patients in a case–control study. Two-hourly blood glucose levels and insulin requirements over the first 10 days after admission were assessed in 95 consecutive feed-intolerant (nasogastric aspirate > 250 ml during feed) critically ill patients and 50 age-matched, feed-tolerant patients who received feeds for at least 3 days. A standard insulin protocol was used to maintain glycemia at 5.0–7.9 mmol/l. The peak glucose levels were higher before and during enteral feeding in feed-intolerant patients. The mean and trough levels were, however, similar between the two groups on admission, at 24 h prior to feeding, and for the first 4 days of feeding. The variations over 24 h before and during enteral feeding were greater in feed-intolerant patients. A blood glucose level greater than 10 mmol/l was more prevalent in patients with feed intolerance. The time taken to develop feed intolerance was inversely related to glycemia at admission. Feed intolerance in critically ill patients is thus associated with a greater degree of glycemic variation and hyperglycemia. The data suggest that more intensive insulin therapy may be required to minimize feed intolerance. Obviously, deciding on cause–effect relationships in this matter is hard, since fluctuation of enteral feeding and absorption could easily affect glucose control as well.

er degree of glycemic variation and hyperglycemia. The data suggest that more intensive insulin therapy may be required to minimize feed intolerance. Obviously, deciding on cause–effect relationships in this matter is hard, since fluctuation of enteral feeding and absorption could easily affect glucose control as well. Apart from confounding factors in glucose control, the manner by and degree to which tight glucose control is achieved is important. Meynaar et al. [60] evaluated the performance of a nurse-driven computerized insulin protocol in combination with bedside glucose measurements to maintain glycemia at 4.5–7.5 mmol/l, in a mixed adult ICU (n = 182 patients). Mean glucose decreased from 9.23 mmol/l prior to the protocol to 7.68 mmol/l with the final protocol aiming at glycemia of 4.5–7.5 mmol/l. Fifty-three percent of glucose measurements were within the target range, one episode of hypoglycemia (glucose ≤ 2.2 mmol/l) occurred, representing 0.5% of patients or 0.05% of glucose measurements. The combined strategy of successively more ambitious nurse-driven (computerized) insulin protocols and bedside glucose measurements resulted in acceptably low glucose levels with very few episodes of hypoglycemia. This is a careful study on how to achieve tighter control of glycemia with the help of the ICU nurses and point-of-care glucose measurements, preferably in arterial blood.

(computerized) insulin protocols and bedside glucose measurements resulted in acceptably low glucose levels with very few episodes of hypoglycemia. This is a careful study on how to achieve tighter control of glycemia with the help of the ICU nurses and point-of-care glucose measurements, preferably in arterial blood. Lacherade et al. [61] studied failure to achieve glycemic control despite intensive insulin therapy in a medical ICU. The authors assessed the efficacy of an insulin treatment strategy in maintaining normoglycemia, and compared ICU mortality in patients (n = 105) who did and did not reach normoglycemia (≤ 7 mmol/l). Failure to control (mean glucose > 7 mmol/l after initial hyperglycemia correction) occurred in 32 patients and was associated with an increase in ICU mortality (56% vs. 23% in patients with successful control). In the multivariate analysis, failure to control independently predicted death in the ICU. This study emphasizes the difficulty of achieving tight control in complex medical cases but does not indicate whether failure is a marker or a mediator of a downhill course. The accuracy of bedside capillary blood glucose measurements in critically ill patients was the subject of study by Critchell et al. [62]. They compared the accuracy of fingerstick with laboratory venous plasma glucose measurements (laboratory glucose) in medical ICU patients (n = 80) and evaluated the factors involved (n = 277). Accuracy was defined as the percentage of paired values not in accord (> 0.83 mmol/l difference for laboratory values < 4.1 mmol/l and > 20% difference for laboratory values > / = 4.1 mmol/l). Outliers (blood glucose difference > 5.6 mmol/l) were excluded from analyses. Mean fingerstick glucose was 7.2 ± 2.5 mmol/l and mean laboratory glucose 6.8 ± 2.4 mmol/l. The correlation coefficient was 0.91 (Clinical and Laboratory Standards Institute threshold 0.9751). The mean difference (bias) between the two methods was 0.48 ± 1.0 mmol/l, and limits of agreement were +2.5 mmol/l and –1.6 mmol/l. Fifty-three paired measurements (19%) in 22 patients were not in accord. In 83% of these measurements, fingerstick was higher than laboratory glucose. The findings suggest that capillary blood glucose measured by fingerstick is inaccurate in critically ill patients, does not meet the CLSI standard and should be used with great caution in protocols of tight glycemic control. This type of study teaches us how to achieve tight glucose control at the bedside.

ry glucose. The findings suggest that capillary blood glucose measured by fingerstick is inaccurate in critically ill patients, does not meet the CLSI standard and should be used with great caution in protocols of tight glycemic control. This type of study teaches us how to achieve tight glucose control at the bedside. In addition to the increasing knowledge on the importance of glucose control, insight into the mechanisms, diagnosis and therapeutic consequences of adrenocortical dysfunction in the critically ill is expanding. The pituitary–adrenal response to human corticotropin-releasing hormone (hCRH) in critically ill, mechanically ventilated patients (n = 37) was studied by Dimopoulou et al. [63]. A morning blood sample was obtained to measure baseline cortisol, corticotropin (ACTH) and cytokines. Patients were then injected with 100 μg hCRH, and plasma cortisol and ACTH were measured over 2 h. Baseline and peak cortisol concentrations following hCRH were 16 ± 5 μg/dl and 21 ± 5 μg/dl, and median baseline and peak ACTH levels 23 pg/ml and 65 pg/ml, respectively. Higher ACTH levels and longer release of cortisol were noted in non-survivors (n = 18) than in survivors (n = 19). Furthermore, non-survivors had higher concentrations of interleukin 8 and 6 than survivors. Hence, critically ill patients with the highest degree of inflammatory profile who ultimately die particularly demonstrate altered pituitary–adrenal axis responses to hCRH. The data in fact argue in favor of activation of the hypothalamus–pituitary–adrenal axis as a function of disease severity and subsequent stress but do not indicate whether this response is sufficient or not.

lammatory profile who ultimately die particularly demonstrate altered pituitary–adrenal axis responses to hCRH. The data in fact argue in favor of activation of the hypothalamus–pituitary–adrenal axis as a function of disease severity and subsequent stress but do not indicate whether this response is sufficient or not. RAI in patients with severe acute pancreatitis was studied by De Waele et al. [64]. This study aimed to analyze the incidence of RAI, to identify risk factors, and to describe how RAI affects outcome. In this prospective study, 25 patients with severe acute pancreatitis were subjected to a short 250 μg ACTH test within 5 days after admission. Median baseline cortisol level was 26.6 μg/dl and increased upon ACTH to 43.2 μg/dl and 48.8 μg/dl after 30 and 60 min, respectively. RAI (increment in cortisol < 9 μg/dl) was found in 16% of all patients and in 27% of patients with organ dysfunction. Patients with RAI were more severely ill and had higher SOFA scores. All patients with RAI developed pancreatic necrosis, and all of them needed surgical intervention. Twenty-eight-day mortality was higher in patients with RAI (75% vs. 5%). That increasing disease severity is a risk factor for RAI has been described before, but not the prognostic significance of RAI in this disease. Conversely, it is likely that RAI was particularly a manifestation of infected pancreatic necrosis with severe disease and dismal outcome.

r in patients with RAI (75% vs. 5%). That increasing disease severity is a risk factor for RAI has been described before, but not the prognostic significance of RAI in this disease. Conversely, it is likely that RAI was particularly a manifestation of infected pancreatic necrosis with severe disease and dismal outcome. The prognostic value of the adrenocortical response to ACTH was indeed further studied by Riché et al. [65] in 118 consecutive septic shock patients undergoing laparotomy or drainage for intra-abdominal infection. Baseline cortisol and delta cortisol to ACTH were measured during the first 24 h following onset of shock. Baseline levels were higher in non-survivors (39%) than in survivors, but the response to ACTH did not differ between outcome groups. Receiver operating characteristic curves showed threshold values for baseline of 32 μg/dl and a delta of 8  μg/dl that best discriminated survivors from non-survivors. There was no predictive value for hospital survival, however. Adrenal function tests and survival did not differ between patients who received etomidate anesthesia (n = 69) and those who did not. This study thus sheds doubts on the prognostic significance of ACTH test results in abdominal sepsis.

rs from non-survivors. There was no predictive value for hospital survival, however. Adrenal function tests and survival did not differ between patients who received etomidate anesthesia (n = 69) and those who did not. This study thus sheds doubts on the prognostic significance of ACTH test results in abdominal sepsis. Dimopoulou et al. [66] report on a prospective study on adrenocortical responses and outcome prediction in 203 mixed ICU patients. Within 24 h of admission to the ICU a morning blood sample was obtained to measure baseline cortisol, ACTH, and dehydroepiandrosterone sulfate (DHEAS). Subsequently a low-dose (1 μg) ACTH test was performed. Overall, 149 patients survived and 54 died. Non-survivors were older and had higher SOFA and APACHE II scores. Non-survivors had a lesser rise in cortisol (5.0 vs. 8.3 μg/dl) and lower DHEAS (1065 vs. 1642 ng/ml) than survivors. The two groups had similar baseline and stimulated cortisol values. Multivariate logistic regression analysis revealed that the incremental rise in cortisol was an independent predictor for poor outcome. In contrast, baseline cortisol or adrenal androgens did not afford prognostic significance. This study underscores the prognostic significance of ACTH-stimulated cortisol but does not reveal whether the type of patient is of importance (sepsis vs. non-sepsis).

in cortisol was an independent predictor for poor outcome. In contrast, baseline cortisol or adrenal androgens did not afford prognostic significance. This study underscores the prognostic significance of ACTH-stimulated cortisol but does not reveal whether the type of patient is of importance (sepsis vs. non-sepsis). Finally, Venkatesh et al. [67] found evidence for altered cortisol metabolism in critically ill patients. Changes in cortisol metabolism may be due to altered activity of the enzyme 11beta-hydroxysteroid dehydrogenase (11beta-HSD). Patients with sepsis (n = 13), multiple trauma (n = 20), and burns (n = 19) were studied concerning serial plasma cortisol:cortisone ratios. Compared with controls, the plasma cortisol:cortisone ratio was elevated in sepsis and trauma on day 1 (22 ± 9, p = 0.01, and 23 ± 19, p = 0.0003, respectively) and remained elevated over the study period. Such a relationship was not demonstrable in burns. The ratio correlated to APACHE II (r = 0.77, p = 0.0008) and SAPS (r = 0.7, p = 0.003) only on day 7 and only in the burns cohort. Hence, there is evidence of altered cortisol metabolism in critical illness due to an increase in 11beta-HSD activity, but the precise relation with ACTH test results and clinical features of RAI remains unclear. Taken together, the optimal definition, clinical significance, risk factors, and mechanisms of adrenocortical suppression in critical illness are still a matter of debate and subject to further study. This review summarizes all articles published in Intensive Care Medicine in 2007, grouped by topic.

Sir: A 38-year-old woman (gravida 2, para 3) was referred to our hospital because of intractable postpartum hemorrhagic shock. During this gemelli pregnancy she was admitted to the hospital three times for vaginal blood loss. Ultrasonography revealed a hematoma and a placenta close to the cervical ostium. Coagulation and platelets were normal. Because of fetal distress during spontaneous labor at 34 weeks an emergency cesarean section was performed. Vaginal hemorrhage began after abdominal closure. Management consisted of uterotonic drugs, intravenous fluids, red blood cells, and fresh frozen plasma without effect. Supravaginal hysterectomy was deemed necessary. Because of pelvic oozing the abdomen was closed after packing of the pelvis. She was still in hemorrhagic shock. Recombinant human activated factor VIIa (rFVIIa; NovoSeven) was administered twice as a 7.2-mg intravenous bolus at an interval of 120 min, but bleeding continued from almost all orifices. The next day she was referred. Hemodynamic instability persisted, and multiorgan failure developed. Angiography of the pelvic arteries resulted in coiling of three small arteries, but hemostasis was not achieved. On the third day tranexamic acid was added and, after lack of effect, a 9-mg intravenous bolus of rFVIIa (90 μg/kg). Coagulation parameters improved and for several hours the hemoglobin level was stable, but she developed blistering of arms, fingers, and feet with slow capillary refill. Toes, heel, and sole of the right foot became black and cold. On laparotomy packings and 9 l blood and cloths were removed. Afterwards there was no bleeding. The ischemia worsened in a few days and spread to the fore foot (Fig. 1) resulting in amputation of the lower extremity. Necrosis also developed on the dorsal side of the left fore arm. Other acra were cyanotic but did not become necrotic. Since delivery she had received 48 U red blood cells, 30 U fresh frozen plasma and 17 U five donors each of platelets. Multiorgan failure improved, and 19 days after admission the patient was discharged to the ward. Fig. 1 Right foot of the patient

he left fore arm. Other acra were cyanotic but did not become necrotic. Since delivery she had received 48 U red blood cells, 30 U fresh frozen plasma and 17 U five donors each of platelets. Multiorgan failure improved, and 19 days after admission the patient was discharged to the ward. Fig. 1 Right foot of the patient rFVIIa use has been reported for severe postpartum hemorrhage [1, 2]. Immediately after administration of the third dose of rFVIIa hemostasis was indeed reached. However, severe thromboembolic complications occurred. Although there seems to be a relationship in time between the onset of complications and moment of drug administration, it is hard to confirm this relationship. rFVIIa overdosing or sepsis may also play a role. However, there was no positive confirmation of infection. The incidence of thrombotic complications in licensed use is 1–2%. Most thromboembolic events have followed off-label use of rFVIIa [3]. Off-label rFVIIa is used principally in desperate situations. However, to make a balanced choice, even then we need to have an impression of the complication rate. Safety and efficacy in off-label use must be studied [4]. Pregnant and postpartum women should receive special attention in these studies because of their particular coagulation condition.

t in decision-making in interhospital transport of a critically ill patient. Additionally, transport facilities are perceived as most important by the majority of medical heads of Dutch ICUs. Neither characteristics of the patient's condition nor the level of supportive care seems to be of significance in this process. The large number of publications on interhospital transport reflects the interest in this complex part of IC medicine but are descriptive and mainly focuses on the technical and organizational aspects of transport [1, 9, 12, 14]. The use of specialized transport teams and appropriate equipment may result in a decrease in transport associated morbidity and mortality by creating an intensive care environment in a vehicle-ground ambulance or aircraft [8, 9, 15–17].

Introduction Interhospital transport of critically ill patient may be indicated if additional care, whether technical, cognitive, or procedural, is not available at the existing location [1]. Regionalization of intensive care medicine in centers with high patient volumes appears to improve outcome of patients and therefore may further increase the need for these transports [2–4]. The risks associated with interhospital transport should be weighted against its potential benefit for each individual critically ill patient [5–7]. The use of specialized teams and appropriate equipment might reduce these risks [8, 9]. Although guidelines have been developed to increase the safety of interhospital transport of critically ill patients, clinical evidence is lacking on factors determining the transportability of these patients [1, 4]. Decision-making in interhospital transport involves appraisal of several determinants including patient characteristics, indication for transport, level of escort, and transport facilities. The process of appraisal of these variables, however, has never been studied [10]. The aim of the present study was to assess the relative importance of clinical and transport-related determinants influencing physicians' decision-making in interhospital transport of critically ill patients.

Introduction Interhospital transport of critically ill patient may be indicated if additional care, whether technical, cognitive, or procedural, is not available at the existing location [1]. Regionalization of intensive care medicine in centers with high patient volumes appears to improve outcome of patients and therefore may further increase the need for these transports [2–4]. The risks associated with interhospital transport should be weighted against its potential benefit for each individual critically ill patient [5–7]. The use of specialized teams and appropriate equipment might reduce these risks [8, 9]. Although guidelines have been developed to increase the safety of interhospital transport of critically ill patients, clinical evidence is lacking on factors determining the transportability of these patients [1, 4]. Decision-making in interhospital transport involves appraisal of several determinants including patient characteristics, indication for transport, level of escort, and transport facilities. The process of appraisal of these variables, however, has never been studied [10]. The aim of the present study was to assess the relative importance of clinical and transport-related determinants influencing physicians' decision-making in interhospital transport of critically ill patients. Methods We sent a national questionnaire survey with paper case descriptions, so-called clinical vignettes, to the medical heads (intensivist or supervising consultant) of all 95 intensive care units (ICUs) in The Netherlands. Neonatal and pediatric ICUs were excluded. Questionnaires were anonymous but coded, and therefore so nonresponders could be followed up with a postal reminder 2 months later. A prepaid envelope was included for its return, and a web-based version was available for digital responses. Of the 95 questionnaires 78 (82%) were returned and all were suitable for analysis. Respondents' mean age was 45 ± 6.6 years (Table 1). Most (n = 66, 86%) were intensivists with either anesthesiology or internal medicine as medical specialty. The median number of interhospital transport leaving their ICU was one per month, with a considerable range (0.01–12). Table 1 Characteristics of the 78 responding intensive care physicians and their hospitals

1). Most (n = 66, 86%) were intensivists with either anesthesiology or internal medicine as medical specialty. The median number of interhospital transport leaving their ICU was one per month, with a considerable range (0.01–12). Table 1 Characteristics of the 78 responding intensive care physicians and their hospitals Mean age (years) 45(±6.6) Medical speciality (%)a Intensive care medicine 66 (86%) Anesthesiology 37 (48%) Internal medicine 34 (44%) Surgery 1 (1%) Other 5 (7%) Type of hospital Academic medical center 13 (17%) Teaching hospital, nonacademic 34 (44%) Regional public hospital 30 (38%) Number of beds in ICU, median (range) 8 (2–42) Number of interhospital transport per month median (range) 1 (0.01–12) aMultiple specialities per physician possible The interhospital critical care transport system in The Netherlands is diverse. The majority of the transports are by ground (standard) ambulances escorted by an advanced life-support paramedic and occasionally complemented by the sending physician. Only a few regions use a dedicated, fully equipped mobile ICU with an escorting team of intensive care (IC) physician and IC nurse. The questionnaire The questionnaire consisted of two parts: (a) characteristics of the respondent and its ICU including frequency of interhospital ICU transport from their hospital; (b) 16 clinical vignettes.

