Browse the corpus

Impaired hemodynamic function is a recognized complication of severe Plasmodium falciparum malaria (1). In African children with severe malaria (SM), clinical markers of impaired perfusion are frequent and associated with increased mortality (2–8). While parasite sequestration is the hallmark of malaria pathogenesis, at a clinical level, the production of inflammatory cytokines and acidosis have strong synergies with bacterial sepsis (9, 10). Impaired cardiac function has long been established as a component of hypoperfusion secondary to septic shock; however, the role of cardiac systolic function in SM has not been well described (11–13). The Fluid Expansion as Supportive Therapy trial demonstrated increased mortality in bolus arms compared with controls (no bolus) in African children with severe febrile illness (including a large subgroup of malaria) (14). There was no evidence to indicate that fluid overload was related to the increased mortality associated with fluid boluses. A subsequent analysis of the terminal clinical events (TCEs) showed that the major difference between bolus and control arms was a higher proportion of cardiogenic shock TCE in bolus arms (n = 123; 4.6% vs 2.6%; p = 0.008) and not due to respiratory or neurologic TCEs (15). What remains unclear is to what extent myocardial dysfunction accompanied the clinical presentation and compromised the response to fluid bolus therapy.

and control arms was a higher proportion of cardiogenic shock TCE in bolus arms (n = 123; 4.6% vs 2.6%; p = 0.008) and not due to respiratory or neurologic TCEs (15). What remains unclear is to what extent myocardial dysfunction accompanied the clinical presentation and compromised the response to fluid bolus therapy. SM is frequently complicated by anemia requiring transfusion. An in-depth understanding of cardiac function in children with SM is essential to the development of management guidelines. The purpose of this study was to assess cardiac systolic function using echocardiography and laboratory markers of myocyte injury in children with severe P. falciparum malaria on a pediatric ward typical of a low-resource setting in Africa. MATERIALS AND METHODS We conducted a prospective observational study of children presenting with severe P. falciparum malaria to Mbale Regional Referral Hospital (MRRH), Eastern Uganda, an area of high perennial malaria transmission.

SM is frequently complicated by anemia requiring transfusion. An in-depth understanding of cardiac function in children with SM is essential to the development of management guidelines. The purpose of this study was to assess cardiac systolic function using echocardiography and laboratory markers of myocyte injury in children with severe P. falciparum malaria on a pediatric ward typical of a low-resource setting in Africa. MATERIALS AND METHODS We conducted a prospective observational study of children presenting with severe P. falciparum malaria to Mbale Regional Referral Hospital (MRRH), Eastern Uganda, an area of high perennial malaria transmission. Population Children 3 months to 12 years old presenting with presence or history of fever with signs or symptoms consistent with SM as defined by modified World Health Organization criteria (16) and a positive rapid diagnostic test (RDT) for P. falciparum malaria (OpitiMAL; Diamed, Fribourg, Switzerland) were eligible for the study. Informed consent was obtained from a parent or guardian. Case definition for “SM” included a malaria positive slide or RDT plus evidence of impaired consciousness, respiratory distress, or severe anemia (hemoglobin < 5 g/dL). Cerebral malaria was defined as unrousable coma (Glasgow Coma Score < 8 [age 5–12 yr] or Blantyre Coma Score < 3 [age 0–4 yr]) lasting more than 30 minutes after a seizure and absence of other causes of coma. The study was approved by the MRRH Institutional Review Committee and the London School of Hygiene and Tropical Medicine Ethics Committee.

Population Children 3 months to 12 years old presenting with presence or history of fever with signs or symptoms consistent with SM as defined by modified World Health Organization criteria (16) and a positive rapid diagnostic test (RDT) for P. falciparum malaria (OpitiMAL; Diamed, Fribourg, Switzerland) were eligible for the study. Informed consent was obtained from a parent or guardian. Case definition for “SM” included a malaria positive slide or RDT plus evidence of impaired consciousness, respiratory distress, or severe anemia (hemoglobin < 5 g/dL). Cerebral malaria was defined as unrousable coma (Glasgow Coma Score < 8 [age 5–12 yr] or Blantyre Coma Score < 3 [age 0–4 yr]) lasting more than 30 minutes after a seizure and absence of other causes of coma. The study was approved by the MRRH Institutional Review Committee and the London School of Hygiene and Tropical Medicine Ethics Committee. Data Collection and Treatment Demographic and clinical data were collected using standardized forms at admission. Nutritional status was assessed by height, weight, and mid-upper arm circumference (MUAC). Hemoglobin (g/dL) was determined by HemoCue (Angelholm, Sweden) at admission and 24 hours. Blood glucose (Acon Labs, San Diego CA) was determined at admission and every 8 hours until the child was conscious. Lactate was determined at admission (Lactate Pro; Arkray Labs, Amsterdam, The Netherlands). Cardiac troponin I (cTnI) and brain natriuretic peptide (BNP) were batch assayed from admission samples using enzyme-linked immunosorbent assay (ELH-CTNI and ELHBNP RayBio Elisa; RayBiotech, Norcoss, GA). All children received standard treatments in accordance to Uganda national guidelines (17). Children with severe anemia were transfused with 20 mL/Kg whole blood. Fluid administration was given at the discretion of the treating clinician.