The interhospital critical care transport system in The Netherlands is diverse. The majority of the transports are by ground (standard) ambulances escorted by an advanced life-support paramedic and occasionally complemented by the sending physician. Only a few regions use a dedicated, fully equipped mobile ICU with an escorting team of intensive care (IC) physician and IC nurse. The questionnaire The questionnaire consisted of two parts: (a) characteristics of the respondent and its ICU including frequency of interhospital ICU transport from their hospital; (b) 16 clinical vignettes. Clinical vignettes The 16 clinical vignettes are showed in Table 2. We identified eight potential determinants in decision making of IC transport which are known from clinical studies and critical care transport experience from the authors [1, 6–9, 11, 12]. The determinants were incorporated in the clinical vignettes: (a) age (30 vs. 60 vs. 80 years); (b) arterial oxygenation pressure (7.5 vs. 16.5 kPa); (c) level of positive expiratory pressure (PEEP) (8 vs. 18 cmH2O); (d) dose of noradrenaline infusion (0.12 vs. 0.60 μg/kg per minute); (e) arrhythmia (self-terminating ventricular tachycardia < 24 h vs. no arrhythmia within 6 h); (f) transport facility (fully equipped mobile ICU, i.e., IC ventilator, IC monitor including invasive blood pressure monitoring and capnography, sufficient number of syringe pumps) vs. standard ambulance (i.e., transport ventilator without IC performance, no invasive and capnography monitoring); (g) escorting personnel paramedic (advanced life support paramedic characterized by, e.g., protocolized advanced life support with medication, cardiopulmonary resuscitation intubation) vs. IC physician and paramedic vs. IC nurse and paramedic vs. team of IC physician and IC nurse and paramedic; (h) indication for transport (shortage of ICU beds in referring hospital vs. essential intervention not available in referring hospital). Table 2 The 16 case vignettes. Basic structure of each case vignette: patient admitted to ICU after initial presentation in the emergency department with severe sepsis (probably pneumococcal), Acute Physiology and Chronic Health Evaluation II of 18, mean arterial pressure of 70 mmHg after adequate fluid-resuscitation, endotracheally intubated and mechanically ventilated with 50% FIO2 and after 6 h in the ICU need for interhospital transport [VT, ventricular tachycardia (self terminating); MICU, mobile ICU]

, Acute Physiology and Chronic Health Evaluation II of 18, mean arterial pressure of 70 mmHg after adequate fluid-resuscitation, endotracheally intubated and mechanically ventilated with 50% FIO2 and after 6 h in the ICU need for interhospital transport [VT, ventricular tachycardia (self terminating); MICU, mobile ICU] Patients characteristics Transport conditions Age (years) paO2 (kPa) PEEP (cmH2O) Noradrenaline (μg/kg per minute) Arrhythmia Equipment Escorting personnel Indication for transport 1 30 16.5 18 0.12 VT < 24 h MICU trolley IC nurse Lack of ICU beds 2 30 16.5 8 0.12 None Basic ambulance IC physician and IC nurse Intervention elsewhere 3 30 16.5 18 0.6 None Basic ambulance Paramedic Lack of ICU beds 4 30 7.5 8 0.6 None Basic ambulance IC physician Lack of ICU beds 5 30 7.5 8 0.12 VT < 24 h MICU trolley IC physician and IC nurse Lack of ICU beds 6 80 7.5 8 0.6 VT < 24 h Basic ambulance IC nurse Intervention elsewhere 7 80 16.5 8 0.12 VT < 24 h Basic ambulance Paramedic Lack of ICU beds 8 60 7.5 8 0.12 none MICU trolley Paramedic Intervention elsewhere 9 30 7.5 18 0.12 none Basic ambulance IC nurse Intervention elsewhere 10 30 7.5 18 0.60 VT < 24 h MICU trolley Paramedic Intervention elsewhere 11 30 16.5 8 0.60 VT < 24 h MICU trolley IC physician Intervention elsewhere 12 60 16.5 18 0.60 VT < 24 h Basic ambulance IC physician and IC nurse Intervention elsewhere 13 80 16.5 18 0.12 none MICU trolley IC physician Intervention elsewhere 14 60 7.5 18 0.12 VT < 24 h Basic ambulance IC physician Lack of ICU beds 15 60 16.5 8 0.60 none MICU trolley IC nurse Lack of ICU beds 16 80 7.5 18 0.60 none MICU trolley IC physician and IC nurse Lack of ICU beds

IC nurse Intervention elsewhere 13 80 16.5 18 0.12 none MICU trolley IC physician Intervention elsewhere 14 60 7.5 18 0.12 VT < 24 h Basic ambulance IC physician Lack of ICU beds 15 60 16.5 8 0.60 none MICU trolley IC nurse Lack of ICU beds 16 80 7.5 18 0.60 none MICU trolley IC physician and IC nurse Lack of ICU beds As 768 case descriptions were needed to present all possible combinations of the eight determinants and their levels, the number of representative clinical vignettes were reduced to 16 using an orthogonal main-effects design [13]. This approach permits the statistical testing by conjoint analysis of a suitable fraction of all possible combinations of the factors (determinants) and their levels. Respondents were asked to rate the degree of transportability, defined as their personal clinical decision, whether they would let this patient be transported, for each of the 16 critically ill patients described in clinical vignettes. A seven point Likert scale was used ranging from 1 (“entirely not transportable”) to 7 (“definitely transportable”). It was emphasized that no true or false answers were sought but their clinical judgment.

would let this patient be transported, for each of the 16 critically ill patients described in clinical vignettes. A seven point Likert scale was used ranging from 1 (“entirely not transportable”) to 7 (“definitely transportable”). It was emphasized that no true or false answers were sought but their clinical judgment. Statistical analysis The means and standard deviations for continuous variables and distributions for frequency of categorical variables were summarized using descriptive statistics. Conjoint analysis was performed with transportability as dependent variable to calculate the relative weights for each level of the determinants [13]. This results in a utility score for each determinant level expressed in β with 95% confidence interval. These utility scores, estimated by least squares regression analogous to regression coefficients, provide a quantitative measure of the preference for each determinant level, with larger values corresponding to greater preference.

n a utility score for each determinant level expressed in β with 95% confidence interval. These utility scores, estimated by least squares regression analogous to regression coefficients, provide a quantitative measure of the preference for each determinant level, with larger values corresponding to greater preference. Considering the individual respondents as random effects took into account that the preference score originating from 16 repeated measurements. Determinants with a negative β indicated preference against transportability. The reference value, by definition β = 0, was defined as the optimal conditions for critical care transport (youngest age, highest PaO2, lowest dose of noradrenaline, no arrhythmia, fully equipped mobile ICU ambulance, escorting team of IC physician and IC nurse, intervention required not available in own facility). The conjoint analysis was repeated in relation to (a) type of hospital the respondents were working in regional hospital or teaching/university hospital, (b) speciality of the respondent, and (c) the method of data collection, either paper or online questionnaire.

, intervention required not available in own facility). The conjoint analysis was repeated in relation to (a) type of hospital the respondents were working in regional hospital or teaching/university hospital, (b) speciality of the respondent, and (c) the method of data collection, either paper or online questionnaire. Results The impact of the determinants in the decision making on transportability is displayed in Fig. 1. Those with the largest negative effects on preference for transportability were the type of escorting personnel [paramedic only: β = –3.1 (–3.7 to –2.5); IC nurse and paramedic: β = –2.1 (–2.5 to –1.7); team of IC physician, nurse, and paramedic: β = –1.0 (–1.2 to –0.8); standard ambulance: β = –1.21 (–1.7 to –0.8)]. Determinants reflecting the critically ill patient's condition and intensity of treatment were scored to be of relative minor importance [dose of noradrenaline: β = –0.6 (–1.0 to –0.1); arterial oxygenation β = –0.8 (–1.3 to –0.4); level of PEEP β = –0.6 (–1.0 to –0.1)]. Age [60 years: β = 0.1 (–0.2 to 0.3); 80 years: β = 0.1 [–0.4 to 0.7)], cardiac arrhythmia [β = 0.1 (–0.4 to 0.5)], and the indication for transport (β = –0.3 (–0.8 to 0.1)] had no significant effect on the preference for transportability (Fig. 1). Fig. 1 Relative weight (expressed in β, 95% confidence interval) of determinants influencing the decision on interhospital IC transport. ref, Reference value; PEEP, positive end-expiratory pressure; ventric, ventricular; IC, intensive care

0.1)] had no significant effect on the preference for transportability (Fig. 1). Fig. 1 Relative weight (expressed in β, 95% confidence interval) of determinants influencing the decision on interhospital IC transport. ref, Reference value; PEEP, positive end-expiratory pressure; ventric, ventricular; IC, intensive care Repeated analyses did not demonstrate significant differences in relative weights of the determinants in relation to respondents' working location (regional hospital vs. large teaching hospital or academic medical center), type of medical speciality, or method of data collection (paper vs. online). Discussion Decision-making in interhospital transport involves appraisal of several determinants such as patient characteristics, indication for transport, level of escort, and transport facilities. The present study shows that the level of escorting personnel is an important determinant in decision-making in interhospital transport of a critically ill patient. Additionally, transport facilities are perceived as most important by the majority of medical heads of Dutch ICUs. Neither characteristics of the patient's condition nor the level of supportive care seems to be of significance in this process.

ly focuses on the technical and organizational aspects of transport [1, 9, 12, 14]. The use of specialized transport teams and appropriate equipment may result in a decrease in transport associated morbidity and mortality by creating an intensive care environment in a vehicle-ground ambulance or aircraft [8, 9, 15–17]. Despite the growth in interhospital transport due to regionalization of intensive care medicine the process by which IC physicians identify patients transportability is not well known [3, 10, 18]. Transportability as a result of a professional analysis of the balance between risks and potential benefits of an individual transport is hard to define. The accumulating literature on improved outcome associated with ICUs treating larger volumes of patients (e.g., with severe sepsis or mechanical ventilation) is not adequately accompanied by research on clinical parameters determining transportability in such conditions [19, 20]. A study by Lee et al. [10] used a questionnaire with clinical scenarios before and after a program, including a 15-min training in the use of interhospital transfer rules [10]. After the start of the program clinical staff were able to make appropriate decisions using these guidelines focusing on diagnosis and physiology. To our knowledge, however, no study has mimicked decision making in interhospital transport with appraisal of several realistic and detailed determinants as in daily clinical practice (i.e., as those in tested in this conjoint analysis) by experienced intensivists who endorse such transports.

diagnosis and physiology. To our knowledge, however, no study has mimicked decision making in interhospital transport with appraisal of several realistic and detailed determinants as in daily clinical practice (i.e., as those in tested in this conjoint analysis) by experienced intensivists who endorse such transports. Age is an important prognostic factor, for mortality rates are higher in elderly than in younger ICU patients [21]. This has not been studied in transported IC patients, but it is conceivable that intensivists would weigh this determinant in their transportability decision. The finding of the present study that age does not influence decision making for transportability is remarkable. The same holds true for the level of PEEP, which seems representative for severity of oxygenation and is known to be a critical factor in transport [11]. IC physicians, however, seem to consider factors associated with severity of illness (age, PEEP, noradrenaline dose, oxygenation) to be less important than to transport conditions. International guidelines underline the importance of these conditions, but clinical transport studies and recommendations are lacking to address the issue of transport-related morbidity and mortality of extreme critical ill patients despite optimal expertise and equipment [7–9, 17, 22].

nt than to transport conditions. International guidelines underline the importance of these conditions, but clinical transport studies and recommendations are lacking to address the issue of transport-related morbidity and mortality of extreme critical ill patients despite optimal expertise and equipment [7–9, 17, 22]. One of the limitations of this study is the intrinsic shortcoming of the vignette method. Paper case descriptions, so-called clinical vignettes, have been recognized as a valid policy capturing tool to assess preferences in clinical practice [18, 23]. However, it is impossible to overcome the sentinel effect in which the physicians know they are being evaluated. Due to this “Hawthorne effect” there may be a discrepancy between physicians' decisions in practice and their answers to vignettes with hypothetical patients. Another limitation is the choice of content of the vignettes with eight determinants of transportability. The content of vignettes survey is limited to a number of determinants with their corresponding levels as an intrinsic element of conjoint analysis to generate an optimal number of vignettes a respondent would still adequately evaluate [13]. The set of determinants used in this study is based only on literature and critical care transport experience and could therefore be biased [1, 6–9, 11, 12]. Other, unknown factors could not be studied as critical in transport. Those factors would only be revealed in clinical transport studies documenting all clinical parameters and relate them with clinical outcome after transport. Finally, this national questionnaire survey is limited by the Dutch situation, where due to geography interhospital transport is carried out by ground ambulance without air medical transport. It is conceivable that the choice of vehicle is a crucial determinant in the decision making in combination with the interhospital distance [24].

ional questionnaire survey is limited by the Dutch situation, where due to geography interhospital transport is carried out by ground ambulance without air medical transport. It is conceivable that the choice of vehicle is a crucial determinant in the decision making in combination with the interhospital distance [24]. Conclusions This policy observing study indicates the importance of optimal escorting and transport facilities in interhospital transport as appreciated by IC physicians. These conditions are considered to be essential and enable even severe critically ill patients to be transported. Further clinical (transport) research should reveal which levels of expertise and transport facilities are indicated for which category of critically ill patients to tailor the use of expensive resources required for those inevitable road trips [9, 17]. Acknowledgements E.J.v.L. designed the study, performed the measurements, assisted in the statistical analyses, and drafted the manuscript. R.d.V. designed the study, performed the statistical analysis, and participated in drafting the manuscript. J.M.B. performed the statistical analysis and participated in drafting the manuscript. R.d.H. participated in the statistical analysis and drafting the manuscript. M.J.S. and M.B.V. participated in the study design and drafting the manuscript. All authors read and approved the final manuscript. We thank all our colleagues intensive care physicians who returned our questionnaire.

drafting the manuscript. R.d.H. participated in the statistical analysis and drafting the manuscript. M.J.S. and M.B.V. participated in the study design and drafting the manuscript. All authors read and approved the final manuscript. We thank all our colleagues intensive care physicians who returned our questionnaire. Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.

Sir: Nowadays, the practice of withholding and withdrawal of life-sustaining treatments in children is medically and ethically acceptable when these measures can no longer yield a good outcome. In North America and Europe, 28–65% of all paediatric intensive care unit (PICU) deaths follow a restriction in care [1]. However, clinicians' attitudes and actions regarding end-of-life decisions may be altered by exposure to a different culture or religion [2]. When clinicians believe that life-sustaining treatment is medically inappropriate or inhumane, they are not necessarily obliged to provide it simply because it is demanded on religious grounds by the parents. Instead, alternative religious interpretations and attempts to reach a consensus on the appropriate limits to life-sustaining treatment should be discussed [3]. A 7-year-old boy was diagnosed with a cerebellar medulloblastoma. Complete remission was initially achieved but 6 months before admission he had a relapse of the tumour without therapeutic options. He developed clinical signs of an upper airway obstruction. It was thought that a viral infection superimposed on his vocal cords paralysis was the main cause. Life-sustaining treatment was started because the viral infection was considered to be an intercurrent, curable event. He was intubated and assisted mechanical ventilation (MV) was initiated, after which he was transferred to our PICU. During the following days, his neurological condition rapidly deteriorated. No signs of pain or discomfort were observed.

because the viral infection was considered to be an intercurrent, curable event. He was intubated and assisted mechanical ventilation (MV) was initiated, after which he was transferred to our PICU. During the following days, his neurological condition rapidly deteriorated. No signs of pain or discomfort were observed. The attending physician informed the parents about these developments. It became apparent that the prolongation of life-sustaining treatment would not contribute to a good outcome. Since death was imminent, the attending physician discussed the possibilities of withdrawing or withholding treatment with the parents. Withdrawing treatment, i. e. stopping MV, was not an option because of the parents' religious beliefs. For them, such action seemed to be intended to hasten death and was therefore prohibited. The attending physician was concerned that the parents would also object to withdrawal of MV after diagnosis of brain death. After careful consideration, the attending physician proposed to switch to PS/CPAP ventilation, thereby respecting the wishes of the parents to continue MV. In addition, the parents accepted the explanation that cessation of the central respiratory drive meant a fatal progression of the underlying disease, so that no change to controlled MV should be made. Shortly thereafter, the boy died in the presence of his parents and sister. He had experienced a fatal apnoea while still intubated and on PS/CPAP. Despite their tragic loss, the parents were pleased at their involvement in the discussion on the end-of-life decision and satisfied that their religious convictions had been respected.

after, the boy died in the presence of his parents and sister. He had experienced a fatal apnoea while still intubated and on PS/CPAP. Despite their tragic loss, the parents were pleased at their involvement in the discussion on the end-of-life decision and satisfied that their religious convictions had been respected. There are no paediatric guidelines for withdrawal of a life-sustaining treatment like MV. The two approaches used are termed “terminal extubation”, i. e. removing the endotracheal tube without weaning ventilatory support, and “terminal weaning”, decreasing ventilator support before extubation [4]. This case report demonstrates that religious beliefs may prohibit both approaches and provides an elegant alternative in a patient with respiratory insufficiency of central origin. Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.

Sir: We read with interest the recent revision of the Surviving Sepsis Campaign (SSC) guidelines by Dellinger et al. [1]. The use of the GRADE system to classify the strength of the recommendations has certainly improved the guidelines. However, we regret that not all guidelines were adjusted according to the current literature. First of all, the absence of a recommendation regarding selective digestive tract decontamination (SDD) is striking. The guidelines group was evenly split, with equal numbers weakly in favor and against recommending the use of SDD. This is remarkable, since SDD is one of the best ever evaluated therapies in intensive care medicine, with more than 50 randomized controlled trials and 10 meta-analyses showing that SDD reduces pneumonia by 65% and mortality by 22% [2]. The authors gave several reasons why they chose not to recommend SDD in their guidelines. They argue that no studies regarding SDD specifically focused on septic patients. However, several other guidelines based on general ICU populations (i.e., stress ulcer prophylaxis, deep vein thrombosis prophylaxis, glucose control and bicarbonate therapy) received strong recommendations. Furthermore, the authors state that studies comparing SDD with non-antimicrobial interventions, such as ventilator bundles, are needed. Are they seriously suggesting that until these studies have been performed a therapy with proven high efficiency should be withheld from patients with severe sepsis? It seems that no scientific arguments, no study whatsoever could change the apparently biased authors.

erventions, such as ventilator bundles, are needed. Are they seriously suggesting that until these studies have been performed a therapy with proven high efficiency should be withheld from patients with severe sepsis? It seems that no scientific arguments, no study whatsoever could change the apparently biased authors. The main argument against the use of SDD is the persistent concern regarding emergence of antimicrobial resistance in critically ill patients. Antimicrobial resistance was not a clinical problem in 10 SDD studies monitoring resistance for 2–9 years [3–11]. SDD even seemed to reduce the resistance of aerobic Gram-negative bacilli, the target microorganisms of SDD [12, 13], possibly because the addition of enteral to parenteral antimicrobials prevents spontaneous mutation of target bacteria and eradicates mutants. In their “rationale” the authors are especially concerned about emergence of resistant Gram-positive infections. The SDD prophylaxis is not active against vancomycin-resistant enterococci (VRE) and methicillin-resistant S. aureus (MRSA) and may promote gut overgrowth of these intrinsically resistant bacteria. Therefore, in ICUs with endemic MRSA enteral vancomycin is required as a component of SDD. VRE did not emerge in any of the studies using enteral vancomycin, and there is no evidence that SDD promotes infection due to Gram-positive bacteria [14–19]. On the contrary, the continued use of only systemic antibiotics may lead to a further rise in drug-resistant Gram-positive bacteria. We propose, therefore, that the authors of the SSC guidelines use the available literature instead of their bias.

idence that SDD promotes infection due to Gram-positive bacteria [14–19]. On the contrary, the continued use of only systemic antibiotics may lead to a further rise in drug-resistant Gram-positive bacteria. We propose, therefore, that the authors of the SSC guidelines use the available literature instead of their bias. Secondly, the strong recommendation in favor of the use of stress ulcer prophylaxis is not, in our view, in line with currently available evidence. This recommendation is, like that in the guidelines of 2004, still mainly based on ancient studies performed in the 1980s [20–23], a meta-analysis from 1991 [24], and a large trial in 1998 [25] without a control arm. However, the most recent meta-analysis [26] shows no reduction of clinical important bleeding – but is somehow completely ignored. Whether the results of these older trials are applicable nowadays is questionable, since the incidence of stress ulcer-related bleeding has significantly decreased over recent decades due to improved ICU treatment [27, 28]. This definitely affects the balance between the benefit of prevention of gastro-intestinal bleeding and the increased risk of ventilator-associated pneumonia due to higher stomach pH [29]. Several recent trials show comparable rates of bleeding and endoscopic evidence of stress-related injury between treatment and placebo groups [30–33]. These results are pathophysiologically plausible, since stress ulcers are caused not by increased secretion of gastric acid, but by splanchnic hypoperfusion. Unfortunately, many recent trials only compare H2 blockers with proton pump inhibitors, without a placebo group. Altogether, according to the most recent meta-analysis and the more recent trials, a strong recommendation not to use stress ulcer prophylaxis would be more appropriate.

cid, but by splanchnic hypoperfusion. Unfortunately, many recent trials only compare H2 blockers with proton pump inhibitors, without a placebo group. Altogether, according to the most recent meta-analysis and the more recent trials, a strong recommendation not to use stress ulcer prophylaxis would be more appropriate. Thirdly, we disagree with the strength of the recommendation to reduce blood glucose levels in patients with severe sepsis. On the current evidence, this should be at most a weak recommendation. The beneficial effect of intensive insulin therapy has been demonstrated only in surgical patients, not in septic patients [34–36]. The benefit versus harm balance of intensive insulin therapy may be quite different for patients with severe sepsis than for the investigated surgical patients. It is not unreasonable to assume that septic patients may be more at risk for hypoglycemia, because sepsis may be associated with a deficiency of counterregulatory hormones. In the study of medical patients by van den Berghe [36], as well as the VISEP study [35] and the Glucontrol study [34], the risk of hypoglycemia was substantially increased, and hypoglycemia was an independent risk factor for mortality. None of these studies followed up the patients with hypoglycemia for neurocognitive impairment. Furthermore, the target glucose level of <150 mg/dl recommended in the guidelines is based solely on expert opinion and is not supported by data from any trial. Therefore, the beneficial effect, the harmlessness, and the target glucose level of intensive insulin therapy remain to be demonstrated in septic patients.

nt. Furthermore, the target glucose level of <150 mg/dl recommended in the guidelines is based solely on expert opinion and is not supported by data from any trial. Therefore, the beneficial effect, the harmlessness, and the target glucose level of intensive insulin therapy remain to be demonstrated in septic patients. In conclusion, the revised SSC guidelines have certainly been improved by the use of the GRADE system to classify the strength of the recommendations. However, a strong recommendation in favor of the use of SDD should have been implemented. The strong recommendations in favor of stress ulcer prophylaxis and glucose control are not in line with current evidence. Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited. An author's reply to this comment is available at: http://dx.doi.org/10.1007/s00134-008-1090-z.

Introduction The development of pediatric intensive care has contributed to improved survival in children with critical illness [1, 2]. Traditional outcome measures such as length of stay, number of ventilation days, and mortality are not the only relevant issues. New disease patterns have emerged due to long-term complications and effects of the original illness and its treatment. Physical sequelae, disability, and functional health have become equally important outcome measures in pediatric intensive care unit (PICU) survivors. Awareness of physical sequelae of underlying diseases and intensive care treatment and subsequently their impact on growth and development could lead to improvement in treatment and support after discharge. Studies on physical sequelae in PICU survivors are scarce. In studies evaluating functional health and quality of life aspects, one third of the evaluated children are in full health 1 year after discharge and 60–80% of the children have good quality of life [3–9]. In 1992 the Pediatric Cerebral Performance Category (PCPC) and the Pediatric Overall Performance Category (POPC) were developed and validated to evaluate outcome in a general PICU population [10, 11]. In 58% and 27% of the studied children, respectively, PCPC and POPC scores at PICU discharge were normal. These scores correlated well with the results of extensive psychometric tests (such as Bayley Scales of Infant Development) performed at one and 6 months after discharge [11, 12].

general PICU population [10, 11]. In 58% and 27% of the studied children, respectively, PCPC and POPC scores at PICU discharge were normal. These scores correlated well with the results of extensive psychometric tests (such as Bayley Scales of Infant Development) performed at one and 6 months after discharge [11, 12]. In multidisciplinary PICU populations reports on outcome are scarce. Historically, outcome research in pediatrics has been based either on an age-specific approach such as follow-up studies of premature infants or on a more disease-oriented approach such as follow-up studies in survivors of cardiothoracic surgery or trauma [13–15]. Evaluative research of adult intensive care survivors has shown the effect of intensive care treatment per se [16–18]. The aim of this study was to evaluate physical and neurocognitive sequelae 3 months after discharge in children who were previously healthy and unexpectedly admitted to the PICU. Materials and methods This study was carried out between December 2002 and October 2005 as part of an on-going explorative research program on physical and psychological sequelae in children and their parents after an acute and unexpected PICU admission. The PICU of the Emma Children's Hospital, Academic Medical Center of Amsterdam is a tertiary PICU with 14 beds, admitting medical, surgical, and trauma patients from the greater Amsterdam area. The local ethics review board approved the study protocol.

ren and their parents after an acute and unexpected PICU admission. The PICU of the Emma Children's Hospital, Academic Medical Center of Amsterdam is a tertiary PICU with 14 beds, admitting medical, surgical, and trauma patients from the greater Amsterdam area. The local ethics review board approved the study protocol. Participants This study included only previously healthy children who were unexpectedly referred to the PICU with an acute life-threatening illness; we excluded children with known underlying illnesses or scheduled elective surgery. We included all previously healthy patients with respiratory or circulatory insufficiency, all trauma patients irrespective of length of PICU stay, and patients admitted to the PICU for other reasons for 7 days or more. Exclusion criteria were admission due to abuse or self-intoxication and the inability to complete Dutch-language questionnaires. Patient characteristics were obtained from medical records and from the patient data management system.

h of PICU stay, and patients admitted to the PICU for other reasons for 7 days or more. Exclusion criteria were admission due to abuse or self-intoxication and the inability to complete Dutch-language questionnaires. Patient characteristics were obtained from medical records and from the patient data management system. Of the 250 children eligible for inclusion 186 participated. The parents of 64 did not participate: in 13 cases they refused to participate, in 33 they accepted participation but did not present at the outpatient clinic, and in 18 they did not respond at all. Reasons for refusal included: “everything is going well,” “we have seen too many hospitals,” “we need some rest,” and “we don't want to remember that time”. Participants and nonparticipants differed only in (a) length of stay, (b) length of ventilation, (c) reasons for admission, and (d) diagnosis at discharge. For example, participants had a longer length of stay and had more ventilation days than nonparticipants; 14 participants stayed in the PICU longer than 21 days and had more than 14 ventilation days (Tables 1, 2). Table 1 Patient characteristics of PICU admission in participants and nonparticipants

gnosis at discharge. For example, participants had a longer length of stay and had more ventilation days than nonparticipants; 14 participants stayed in the PICU longer than 21 days and had more than 14 ventilation days (Tables 1, 2). Table 1 Patient characteristics of PICU admission in participants and nonparticipants Participants (n = 186) Nonparticipants (n = 64) p-value Age of child, median (years; range) 1.4 (0.1–17.3) 1.5 (0.2–15.9) 0.423 Length of stay in PICU, median (days; range) 6.0 (1.0–336.0) * 4.0 (1.0–20.0) 0.006 Length of artificial ventilation, median (days; range) 5.0 (0.0–119.0) * 3.0 (0.0–19.0) 0.004 Risk of mortalitya, median (%; range) 4.2 (0.2–80.7) 3.8 (0.2–26.6) 0.293 Female 68 (37%) 23 (36%) 0.929 Reason for PICU admission * 0.001 Respiratory insufficiency 77 (41%) 33 (52%) Circulatory insufficiency 31 (17%) 13 (21%) Trauma 44 (24%) 9 (13%) Neurological disorder 22 (12%) 8 (12%) Metabolic disorder 6 (3%) 2 (3%) Gastrointestinal disorder 6 (3%) 0 (0%) Treatment characteristics (yes) Artificial ventilation 151 (81%) 55 (86%) 0.389 Circulatory support 46 (25%) 16 (25%) 0.966 Neuromuscular blocking agents 37 (20%) 18 (28%) 0.170 Corticosteroids 69 (37%) 21 (33%) 0.757 Central venous catheter 67 (36%) 18 (28%) 0.250 * p < 0.05 participants vs. nonparticipants; a Paediatric Index of Mortality II Table 2 Diagnosis at PICU discharge in participants and nonparticipants

Participants (n = 186) Nonparticipants (n = 64) p-value Age of child, median (years; range) 1.4 (0.1–17.3) 1.5 (0.2–15.9) 0.423 Length of stay in PICU, median (days; range) 6.0 (1.0–336.0) * 4.0 (1.0–20.0) 0.006 Length of artificial ventilation, median (days; range) 5.0 (0.0–119.0) * 3.0 (0.0–19.0) 0.004 Risk of mortalitya, median (%; range) 4.2 (0.2–80.7) 3.8 (0.2–26.6) 0.293 Female 68 (37%) 23 (36%) 0.929 Reason for PICU admission * 0.001 Respiratory insufficiency 77 (41%) 33 (52%) Circulatory insufficiency 31 (17%) 13 (21%) Trauma 44 (24%) 9 (13%) Neurological disorder 22 (12%) 8 (12%) Metabolic disorder 6 (3%) 2 (3%) Gastrointestinal disorder 6 (3%) 0 (0%) Treatment characteristics (yes) Artificial ventilation 151 (81%) 55 (86%) 0.389 Circulatory support 46 (25%) 16 (25%) 0.966 Neuromuscular blocking agents 37 (20%) 18 (28%) 0.170 Corticosteroids 69 (37%) 21 (33%) 0.757 Central venous catheter 67 (36%) 18 (28%) 0.250 * p < 0.05 participants vs. nonparticipants; a Paediatric Index of Mortality II Table 2 Diagnosis at PICU discharge in participants and nonparticipants Participants (n = 186) Nonparticipants (n = 64) p-value n % n % Respiratory insufficiency 0.001 Pneumonia 41 22 19 30 Upper airway obstruction 19 10 13 20 Asthma 13 7 0 – Other 4 2 1 2 Circulatory insufficiency 0.001 Congenital heart disease 14 8 0 – Other cardiac disease 4 2 1 2 Meningococcal disease 11 6 8 12 Toxic shock syndrome 0 – 4 6 Other septic shock 2 1 0 – Trauma 0.001 Head trauma 35 19 7 11 Other trauma 9 5 1 2 Neurological disorder 0.001 Status epilepticus 15 8 5 8 Meningitis 3 2 3 5 Other 4 2 0 – Metabolic disorder 6 3 2 3 – Gastrointestinal disorder 6 3 0 – – * p < 0.05 participants vs. nonparticipants

11 6 8 12 Toxic shock syndrome 0 – 4 6 Other septic shock 2 1 0 – Trauma 0.001 Head trauma 35 19 7 11 Other trauma 9 5 1 2 Neurological disorder 0.001 Status epilepticus 15 8 5 8 Meningitis 3 2 3 5 Other 4 2 0 – Metabolic disorder 6 3 2 3 – Gastrointestinal disorder 6 3 0 – – * p < 0.05 participants vs. nonparticipants Definitions The term previously healthy was defined as having no need for medical supervision at anytime before PICU admission. Included were children presenting at the emergency room and directly admitted to the PICU as well as children first admitted to the general ward who deteriorated and were subsequently admitted to the PICU. Physical sequelae are defined as any physical complaints or abnormalities found at the outpatient follow-up clinic by physical examination. Since all children were previously healthy, they had no physical complaints or abnormalities until shortly before PICU admission. Thus all residual physical complaints or abnormalities had to be related to either the underlying illness or to complications of PICU procedures. Study protocol After discharge from the PICU each family received a letter at home explaining the aim and content of the research program. Families were contacted by telephone to invite participation. In cases of failure follow-up letters with tear-off reply slip inviting participation were sent. Written informed consent was obtained from all participating families.

h family received a letter at home explaining the aim and content of the research program. Families were contacted by telephone to invite participation. In cases of failure follow-up letters with tear-off reply slip inviting participation were sent. Written informed consent was obtained from all participating families. Three months after discharge, all families were invited to visit the outpatient follow-up clinic for structured medical examination of the child. Follow-up at 3 months after discharge was chosen because of psychological follow-up on posttraumatic stress disorder; the results of this assessment will be reported elsewhere. Structured medical examination was performed by semistructured history taking (current health status, well being, social, cognitive and physical functioning and development) and physical examination by one author (H.K.). Complications from PICU procedures such as scars, upper airway obstruction, postthrombotic syndrome and hypoxic-ischemic injury were structurally evaluated [19, 20]. Since many parents mentioned cognitive and behavioral problems spontaneously, these questions were added to the standard interview. Physical and neurocognitive sequelae found at the outpatient follow-up clinic were categorized into two groups: (a) previously unknown pre-PICU morbidity (e. g., a patient admitted to the PICU because of cyanosis caused by an until then not recognized congenital heart defect), (b) acquired morbidity (e. g., a patient with meningococcal infection, suffering from scars and postthrombotic syndrome following central venous catheterization). Acquired morbidity can be related either to the underlying illness for which PICU admission was necessary (scars due to meningococcal infection), to complications of PICU procedures (postthrombotic syndrome), or to a combination of the two. As no validated questionnaires exist to evaluate physical sequelae, we used PCPC and POPC scores to determine the extent of physical and neurocognitive sequelae. PCPC and POPC were determined at PICU discharge, 3 months after discharge, and retrospectively 24 h before admission to the PICU (baseline values) by one of the authors (H.K.).

estionnaires exist to evaluate physical sequelae, we used PCPC and POPC scores to determine the extent of physical and neurocognitive sequelae. PCPC and POPC were determined at PICU discharge, 3 months after discharge, and retrospectively 24 h before admission to the PICU (baseline values) by one of the authors (H.K.). Data were analyzed using the Statistical Package for Social Sciences, Windows version 11.5. Mann–Whitney and χ2 tests were performed to compare participants and nonparticipants with regard to patient characteristics, reason for PICU admission, and diagnosis at discharge. Results Structured history taking at the outpatient follow-up clinic revealed concentration problems (9%), behavioral problems (15%), delayed psychomotor development (13%), temporary voice changes (7%), eating problems (9%), sleeping problems (9%), and withdrawal symptoms (9%). In 9 of 40 school-age children problems at school were reported. Structured physical examination revealed abnormalities in weight gain (9%), pulmonary auscultation (9%) and neurological (23%) examination, hoarseness after endotracheal intubation (4%), and postthrombotic syndrome after central venous catheterization (4%) and scars (14%).

age children problems at school were reported. Structured physical examination revealed abnormalities in weight gain (9%), pulmonary auscultation (9%) and neurological (23%) examination, hoarseness after endotracheal intubation (4%), and postthrombotic syndrome after central venous catheterization (4%) and scars (14%). Of the 186 patients 128 (69%) had persisting complaints 3 months after discharge and 58 (31%) were healthy. Of the 128 children 55 (43%) had a previously unknown underlying illness (pre-PICU morbidity) that was diagnosed during PICU admission. This included patients with congenital anomalies (congenital heart defect, n = 14; metabolic disorder, n = 4), epilepsy (n = 6), and first-attack asthma patients (n = 13; Table 3). Table 3 Physical sequelae 3 months after discharge in 186 participants (more than one problem per child possible): pre-PICU morbidity is a previously unknown underlying illness that was diagnosed during PICU admission; acquired morbidity is morbidity in a child that was healthy before PICU admission Pre-PICU morbidity (n = 55) Acquired morbidity (n = 73) n % n % Respiratory problems 21 11 23 12 Circulatory problems 14 8 0 – Neurological problems 6 3 48 26 Metabolic disorder 6 3 Miscellaneous problems 8 4 14 8 Tracheotomy – – 3 2 Scars – – 17 9 Hoarseness – – 5 3 Postthrombotic syndrome – – 7 4

Of the 186 patients 128 (69%) had persisting complaints 3 months after discharge and 58 (31%) were healthy. Of the 128 children 55 (43%) had a previously unknown underlying illness (pre-PICU morbidity) that was diagnosed during PICU admission. This included patients with congenital anomalies (congenital heart defect, n = 14; metabolic disorder, n = 4), epilepsy (n = 6), and first-attack asthma patients (n = 13; Table 3). Table 3 Physical sequelae 3 months after discharge in 186 participants (more than one problem per child possible): pre-PICU morbidity is a previously unknown underlying illness that was diagnosed during PICU admission; acquired morbidity is morbidity in a child that was healthy before PICU admission Pre-PICU morbidity (n = 55) Acquired morbidity (n = 73) n % n % Respiratory problems 21 11 23 12 Circulatory problems 14 8 0 – Neurological problems 6 3 48 26 Metabolic disorder 6 3 Miscellaneous problems 8 4 14 8 Tracheotomy – – 3 2 Scars – – 17 9 Hoarseness – – 5 3 Postthrombotic syndrome – – 7 4 Seventy-three (57%) children were healthy before PICU admission and had acquired morbidity (Table 3). The complaints in these children were diverse, and some children had a combination of problems. Complaints consisted of (a) pulmonary complaints (after admission due to RSV infection) or upper airway problems (tracheotomy) after endotracheal intubation; (b) neurological or neurocognitive problems caused by among other factors, hypoxic-ischemic brain injury, traumatic brain injury, meningitis, or intracerebral bleeding; (c) scars after meningococcal disease, trauma, operations, pleural drains, and (central) venous lines; (d) hoarseness 3 months after extubation (e) postthrombotic syndrome after central venous catheterization; (f) miscellaneous illnesses such as renal insufficiency, adrenal insufficiency, and gastroenterological problems (Table 3). In at least 15 (8%) children morbidity was related to complications from PICU procedures. In the other 61 children with acquired morbidity it was not possible to differentiate between morbidity related to underlying illness, to complications from PICU procedures, or both (Table 3).

stroenterological problems (Table 3). In at least 15 (8%) children morbidity was related to complications from PICU procedures. In the other 61 children with acquired morbidity it was not possible to differentiate between morbidity related to underlying illness, to complications from PICU procedures, or both (Table 3). Twenty-four hours before PICU admission 177 of the 186 evaluated children (95%) had in retrospect normal PCPC scores and 135 (73%) normal POPC scores (baseline values). The 9 children (5%) with abnormal PCPC scores and 51 (27%) with abnormal POPC scores had presenting symptoms which were not recognized more than 24 h before PICU admission. Figures 1 and 2 show that at discharge 46 children (25%) had normal PCPC scores and 2 (1%) normal POPC scores. Three months after discharge 143 children (77%) had normal PCPC scores and 58 (31%) normal POPC scores. Three months after discharge, respectively, 147 (79%) and 93 children (50%) showed PCPC and POPC scores that were the same or improved compared to scores before admission. In respect 39 (21%) and 93 (50%) children's scores deteriorated (Figs. 1, 2). Fig. 1 Pediatric Cerebral Performance Category in 186 evaluated children 24 h before PICU admission, at PICU discharge, and 3 months after PICI discharge. Columns from left to right:black, normal; light gray, mild disability; dark gray, moderate disability; white, severe disabililty; medium gray, coma/vegatative state

Fig. 1 Pediatric Cerebral Performance Category in 186 evaluated children 24 h before PICU admission, at PICU discharge, and 3 months after PICI discharge. Columns from left to right:black, normal; light gray, mild disability; dark gray, moderate disability; white, severe disabililty; medium gray, coma/vegatative state Fig. 2 Pediatric Overall Performance Category in 186 evaluated children. POPC scores 24 h before PICU admission, at PICU discharge and 3 months after PICI discharge. Columns from left to right:black, normal; light gray, mild disability; dark gray, moderate disability; white, severe disabililty; medium gray, coma/vegatative state Discussion This is one of the first studies to report the nature and extent of physical sequelae in PICU survivors. Three months after discharge the evaluation of physical and neurocognitive sequelae in PICU survivors showed that only 31% of the evaluated children were healthy while 30% had a previously unknown underlying illness (pre-PICU morbidity), and 39% an acquired morbidity. In at least 8% of these children morbidity was related to complications from PICU procedures. Upper airway disease and pulmonary and neurological problems accounted for some 50% of persisting complaints. Upper airway disease, consisting of upper airway obstruction after endotracheal intubation, or lower airway disease after RSV infection and asthma was found in 47 of 186 evaluated children.

rom PICU procedures. Upper airway disease and pulmonary and neurological problems accounted for some 50% of persisting complaints. Upper airway disease, consisting of upper airway obstruction after endotracheal intubation, or lower airway disease after RSV infection and asthma was found in 47 of 186 evaluated children. Neurological and neurocognitive sequelae, consisting of delayed psychomotor development, epilepsy, pareses, and concentration and behavioral disturbances, were found in 54 of 186 evaluated children. Studies evaluating neurocognitive outcome after PICU survival in a structured way are limited [11, 21–23]. To evaluate the extent of neurocognitive sequelae we used PCPC and POPC scores at discharge and at 3 months' follow-up. At discharge PCPC and POPC scores in our patients were reduced in, respectively, 73% and 91%. Three months later scores had substantially improved. This leads us to believe that long-term follow-up is necessary to evaluate neurocognitive development by structured psychometric testing. In at least 8% of patients morbidity was associated with complications of PICU procedures. Examples include hoarseness due to mucosal damage after endotracheal intubation and impaired growth of extremities due to vascular damage after central venous catheterization. Neither hoarseness nor impaired growth of extremities was ever mentioned spontaneously. In addition, use of medications (corticosteroids, ototoxic drugs, sedatives, and neuromuscular blocking agents) also may lead to long-term sequelae and were not evaluated [24–27].