immunosorbent assay (ELH-CTNI and ELHBNP RayBio Elisa; RayBiotech, Norcoss, GA). All children received standard treatments in accordance to Uganda national guidelines (17). Children with severe anemia were transfused with 20 mL/Kg whole blood. Fluid administration was given at the discretion of the treating clinician. Echocardiography A portable Phillips CX50 Ultrasound System (Philips Healthcare, Andover, MA) equipped with an S5-1 phased array cardiac probe (1–5 MHz) was used. Sonographic measurements were obtained at admission (T0) and 24 hours (T1) by the study author (S.K.). Measurements were recorded and processed using preset software. All images were digitally recorded in Digital Imaging and Communications in Medicine loops and still frames and stored for review by a second reviewer (C.L.M.). Measurements for the estimation of left ventricular (LV) ejection fraction (EF) were made in the apical four-chamber view by obtaining three separate readings in consecutive cardiac cycles for LV end-diastolic diameter (LVEDD) and LV end-systolic diameter (LVESD). LV EF was calculated using the Simpson method: EF = ([LVEDD3 – LVESD3]/LVEDD3) %.

ation of left ventricular (LV) ejection fraction (EF) were made in the apical four-chamber view by obtaining three separate readings in consecutive cardiac cycles for LV end-diastolic diameter (LVEDD) and LV end-systolic diameter (LVESD). LV EF was calculated using the Simpson method: EF = ([LVEDD3 – LVESD3]/LVEDD3) %. Parasternal long axis views were used to obtain LV outflow tract (LVOT) diameter at mid-systole. Pulsed wave Doppler interrogation of the aortic valve with the sample gate positioned at the valve leaflets was used to obtain a velocity time integral (VTI) at the LVOT. Measurements were made for three consecutive cardiac cycles and then averaged. Stroke volume (SV) was calculated using the equation: SV = VTI × cross-sectional area of the aorta, with (cross-sectional area of the aorta = 0.785 × aortic valve diameter2). Cardiac output (SV × heart rate [HR]) was then corrected for total body surface area (TBSA), and a cardiac index (CI) was calculated as: CI = cardiac output/TBSA.

) was calculated using the equation: SV = VTI × cross-sectional area of the aorta, with (cross-sectional area of the aorta = 0.785 × aortic valve diameter2). Cardiac output (SV × heart rate [HR]) was then corrected for total body surface area (TBSA), and a cardiac index (CI) was calculated as: CI = cardiac output/TBSA. Analysis and Statistics Data were entered on Filemaker Pro (Santa Clara, CA) and exported into the Statistical Package for the Social Science (SPSS v22, Chicago, IL) for analysis. TBSA was calculated using the methodology described by Mosteller (18). Significance testing for categorical variables was carried out using chi-square testing for large samples and Fisher exact for small sample sizes. Continuous variables were not normally distributed, and medians with interquartile range (IQR) are reported with nonparametric Mann-Whitney U testing for significance. Pearson correlation coefficients for CI and clinical metrics were calculated for bivariate linear correlation using two-tailed significance testing. Standardized values for cTnI (0.1 ng/mL) and BNP (100 pg/mL) in healthy controls were used as cutoff values for normal (19–27).

ic Mann-Whitney U testing for significance. Pearson correlation coefficients for CI and clinical metrics were calculated for bivariate linear correlation using two-tailed significance testing. Standardized values for cTnI (0.1 ng/mL) and BNP (100 pg/mL) in healthy controls were used as cutoff values for normal (19–27). RESULTS A total of 104 children with SM were enrolled. Median age was 23.3 months (IQR, 8.6–33.1 mo), 57% were males. Summaries of baseline clinical and laboratory data are presented in Table 1 overall and by subgroup SM anemia (SMA n = 61) and other SM (hemoglobin > 5 g/dL) (SM n = 43). Fifteen children (14.4%) were classified as having cerebral malaria, of which seven (46.7%) also had SMA. Significant clinical differences between the two subgroups were noted (Table 1). Impaired perfusion (systolic blood pressure < 75 mm Hg in children age 3–12 mo, < 85 mm Hg in children age 1–5 yr, and < 95 mm Hg in children age > 5 yr) was present in 19% of patients. Lactate was greater than or equal to 4 mmol/dL in 54% of children on presentation. TABLE 1. Demographic, Clinical, and Laboratory Baseline Data Echocardiographic Measurements Of the initial 104 patients who underwent echocardiography at T0, 93 (89.4%) were available for repeat assessment at T1. Of 11 patients who did not receive follow-up echocardiography, six died before 24 hours, four absconded, and one was discharged (recovered) prior to follow-up. Echocardiographic metrics at T0 and T1 are summarized in Figure 1 with variables used to calculate CI and their summary statistics included for comparison.