ascular damage after central venous catheterization. Neither hoarseness nor impaired growth of extremities was ever mentioned spontaneously. In addition, use of medications (corticosteroids, ototoxic drugs, sedatives, and neuromuscular blocking agents) also may lead to long-term sequelae and were not evaluated [24–27]. A number of limitations may have biased the results of this study. First, a considerable number of children were lost due to nonresponse and refusal. Although other follow-up studies in PICU patients had similar response rates, this may have biased our results [4, 7]. We probably missed a number of patients of whom parents were experiencing psychological problems, such as posttraumatic stress disorder [28–30]. Furthermore, the distinct subgroups of participants were not equally distributed; relatively more trauma patients and children with cardiac disorders are evaluated and fewer children with septic shock. Second, structured history taking by one observer involved in PICU care may have biased our results. Third, participants had significantly longer length of stay and more ventilation days than nonparticipants. This might be associated with an increased number of complications and severity of sequelae. Fourth, follow-up time was only 3 months; therefore conclusions on sequelae over an extended period cannot be drawn. Finally, PCPC and POPC scores have been validated for PICU populations; it remains questionable whether these scores can be used to predict problems in the individual patient.

s and severity of sequelae. Fourth, follow-up time was only 3 months; therefore conclusions on sequelae over an extended period cannot be drawn. Finally, PCPC and POPC scores have been validated for PICU populations; it remains questionable whether these scores can be used to predict problems in the individual patient. Despite these limitations we believe that these outcome data of a mixed pediatric and surgical PICU population are important. Our findings are comparable to the few existing studies on physical sequelae in PICU survivors. In an Australian study of 974 children 42% of survivors were normal and 17% functionally normal but required medical supervision. Of the remaining 41% functionally nonnormal survivors 32% will probably be able to lead an independent life [3]. The Health Utilities Index 2 used in four studies showed 27–37% of survivors in full health 1 year after discharge [5–7].

f survivors were normal and 17% functionally normal but required medical supervision. Of the remaining 41% functionally nonnormal survivors 32% will probably be able to lead an independent life [3]. The Health Utilities Index 2 used in four studies showed 27–37% of survivors in full health 1 year after discharge [5–7]. Suggestions for future research Cohort studies of PICU survivors evaluating patient outcome (physical and neurocognitive sequelae, functional status, and quality of life) are essential. Awareness of long-term sequelae may result in changes in treatment during the acute phase and in supportive programs after discharge [17, 18, 31–33]. Long-term follow-up clinics of PICU survivors and rehabilitation programs in exactly the same manner as in neonatal and trauma patients should be developed to detect, support, and treat children with neurocognitive, developmental, and psychological problems. These programs are expected to improve daily life [34, 35]. Pediatric intensivists should be core-members of the multidisciplinary follow-up team as they are familiar with possible risks and complications of PICU treatment. In addition, notifying complications of PICU procedures may serve as a valuable tool of providing feedback on procedures in the acute phase. Proper cohort studies in a pediatric population should be multidisciplinary, multicenter long-term follow-up studies applying specific measurement tools to evaluate sequelae in different age groups.

ing complications of PICU procedures may serve as a valuable tool of providing feedback on procedures in the acute phase. Proper cohort studies in a pediatric population should be multidisciplinary, multicenter long-term follow-up studies applying specific measurement tools to evaluate sequelae in different age groups. Conclusion PICU survival leads to substantial physical sequelae related to underlying illnesses, complications from PICU procedures, or both. We believe that preferably multicenter, long-term follow-up research in a structured and validated way is warranted in PICU survivors. This should include (a) multidisciplinary evaluation of physical and neurocognitive sequelae of the underlying illness and the intensive care treatment per se and subsequently their impact during growth and development, (b) evaluation of risk factors for sequelae, and (c) support after discharge if needed. To guarantee all this, follow-up is needed for a sufficient and extended period of time after discharge. Acknowledgements The authors thank J. H. van der Lee, MD, PhD, Department of Pediatric Clinical Epidemiology, Emma Children's Hospital, AMC Amsterdam, for advice and statistical expertise. Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.

Introduction The incidence of severe sepsis is increasing and mortality rates remain high. Whilst the case fatality rate has decreased modestly, the overall number of deaths is still increasing [1, 2]. Severe sepsis affects approximately 10 ± 4% of ICU patients, and around 100 ± 50 persons per 100,000 population [2–10]. The burden of disease and economic impact of sepsis are considerable and probably comparable to that of ischaemic heart disease [4]. As new treatments for patients with severe sepsis have proven remarkably elusive [11], the publication of the PROWESS trial reporting that drotrecogin alfa (activated) (DAA) reduced mortality was met with considerable enthusiasm [3–10, 12, 13]. Subsequently, much of the initial enthusiasm has waned and opinion over DAA has become polarized [14, 15]. At its annual review of DAA in February 2007, the European Medicines Agency (EMEA) stated: “Taking all the available data together, the Committee for Medicinal Products for Human Use (CHMP) considered that the benefit/risk balance of Xigris required additional clarification and that there was a need for another clinical trial to further prove the efficacy of Xigris in the target population….therefore a placebo-controlled study in patients … with severe sepsis and documented organ failure (e.g. MOD or vasopressor-dependent septic shock) when treated within a strictly defined time window, should be performed to assert the benefit/risk profile of Xigris.” [16].

he efficacy of Xigris in the target population….therefore a placebo-controlled study in patients … with severe sepsis and documented organ failure (e.g. MOD or vasopressor-dependent septic shock) when treated within a strictly defined time window, should be performed to assert the benefit/risk profile of Xigris.” [16]. Eli Lilly and Company (Lilly) agreed to sponsor the study and subsequently assembled an academic steering committee to advise on the design the trial and interpret its results. This article, written by the academic steering committee, describes the scientific and academic background to the study and is accompanied by publication of the full study protocol as an on-line supplement. Additionally, it presents the memorandum of understanding (MoU) delineating the relationship between the steering committee and the sponsor. The MoU recounts the processes by which the steering committee, the independent academic statistical centre (ASC) and the data monitoring committee (DMC) were established and formalizes the respective roles of the sponsor and these committees in the conduct of the trial. The processes and relationships described in the MoU are intended to ensure that the study will be designed, conducted, and analysed to the highest possible scientific standards, and be reported transparently. The primary goal for the SC is to supervise the design, conduct and analysis of a clinical trial that seeks to answer the crucial question: “Does DAA reduce mortality in patients with severe septic shock?”

be designed, conducted, and analysed to the highest possible scientific standards, and be reported transparently. The primary goal for the SC is to supervise the design, conduct and analysis of a clinical trial that seeks to answer the crucial question: “Does DAA reduce mortality in patients with severe septic shock?” Background The PROWESS study, the first phase III placebo-controlled randomised controlled trial of DAA, was stopped early following a second planned interim analysis when 1,690 of the planned 2,280 subjects had been enrolled. The trial reported an absolute reduction in 28-day all-cause mortality of 6.1% (a relative risk reduction [RRR] of 19.4%) [12]. Although the overall trial result was statistically significant, an FDA advisory committee noted several concerns; in particular, approximately half way through recruitment the protocol had been amended and the drug manufacturing process changed and treatment with DAA appeared more beneficial after these changes (Fig. 1).Fig. 1 Cumulative mortality rate over time in the PROWESS trial. Solid black lines, drotrecogin alfa (activated) group; solid grey lines, placebo group. The amended version of the protocol was introduced at Line A, first interim analysis occurred at Line B and the second interim analysis at Line C. (Reproduced from Critical Care Medicine 2004;32(12):2388 with permission)

WESS trial. Solid black lines, drotrecogin alfa (activated) group; solid grey lines, placebo group. The amended version of the protocol was introduced at Line A, first interim analysis occurred at Line B and the second interim analysis at Line C. (Reproduced from Critical Care Medicine 2004;32(12):2388 with permission) Additionally, the FDA advisory committee evaluated the treatment effect within subgroups and concluded that there was no apparent benefit in subjects at lower risk of death as indicated by lower APACHE II scores or in those patient with single organ dysfunction at baseline [17]. The FDA advisory committee split 10:10 on whether DAA should be approved for clinical use [18]. The FDA received statistical advice that the variable treatment effect observed over the course of the trial would be expected in approximately 8% of clinical trials and that the trial results did not appear to be due to the mid-trial protocol amendment or the change in the drug manufacturing process [19]. On 21 November 2001, the FDA licensed DAA but only for the treatment of patients with severe sepsis and a high risk of death (as determined, for example, by the APACHE II score), and requested additional studies in children and in adult patients with severe sepsis and a lower risk of death [20]. In Europe, the European Medicines Agency (EMEA) authorized DAA to be used for the treatment of adult patients with severe sepsis with multiple organ failure when added to best standard care; this authorization was subject to annual review [16].

en and in adult patients with severe sepsis and a lower risk of death [20]. In Europe, the European Medicines Agency (EMEA) authorized DAA to be used for the treatment of adult patients with severe sepsis with multiple organ failure when added to best standard care; this authorization was subject to annual review [16]. The additional trials requested by the FDA Two additional placebo-controlled trials have been conducted. The ADDRESS study in adult patients with severe sepsis who were at a lower risk of death was stopped early after 2,640 participants had been enrolled (Table 1) [21]. The independent DMC estimated the likelihood of demonstrating a reduction in 28-day mortality to be less than 5%. Within the ADDRESS Trial mortality data were available for 321 subjects (12.3%) who had a baseline APACHE II score of 25 or more, the subgroup that appeared to benefit in PROWESS; within this subgroup, 28-day mortality was 29.5% in those assigned to DAA versus 24.7% in those assigned placebo, in-hospital mortality rates were 32.3 and 32.7%, respectively [21]. Similarly, there was no reduction in mortality in the subgroup of patients within ADDRESS who had multiple organ dysfunction at baseline [21].Table 1 Results of placebo controlled randomised trials of drotrecogin alfa (activated)

hose assigned placebo, in-hospital mortality rates were 32.3 and 32.7%, respectively [21]. Similarly, there was no reduction in mortality in the subgroup of patients within ADDRESS who had multiple organ dysfunction at baseline [21].Table 1 Results of placebo controlled randomised trials of drotrecogin alfa (activated) Trial Actual/Planned recruitment % of planned recruitment % Assigned DAA who Died % Assigned Placebo who Died RR DAA vs. Placebo 95% CI for RR PROWESS All patients 1,690/2,280 74.1% 24.7% 30.8% 0.80 0.69–0.94 PROWESS APACHE II < 25 873 NA 18.8 19.0 0.99 0.75–1.30 PROWESS APACHE II > 24 817 NA 30.9% 43.7% 0.71 0.59–0.85 PROWESS single organ failure 418 NA 19.5% 21.2% 0.92 0.63–1.35 PROWESS > 1 organ failure 1,271 NA 26.5% 33.9% 0.78 0.66–0.93 ADDRESS: All patients 2,640/11,444 23.1% 18.5% 17.0% 1.09 0.92–1.28 ADDRESS APACHE II > 24 321 NA 29.5% 24.7% 1.22 0.85–1.74 ADDRESS > 1 Organ failure 862 NA 20.7% 21.9% 0.94 0.73–1.22 RESOLVE 477/6,000 8.0% 17.2% 17.5% 0.98 0.66–1.46 DAA Drotrecogin alfa (activated), RR relative risk, 95% CI = 95% confidence interval, NA not applicable Between November, 2002 and April, 2005, the RESOLVE trial randomly assigned children with sepsis-induced cardiovascular and respiratory failure to receive placebo or DAA [22]. The study was stopped after the second interim analysis when 477 participants had been enrolled when the independent DMC advised that the likelihood of demonstrating a beneficial effect of DAA was low (Table 1) [22].

ned children with sepsis-induced cardiovascular and respiratory failure to receive placebo or DAA [22]. The study was stopped after the second interim analysis when 477 participants had been enrolled when the independent DMC advised that the likelihood of demonstrating a beneficial effect of DAA was low (Table 1) [22]. Evidence of harm Whilst the efficacy of DAA is now the subject of much debate, the evidence that DAA increases the risk of clinically significant bleeding is consistent across trials (Table 2). In ENHANCE, an open label clinical evaluation where DAA was given to patients who satisfied the entry criteria for the PROWESS Study, the serious bleeding rate was 5.5%, somewhat higher than that seen in the subjects who received DAA in the PROWESS study [23]. Outside clinical trials off-label use may be common and the incidence of bleeding may be increased. In one clinical series, the incidence of bleeding was 10.9% and patients who received DAA and suffered bleeding were more likely to die than those who received DAA and did not bleed [24]; in the absence of a control group assessment of how much bleeding is attributable to the administration of DAA is not possible.Table 2 Serious bleeding rates in clinical trials of drotrecogin alfa (activated)

received DAA and suffered bleeding were more likely to die than those who received DAA and did not bleed [24]; in the absence of a control group assessment of how much bleeding is attributable to the administration of DAA is not possible.Table 2 Serious bleeding rates in clinical trials of drotrecogin alfa (activated) Study Placebo n (%) DAA n (%) P PROWESS 17(2.0) 30(3.5) 0.06 PROWESS (CNS) 1(0.1) 2(0.2) NS ADDRESS 28(2.2) 51(3.9) 0.01 ADDRESS (Day 0–6) 15(1.2) 31 (2.4) 0.02 ADDRESS (CNS) 5(0.4) 6 (0.5) 0.72 RESOLVE 16(6.8) 16 (6.7) 0.97 RESOLVE (Day 0–6) 8(3.4) 98 (3.8) 0.83 RESOLVE (CNS) 5(2.1) 11 (4.6) 0.13 ENHANCE – 155 (6.5) – Day 0–6: any serious bleeding event occurring during the DAA infusion period CNS central nervous system bleeding Expert commentary and the implications for patients Results of these studies and the split vote of the FDA advisory committee generated considerable controversy and calls for a confirmatory trial were made as early as 2002 [25]. The results of the ADDRESS and RESOLVE studies resulted in further calls for another placebo controlled trial [26–30]. The recent Cochrane review and meta-analysis also questioned whether DAA should be used to treat patients with severe sepsis and a high risk of death, describing the evidence of efficacy as “very weak” [30]. Furthermore, all three placebo-controlled trials were stopped early; the PROWESS study was stopped for efficacy, an approach that tends to overestimate the beneficial effects of treatments [31, 32].

Expert commentary and the implications for patients Results of these studies and the split vote of the FDA advisory committee generated considerable controversy and calls for a confirmatory trial were made as early as 2002 [25]. The results of the ADDRESS and RESOLVE studies resulted in further calls for another placebo controlled trial [26–30]. The recent Cochrane review and meta-analysis also questioned whether DAA should be used to treat patients with severe sepsis and a high risk of death, describing the evidence of efficacy as “very weak” [30]. Furthermore, all three placebo-controlled trials were stopped early; the PROWESS study was stopped for efficacy, an approach that tends to overestimate the beneficial effects of treatments [31, 32]. With expert opinion divided on the balance of the benefits and risks of treating patients with DAA, clinical use has varied markedly between individual hospitals and clinicians with many eligible patients not receiving the drug [33–35]. The variable use of DAA has serious implications for patients; if DAA does reduce the risk of death then many patients are being denied an effective treatment and dying unnecessarily; conversely, if DAA does not reduce the risk of death, other patients are being treated with an ineffective and expensive agent that may cause clinically significant bleeding.

plications for patients; if DAA does reduce the risk of death then many patients are being denied an effective treatment and dying unnecessarily; conversely, if DAA does not reduce the risk of death, other patients are being treated with an ineffective and expensive agent that may cause clinically significant bleeding. Can an industry-sponsored trial resolve this controversy? It seems clear that the current controversy can only be resolved by additional randomised placebo-controlled trials. As industry sponsored trials tend to report more favourable results than independent research [36, 37], and as some clinicians will be concerned by allegations that Lilly funded the Surviving Sepsis Campaign and used it to market DAA [38, 39], the most credible trial for all concerned would not be funded or conducted by Lilly. Ideally, the trial would be conducted by independent academic clinical trialists who would be responsible for the data, conduct the statistical analysis, interpret the results and write the study manuscripts. One such investigator initiated and trial of DAA, funded by the French government, is planned to run in parallel with the Lilly-sponsored PROWESS-SHOCK trial [40]; both trials will focus on patients with persistent septic shock. These two trials in combination, be they positive or negative, have great potential to settle this controversy. As the PROWESS-SHOCK trial is being sponsored by Lilly at the behest of the EMEA, it will fall short of some commentators’, and our, vision of the most credible trial. With this in mind, the steering committee has sought to develop trial processes that can give clinicians and patients confidence in the trial results; we describe below the process by which the trial was developed, some of the major decisions made regarding trial design, and plans for an independent analysis and reporting of the data.

d, the steering committee has sought to develop trial processes that can give clinicians and patients confidence in the trial results; we describe below the process by which the trial was developed, some of the major decisions made regarding trial design, and plans for an independent analysis and reporting of the data. The PROWESS-SHOCK Trial Organizational structure and role of the study sponsor (Lilly) Lilly has appointed a contract research organization (Parexel International) to be responsible for the day-to-day conduct of the study at each participating centre. At Lilly’s invitation, Drs Taylor Thompson and Marco Ranieri agreed to act as co-principal investigators for PROWESS SHOCK and co-chair an academic steering committee. Additionally, patient recruitment and protocol adherence will be scrutinized by a clinical coordinating centre; the clinical coordinating centre is located at Vanderbilt University and is commercially contracted by Lilly to provide these services. The steering committee is not privy to the terms of the contract but is satisfied that the study design prevents the clinical coordinating centre from introducing bias into the study. The statistical analysis will be performed by both Eli Lilly and an academic statistical centre and the safety of trial participants will be overseen by an independent data monitoring committee (DMC) (see Appendix 1), The members of the DMC will be reimbursed at an hourly rate ($250–$350/h) for participation in conference calls, and face to face meetings addressing data monitoring issues related to the PROWESS-Shock trial. The relationship between Lilly and the study committees is governed by a signed MoU (see text box).Memorandum of Understanding 1. The two principal investigators for the PROWESS–SHOCK study, Dr Thompson and Dr Ranieri, were invited to Co-Chair the committee by Eli Lilly and Company (Lilly). The principal investigators selected the other seven members of the steering committee (SC) in cooperation with Lilly. The principal investigators independently approved the members and had the right to veto members proposed by Lilly. Lilly representation in the SC will be limited to three ex officio non-voting members. 2. The steering committee in collaboration with medical experts from Lilly designed the protocol for PROWESS-SHOCK, after Lilly had come to an initial agreement with the Committee for Medicinal Products for Human Use (CHMP) of the European Medicines Agency (EMEA) to study patients with septic shock.

non-voting members. 2. The steering committee in collaboration with medical experts from Lilly designed the protocol for PROWESS-SHOCK, after Lilly had come to an initial agreement with the Committee for Medicinal Products for Human Use (CHMP) of the European Medicines Agency (EMEA) to study patients with septic shock. Lilly provided analyses from their sepsis database to assist protocol design and provided input regarding statistical and regulatory aspects of the study. The two principal investigators led the protocol design meetings and approved the final protocol. Lilly did not have the right to veto any steps of this process. 3. Dr. Ranieri recommended the chair of the Data Monitoring Committee (DMC) and invited Dr. Slutsky to participate. Dr. Slutsky, the two principal investigators, and Lilly recommended members of this committee. Dr. Slutsky approved the final membership of the committee and had the right to veto members proposed by Lilly. 4. The two principal investigators identified an academic statistical center that (subject to contractual agreement) will be responsible for reviewing and approving the statistical analysis plan, including all of the prospectively defined analyses for efficacy and safety. The plan will be filed with the FDA prior to the data being unblinded and at the same time it will be made freely and publicly available on an academic website. The academic statistical center will be responsible for conducting the primary analysis of the study data, and preparing the main study report for the SC to create the primary manuscript, and any additional analyses of the study data for subsequent manuscripts that in the view of the SC are material to understanding the efficacy and safety of drotrecogin alfa (activated). The academic statistical group will also be responsible for performing any post-hoc analyses requested by the steering committee. The members of the academic statistical center will have unfettered access to the full study database and to the randomisation code at the time the study randomisation code is broken. Eligible members of the academic statistical center will be co-authors on study manuscripts. Statisticians appointed by Lilly will independently conduct the same analyses and any discrepancies will be resolved to the satisfaction of both parties. All documents required for regulatory purposes will be prepared by Lilly. 5.