T1. Of 11 patients who did not receive follow-up echocardiography, six died before 24 hours, four absconded, and one was discharged (recovered) prior to follow-up. Echocardiographic metrics at T0 and T1 are summarized in Figure 1 with variables used to calculate CI and their summary statistics included for comparison. Figure 1. Ejection fraction (EF) (%) and cardiac index (CI) (L/min/m2) at time-0 (admission) and time-1 (24 hr). Median and interquartile range (IQR) for echocardiographic variables in severe malaria (SM) and severe malarial anemia (SMA) groups. HR = heart rate, TBSA = total body surface area (m2), VTI = velocity time integral.

Ejection fraction (EF) (%) and cardiac index (CI) (L/min/m2) at time-0 (admission) and time-1 (24 hr). Median and interquartile range (IQR) for echocardiographic variables in severe malaria (SM) and severe malarial anemia (SMA) groups. HR = heart rate, TBSA = total body surface area (m2), VTI = velocity time integral. Myocardial Function at Admission and at 24 Hours At T0, median CI was 6.4 L/min/m2 (IQR 5.0–7.6 L/min/m2) in the overall group and elevated above the reference range of 4.5 L/min/m2 in 87 of 104 (84%) of the cohort (28). CI was significantly higher in the SMA group than in the SM group at T0, 6.89 versus 5.28 L/m/m2 (p < 0.001). The primary component of this increase was an increased VTI (a surrogate of SV) in the SMA group, suggesting increased cardiac output. At T0, median EF was 58% (IQR, 53%–62%) and similar in the SM and SMA groups (p = 1.00). This admission EF was within normal range (that being greater than 45%) in 100 of 104 of participants (96.2%). Of the four with low EF, one child had severe anemia. One patient had EF equals to 31% with evidence of global hypokenisis (hemoglobin = 6.5) and expired shortly after arrival, another (EF 40%) had high levels of cTnI (7.35 ug/mL) and BNP (386 pg/mL) and had an uneventful recovery. The other two children (EF = 37% and 43%) recovered following resuscitation with normalization of cardiac function (EF = 54%, 56%).

idence of global hypokenisis (hemoglobin = 6.5) and expired shortly after arrival, another (EF 40%) had high levels of cTnI (7.35 ug/mL) and BNP (386 pg/mL) and had an uneventful recovery. The other two children (EF = 37% and 43%) recovered following resuscitation with normalization of cardiac function (EF = 54%, 56%). Median CI in the SMA group significantly decreased (from 6.9 L/min/m2 [IQR, 6.1–7.9] at T0) to 5.6 L/min/m2 [IQR, 4.1–6.3] at T1 (p > 0.001). Median CI in the SM group remained constant at 5.3 L/min/m2 (IQR, 5.7–6.9) at T0 and 5.1 L/min/m2 (IQR, = 4.4–6.7) at T1 (p = 0.479).

idence of global hypokenisis (hemoglobin = 6.5) and expired shortly after arrival, another (EF 40%) had high levels of cTnI (7.35 ug/mL) and BNP (386 pg/mL) and had an uneventful recovery. The other two children (EF = 37% and 43%) recovered following resuscitation with normalization of cardiac function (EF = 54%, 56%). Median CI in the SMA group significantly decreased (from 6.9 L/min/m2 [IQR, 6.1–7.9] at T0) to 5.6 L/min/m2 [IQR, 4.1–6.3] at T1 (p > 0.001). Median CI in the SM group remained constant at 5.3 L/min/m2 (IQR, 5.7–6.9) at T0 and 5.1 L/min/m2 (IQR, = 4.4–6.7) at T1 (p = 0.479). Fluid and Blood Transfusions Received A total of 71 patients (68.3%) received either blood transfusion only (n = 61; 58.7%), crystalloid bolus only (n = 2; 1.9%), or both (n = 8; 7.7%). Of the 61 patients with SMA, 57 (93.4%) received blood transfusion on day 1. Of the four SMA patients who did not get blood transfusion, two died, one absconded, and one child survived receiving delayed transfusion at 72 hours when blood became available. In addition, 14 children (32.6%) with SM received blood transfusion due to a drop in hemoglobin following admission. Mean transfusion volume was 15.7 mL/kg (sd, 4.91) with no difference between groups (p = 0.987). Mean crystalloid volume over 24 hours was similar in both groups, 66.9 mL/kg (sd, 57.3) and 63.2 mL/kg (sd, 32.3) in the SM and SMA groups respectively (p = 0.914). No child received diuretics or inotropes. Clinically, we found no evidence of pulmonary edema during hospitalization. Overall six children died (5.8%), three in the SM and three in the SMA group. All six deaths occurred prior to 24-hour follow-up.

63.2 mL/kg (sd, 32.3) in the SM and SMA groups respectively (p = 0.914). No child received diuretics or inotropes. Clinically, we found no evidence of pulmonary edema during hospitalization. Overall six children died (5.8%), three in the SM and three in the SMA group. All six deaths occurred prior to 24-hour follow-up. Clinical Correlations With CI Pearson correlation coefficients for CI and key clinical and laboratory variables are presented in Table 2. The only significant correlation was with [hemoglobin]. This was a negative correlation, increase in CI for a decrease in [hemoglobin], with r equals to –0.380 (p < 0.001) Figure 2. We found no association between CI and lactate or mean arterial pressure (MAP). TABLE 2. Pearson Correlation Coefficients for Cardiac Index and Clinical Variables Figure 2. Scatter plot of cardiac index versus hemoglobin (Hb) at admission (T0). Nonparametric locally weighted scatter plot smoothing. Cardiac Biomarker Data Using a standard cTnI cutoff value of 0.1 ng/mL, 48% of patients overall (n = 50), 42% of SM patients (n = 18), and 53% of SMA patients (n = 32) had elevated levels of cTnI at admission (20, 22, 29). Using a BNP cutoff value of 100 pg/mL, 19% of all participants (n = 20), 7% of SM patients (n = 3), and 28% of SMA patients (n = 17) had elevated levels of BNP at T0 (19, 23, 24, 30–32).

verall (n = 50), 42% of SM patients (n = 18), and 53% of SMA patients (n = 32) had elevated levels of cTnI at admission (20, 22, 29). Using a BNP cutoff value of 100 pg/mL, 19% of all participants (n = 20), 7% of SM patients (n = 3), and 28% of SMA patients (n = 17) had elevated levels of BNP at T0 (19, 23, 24, 30–32). We examined clinical and laboratory correlates with cardiac biomarkers and found no correlation with age, MUAC, respiratory or HR, pulse oximetry, blood pressure, lactate, hemoglobin, CI, or EF. cTnI and BNP levels correlated well (r = 0.850; p < 0.001) (Supplemental Table 1, Supplemental Digital Content 1, http://links.lww.com/PCC/A580).

and laboratory correlates with cardiac biomarkers and found no correlation with age, MUAC, respiratory or HR, pulse oximetry, blood pressure, lactate, hemoglobin, CI, or EF. cTnI and BNP levels correlated well (r = 0.850; p < 0.001) (Supplemental Table 1, Supplemental Digital Content 1, http://links.lww.com/PCC/A580). DISCUSSION In this study, we assessed cardiac function using echocardiography and cardiac biomarkers in children presenting with severe P. falciparum malaria. The data represent the largest sample size to date describing cardiac physiology in SM in children. Our echocardiographic data demonstrate that most children (96.2%) with severe P. falciparum malaria have normal EF despite some elevation of the cardiac biomarkers. In addition, we demonstrated that CI was within normal range for children with SM but moderately elevated in cases of SMA at admission but then normalized following treatment (whole blood transfusion and antimalarials). Given that the primary difference in CI was a function of the VTI, rather than HR, which was actually higher in SM, we conclude that this observed difference in CI for patients with SMA is a function of increased SV in children with SMA. Children with SMA had significantly higher median lactate, increased respiratory rate, and lower MAP, thus lower systemic vascular resistance. Thus, decreased afterload also may account for the higher SV in the SMA compared with SM group.

e in CI for patients with SMA is a function of increased SV in children with SMA. Children with SMA had significantly higher median lactate, increased respiratory rate, and lower MAP, thus lower systemic vascular resistance. Thus, decreased afterload also may account for the higher SV in the SMA compared with SM group. Children in both the SM and SMA groups were matched for age and gender; however, there was a small difference in median MUAC, 15.8 versus 13.5 cm. Eleven of 104 children were in the “orange” zone (moderate acute malnutrition), and the remainder were in the “green” zone (normal range) for MUAC. No patients were severely malnourished. Prior data have demonstrated a preserved CI in malnutrition when adjusted for TBSA (33). As such, we believe that nutritional status did not significantly affect our results.

” zone (moderate acute malnutrition), and the remainder were in the “green” zone (normal range) for MUAC. No patients were severely malnourished. Prior data have demonstrated a preserved CI in malnutrition when adjusted for TBSA (33). As such, we believe that nutritional status did not significantly affect our results. We found mild elevation (> 100 pg/mL) of BNP levels in 19% of patients (n = 20) with SM. No patients had markedly elevated levels of BNP (range, 22–386 pg/mL). BNP values in our cohort were above the reference range for healthy children and were consistent with levels seen in critical illness (19, 21, 23, 24, 30, 32, 34, 35). Commonly used cutoff values for BNP in children with cardiac dysfunction are variable; however, the elevation in BNP seen in our cohort was below than those commonly used to identify LV dysfunction (25, 26, 36, 37). Given that very few children had echocardiographic evidence of depressed EF, and the fact that BNP levels were below than those commonly used to identify LV dysfunction, we conclude that LV failure is not a significant contributor to the observed mortality in children with SM.

identify LV dysfunction (25, 26, 36, 37). Given that very few children had echocardiographic evidence of depressed EF, and the fact that BNP levels were below than those commonly used to identify LV dysfunction, we conclude that LV failure is not a significant contributor to the observed mortality in children with SM. The frequency of cTnI elevation in our cohort was notable, 48% having elevated levels of cTnI. The level of troponin elevation was similar to that of reported values in pediatric patients with severe sepsis (29, 38, 39). In our patients, there was no correlation with either cTnI or BNP and any clinical or echocardiographic variables. We suggest that BNP and cTnI elevations, well described in both adults and children with severe sepsis, are likely due to myocardial stress in the context of critical illness and severe anemia rather than indices of a failing myocardium (21, 29, 38). As a significant proportion of children (48%) had elevated cTnI levels and only 3.8% had evidence of depressed EF, we conclude that cardiac dysfunction and myocardial injury (as evidenced by elevated troponin) are not related in a direct causal pathway in pediatric patients with SM and instead reflect a state of hypoperfusion secondary to shock and anemia. In the context of SM, cTnI elevation may also be due to RBC sequestration in the coronary microvasculature.