gible members of the academic statistical center will be co-authors on study manuscripts. Statisticians appointed by Lilly will independently conduct the same analyses and any discrepancies will be resolved to the satisfaction of both parties. All documents required for regulatory purposes will be prepared by Lilly. 5. The monthly SC meetings via teleconference and/or all face-to-face meetings will include an open session with Lilly experts and closed meetings independent of Lilly. The principal investigators are responsible for documenting any discussions that occur during the closed meetings of the SC. 6. The sponsor has pledged to make public, after publication of the main paper, the final study report for the study which will include the following components: Introduction, Full Protocol, Investigational Plan, Statistical Methods, Disposition of Patients, Protocol Violations, Efficacy Evaluations, Safety Evaluations, and Tables of baseline and on study variables as well as analyses of all outcomes, relevant Figures and Graphics, and Final Conclusions. In the two years following publication of the main paper, the academic statistical group will provide additional analyses of the full database as requested by any member of the public and approved by the SC. The SC will develop procedures for reviewing public requests, judging the merit of these requests for additional analyses, and adjudicating duplicate requests for access to the database. Lilly will not have the right to veto any requests the SC approves. The cost of any additional analysis will be paid by the person or group requesting the analysis. 7. The two principal investigators are responsible for the main publications from this study. There will be no scientific writer from Eli Lilly and Company for the principal manuscript or subsequent manuscripts approved by the SC. Appropriate co-authors from the sponsor will be included if they meet standard criteria for authorship. However, (a) the two principal investigators and the independent members of the steering committee will be responsible for final approval of the content and conclusions of the manuscripts; (b) the manuscripts will be submitted by the two principal investigators not by Lilly; (c) Lilly will not have the right to veto any steps of this process. 8.

pal investigators and the independent members of the steering committee will be responsible for final approval of the content and conclusions of the manuscripts; (b) the manuscripts will be submitted by the two principal investigators not by Lilly; (c) Lilly will not have the right to veto any steps of this process. 8. Lilly will remunerate the members of the SC and DMC or their employing institutions for time spent performing committee activities and the amounts paid to each committee member will be stated at any major presentation of study results and in all manuscripts reporting study results that are authored by the SC. Time spent in preparation of manuscripts and lecture materials will not be remunerated by Lilly. Any lecture materials regarding the conduct or results of the study will be clearly identified as either prepared and approved by the Steering Committee or prepared by Lilly independent of the steering committee. 9. Following publication of the primary manuscript SC and DMC members will have an absolute right to make public comment in any medium should they have concerns about the design, conduct or analysis of the study, or the interpretation or presentation of the study data.

ed by Lilly independent of the steering committee. 9. Following publication of the primary manuscript SC and DMC members will have an absolute right to make public comment in any medium should they have concerns about the design, conduct or analysis of the study, or the interpretation or presentation of the study data. Study design and goals PROWESS-SHOCK will be a concealed, randomised placebo-controlled trial in which patients, care-givers, data collectors, statisticians, the study steering committee, and the clinical coordinating centre will be blinded. Whilst mindful of the results of the PROWESS, ADDRESS and RESOLVE studies, and Lilly’s obligations to drug registration agencies, the primary goal of the trial is to provide clinicians with robust evidence regarding the efficacy and safety of DAA in a clearly defined and clinically important patient population. Proposed time-lines for the trial are to start recruitment in March 2008, complete recruitment in July 2010 and have 28-day mortality data available for public dissemination by the end of 2010.

Study design and goals PROWESS-SHOCK will be a concealed, randomised placebo-controlled trial in which patients, care-givers, data collectors, statisticians, the study steering committee, and the clinical coordinating centre will be blinded. Whilst mindful of the results of the PROWESS, ADDRESS and RESOLVE studies, and Lilly’s obligations to drug registration agencies, the primary goal of the trial is to provide clinicians with robust evidence regarding the efficacy and safety of DAA in a clearly defined and clinically important patient population. Proposed time-lines for the trial are to start recruitment in March 2008, complete recruitment in July 2010 and have 28-day mortality data available for public dissemination by the end of 2010. Choice and rationale for study population (Protocol Sect. 4.1) The trial must satisfy regulatory requirements but equally it must provide clinicians with a clearly identifiable patient population to which the study results, be they positive or negative, can be applied. The EMEA obliged Lilly to complete a trial in patients “with severe sepsis and documented organ failure (e.g. multiple organ dysfunction [MOD] or vasopressor dependent septic shock)”. The steering committee has elected to make persistent vasopressor-dependent septic shock (defined as the continuous requirement for a vasopressor above a minimum threshold dose for at least four hours) the key inclusion criterion as such patients are clearly identifiable in clinical practice and appeared to benefit in the PROWESS trial. They also represent the majority of patients that clinicians who use DAA treat with the drug [41]. Standard definitions will be used to identify patients with severe sepsis [42]. In addition, treatment with the vasopressor must be continued through the time required to obtain informed consent and for the study pharmacist to be ready to prepare the study drug; patients’ whose septic shock has resolved during this time will not be eligible for randomisation. Furthermore, patients must have one further clinical sign consistent with hypoperfusion (Table 3). The patients within the PROWESS study who most closely satisfied these inclusion criteria had a 28-day mortality of 29.3% for those assigned to DAA and 37.6% for those assigned placebo (RR 0.78, 95% CI 0.63–0.96; Lilly—data on file). The inclusion criteria are listed in Table 3; based on these inclusion criteria, we estimate a placebo mortality rate of 35%. Thus a study of 1,500 patients will provide 80% power to detect a relative risk reduction of 20% with an α of 0.05.Table 3 Inclusion criteria

ebo (RR 0.78, 95% CI 0.63–0.96; Lilly—data on file). The inclusion criteria are listed in Table 3; based on these inclusion criteria, we estimate a placebo mortality rate of 35%. Thus a study of 1,500 patients will provide 80% power to detect a relative risk reduction of 20% with an α of 0.05.Table 3 Inclusion criteria Inclusion criteria to obtain informed consent 1. Aged ≥ 18 years old 2. Must have an infection requiring intravenous antimicrobial therapy 3. Must meet at least two of the four systemic inflammatory response syndrome (SIRS) criteria. 4. Must have septic shock, defined as: (a) The patient must have received intravenous fluid resuscitation of ≥ 30 mL/kg administered within the time period spanning the 4 hours before and 4 hours after initiation of vasopressor therapy. (b) The patient must have a continuous requirement for at least one of the vasopressors listed below at the dose shown below for at least four hours: Norepinephrine ≥ 5 mcg/min Dopamine ≥ 10 mcg/kg/min Phenylephrine ≥ 25 mcg/min Epinephrine ≥ 5 mcg/min Vasopressin ≥ 0.03 units/min (c) The patient must meet at least 1 of the following criteria consistent with hypoperfusion during the 36 hours prior to study entry: Metabolic acidosis: base deficit ≥ 5.0 mmol/L or venous bicarbonate < 18 mmol/L or lactate ≥ 2.5 mMol/L. Urine output < 0.5 mL/kg h−1 for 1 hour or a 50% increase in creatinine from a known baseline level. Acute hepatic dysfunction: AST or ALT > 500 IU/dL or bilirubin > 2 g/dL. Inclusion criterion to proceed to randomisation 5. Patients must remain vasopressor dependent throughout the pretreatment period and through the time of randomisation with the goal of maintaining a systolic blood pressure of approximately 90 mm Hg or higher or a mean arterial pressure of 65 mm Hg or higher with reasonable attempts made to wean the patient from vasopressor support, if applicable. (Note: dopamine at doses < 5 mcg/kg/min does not fulfil the criteria for vasopressor dependency.)

goal of maintaining a systolic blood pressure of approximately 90 mm Hg or higher or a mean arterial pressure of 65 mm Hg or higher with reasonable attempts made to wean the patient from vasopressor support, if applicable. (Note: dopamine at doses < 5 mcg/kg/min does not fulfil the criteria for vasopressor dependency.) As a further safeguard to ensure as far as possible that only eligible patients enter the trial, all patients will be screened by a clinical coordinating centre prior to randomisation. The staff of the clinical coordinating centre (CCC) will review a checklist of inclusion and exclusion criteria with site investigators or coordinators and must approve randomisation. Reasons for advising against randomisation will be recorded and reviewed monthly by the academic steering committee. As should be the case in any randomised controlled trial, treatment allocation will be concealed with local investigators and CCC staff blinded to the randomisation sequence so that decisions to include or exclude patients will be made with no knowledge of the patient’s potential treatment group. To minimize the risk of a centre effect and reduce the effect of an imbalance in concomitant treatments between the two groups, randomisation will be stratified by centre. As far as is practical, the use of concomitant therapies that are likely to influence the outcome from septic shock will be captured in the case report form and reported.

f a centre effect and reduce the effect of an imbalance in concomitant treatments between the two groups, randomisation will be stratified by centre. As far as is practical, the use of concomitant therapies that are likely to influence the outcome from septic shock will be captured in the case report form and reported. To reduce the risk of recruitment bias, participating centres conducting competing trials in patients with septic shock will use a predetermined random or sequential selection procedure for deciding which trial a patient enters when a patient is eligible for more than one trial. Short term and longer term outcomes (Protocol Sect. 6.1) The primary outcome measure will be all-cause mortality in the intention-to-treat population (all patients randomised regardless of whether they received all or any study treatment) 28-days after randomisation. Survival and quality of life will be assessed for up to 180 days after randomisation. The secondary outcome measures are listed in the protocol.

all-cause mortality in the intention-to-treat population (all patients randomised regardless of whether they received all or any study treatment) 28-days after randomisation. Survival and quality of life will be assessed for up to 180 days after randomisation. The secondary outcome measures are listed in the protocol. Adaptive design to allow an increase in sample size if overall mortality is lower than expected (Protocol Sect. 8.1.1) Intensive care treatment of patients with severe sepsis and septic shock is changing and the case fatality rate is decreasing [1, 2, 43]. Against this background, an accurate prediction of the placebo group mortality is difficult. Therefore, the steering committee has made provision for the study statisticians to calculate the blinded mortality rate when approximately 750 patients have been enrolled, if the overall mortality rate is less than 30% the sample size may be increased by up to 500 subjects. Rationale for conservative efficacy stopping rule and absence of “futility” stopping rule (Protocol Sect. 8.2.8) The DMC and the steering committee separately discussed the principles for interim analyses and stopping rules; both committees were in favour of very conservative stopping rules and that the trial should not be stopped for futility. The DMC recommended the final stopping rules for the study and these were accepted by the steering committee.

g committee separately discussed the principles for interim analyses and stopping rules; both committees were in favour of very conservative stopping rules and that the trial should not be stopped for futility. The DMC recommended the final stopping rules for the study and these were accepted by the steering committee. Two interim analyses are planned; the first after approximately one-third and the second after two-thirds of the planned number of subjects have completed 28 days of follow up. The DMC will be supplied with unblinded data; these data will not be seen by the steering committee or the sponsor. For the first interim analysis, no efficacy stopping rules are planned. We reasoned that only an implausibly large benefit of DAA over placebo would stop the trial after recruitment of 500 subjects and that such an effect should be regarded with great scepticism [44]. In addition, it is highly likely that the number of deaths accrued would fall below that recommended for stopping such a trial [31]. It is clear that seemingly convincing treatment effects observed early in trials may not be sustained when trials are allowed to complete recruitment as planned [32, 44].

scepticism [44]. In addition, it is highly likely that the number of deaths accrued would fall below that recommended for stopping such a trial [31]. It is clear that seemingly convincing treatment effects observed early in trials may not be sustained when trials are allowed to complete recruitment as planned [32, 44]. As one experienced commentator stated “Decisions on early stopping (or not) need to be based on wise judgments interpreting the totality of available evidence, both in the current trial (considering primary and other efficacy outcomes and safety issues) and in other external evidence (especially from related trials). Accordingly, a statistical stopping boundary is only one useful objective component in an inevitably more challenging decision making process.” [32] Additionally, in the event that the mortality in the DAA group was lower at the first interim analysis, we consider it ethical to continue the trial as the proportion of eligible patients receiving DAA outside the trial is likely to be much lower than the 50% within the trial [24, 33]. For the second interim analysis at 1,000 patients, an efficacy stopping guideline is proposed if DAA is superior to placebo at a p value of ≤0.001 and if at least 250 deaths have occurred [31]. At this point, we reason that such an effect would be compelling and would be sufficient to convince many clinicians to change their practice.

nterim analysis at 1,000 patients, an efficacy stopping guideline is proposed if DAA is superior to placebo at a p value of ≤0.001 and if at least 250 deaths have occurred [31]. At this point, we reason that such an effect would be compelling and would be sufficient to convince many clinicians to change their practice. We decided against a formal futility stopping guideline because of the importance to determine with as much certainty as possible whether DAA is ineffective and thus discontinue its use in usual care. However, the independent DMC may recommend the trial be stopped for safety concerns at any time. A comparison of the PROWESS and PROWESS-SHOCK trials is presented in Table 4.Table 4 Comparison of PROWESS and PROWESS-SHOCK trials

We decided against a formal futility stopping guideline because of the importance to determine with as much certainty as possible whether DAA is ineffective and thus discontinue its use in usual care. However, the independent DMC may recommend the trial be stopped for safety concerns at any time. A comparison of the PROWESS and PROWESS-SHOCK trials is presented in Table 4.Table 4 Comparison of PROWESS and PROWESS-SHOCK trials PROWESS PROWESS-SHOCK Inclusion criteria Severe Sepsis (62.5% in shock) Persistent septic shock Initial fluid resuscitation ≥500 mL ≥30 mL/kg Primary end point 28-day all-cause mortality 28-day all-cause mortality Primary analysis “As Treated” conducted by Sponsor “Intention-to-treat” conducted by Independent ASC and Sponsor Selected secondary end points Changes in plasma D-dimer and Serum IL-6 28 Day mortality by protein-C class; organ failure; 90 and 180 day mortality Planned sample size 2,280 patients (stopped @ 2nd interim analysis for efficacy N = 1,690 patients) 1,500 (option to enroll 2000 if aggregate mortality @ 750 patients <30%) Efficacy stopping (By independent DMC) At both interim analyses Guidelines per O’Brien-Fleming; Lan and Demets No efficacy guidelines at 1st interim; consider stopping at 2nd interim if P ≤ 0.001 and ≥ 250 deaths overall * The “as treated” analysis included all patients who received the infusion for any length of time; ASC Academic Statistical Center

t DMC) At both interim analyses Guidelines per O’Brien-Fleming; Lan and Demets No efficacy guidelines at 1st interim; consider stopping at 2nd interim if P ≤ 0.001 and ≥ 250 deaths overall * The “as treated” analysis included all patients who received the infusion for any length of time; ASC Academic Statistical Center Safety monitoring (Protocol Sect. 6.2) Site investigators are responsible for reporting all serious adverse events and any non-serious bleeding or thrombotic events that occur within 28 days of randomisation. After 28 days, only serious adverse events that the site investigator considers to be related to the study drug, the drug delivery system, or a protocol procedure are reported. The DMC will be provided with monthly reports regarding safety data. If after reviewing these reports the DMC has concern regarding the incidence of adverse or serious adverse events in either arm of the study, the DMC can request and will be granted an unscheduled review of unblinded safety data. The DMC will also review unblinded data at all scheduled interim analyses.

ng safety data. If after reviewing these reports the DMC has concern regarding the incidence of adverse or serious adverse events in either arm of the study, the DMC can request and will be granted an unscheduled review of unblinded safety data. The DMC will also review unblinded data at all scheduled interim analyses. Ethics of a placebo-controlled trial of licensed drug in high risk population: identifying study sites with equipoise/uncertainty There are a number ethical issues and practical challenges in conducting a placebo controlled trial of a licensed drug within the current indication for use and in a vulnerable patient population. Central to their consideration is that the evidence in favour of DAA must now be considered equivocal, numerous respected commentators have called for another such trial [25–30], and currently relatively few patients with severe sepsis receive DAA [24, 33–35]. From an individual patient perspective, any patient for whom the treating clinician considers DAA to be clearly indicated or contra-indicated will be excluded from the trial, thus the trial will only recruit patients where the senior treating clinician has clinical equipoise or substantial uncertainty over the balance of the benefits and risks associated with treatment with DAA. An important criterion for the selection of trial centres is that the local clinicians are uncertain about the benefits and risks of treatment with DAA and consider it ethical to treat eligible patients with a blinded study drug that may be either DAA or placebo.

e benefits and risks associated with treatment with DAA. An important criterion for the selection of trial centres is that the local clinicians are uncertain about the benefits and risks of treatment with DAA and consider it ethical to treat eligible patients with a blinded study drug that may be either DAA or placebo. Another risk to the conduct of the trial is that clinicians may wish to treat patients who are deteriorating whilst receiving study drug with commercially available DAA; this risk will be minimized by selecting centres where clinicians are uncertain about the benefits and risks of treatment with DAA and where clinicians report that this uncertainty has resulted in minimal use of DAA. These aspects were assessed in a questionnaire as part of the participating centre evaluation, the document evaluating potential participating centres is reproduced in Appendix 2. Additionally, the physical properties of DAA mean that it is possible for caregivers to become “unblinded” to treatment allocation and this might theoretically influence the use of concomitant treatments (such as commercially available DAA) or willingness to continue active medical treatment. The use of commercially available DAA and the issuing of “DNR” orders will be monitored during the trial and reported.

me “unblinded” to treatment allocation and this might theoretically influence the use of concomitant treatments (such as commercially available DAA) or willingness to continue active medical treatment. The use of commercially available DAA and the issuing of “DNR” orders will be monitored during the trial and reported. Maximizing clinician acceptance of the PROWESS SHOCK results The controversy over the PROWESS study results has been documented elsewhere [19, 25, 26, 38]. We hope to reduce the risk of similar controversy following the current trial; we created conservative efficacy stopping guidelines to assure a more precise estimate of the overall treatment effect and reduce the concern that trials stopped early overestimate treatment effects [31, 45]. The trial will use an academic clinical coordinating centre to assist investigators in recruiting appropriate patients and in following the protocol precisely. The clinical coordinating centre will serve as a resource for study procedures which should minimize the learning curve and its potential impact on the trial results [46]. Finally, we hope that publishing the trial protocol, the memorandum of understanding governing the relationship between Lilly and the study committees and full disclosure of all prior and present relationships relevant to the trial, together with plans for independent analysis of the data, we will provide clinicians and regulators with sufficient information to allow them to fully evaluate the conduct of the trial and to interpret the results.

nd the study committees and full disclosure of all prior and present relationships relevant to the trial, together with plans for independent analysis of the data, we will provide clinicians and regulators with sufficient information to allow them to fully evaluate the conduct of the trial and to interpret the results. Summary Although the initial report of the PROWESS study suggested a significant breakthrough in the search for treatments for patients with severe sepsis, the subsequent controversy surrounding the conduct of that trial and the subsequent negative trials, have left many clinicians uncertain whether to treat their patients with DAA or not. There have been a number of calls for another placebo controlled trial some of which have specifically called for a trial run by a not-for-profit organisation [24]. Whilst Lilly will sponsor a new trial, the steering committee has put in place processes which it hopes will increase the transparency with which the trial is conducted so that the trial can provide robust evidence that will be acceptable to clinicians. This is an ethical imperative, as only by providing credible evidence and resolving this controversy can we serve the best interests of our patients. Below is the link to the electronic supplementary material. Supplemantary material (DOC 379 kb) Appendix 1 Members of the Data Monitoring Committee Arthur S. Slutsky (Chair) St. Michael’s Hospital; University of Toronto, Toronto, Canada Derek C. Angus University of Pittsburgh School of Medicine, Pittsburgh, USA Raymond J. Carroll Texas A&M University, College Station, USA Timothy W. Evans

Below is the link to the electronic supplementary material. Supplemantary material (DOC 379 kb) Appendix 1 Members of the Data Monitoring Committee Arthur S. Slutsky (Chair) St. Michael’s Hospital; University of Toronto, Toronto, Canada Derek C. Angus University of Pittsburgh School of Medicine, Pittsburgh, USA Raymond J. Carroll Texas A&M University, College Station, USA Timothy W. Evans Royal Brompton Hospital and Imperial College London, London, UK Gordon Guyatt McMaster University, Hamilton, Canada Charles Weijer, University of Western Ontario, London, Canada Appendix 2 Potential investigator questionnaire Eli Lilly and Company is sponsoring a placebo controlled study to further define the benefits and safety profile of Xigris®, drotrecogin alfa (activated), in adult patients with persistent septic shock. PAREXEL is the contract research organization for this study. PAREXEL and Lilly are in the process of selecting qualified investigators and investigative sites in select countries around the world. Please review the protocol highlights and complete and return the questionnaire if you are interested in participating in this study. Investigator selection will be completed this fall. Patient enrollment will begin in the first quarter of 2008. Study drug drotrecogin alfa (activated) or placebo. Study objectives The study objectives are to investigate efficacy (28-day all-cause mortality) and safety in adult patients with septic shock that are treated with either drotrecogin alfa (activated) or placebo. Study design Patients will be randomly (1:1) assigned to either the drotrecogin alfa (activated) or placebo treatment group.