ction and myocardial injury (as evidenced by elevated troponin) are not related in a direct causal pathway in pediatric patients with SM and instead reflect a state of hypoperfusion secondary to shock and anemia. In the context of SM, cTnI elevation may also be due to RBC sequestration in the coronary microvasculature. Reference values of CI in healthy children indicate a median of 4.5 L/min/m2 (28). Median values for CI in our cohort were slightly elevated in children with SM and significantly elevated in children with SMA at T0 (Fig. 1). At T1, following resuscitation, CI was close to reference range in both groups. Previous studies describing cardiac function in SM have shown discordant results. Yacoub et al (11) demonstrated mild depression in cardiac function that was more pronounced in children with acidosis. In this study, SV improved after fluid bolus and normalized at discharge; however, sample size was small, n equals to 30. Murphy et al (12) found no qualitative evidence of myocardial dysfunction in children with and without severe anemia, and Mocumbi et al (40) also reported normal qualitative ventricular function in children with SM. Sample sizes in both of these studies were small, 26 and 45, and both studies used qualitative measures. Our data suggest that cardiac output in severe P. falciparum with SMA is increased and is a function in increased SV with normal to high EF. This increase in cardiac output normalized following transfusion and antimalarial therapy and is consistent with an appropriate response to anemia and shock.

qualitative measures. Our data suggest that cardiac output in severe P. falciparum with SMA is increased and is a function in increased SV with normal to high EF. This increase in cardiac output normalized following transfusion and antimalarial therapy and is consistent with an appropriate response to anemia and shock. In pediatric bacterial sepsis, distinct hemodynamic patterns characterized as either warm or cold shock have been observed with the type of bacterial pathogen determining the nature of cardiac dysfunction (41). In pediatric sepsis with refractory shock, cardiac function is often depressed and correlates with poor outcomes (42, 43). Although inflammatory mediators and cytokine release, with resultant vasodilatory shock, with or without “myocardial stunning” is well described in bacterial sepsis, the role of proinflammatory cytokines in SM are not well defined (41, 42, 44–48). As previous studies in pediatric sepsis have found increased survival when CI was between 3.3 and 6 L/min/m2, the ranged deemed to be “normal,” we conclude that children with SM have preserved cardiac function. Only two children in our cohort had a CI less than 3.3 L/min/m2 (3.0 and 3.1 L/min/m2). Both of these children had SMA and recovered without complication.

found increased survival when CI was between 3.3 and 6 L/min/m2, the ranged deemed to be “normal,” we conclude that children with SM have preserved cardiac function. Only two children in our cohort had a CI less than 3.3 L/min/m2 (3.0 and 3.1 L/min/m2). Both of these children had SMA and recovered without complication. Our data provide the first comprehensive clinical, laboratory, and echocardiographic description of cardiac physiology in SM. Our data demonstrated that children with SMA had significantly elevated CI, which normalized following blood transfusion and antimalarial therapy. We also detected a moderate correlation between [hemoglobin] and CI, such that as [hemoglobin] decreases, CI increases. We suggest that the primary physiologic driver of the observed increase in CI is likely that of severe anemia and is a normal adaptive response to anemia and shock with an increase in SV. As we were not able to definitively identify children with concomitant bacterial infection, which is likely, we cannot draw conclusions regarding the contribution of bacterial coinfection. The use of VTI to calculate CI depends on accurate and reproducible measurement of the LVOT. As such, any inaccuracies in the measurement of LVOT may be reflected in the comparison of CI at T0 and T1.

t bacterial infection, which is likely, we cannot draw conclusions regarding the contribution of bacterial coinfection. The use of VTI to calculate CI depends on accurate and reproducible measurement of the LVOT. As such, any inaccuracies in the measurement of LVOT may be reflected in the comparison of CI at T0 and T1. CONCLUSIONS Children with SM have preserved myocardial systolic function. CI is elevated in SMA and correlates inversely with [hemoglobin]. cTnI is elevated in a significant proportion of children with SM but does not correlate with myocardial dysfunction. The physiologic mechanism for the increase in CI in SMA is increased SV and is an appropriate physiologic response. Following blood transfusion and antimalarial therapy, CI in SMA returns to baseline. ACKNOWLEDGMENTS We would like to acknowledge the support of Philips Medical in providing loaned ultrasound equipment for this study. We would also like to thank the clinical nursing team at Mbale Regional Referral Hospital for their diligence and dedication. Supplementary Material *See also p. 262. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (http://journals.lww.com/pccmjournal).

ACKNOWLEDGMENTS We would like to acknowledge the support of Philips Medical in providing loaned ultrasound equipment for this study. We would also like to thank the clinical nursing team at Mbale Regional Referral Hospital for their diligence and dedication. Supplementary Material *See also p. 262. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (http://journals.lww.com/pccmjournal). The FEAST trial was supported by a grant (G0801439) from the Medical Research Council (MRC), United Kingdom provided through the (MRC) Department for International Development concordat, Centres for Global Health Research, Imperial College Wellcome Trust Center for Global Health, United Kingdom (100693/Z/12/Z). Dr. Kotlyar received funding from London School of Hygiene and Tropical Medicine (research support grant), and he received support for article research from Wellcome Trust/COAF. Dr. Maitland received support for article research from Wellcome Trust/COAF and Research Councils United Kingdom. Dr. Moore disclosed that this project was supported by a loan of ultrasound equipment from Philips Healthcare to Yale University for collection of echo data, and his institution received funding from Philips Healthcare. The remaining authors have disclosed that they do not have any potential conflicts of interest.

In the United States of America and Japan, approximately 40,000 and 9,000 children, respectively, undergo pediatric cardiac surgery, annually, and the overall mortalities are 3.1% and 2.3%, respectively; these values have significantly improved in recent decades (1, 2). However, postoperative healthcare-associated infections (HAIs) after pediatric cardiac surgery remain significant causes of morbidity and mortality (3–8). Despite the wide acceptance of guidelines and prevention bundles, the frequency of HAIs remains high, at 6.0–30.8%, after pediatric cardiac surgery (3, 5, 6, 9). Therefore, it is essential to identify and modify the risk factors for postoperative HAIs. Several studies have reported various risk factors for postoperative HAIs (3–5, 9–11). However, none of those studies evaluated the risk factors throughout the perioperative period for consecutive pediatric patients and for all four common HAIs, including bloodstream infection (BSI), surgical site infection (SSI), pneumonia, and urinary tract infection (UTI) (6, 9). Furthermore, dopamine has never been assessed as a potential risk factor for HAIs after pediatric cardiac surgery despite its association with HAIs in children with sepsis (12). Therefore, we included dopamine and other inotropes as potential risk factors for postoperative HAIs. In the present study, we aimed to identify pre-, intra-, and early postoperative risk factors for these four common HAIs after pediatric cardiac surgery, including the use of dopamine and other inotropes.

Several studies have reported various risk factors for postoperative HAIs (3–5, 9–11). However, none of those studies evaluated the risk factors throughout the perioperative period for consecutive pediatric patients and for all four common HAIs, including bloodstream infection (BSI), surgical site infection (SSI), pneumonia, and urinary tract infection (UTI) (6, 9). Furthermore, dopamine has never been assessed as a potential risk factor for HAIs after pediatric cardiac surgery despite its association with HAIs in children with sepsis (12). Therefore, we included dopamine and other inotropes as potential risk factors for postoperative HAIs. In the present study, we aimed to identify pre-, intra-, and early postoperative risk factors for these four common HAIs after pediatric cardiac surgery, including the use of dopamine and other inotropes. MATERIALS AND METHODS Design and Setting We conducted a retrospective observational study of pediatric patients admitted to the PICU at the Osaka Women’s and Children’s Hospital in Osaka, Japan, after cardiac surgery, from January 1, 2013, to December 31, 2015. The PICU had 8–12 beds, during the study period, for medical and surgical pediatric patients. Approximately 400 pediatric patients were admitted to the PICU annually, with approximately 190 of these patients being admitted after cardiac surgery. Heart transplantation was not performed at our hospital. This study was approved by the hospital’s Ethics Committee, and the need for informed consent was waived due to its retrospective nature.

MATERIALS AND METHODS Design and Setting We conducted a retrospective observational study of pediatric patients admitted to the PICU at the Osaka Women’s and Children’s Hospital in Osaka, Japan, after cardiac surgery, from January 1, 2013, to December 31, 2015. The PICU had 8–12 beds, during the study period, for medical and surgical pediatric patients. Approximately 400 pediatric patients were admitted to the PICU annually, with approximately 190 of these patients being admitted after cardiac surgery. Heart transplantation was not performed at our hospital. This study was approved by the hospital’s Ethics Committee, and the need for informed consent was waived due to its retrospective nature. Inclusion and Exclusion Criteria We included all consecutive cases of pediatric cardiac surgery performed in patients less than or equal to 18 years old admitted to the PICU. We excluded surgical procedures if they were performed in patients in the neonatal ICU. We excluded patients who underwent pacemaker implantation, aortopexy, tracheoinnominate artery fistula ligation, tumor excision, and cardiac surgery after noncardiac thoracic surgery during the same PICU stay. We excluded patients who died within 48 hours following surgery. Clinical Endpoint The primary endpoint was the occurrence of an HAI, such as BSI, SSI, pneumonia, and UTI, after pediatric cardiac surgery.