Study objectives The study objectives are to investigate efficacy (28-day all-cause mortality) and safety in adult patients with septic shock that are treated with either drotrecogin alfa (activated) or placebo. Study design Patients will be randomly (1:1) assigned to either the drotrecogin alfa (activated) or placebo treatment group. The study will enroll adult patients who meet the following criteria:Evidence of an infection and the presence of SIRS Presence of septic shock The patient must have received ≥ 30 mL/kg intravenous fluid resuscitation The patient must have a continuous requirement for vasopressor support for at least 4 hours at a minimum dose (Norepinephrine ≥ 5 mcg/min, or similar dose of dopamine, phenylephrine, epinephrine, or vasopressin) At least one clinical sign consistent with hypoperfusion (sepsis-induced) Metabolic acidosis (base deficit ≥ 5.0 mEq/L, venous bicarbonate < 18 mEq/dL or lactate ≥ 2.5 mMol/L) Renal injury (urine output < 0.5 mL/kg/h for 1 hour or a 50% increase in creatinine) Acute hepatic dysfunction (AST or ALT > 500 IU/dL or bilirubin > 2 g/dL). Study drug infusion must begin within 24 h of septic shock onset (i.e. after initiation of vasopressor therapy) and within 36 h of any non-cardiovascular sepsis-induced organ dysfunction. If you are not the proper person to review and complete this questionnaire, kindly forward this to the appropriate person at your institution. Conflict of Interest A statement by the authors concerning their financial relationship with regard to the subject of this paper is available as Electronic Supplementary Material.

Study drug infusion must begin within 24 h of septic shock onset (i.e. after initiation of vasopressor therapy) and within 36 h of any non-cardiovascular sepsis-induced organ dysfunction. If you are not the proper person to review and complete this questionnaire, kindly forward this to the appropriate person at your institution. Conflict of Interest A statement by the authors concerning their financial relationship with regard to the subject of this paper is available as Electronic Supplementary Material. This article is discussed in the editorials available at: doi:10.1007/s00134-008-1274-6; 10.1007/s00134-008-1302-6 and 10.1007/s00134-008-1303-5. An erratum to this article can be found at http://dx.doi.org/10.1007/s00134-010-2081-4

Dear Editor-in-Chief, Dr. Petros et al. noted in their comment to our case-report about accumulation of oral antibiotics in the digestive tract [1] that, we did not estimate the incidence of this side effect. No published case-reports of selective decontamination of the digestive tract (SDD) accumulation were found and we are not aware of similar cases besides our report. Therefore, SDD accumulation seems very rare. It may remain unrecognized, but apparently an accumulation of clinical importance has not been described earlier. Dr. Petros et al. also requested us to report the composition of the SDD paste and suspension. The SDD liquid suspension is composed of polymyxin E, tobramycin-sulphate, water, methylhydroxybenzoate and amphotericin B. The latter suspension contains carboxymethylcellulose. In case 1, the SDD paste includes liquid paraffin and hypromellose besides the antibiotics. In cases 2 and 3, Orabase (Convatec) is used instead of hypromellose. Furthermore, in the ICU departments of cases 1 and 3, nasogastric tubes were manufactured from transparent polyvinyl chloride and not changed on a routine base.

Dr. Petros et al. also requested us to report the composition of the SDD paste and suspension. The SDD liquid suspension is composed of polymyxin E, tobramycin-sulphate, water, methylhydroxybenzoate and amphotericin B. The latter suspension contains carboxymethylcellulose. In case 1, the SDD paste includes liquid paraffin and hypromellose besides the antibiotics. In cases 2 and 3, Orabase (Convatec) is used instead of hypromellose. Furthermore, in the ICU departments of cases 1 and 3, nasogastric tubes were manufactured from transparent polyvinyl chloride and not changed on a routine base. The alternative diagnoses suggested by Dr. Petros et al. are highly appreciated. However, esophageal obstruction by solidification of enteral feed is very rare and seldom reported in the literature. Interactions with drugs like sucralfate may play a role [2], but in none of our cases sucralfate was administered. In our cases, SDD compounds were identified in the clots by pharmaceutical analysis in different parts of the gastro-intestinal tract. Awareness of potential accumulation of SDD in the digestive tract may lead to early recognition and prevention of this complication. Thorough cleansing of the oropharynx before applying the next dose of SDD paste might prevent this complication. However, this apparently seldom occuring complication should not prohibit application of SDD as a proven strategy to decrease nosocomial infections and mortality in critically ill patients.

his complication. Thorough cleansing of the oropharynx before applying the next dose of SDD paste might prevent this complication. However, this apparently seldom occuring complication should not prohibit application of SDD as a proven strategy to decrease nosocomial infections and mortality in critically ill patients. Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited. This reply refers to the comment available at: doi:10.1007/s00134-008-1136-2.

Introduction Non-invasive ventilation (NIV) is an important complement to invasive mechanical ventilation [1–4] in patients with acute exacerbations of chronic obstructive pulmonary disease (COPD) [5], and with severe cardiogenic pulmonary edema [6]. Problems with the commonly used interfaces for NIV application include air leakage [7, 8], patient discomfort [9] and pressure-related ulcerations of the nose [10], which can limit the duration of use and account for a large proportion of NIV failures [11]. Navalesi and co-workers [8] demonstrated that the type of interface used for NIV is important with respect to patient tolerance and the time of NIV application. A new NIV interface, the helmet, has been introduced recently and was tested in different clinical situations [12–15]. The results are promising in terms of user acceptance [13]. However, due to the large collapsible and compliant chamber encompassing the patient’s head it significantly impairs patient–ventilator synchrony with conventional pneumatic systems [16, 17] and is less effective in reducing the work of breathing than the face mask [16, 18].

promising in terms of user acceptance [13]. However, due to the large collapsible and compliant chamber encompassing the patient’s head it significantly impairs patient–ventilator synchrony with conventional pneumatic systems [16, 17] and is less effective in reducing the work of breathing than the face mask [16, 18]. New methods for neural triggering and cycling-off, using the diaphragm electrical activity (EAdi), can be used to initiate and terminate ventilatory assist in synchrony with inspiratory efforts and, hence, may overcome some of the shortcomings of conventional pneumatic triggering and cycling-off with the helmet [19]. The aim of this study was to compare synchrony between inspiratory effort and ventilator assist, as well as breathing comfort during neurally or pneumatically triggered and cycled-off non-invasive pressure support ventilation delivered with the helmet interface in healthy volunteers. Some of the data have been presented at the Congress of the European Society of Intensive Care Medicine in 2006 [20]. Materials and methods Subjects Seven healthy subjects (four female) were studied. Their mean (±SD) age, height and weight were 37 ± 5 years, 173 ± 10 cm and 67 ± 14 kg, respectively. Six subjects had prior knowledge about mechanical ventilation. The study was approved by the Research Ethics Committee of St. Michael’s Hospital in Toronto. Subjects gave their informed consent in writing.

were studied. Their mean (±SD) age, height and weight were 37 ± 5 years, 173 ± 10 cm and 67 ± 14 kg, respectively. Six subjects had prior knowledge about mechanical ventilation. The study was approved by the Research Ethics Committee of St. Michael’s Hospital in Toronto. Subjects gave their informed consent in writing. Instrumentation An 8-Fr catheter equipped with sensors for measurement of EAdi [21], and balloons for measurement of esophageal (Pes) and gastric (Pga) pressure [22] were inserted transnasally into the stomach [21, 23]. EAdi sensors were positioned at the level of the diaphragm by online supervision of recorded electrical and pressure signals [23]. Airway pressure was measured inside the helmet close to the subject’s mouth. All signals were digitized at 2 kHz and stored on a personal computer for off-line evaluation. Expiratory muscle activity was monitored by the differential EMG signal from two surface electrodes placed 30-cm apart on either side of the mid-line of the upper abdominal wall. All subjects received NIV with the helmet (Starmed Castar “R”, Mirandola, Italy) via a modified conventional ICU ventilator (Servo 300, Maquet Critical Care, Solna, Sweden).

Instrumentation An 8-Fr catheter equipped with sensors for measurement of EAdi [21], and balloons for measurement of esophageal (Pes) and gastric (Pga) pressure [22] were inserted transnasally into the stomach [21, 23]. EAdi sensors were positioned at the level of the diaphragm by online supervision of recorded electrical and pressure signals [23]. Airway pressure was measured inside the helmet close to the subject’s mouth. All signals were digitized at 2 kHz and stored on a personal computer for off-line evaluation. Expiratory muscle activity was monitored by the differential EMG signal from two surface electrodes placed 30-cm apart on either side of the mid-line of the upper abdominal wall. All subjects received NIV with the helmet (Starmed Castar “R”, Mirandola, Italy) via a modified conventional ICU ventilator (Servo 300, Maquet Critical Care, Solna, Sweden). NIV was delivered either with pneumatic trigger (Ptr) and cycling-off (Poff) or with neural trigger (Ntr) and cycling-off (Noff). Ptr and Ntr were adjusted to avoid auto-cycling, verified by performing a 10-s period of apnea (group mean value for Ptr was −1.3 ±SD 0.6 cm H2O). Conventional flow cycling-off (Poff) was 5% of peak inspiratory flow (default on Servo300). Ntr was set to trigger when the EAdi exceeded the random noise-variability. Noff occurred when EAdi fell to 60% of peak EAdi. During all runs, a PEEP of 5 cm H2O was applied.

a (group mean value for Ptr was −1.3 ±SD 0.6 cm H2O). Conventional flow cycling-off (Poff) was 5% of peak inspiratory flow (default on Servo300). Ntr was set to trigger when the EAdi exceeded the random noise-variability. Noff occurred when EAdi fell to 60% of peak EAdi. During all runs, a PEEP of 5 cm H2O was applied. Protocol All subjects were studied while seated. The maximum EAdi\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{(EA}}_{{\text{di}}_{{ \max }} } ) $$\end{document} was obtained by a maximal inspiratory maneuver [23]. Subsequently, during a practice period, subjects were instructed to follow a specific respiratory rate with a predetermined rate displayed as a time-line on a computer screen until they were able to maintain the predetermined respiratory rate (RR). Afterward the subjects repeated periods of breathing on the ventilator with RR of 10, 20, or 30 breaths per min (bpm), at pressure support (PSV) levels of 5, 10, or 20 cm H2O above PEEP (5 cm H2O). NIV was either neurally (Ntr and Noff) or pneumatically (Ptr and Poff) triggered and cycled-off. A total of 18 combinations of 2-min duration, were randomly performed using ballots. Before the 2-min period of data recording, each subject breathed until the new target respiratory rate and assist level was reached. Each measurement period was followed by 2 min of rest with CPAP of 5 cm H2O. Subjects were not informed about the applied PSV level, or the trigger type.

randomly performed using ballots. Before the 2-min period of data recording, each subject breathed until the new target respiratory rate and assist level was reached. Each measurement period was followed by 2 min of rest with CPAP of 5 cm H2O. Subjects were not informed about the applied PSV level, or the trigger type. To limit the subjects’ breathing efforts, the inspiratory effort was monitored by the investigators throughout the protocol period, and subjects were instructed to lower their inspiratory effort if EAdi exceeded 20% of the voluntary maximum EAdi. Subjects were also instructed to not use expiratory muscles (supervised by monitoring the abdominal EMG). Subjects comfort of breathing was assessed by a visual analogue scale (VAS) (0 mm = maximal comfort to 100 mm = unbearable) and marked by the subjects themselves at the end of each study period.

To limit the subjects’ breathing efforts, the inspiratory effort was monitored by the investigators throughout the protocol period, and subjects were instructed to lower their inspiratory effort if EAdi exceeded 20% of the voluntary maximum EAdi. Subjects were also instructed to not use expiratory muscles (supervised by monitoring the abdominal EMG). Subjects comfort of breathing was assessed by a visual analogue scale (VAS) (0 mm = maximal comfort to 100 mm = unbearable) and marked by the subjects themselves at the end of each study period. Off-line data analysis Ntr and Ptr delays were determined by measuring the time between the onset of EAdi and the onset of ventilatory assist using the internal signal of the ventilator. Trigger delays during Ntr NIPSV could vary (see ESM for details). Noff and Poff were determined by measuring the time between the point where EAdi was reduced to 60% of its peak value and the onset of expiratory flow. Mean EAdi, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{EA}}_{{\text{di}}_{{ \min }} } , $$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{EA}}_{{\text{di}}_{{ \max }} } $$\end{document} and the corresponding pressure time products were calculated for the unassisted period of inspiration \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{(EA}}_{{\text{di}}_{{\text{TR}}} } ,{\kern 1pt} \;{\text{PTPEA}}_{{\text{DI}}_{{\text{TR}}} } ) $$\end{document} and for the entire neural inspiration \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{(EA}}_{{\text{di}}_{{\text{insp}}} } ,{\kern 1pt} \;{\text{PTPEA}}_{{\text{di}}_{{\text{Insp}}} } ). $$\end{document} The peak EAdi was calculated for each breath and was expressed as percentage of the maximum EAdi obtained during the maximum inspiratory maneuvers.

ength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{(EA}}_{{\text{di}}_{{\text{insp}}} } ,{\kern 1pt} \;{\text{PTPEA}}_{{\text{di}}_{{\text{Insp}}} } ). $$\end{document} The peak EAdi was calculated for each breath and was expressed as percentage of the maximum EAdi obtained during the maximum inspiratory maneuvers. Mean tidal excursion of Pes (ΔPes) and the pressure time product for the Pes (PTPes) were calculated for the unassisted pre-trigger inspiratory phase \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ (P_{{\text{es}}_{{\text{tr}}} } {\text{ and PTP}}_{{\text{es}}_{{\text{tr}}} } ) $$\end{document} and for the entire neural inspiratory period \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ (P_{{\text{es}}_{{\text{tot}}} } {\text{ and PTP}}_{{\text{es}}_{{\text{tot}}} } ). $$\end{document} The number of wasted inspiratory efforts, i.e., failure to initiate PSV in the presence of a neural inspiratory effort is expressed as percentage of all neural efforts within the same time period.

Mean tidal excursion of Pes (ΔPes) and the pressure time product for the Pes (PTPes) were calculated for the unassisted pre-trigger inspiratory phase \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ (P_{{\text{es}}_{{\text{tr}}} } {\text{ and PTP}}_{{\text{es}}_{{\text{tr}}} } ) $$\end{document} and for the entire neural inspiratory period \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ (P_{{\text{es}}_{{\text{tot}}} } {\text{ and PTP}}_{{\text{es}}_{{\text{tot}}} } ). $$\end{document} The number of wasted inspiratory efforts, i.e., failure to initiate PSV in the presence of a neural inspiratory effort is expressed as percentage of all neural efforts within the same time period. For successfully triggered breaths, asynchrony was calculated over the whole breathing cycle on a breath by breath basis [24, 25] and was presented as a percentage of the duration of the neural breath (neural Ttot). Wasted efforts were counted as 100% asynchrony.

The number of wasted inspiratory efforts, i.e., failure to initiate PSV in the presence of a neural inspiratory effort is expressed as percentage of all neural efforts within the same time period. For successfully triggered breaths, asynchrony was calculated over the whole breathing cycle on a breath by breath basis [24, 25] and was presented as a percentage of the duration of the neural breath (neural Ttot). Wasted efforts were counted as 100% asynchrony. Statistical analysis Data are presented as median and 25th and 75th percentiles if not stated otherwise. Analysis was performed using non-parametric repeated measures analysis [26], corrected for multiple testing by Bonferroni correction and pair-wise comparisons (Wilcoxon test). A p value <0.05 was considered to be significant. Correlation between comfort of breathing and the percentage of asynchrony was calculated using the Spearman’s R statistic (see ESM for details).

ed measures analysis [26], corrected for multiple testing by Bonferroni correction and pair-wise comparisons (Wilcoxon test). A p value <0.05 was considered to be significant. Correlation between comfort of breathing and the percentage of asynchrony was calculated using the Spearman’s R statistic (see ESM for details). Results All subjects successfully followed the breathing instructions and reached the targeted breathing frequency, i.e., there was neither significant difference between the neural breathing frequency during Ptr and Ntr at the same PSV level nor was there a difference in the neural inspiratory times (Table 1). \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{EA}}_{{\text{di}}_{{ \min }} } $$\end{document} (Ntr: 26.0, 24.1–28.3; Ptr: 26.6, 24.8–28.8, p = 0.08) and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{EA}}_{{\text{di}}_{{ \max }} } $$\end{document} (Ntr: 64.0, 35.5–92.4; Ptr: 69.7, 47.1–96.5, p = 0.43) were comparable during Ptr and Ntr. The EAdi levels during inspiration were maintained at 9.5% (5.2–12.6) and 10.1% (6.1–16.0) of the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{EA}}_{{\text{di}}_{{ \max }} } $$\end{document} during Ntr and Ptr, respectively.

\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{EA}}_{{\text{di}}_{{ \max }} } $$\end{document} during Ntr and Ptr, respectively. There was no difference in \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{EA}}_{{\text{di}}_{{\text{insp}}} } $$\end{document} between Ntr and Ptr at PSV levels of 5 cm H2O (p = 0.74) and 10 cm H2O (p = 0.33), whereas at PSV levels of 20 cm H2O the height of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{EA}}_{{\text{di}}_{{\text{insp}}} } $$\end{document} differed significantly (p = 0.0098) Thus, overall \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{PTPEA}}_{{\text{di}}_{{\text{insp}}} } $$\end{document} was slightly higher during Ptr (Ntr: 38.6, 22.5–52.9; Ptr: 43.1, 28.5–59.9, p = 0.045) (Table 1).Table 1 Pneumatic-triggered versus neural-triggered NIV at different levels of PSV and respiratory rates

k} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{PTPEA}}_{{\text{di}}_{{\text{insp}}} } $$\end{document} was slightly higher during Ptr (Ntr: 38.6, 22.5–52.9; Ptr: 43.1, 28.5–59.9, p = 0.045) (Table 1).Table 1 Pneumatic-triggered versus neural-triggered NIV at different levels of PSV and respiratory rates There was no difference between mean measured respiratory rate (RR), neural inspiratory time, total inspiratory diaphragmatic activity \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{(EA}}_{{\text{di}}_{{\text{insp}}} } ),{\text{ EA}}_{{\text{di}}_{{ \min }} } {\text{ and EA}}_{{\text{di}}_{{ \max }} } . $$\end{document} Inspiratory delay time (delay-on) (p < 0.001), expiratory cycling-off (delay-off) (p < 0.001) \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{PTPEA}}_{{\text{di}}_{{\text{Tr}}} } $$\end{document} (p < 0.001), \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{PTPEA}}_{{\text{di}}_{{\text{insp}}} } $$\end{document} (p = 0.045) and the unassisted inspiratory effort \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{(PTP}}_{{\text{es}}_{{\text{Tr}}} } ) $$\end{document} (p < 0.001) differed significantly. Wasted inspiratory efforts (WE) (expressed in percent of neural efforts) occurred at higher respiratory rates and levels of pressure support during Ptr while there were none during Ntr. Data are presented as median values (25th and 75th percentiles)

{{\text{Tr}}} } ) $$\end{document} (p < 0.001) differed significantly. Wasted inspiratory efforts (WE) (expressed in percent of neural efforts) occurred at higher respiratory rates and levels of pressure support during Ptr while there were none during Ntr. Data are presented as median values (25th and 75th percentiles) NIV non-invasive ventilation, PSV pressure support ventilation, RR respiratory rate, bpm breaths per minute, Ti neural inspiratory time, Tot total neural inspiratory and expiratory time, EAdi electrical diaphragmatic activity, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ EA_{di_{\min } } {\text{ and }}EA_{di_{\max } } $$\end{document} minimal and maximal EAdi during inspiration, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ EA_{di_{insp} } $$\end{document} mean EAdi during inspiration, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ EA_{di_{Tr} } $$\end{document} mean EAdi during unassisted phase of inspiration, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ PTPEA_{di_{Tr} } {\text{ and }}PTPEA_{di_{insp} } $$\end{document} pressure time product of unassisted inspiratory phase and inspiration, delay-on time between onset of EAdi and the onset of ventilatory assist, delay-off time between end of neural Ti defined as reduction to 60% of peak activity and end of ventilator assist, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ PTP_{es_{Tr} } $$\end{document} pressure time product of esophageal pressure during unassisted phase of inspiration, WE wasted inspiratory efforts

ckage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ PTP_{es_{Tr} } $$\end{document} pressure time product of esophageal pressure during unassisted phase of inspiration, WE wasted inspiratory efforts Subject–ventilator asynchrony Figure 1 depicts tracings of EAdi and ventilator assist during neural and pneumatic triggering and cycling-off of NIV with the helmet in one healthy subject. Figure 2 shows differences between the delays during pneumatic and neural triggering. Ptr was delayed relative to Ntr (p < 0.001, Table 1), and the difference between the neural and pneumatic trigger delays increased (p < 0.001) as the level of PSV and RR increased (Fig. 2). During pneumatic triggering, the number of wasted inspiratory efforts increased at higher respiratory rates and with higher pressure support levels (Table 1). At a respiratory rate of 10 bpm, no wasted inspiratory efforts occurred regardless of the level of PSV, whereas at a PSV of 20 cm H2O, at a respiratory rate of 30 bpm, there was a significant increase of wasted efforts to 52.0% (29.3–58.6; p = 0.001) of all neural inspiratory efforts. No wasted inspiratory efforts occurred during neural control of NIV.Fig. 1 Example of diaphragm electrical activity (EAdi), and ventilatory assist during pneumatically (left panel) and neurally (right panel) triggered and cycled-off NIV in one healthy subject breathing with the helmet interface. In this example, respiratory rate was 30 bpm, PSV level was 20 cm H2O. The dashed line shows the start of neural inspiration and the continuous line shows the end of neural inspiration. Note excessive asynchrony during pneumatically triggered and cycled-off NIV

n one healthy subject breathing with the helmet interface. In this example, respiratory rate was 30 bpm, PSV level was 20 cm H2O. The dashed line shows the start of neural inspiration and the continuous line shows the end of neural inspiration. Note excessive asynchrony during pneumatically triggered and cycled-off NIV Fig. 2 Differences in inspiratory delays (delay time Ptr − delay time Ntr) during NIV with a pressure support of 5, 10 and 20 cm H2O and respiratory rates of 10, 20 and 30 bpm were shown. During all combinations of respiratory rates and pressure support levels, there was a significant difference in inspiratory delays being highly increased during pneumatically triggered NIV. Symbols represent group median values and the bars indicate 25th and 75th percentiles. Delay-on inspiratory delay between the onset of the volunteers’ inspiratory effort and the start of the ventilatory support, Ptr pneumatic trigger, Ntr neural trigger, NIV non-invasive ventilation

during pneumatically triggered NIV. Symbols represent group median values and the bars indicate 25th and 75th percentiles. Delay-on inspiratory delay between the onset of the volunteers’ inspiratory effort and the start of the ventilatory support, Ptr pneumatic trigger, Ntr neural trigger, NIV non-invasive ventilation The expiratory delay for all combinations of PSV levels and RR during Poff was longer than during Noff (p < 0.001) (Table 1), and the expiratory delays during Poff became longer with increasing PSV levels and target RR compared to Noff (p < 0.001) as depicted in Fig. 3.Fig. 3 Differences in expiratory delays (delay time Poff − delay time Noff) during NIV with PSV of 5, 10 and 20 cm H2O and respiratory rates of 10, 20 and 30 bpm were shown. During all combinations of respiratory rates and pressure support levels, there was a significant difference in expiratory delays that were highly increased during pneumatically triggered NIV. Symbols represent group median values and the bars indicate 25th and 75th percentiles. Delay-off delay between the onset neural end of inspiration and the end of the ventilatory support, Poff pneumatic cycling-off, Noff neural cycling-off, NIV non-invasive ventilation

ighly increased during pneumatically triggered NIV. Symbols represent group median values and the bars indicate 25th and 75th percentiles. Delay-off delay between the onset neural end of inspiration and the end of the ventilatory support, Poff pneumatic cycling-off, Noff neural cycling-off, NIV non-invasive ventilation Overall, the subject–ventilator synchrony was progressively impaired with increasing respiratory rate and levels of PSV during pneumatic triggering and cycling-off as depicted in Fig. 4, (p < 0.001). In contrast, increasing respiratory rate and levels of assists only had negligible influence on the asynchrony during neural triggering and cycling-off (Fig. 4).Fig. 4 Percentage of asynchrony over the whole breath duration as calculated during Ntr and Ptr NIV: (delay-on + delay-off/neural Ttot × 100). Wasted inspiratory efforts were counted as 100% asynchrony. Percentage of asynchrony during Ptr was markedly higher compared to Ntr. Symbols represent group median values and the bars indicate 25th and 75th percentiles. Ntr neurally triggered, Ptr pneumatically triggered, NIV non-invasive ventilation, PSV pressure support ventilation

ory efforts were counted as 100% asynchrony. Percentage of asynchrony during Ptr was markedly higher compared to Ntr. Symbols represent group median values and the bars indicate 25th and 75th percentiles. Ntr neurally triggered, Ptr pneumatically triggered, NIV non-invasive ventilation, PSV pressure support ventilation Unassisted “pre-trigger” inspiratory effort during PSV Depending on the combination between PSV and target RR, the efforts to trigger the ventilator were 5–60 times higher during Ptr than during Ntr. Median \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{PTPEA}}_{{\text{di}}_{{\text{Tr}}} } $$\end{document} during Ntr was 3.4 (2.7–4.2) compared to 16.9 (10.2–26.6) during Ptr (p < 0.001). For all combinations, the \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{PTP}}_{{\text{es}}_{{\text{tr}}} } $$\end{document} during Ptr was −264 (−470 to −130) cm H20 ms per breath compared to −11.7 (−30.6 to −6.2; p < 0.001) during Ntr (Table 1). The median \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{PTP}}_{{\text{es}}_{{\text{tr}}} } $$\end{document} expressed as percent of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{PTP}}_{{\text{es}}_{{\text{tot}}} } $$\end{document} was 21.5% (11.1–23.5) during Ptr and significantly lower during Ntr, 1.4% (0.9–1.6) (p < 0.001).

sym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{PTP}}_{{\text{es}}_{{\text{tot}}} } $$\end{document} was 21.5% (11.1–23.5) during Ptr and significantly lower during Ntr, 1.4% (0.9–1.6) (p < 0.001). Comfort of breathing Breathing comfort was lower during Ptr than during Ntr in 84.4% of all runs (p < 0.001) (Fig. 5). Overall, comfort of breathing was more than two times worse during pneumatic triggering compared to neural triggering [Ptr median 43.0 mm (20.2–75.6) vs. Ntr 19.2 mm (10.1–28.5)]. Increasing the respiratory rate resulted in minor changes of breathing comfort during neural triggering (p = 0.003), while there was a reduction in breathing comfort during pneumatic triggering (p = 0.953). Increasing PSV at a given respiratory rate tended to decrease breathing comfort during both neural triggering and pneumatic triggering, although the reduction in comfort tended to be larger with pneumatic triggering (Fig. 5). During pneumatic triggering, comfort of breathing was significantly correlated with the amount of asynchrony (r = 0.59, p < 0.001), whereas no such correlation was observed during neural triggering (r = −0.146, p = 0.253).Fig. 5 Breathing comfort in mm as assessed by a visual analogue scale during neurally and pneumatically triggered and cycled-off NIV at different levels of assist and respiratory rates (0 very comfortable, 100 unbearable). During all combinations of respiratory rates and PSV levels, comfort of breathing was better during neurally triggered and cycled-off NIV. Symbols represent group median values and the bars indicate 25th and 75th percentiles. NIV non-invasive ventilation, PSV pressure support ventilation, VAS visual analogue scale

le). During all combinations of respiratory rates and PSV levels, comfort of breathing was better during neurally triggered and cycled-off NIV. Symbols represent group median values and the bars indicate 25th and 75th percentiles. NIV non-invasive ventilation, PSV pressure support ventilation, VAS visual analogue scale Discussion The present study demonstrates that the airway pressure generated by healthy subjects using ~10% of their maximal EAdi was less efficient to pneumatically control PSV with the helmet interface at increasing breathing frequencies and pressure support levels compared to a neural trigger and cycling-off algorithm using EAdi. Due to the trigger asynchrony during Ptr, also the neural and mechanical efforts during Ptr became several times higher than during Ntr. This suggests that during NIV with pneumatic triggering, it is necessary to force the ventilator into synchrony by increasing inspiratory effort, which would defeat the purpose of providing ventilatory assist. Previous studies comparing the helmet to the face mask interface during pneumatically triggered and cycled-off PSV show that the helmet is less effective in unloading the respiratory muscles, which was partially explained by inspiratory trigger delays and the impaired pressurization rate [16–18]. The difference in trigger delays between Ntr and Ptr could not be due to mechanical response of the ventilator since NIV was delivered with the same ventilator throughout the study. Hence, apart from changes in trigger delays, no changes in raised time or pressure delivery could occur.

Previous studies comparing the helmet to the face mask interface during pneumatically triggered and cycled-off PSV show that the helmet is less effective in unloading the respiratory muscles, which was partially explained by inspiratory trigger delays and the impaired pressurization rate [16–18]. The difference in trigger delays between Ntr and Ptr could not be due to mechanical response of the ventilator since NIV was delivered with the same ventilator throughout the study. Hence, apart from changes in trigger delays, no changes in raised time or pressure delivery could occur. A critique on the present study was that a pressure trigger of −1.3 ± 0.6 cm H2O was required to avoid auto-triggering. This likely increased trigger delays and the fraction of the pressure time product necessary to pneumatically trigger the ventilator [27]. In the absence of leaks, studies suggest that flow triggering is more efficient than pressure triggering [28, 29]. However, there are also results in favor of pressure triggering [30]. The reported improvements of delays between flow and pressure triggering (40–43) are not of a magnitude that can match the improvement of implementing neural trigger relative to pressure trigger using a helmet interface in the present study. In fact, the trigger delays during Ptr were similar to those observed with helmet interface in previous studies, whereas the trigger delays during Ntr in the present study are within the range of those reported with face mask in previous studies [16–18]. Since the helmet interface is presumed to be frequently associated with leaks, which makes triggering and cycling very complex [31], it is questionable whether flow triggering is recommendable. It should be noted that in the present study, increasing breathing frequency and assist levels only increased trigger delays and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{PTP}}_{{\text{es}}_{{\text{tr}}} } $$\end{document} during Ptr and not during Ntr.

\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\text{PTP}}_{{\text{es}}_{{\text{tr}}} } $$\end{document} during Ptr and not during Ntr. With regard to wasted inspiratory efforts, which are the worst type of trigger asynchrony, a previous study did not report wasted efforts during NIV (5 cm H2O) with helmet and face mask in a study with healthy volunteers who kept their breathing frequency around 15 bpm [12]. In a lung model study using the helmet interface, wasted inspiratory efforts occurred at a respiratory rate above 20 bpm (PSV level of 18 cm H2O) and were negatively influenced by increasing PEEP and PSV levels [17]. In the present study, where both inspiratory and expiratory efforts were monitored and restricted, wasted inspiratory efforts prevailed at highest PSV (20 cm H2O) or RR (30 bpm) levels. Looking at the baseline characteristics of patients requiring non-invasive ventilatory support due to ARF respiratory rates above 30 would be expected [32–34]. Thus, compared to our results where RR did not exceed 30 bpm, the amount of wasted efforts might be even higher. With regard to Ntr, increasing frequencies above 30 bpm should not increase the risk of wasted inspiratory efforts.

vasive ventilatory support due to ARF respiratory rates above 30 would be expected [32–34]. Thus, compared to our results where RR did not exceed 30 bpm, the amount of wasted efforts might be even higher. With regard to Ntr, increasing frequencies above 30 bpm should not increase the risk of wasted inspiratory efforts. In terms of cycling-off the PSV, the ventilator used in the present study included a fixed cycling-off algorithm that terminates assist when flow has dropped to 5% of peak inspiratory flow. Given that the addition of the compliant helmet into the respiratory circuit increases the time constant of the total respiratory system (helmet, respiratory circuit and respiratory system) the late pneumatic cycling-off at 5% of peak inspiratory flow is likely not ideal. To match flow cycling-off with neural inspiratory termination, Du et al. [35] demonstrated that prolonging the time constant of the total respiratory system requires that flow cycles-off at a higher percentage of peak inspiratory flow. The need for varying flow cycling-off criteria suggested to match neural breath termination has been demonstrated in obstructive patients where the flow cycling-off had to take place at about 50% of peak inspiratory flow [36], whereas 5% appears sufficient in patients with restrictive lung disease [37]. However, since the time constant of the respiratory system (and the helmet) changes as the pressure assist level and/or breathing frequency change—and that there is currently no standard to what percentages of peak flow should be used to cycle-off at various levels of assist—the authors chose to let the cycling-off criteria be dictated by the ventilator used.

espiratory system (and the helmet) changes as the pressure assist level and/or breathing frequency change—and that there is currently no standard to what percentages of peak flow should be used to cycle-off at various levels of assist—the authors chose to let the cycling-off criteria be dictated by the ventilator used. Since the present study compared different levels of PSV and breathing frequency and that the results are presented as the difference in cycling-off delays between pneumatic and neural cycling-off at the same level of PSV and RR, our findings should be adequate in terms of comparing changes in time delays for cycling-off during Poff compared to Noff with increasing respiratory rates and PSV levels. It should be noted that the findings of increasing cycling-off delays during increasing levels of pneumatically controlled PSV observed in the present study agree with previous studies on PSV in intubated patients [27, 38]. The findings of the present study substantiate that neural off-cycling with EAdi can reduce the problem of excessive prolongation of assist into neural exhalation and its associated influence on breathing pattern [24, 38].

observed in the present study agree with previous studies on PSV in intubated patients [27, 38]. The findings of the present study substantiate that neural off-cycling with EAdi can reduce the problem of excessive prolongation of assist into neural exhalation and its associated influence on breathing pattern [24, 38]. As indicated by the asynchrony percentage as high as 30% during pneumatically triggered and cycled NIV, which was reduced to 5% during neurally triggered and cycled-off NIV, the advantage of the latter appears undisputable with regard to helmet ventilation. Our data confirm those of Racca et al. [39] that wasted efforts and impaired trigger synchrony are likely due to the properties of the helmet itself, being a collapsible device that will dampen the transmission of pressures delivered to the patient as well as reduce the ability to sense the pressure generated by the patient. Thus, the present study suggests that neural triggering can achieve an important reduction in the trigger effort with a helmet device. In the present study, breathing comfort was rated closer to maximal comfort when PSV and RR were low, which was in contrast to observations in patients with ARF. Vitacca et al. [40] reported that the highest level of comfort in patients on NIV with acute exacerbation of COPD occurred at a PSV level of 17 + 6 cm H2O and a breathing frequency of 18 + 6 bpm.

rt was rated closer to maximal comfort when PSV and RR were low, which was in contrast to observations in patients with ARF. Vitacca et al. [40] reported that the highest level of comfort in patients on NIV with acute exacerbation of COPD occurred at a PSV level of 17 + 6 cm H2O and a breathing frequency of 18 + 6 bpm. During neurally triggered and cycled-off PSV, our findings showed that comfort decreased with increasing PSV levels at similar breathing frequency despite no changes in subject–ventilator asynchrony was observed. Thus, it is reasonable to assume that in healthy subjects “no assist” equals maximal comfort and increased PSV is associated with a decrease in comfort. Hence, one should be careful in interpreting the findings of the present study in relation to a patient population with respiratory failure. However, as evidenced by the strong correlations found between asynchrony and comfort during pneumatically triggered and cycled-off NIV, which was abolished during neurally controlled NIV, subject–ventilator asynchrony plays an important role in the perception of breathing comfort. It should be noted that adding a nasogastric tube is uncomfortable. However, given the complications of endotracheal intubation, it appears reasonable to examine the possible advantage of trading tracheal for esophageal invasiveness, using non-invasive interfaces.

important role in the perception of breathing comfort. It should be noted that adding a nasogastric tube is uncomfortable. However, given the complications of endotracheal intubation, it appears reasonable to examine the possible advantage of trading tracheal for esophageal invasiveness, using non-invasive interfaces. Conclusion The present study demonstrates in healthy subjects that subject–ventilator synchrony, trigger effort, and breathing comfort with a helmet interface is considerably less impaired during increasing levels of PSV and respiratory rates with neural triggering and cycling-off, compared to conventional pneumatic triggering and cycling-off. Electronic supplementary material Below is the link to the electronic supplementary material. ESM (DOC 760 kb)

Conclusion The present study demonstrates in healthy subjects that subject–ventilator synchrony, trigger effort, and breathing comfort with a helmet interface is considerably less impaired during increasing levels of PSV and respiratory rates with neural triggering and cycling-off, compared to conventional pneumatic triggering and cycling-off. Electronic supplementary material Below is the link to the electronic supplementary material. ESM (DOC 760 kb) We are indebted to the volunteers for their participation, to Norman Comtois for his invaluable assistance, and to Orla Smith, RN and her team for the coordination given during the study. Statistical analysis of the data was performed after advisory service of the Department of Medical Statistics, University of Göttingen, Germany, with special gratitude to Dr. E. Kahler. This study was financed by departmental funds from the Department of Critical Care Medicine, St. Michael’s Hospital, University of Toronto, Canada and the Department of Anesthesiology, Emergency and Critical Care Medicine, University of Göttingen, Germany. Onnen Moerer holds a research grant from a research program, Faculty of Medicine, Georg-August-University Göttingen. Lukas Brander hold postdoctoral fellowships from the Swiss Foundation for Fellowships in Medicine and Biology (SSMBS) provided by Novartis AG and from the Division of Respirology at the University of Toronto provided by Merck-Frosst. Dr. Sinderby was financially supported by “R. Samuel McLaughlin Foundation”.

öttingen. Lukas Brander hold postdoctoral fellowships from the Swiss Foundation for Fellowships in Medicine and Biology (SSMBS) provided by Novartis AG and from the Division of Respirology at the University of Toronto provided by Merck-Frosst. Dr. Sinderby was financially supported by “R. Samuel McLaughlin Foundation”. Conflict of interest Dr. Beck and Dr. Sinderby have made inventions related to neural control of mechanical ventilation that are patented. The license for these patents belongs to Maquet Critical Care. Future commercial uses of this technology may provide financial benefit to Dr. Beck and Dr. Sinderby through royalties. Dr. Beck and Dr. Sinderby each own 50% of Neurovent Research Inc (NVR). NVR is a research and development company that builds the equipment and catheters for research studies. NVR has a consulting agreement with Maquet Critical Care. Dr. Slutsky consults for companies that make ventilators, specifically, Maquet Critical Care and Hamilton Medical, and is compensated for these consultations St. Michael's Hospital will get a small royalty if Maquet includes SMH-patented discoveries in the ventilator. Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited. This article is discussed in the editorial available at: doi:10.1007/s00134-008-1164-y. Electronic supplementary material

Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited. This article is discussed in the editorial available at: doi:10.1007/s00134-008-1164-y. Electronic supplementary material The online version of this article (doi:10.1007/s00134-008-1163-z) contains supplementary material, which is available to authorized users.