Inclusion and Exclusion Criteria We included all consecutive cases of pediatric cardiac surgery performed in patients less than or equal to 18 years old admitted to the PICU. We excluded surgical procedures if they were performed in patients in the neonatal ICU. We excluded patients who underwent pacemaker implantation, aortopexy, tracheoinnominate artery fistula ligation, tumor excision, and cardiac surgery after noncardiac thoracic surgery during the same PICU stay. We excluded patients who died within 48 hours following surgery. Clinical Endpoint The primary endpoint was the occurrence of an HAI, such as BSI, SSI, pneumonia, and UTI, after pediatric cardiac surgery. Definitions of HAIs Postoperative HAIs were defined as cases of HAI diagnosed after surgery, with no evidence that the infection was present or incubating at the time of surgery. In the present study, BSI, SSI, pneumonia, and UTI were retrospectively diagnosed, according to the 2008 definitions of the Centers for Disease Control and Prevention and National Healthcare Safety Network (CDC/NHSN) (13). All the patients were surveyed up to 48 hours after discharge from the PICU. With regard to SSI, patients were surveyed up to 30 days for superficial incisional SSI and up to 90 days for deep incisional SSI and organ/space SSI, after surgery. The HAIs were diagnosed independently by two doctors.

etwork (CDC/NHSN) (13). All the patients were surveyed up to 48 hours after discharge from the PICU. With regard to SSI, patients were surveyed up to 30 days for superficial incisional SSI and up to 90 days for deep incisional SSI and organ/space SSI, after surgery. The HAIs were diagnosed independently by two doctors. BSI included both laboratory-confirmed BSI (LCBSI) and clinical sepsis. LCBSI was defined when a pathogen, not related to another site, was cultured from greater than or equal to one blood cultures. In addition, when a pathogen was a common skin contaminant (e.g., coagulase-negative staphylococci), greater than or equal to two positive cultures were needed to define LCBSI. Clinical sepsis was defined when a patient less than or equal to 1 year old exhibited signs or symptoms of infection; a physician then instituted treatment for sepsis where the blood culture test was either not performed or indicated negative results. SSIs included incisional SSI and organ/space SSI. Incisional SSI consisted of superficial SSI, which involved only the skin and subcutaneous tissue, and deep incisional SSI, which involved the deep soft tissues. Organ/space SSIs involved organs or spaces that were exposed during surgery. Pneumonia was defined as cases in which 1) infiltrate or consolidation was observed on a chest radiograph; 2) signs or symptoms (fever, leukopenia, or leukocytosis) were detected; and 3) worsening sputum or respiratory function was observed. UTI was confirmed when a patient had at least one of the signs or symptoms and one of the test results shown below.

Pneumonia was defined as cases in which 1) infiltrate or consolidation was observed on a chest radiograph; 2) signs or symptoms (fever, leukopenia, or leukocytosis) were detected; and 3) worsening sputum or respiratory function was observed. UTI was confirmed when a patient had at least one of the signs or symptoms and one of the test results shown below. Signs or symptoms: Fever (body temperature > 38°C), urgency, frequency, dysuria, suprapubic tenderness, or costovertebral angle pain or tenderness. Test results are as follows: 1) A positive urine culture of greater than or equal to 105 colony-forming units per mL. 2) At least one of the following: 1) positive dipstick for leukocyte esterase and/or nitrite, 2) pyuria, or 3) microorganisms observed on Gram stain. Catheter-associated UTI was defined as UTI in the presence of a urinary catheter, within 48 hours.

1) A positive urine culture of greater than or equal to 105 colony-forming units per mL. 2) At least one of the following: 1) positive dipstick for leukocyte esterase and/or nitrite, 2) pyuria, or 3) microorganisms observed on Gram stain. Catheter-associated UTI was defined as UTI in the presence of a urinary catheter, within 48 hours. Prevention and Control Strategies for HAIs Patients received prophylactic antibiotic therapy with cefazolin (60 mg/kg a day), which was initiated during the induction of anesthesia and continued until 48 hours following surgery. Patients proven to have nasal colonization of methicillin-resistant Staphylococcus aureus (MRSA) received intranasal mupirocin for 3 days before surgery, prophylactic vancomycin (15 mg/kg per dose) during the operative period, and cefmetazole (60 mg/kg a day) until 48 hours after surgery. Patients who experienced delayed sternal closure (DSC) from January 2013 to December 2014 received prophylactic vancomycin (15 mg/kg per dose) and meropenem (60 mg/kg a day) until 48 hours of sternal closure, whereas those with this condition from January 2015 onwards received only cefazolin (60 mg/kg a day) until 48 hours of sternal closure. Preoperative disinfection of the skin was performed with 70% alcohol, followed by povidone iodine. A single dose of methylprednisolone (30 mg/kg) was administered to all patients who underwent cardiopulmonary bypass (CPB). The insertion of and maintenance bundles for the central venous catheter and urinary catheter and a ventilator-associated pneumonia prevention program were conducted based on the CDC/NHSN guidelines. Institutional guidelines were provided, and healthcare personnel were educated about the indication, insertion, and maintenance of the devices. For BSI prevention, trained personnel inserted central venous catheters using aseptic techniques and maximum sterile barrier precautions. For skin preparation, povidone iodine was used as alcoholic chlorhexidine antiseptic was not available. For UTI prevention, a closed drainage system for urinary catheters was implemented. For ventilator-associated pneumonia prevention, the readiness to extubate was assessed daily. Oral care was provided, and the head of the patient was elevated. We did not use a prospective surveillance program, to monitor the adherence to the guideline.