Introduction Acute renal failure (ARF) occurs in 20% of patients with severe sepsis and in the 50% of patients with septic shock [1, 2]. The combination of ARF and sepsis is associated with a 70% mortality rate as compared with a 45% mortality rate in patients with ARF alone [1, 2]. Experimental [3–5] and clinical [6] studies suggested that circulating lipopolysaccharide (LPS), a cell wall component of Gram-negative bacteria, is the key element of the sepsis-induced ARF [7]. Direct hemoperfusion through a column of immobilized polymyxin B (PMX-B) to reduce plasma levels of LPS [8–10] has been therefore proposed to improve hemodynamics and oxygenation [11] and prevent ARF [12]. Apoptosis, an energy-dependent process whereby cells carry out programmed death [13], contributes to the pathogenesis of ARF [14]. Recently, Jo et al. [4] and Bordoni et al. [15] suggested that Fas-mediated and caspase-mediated apoptosis of tubular cells might be one of the possible mechanisms involved in the endotoxemia-induced renal dysfunction. Consistent with these findings, several studies have shown that circulating LPS may cause an inappropriate activation of proapoptotic pathways in immune, epithelial, and endothelial cells [6, 16, 17]. Moreover, LPS can directly act on kidney-resident cells such as podocytes and tubular epithelium, stimulating the synthesis of inflammatory mediators [18–20].

studies have shown that circulating LPS may cause an inappropriate activation of proapoptotic pathways in immune, epithelial, and endothelial cells [6, 16, 17]. Moreover, LPS can directly act on kidney-resident cells such as podocytes and tubular epithelium, stimulating the synthesis of inflammatory mediators [18–20]. The present study was designed to test the hypothesis that extracorporeal therapy with PMX-B may prevent Gram-negative sepsis-induced ARF by reducing the activity of proapoptotic circulating factors on tubular cells and glomerular podocytes. The primary outcome variable was viability of renal cell cultures exposed to the plasma of septic patients. Secondary outcome variables were different markers of renal function. Parts of these data were published as an abstract [21]. Methods Patients admitted to the intensive care units of the Molinette (University of Turin) and Niguarda (Milan) hospitals were recruited between January 2006 and April 2007. Ethics committees approved the protocol and written informed consent was obtained from all patients. A detailed version of the methods is available in the Electronic Supplementary Material (ESM).

ts of the Molinette (University of Turin) and Niguarda (Milan) hospitals were recruited between January 2006 and April 2007. Ethics committees approved the protocol and written informed consent was obtained from all patients. A detailed version of the methods is available in the Electronic Supplementary Material (ESM). Patients and treatments Patients were eligible to participate in the study if they had a positive culture for Gram-negative bacteria and randomization could be performed within 24 h of matching study criteria. Inclusion criteria were as follows: (1) three of the systemic inflammatory response system (SIRS) and (2) presence of one organ dysfunction [22, 23]. Exclusion criteria were as follows: (1) presence of two or more failing organs; life expectancy ≤ 30 days; (2) presence of a ‘‘do not resuscitate’’ order; (3) HIV infection; (4) uncontrolled hemorrhage within 24 h before study entry; (5) organ transplantation during the year before study entry; (6) history of sensitivity to polymyxin-B or anticoagulant and/or extracorporeal circulation; (7) severe thrombocytopenia (<30,000 cells/mm3) and/or granulocytopenia (<500 cells/mm3); (8) an acute physiology and chronic health evaluation (APACHE) II score > 30 [23].

n transplantation during the year before study entry; (6) history of sensitivity to polymyxin-B or anticoagulant and/or extracorporeal circulation; (7) severe thrombocytopenia (<30,000 cells/mm3) and/or granulocytopenia (<500 cells/mm3); (8) an acute physiology and chronic health evaluation (APACHE) II score > 30 [23]. Patients were randomized to receive conventional treatments according to the Surviving Sepsis Campaign guidelines [24] (CONV group) or to undergo, in addition to standard care, a hemoperfusion session with Polymyxin-B Fiber cartridge at the moment of inclusion and after 24 h. Each hemoperfusion session lasted 2 h (PMX group). Sequential organ failure assessment (SOFA) [25] and Acute-Dialysis-Quality Initiative (ADQI)-RIFLE [scoring system that defines three grades of increasing severity of ARF (risk, injury, and failure, respectively, R, I, and F) and two outcome variables (loss and end-stage kidney disease, respectively, L and E)] [26] scores were evaluated at T0 and T72. Amount and kind of catecholamine support were also recorded at T0 and T72.

system that defines three grades of increasing severity of ARF (risk, injury, and failure, respectively, R, I, and F) and two outcome variables (loss and end-stage kidney disease, respectively, L and E)] [26] scores were evaluated at T0 and T72. Amount and kind of catecholamine support were also recorded at T0 and T72. Plasma and urine determinations Blood and urine samples were collected at the time of inclusion (T0) and 24 h (T24) and 72 h (T72) later. In the PMX group, samples were also collected at the end of each hemoperfusion (T2 and T26). Clinical variables were followed for 28 days. Blood samples were stored at −80°C. Plasma levels of LPS were evaluated by Limulus Amebocyte Lysate (LAL) kinetic test (Cambrex, Walkersville, MD). Plasma concentrations of tumor necrosis factor-α (TNF-(α) were determined by an enzyme-linked immunoabsorbent assay (ELISA; R&D Systems, Minneapolis, MN). Urine samples were collected for the evaluation of proteins and the tubular enzymes NAG (N-acetyl-glucosaminidase), GGT (gamma-glutamyl transpeptidase), and AAP (Ala-Leu-Gly aminopeptidase). All values were expressed as ratio to urine creatinine [17] to avoid the tubular concentration factor [27].

ms, Minneapolis, MN). Urine samples were collected for the evaluation of proteins and the tubular enzymes NAG (N-acetyl-glucosaminidase), GGT (gamma-glutamyl transpeptidase), and AAP (Ala-Leu-Gly aminopeptidase). All values were expressed as ratio to urine creatinine [17] to avoid the tubular concentration factor [27]. Kidney cell cytotoxicity and apoptosis Human proximal tubular epithelial cells and glomerular podocytes were isolated from the renal cortex and cultured as previously described [28]. Cells were incubated with plasmas obtained from the CONV group and the PMX group. Plasma from healthy volunteers or vehicle alone was used as controls. The cytotoxic effect of different plasmas was evaluated by a colorimetric assay based on the ability of XTT sodium salts to bind to mitochondrial dehydrogenases of viable cells [29]. Podocytes and tubular cell cultures were subjected to terminal deoxynucleotidyltransferase (TdT)-mediated dUTP nick end labeling (TUNEL) assay [29] to identify chromatin fragmentation in apoptotic cells. In a pilot study, that included five patients that matched study inclusion and exclusion criteria, we found that the collected plasma induced apoptosis in 45–50 cells per field of cultured tubular and podocytes. This trial was hence designed to enroll 16 patients to demonstrate at least a 30% reduction in proapoptotic activity, with a 5% risk of type I error and a power of 90%. Fas (CD95) was detected by immunofluorescence and by FACS analysis as previously described [4]. Bax and Bcl-2 family proteins were investigated by Western blot [29]. The activity of caspase-3, caspase-8, and caspase-9 was assessed by ELISA.

on in proapoptotic activity, with a 5% risk of type I error and a power of 90%. Fas (CD95) was detected by immunofluorescence and by FACS analysis as previously described [4]. Bax and Bcl-2 family proteins were investigated by Western blot [29]. The activity of caspase-3, caspase-8, and caspase-9 was assessed by ELISA. In selected experiments, tubular cells were seeded on six-well plates and TNF-receptor 1 (TNF-R1) short interfering RNA (siRNA) or relative control siRNA (80 pM) was introduced according to manufacturer’s instructions (Santa Cruz Biotech., Santa Cruz, CA). After 48 h, TNF-R1 knockdown was verified by RT-PCR, immunofluorescence, and Western blot analysis. Subsequently, engineered tubular cells were incubated with different plasma and subjected to TUNEL assay. Kidney cell functions Transepithelial electrical resistance (TER) was used to estimate alteration of selective cell polarity [30]. Permeability was evaluated by diffusion of Trypan blue-albumin complexes on podocyte monolayer. Adhesion to extracellular matrixes and morphogenesis assay was studied as previously described [30]. Expression of intercellular adhesion molecules-1 (ICAM-1) and CD40 on tubular epithelial cells was evaluated by cytofluorimetric analysis. Detection of ZO1, megalin, nephrin, and B7.1 was performed by immunofluorescence [31, 32].

o extracellular matrixes and morphogenesis assay was studied as previously described [30]. Expression of intercellular adhesion molecules-1 (ICAM-1) and CD40 on tubular epithelial cells was evaluated by cytofluorimetric analysis. Detection of ZO1, megalin, nephrin, and B7.1 was performed by immunofluorescence [31, 32]. Statistical analysis The data were expressed as mean ± SD or median and range when appropriate. Comparisons were performed using the unpaired t test, the Mann–Whitney, the chi-square and the Fisher exact test when appropriate. Comparison among different time points within each group was evaluated using a two-way analysis of variance for repeated measurements (ANOVA) with Newman–Keuls correction; if significant (P < 0.05), values obtained at different levels of exposure to plasma were compared with T0 using the paired t test or Wilcoxon’s signed-rank test when appropriate (Statview 5.0, CA). Results Twenty patients with Gram-negative infection and sepsis were assessed for eligibility. One was excluded for HIV infection, three for presence of two or more failing organs. Sixteen consecutive patients were enrolled, eight patients in the PMX-B group and eight in the CONV group. Some of the results are available on the ESM.

y patients with Gram-negative infection and sepsis were assessed for eligibility. One was excluded for HIV infection, three for presence of two or more failing organs. Sixteen consecutive patients were enrolled, eight patients in the PMX-B group and eight in the CONV group. Some of the results are available on the ESM. Effects on clinical and biochemical parameters On admission, all patients were treated according to the Surviving Sepsis Campaign guidelines and required catecholamine support. Patients were equally distributed for age, gender, APACHE II, sites of infection, antibiotic therapy and organ dysfunctions (Table 1). After 72 h, SOFA decreased (P = 0.02) and RIFLE improved (P = 0.04) in the PMX group, while it did not change in the CONV group; five out of eight patients were weaned from catecholamine support in the PMX group (P = 0.02), whereas in the CONV group all patients but one were still on catecholamines (Table 1). The need for renal replacement therapy occurred in one patient in the PMX group (17%) and in three patients (37%) in the CONV group (P = 0.02). In the PMX group but not in CONV group, a significant reduction in plasmatic levels of LPS and TNF-alpha was observed at T72. Table 1 Patients’ demographic, clinical, and biochemical data

r renal replacement therapy occurred in one patient in the PMX group (17%) and in three patients (37%) in the CONV group (P = 0.02). In the PMX group but not in CONV group, a significant reduction in plasmatic levels of LPS and TNF-alpha was observed at T72. Table 1 Patients’ demographic, clinical, and biochemical data PMX CONV Age (years) 61 ± 10 59 ± 13 Male N (%) 6 (75) 6 (75) APACHE II 21.2 ± 3.6 20.4 ± 4.4 Site of infection N (%) Abdominal 7 (87) 5(65) Urinary tract 1 (13) 3(35) Antibiotics N (%) Glycopeptides 4 (50) 5 (62) Aminoglycosides 3 (37) 4 (50) SOFA score T0 11 (13–7) 9 (20–7) T72 5.5 (11–2)a 10.5 (23–6)d RIFLE N (%) Risk T0 5 (62.5) 5 (62.5) T72 2 (25) 2 (25) Injury T0 2 (25) 1 (12.5) T72 0 1 (12.5) Failure T0 0 0 T72 0 1(12.5) Risk + injury + failure T0 7 (87) 6 (75) T72 2 (25)b 4 (50) Patients on catecholamine N (%) T0 8 (100) 8(100) T72 3 (37)c 7 (87) Noradrenaline (μg/kg/min) T0 0.5 (1.0–0.1) 0.45 (0.8–0.08) T72 0 (0.4–0)a 0.12 (0.8–0) LPS (pg/ml) T0 187.83 ± 21.45 162.57 ± 37.87 T72 87.74 ± 19.45a 184.63 ± 30.56 TNF-alpha (pg/ml) T0 84.53 ± 19.85 76.53 ± 16.32 T72 21.34 ± 7.98a 95.34 ± 19.64 Data are expressed as mean ± SD and median (range) PMX polymyxin-B, CONV conventional, APACHE II acute physiology and chronic health evaluation, SOFA sequential organ failure assessment, RIFLE risk, injury, failure, loss of function, end stage aWilcoxon Signed Rank test PMX: T0 versus T72, P = 0.02 bFisher Exact test PMX: T0 versus T72, P = 0.04 cFisher Exact test: PMX versus CONV at T72, P = 0.02 dMann–Whitney test: PMX T72 versus CONV T72, P = 0.02

PMX polymyxin-B, CONV conventional, APACHE II acute physiology and chronic health evaluation, SOFA sequential organ failure assessment, RIFLE risk, injury, failure, loss of function, end stage aWilcoxon Signed Rank test PMX: T0 versus T72, P = 0.02 bFisher Exact test PMX: T0 versus T72, P = 0.04 cFisher Exact test: PMX versus CONV at T72, P = 0.02 dMann–Whitney test: PMX T72 versus CONV T72, P = 0.02 Effect on cellular cytotoxicity and apoptosis Compared to incubation with vehicle alone or healthy plasma, septic plasma (concentration 5% diluted in normal culture medium) derived from both PMX and CONV groups reduced the viability of glomerular podocytes (not shown) and tubular cells, as detected by the XTT-based assay (Figure 1, A and B in ESM). In the PMX group, a significant reduction of tubular injury was observed after challenging with plasma collected at the end of both PMX-B hemoperfusion sessions (PMX T2 and PMX T26) and after 72 h of the enrollment in the study (PMX T72). By contrast, in the CONV group, plasma-induced tubular injury remained constant over time (Figure 1, A and B in ESM).

ction of tubular injury was observed after challenging with plasma collected at the end of both PMX-B hemoperfusion sessions (PMX T2 and PMX T26) and after 72 h of the enrollment in the study (PMX T72). By contrast, in the CONV group, plasma-induced tubular injury remained constant over time (Figure 1, A and B in ESM). Plasma from patients of CONV T0 and PMX T0 groups induced significant tubular cell apoptosis, as detected by TUNEL assay (Fig. 1a). While plasma-induced apoptosis remained significantly higher in the CONV T72 group, it was significantly reduced in the PMX T72 group compared to PMX T0. Furthermore, addition of 5 μg/ml PMX-B, a dose found to significantly inhibit LPS biological activity without inducing tubular cell death, to the culture significantly reduced the proapoptotic activity of PMX T0, CONV T0, and CONV T72 plasma but not of PMX T72 plasma. However, plasma from the PMX-B hemoperfusion with or without the addition of Polymyxin-B did not completely abrogate the proapoptotic effect of plasma (Figure 1,C in ESM). In all apoptosis experiments, LPS (30 ng/ml) was used as a positive control. Fig. 1 a Evaluation of tubular apoptosis (TUNEL) induced by incubation for 48 h with CONV or PMX plasmas in tubular cells. All PMX and CONV plasmas induced a significant increase of tubular apoptosis (*P < 0.05). Incubation of tubular cells with PMX T72 plasma resulted in a significant decrease of apoptosis compared to PMX T0 plasma (†P < 0.05). LPS (30 ng/mL) was used as positive control. b Evaluation of tubular apoptosis (TUNEL) in tubular cells subjected to short interfering RNA (siRNA) for tumor necrosis factor-receptor 1 (TNF-R1) or for a noncoding control after incubation with CONV and PMX plasma. Compared to control siRNA, a significant decrease of tubular apoptosis was observed in siRNA TNF-R1 tubular cells incubated with CONV and PMX plasma (*P < 0.05). LPS (30 ng/mL) was used as positive control

or tumor necrosis factor-receptor 1 (TNF-R1) or for a noncoding control after incubation with CONV and PMX plasma. Compared to control siRNA, a significant decrease of tubular apoptosis was observed in siRNA TNF-R1 tubular cells incubated with CONV and PMX plasma (*P < 0.05). LPS (30 ng/mL) was used as positive control TNF-R1 knockdown by siRNA induced a significant decrease of tubular apoptosis at T0 and T72 either in PMX or CONV group (Fig. 1b).

or tumor necrosis factor-receptor 1 (TNF-R1) or for a noncoding control after incubation with CONV and PMX plasma. Compared to control siRNA, a significant decrease of tubular apoptosis was observed in siRNA TNF-R1 tubular cells incubated with CONV and PMX plasma (*P < 0.05). LPS (30 ng/mL) was used as positive control TNF-R1 knockdown by siRNA induced a significant decrease of tubular apoptosis at T0 and T72 either in PMX or CONV group (Fig. 1b). Activities of caspase-3, caspase-8, and caspase-9 were significantly increased in tubular cells after incubation with PMX T0 and CONV T0 plasma (Fig. 2). All caspases remained activated after incubation of tubular cells with CONV T72 plasma. By contrast, we observed a significant reduction of all caspase activities after PMX T72 plasma challenge. Fig. 2 Top: Enzyme-linked immunoabsorbent (ELISA) evaluation of caspase-3, caspase-8, and caspase-9 activities on tubular cells cultured for 48 h with 5% CONV or PMX plasma. PMX T0 and CONV T0 and T72 plasmas induced a significant increase of all caspase activities (*P < 0.05 vs. healthy plasma). A significant decrease of all caspase activities was found with PMX T72 plasma compared to PMX T0 (†P < 0.05 PMX T72 vs. PMX T0); however, caspase-3 and caspase-9 activities remained significantly higher than healthy plasma (*P < 0.05 PMX T72 vs. healthy plasma). Middle: Representative images of FACS and immunofluorescence (insets) analysis of Fas (CD95) expression on tubular cell surface after exposure to CONV or PMX plasma. PMX T0 and CONV T0 and T72 plasmas all induced a marked upregulation of Fas, which was significantly reduced in presence of PMX T72 plasma. ×400 magnification. Bottom: Representative western blot analysis of the mitochondrial proteins Bax and Bcl2 in tubular cells exposed to CONV or PMX plasma, and related densitometric analysis expressed as Bax/Bcl2 ratio. PMX T0, CONV T0, and CONV T72 plasmas induced a marked upregulation of the Bax/Bcl2 ratio that was reduced in the presence of PMX T72 plasma. (Lane 1 Vehicle; Lane 2 Healthy; Lane 3 PMX T0; Lane 4 PMX T72; Lane 5 CONV T0; Lane 6 CONV T72). Beta-actin was used as reference for protein loading

xpressed as Bax/Bcl2 ratio. PMX T0, CONV T0, and CONV T72 plasmas induced a marked upregulation of the Bax/Bcl2 ratio that was reduced in the presence of PMX T72 plasma. (Lane 1 Vehicle; Lane 2 Healthy; Lane 3 PMX T0; Lane 4 PMX T72; Lane 5 CONV T0; Lane 6 CONV T72). Beta-actin was used as reference for protein loading Compared to controls, FACS and immunofluorescence demonstrated a marked upregulation of Fas expression in presence of PMX T0 and CONV T0 plasma, which remained upregulated after incubation of tubular cells with CONV T72 plasma. A significant decrease in Fas expression was observed after incubation with PMX T72 plasma (Fig. 2). Moreover, tubular cells exposed to PMX T72 plasma showed a significant decrease of the ratio between the mitochondrial proteins Bax and Bcl-2 (Fig. 2). Effects on cellular function After incubation with PMX T0, CONV T0, or CONV T72 plasmas, tubular cells exhibited significantly lower TER values in comparison to stimulation with vehicle alone or healthy plasma. By contrast, incubation with PMX T72 plasma induced a significant increase of TER levels in comparison to PMX T0, CONV T0, or CONV T72 (Figure 2, A in ESM).

ith PMX T0, CONV T0, or CONV T72 plasmas, tubular cells exhibited significantly lower TER values in comparison to stimulation with vehicle alone or healthy plasma. By contrast, incubation with PMX T72 plasma induced a significant increase of TER levels in comparison to PMX T0, CONV T0, or CONV T72 (Figure 2, A in ESM). Incubation of tubular cells with PMX T0, CONV T0, or CONV T72 plasmas resulted in a significant reduction of cell adhesion to type IV collagen-coated, fibronectin-coated, and Matrigel-coated surfaces, which was significantly abated in presence of PMX T72 plasma (Figure 2, B in ESM). Furthermore, we found that CONV plasmas inhibited morphogenesis of tubular cells cultured onto Matrigel and that PMX-B hemoperfusion restored proper tubular formation (Figure 2, C in ESM). Incubation with PMX T0 and CONV T0 plasmas induced a significant upregulation of the costimulatory molecule CD40 and of the adhesion receptor ICAM-1 on tubular cell surface, in comparison to exposure to vehicle alone or healthy plasma. The overexpression of both proteins was reduced after incubation with PMX T72 (Figure 3 in ESM).