n, a closed drainage system for urinary catheters was implemented. For ventilator-associated pneumonia prevention, the readiness to extubate was assessed daily. Oral care was provided, and the head of the patient was elevated. We did not use a prospective surveillance program, to monitor the adherence to the guideline. Data Collection All the data were retrospectively collected from the medical records of patients. If a patient had undergone greater than or equal to two surgeries during a single PICU stay, only the first surgery was assessed. We recorded the number of surgeries, number of patients, and patients’ demographic data. We defined surgical complexity using the Risk Adjustment for Congenital Heart Surgery (RACHS)–1 category (14). Each RACHS-1 category included cardiac surgeries that had a similar mortality, with group 1 having the lowest risk of mortality and group 6 having the highest risk. We calculated the number of patients who underwent CPB during the surgical procedure, length of mechanical ventilation, length of postoperative PICU stay, length of postoperative hospital stay, and mortality up to 28 days after discharge from the PICU. We also identified the number of preoperative and postoperative HAIs. In addition, we identified the HAI sites, pathogens, and the day of diagnosis postoperatively. Potential Risk Factors for Postoperative HAIs The screened potential risk factors for postoperative HAIs, based on the literature and our consideration, were as follows.

Data Collection All the data were retrospectively collected from the medical records of patients. If a patient had undergone greater than or equal to two surgeries during a single PICU stay, only the first surgery was assessed. We recorded the number of surgeries, number of patients, and patients’ demographic data. We defined surgical complexity using the Risk Adjustment for Congenital Heart Surgery (RACHS)–1 category (14). Each RACHS-1 category included cardiac surgeries that had a similar mortality, with group 1 having the lowest risk of mortality and group 6 having the highest risk. We calculated the number of patients who underwent CPB during the surgical procedure, length of mechanical ventilation, length of postoperative PICU stay, length of postoperative hospital stay, and mortality up to 28 days after discharge from the PICU. We also identified the number of preoperative and postoperative HAIs. In addition, we identified the HAI sites, pathogens, and the day of diagnosis postoperatively. Potential Risk Factors for Postoperative HAIs The screened potential risk factors for postoperative HAIs, based on the literature and our consideration, were as follows. Preoperative. The preoperative potential risk factors included age less than 6 months (3–6, 9–11), sex (3–5, 10), genetic abnormality (3–5, 10), absence of splenic function (10), preoperative PICU stay (3, 4, 9, 10), and nasal colonization of MRSA (15, 16). Race was not included because Japan is almost a racially homogeneous nation.

The preoperative potential risk factors included age less than 6 months (3–6, 9–11), sex (3–5, 10), genetic abnormality (3–5, 10), absence of splenic function (10), preoperative PICU stay (3, 4, 9, 10), and nasal colonization of MRSA (15, 16). Race was not included because Japan is almost a racially homogeneous nation. Intraoperative. The intraoperative potential risk factors included RACHS-1 score greater than or equal to 3 (4, 5, 9, 11), American Society of Anesthesiologists score greater than or equal to 3 (10), surgery duration greater than 3 hours (3, 9, 10), CPB (3–5, 9, 10), and lowest core temperature during surgery less than 32°C (3, 9, 10).

he intraoperative potential risk factors included RACHS-1 score greater than or equal to 3 (4, 5, 9, 11), American Society of Anesthesiologists score greater than or equal to 3 (10), surgery duration greater than 3 hours (3, 9, 10), CPB (3–5, 9, 10), and lowest core temperature during surgery less than 32°C (3, 9, 10). Postoperative. The postoperative potential risk factors included DSC (3, 5, 6, 9–11), extracorporeal membrane oxygenation (ECMO) (3, 5), mechanical ventilation for greater than or equal to 3 days (3, 5, 6, 9, 10), PICU stay for greater than or equal to 3 days (6, 9, 11), peritoneal dialysis or continuous hemodiafiltration (PD/CHDF), RBC transfusion within 2 days of surgery (3, 9), use of postoperative steroids (3, 10), peak glucose greater than 360 mg/dL within 2 days of surgery (9, 10), multiple surgeries during a single PICU stay (4), and the use of dopamine (12) and other inotropes including epinephrine, norepinephrine, and dobutamine within 3 days of surgery. ECMO and PD/CHDF were considered when they were performed within 3 days of surgery. Postoperative steroids, including hydrocortisone, prednisolone, and dexamethasone, were considered when they were administered within 3 days of surgery. Multiple surgeries during a single PICU stay were defined when greater than or equal to two separate cardiovascular surgeries were performed during the same PICU stay, prior to the development of HAIs; these did not include surgery to stop bleeding or surgery for sternal closure.

administered within 3 days of surgery. Multiple surgeries during a single PICU stay were defined when greater than or equal to two separate cardiovascular surgeries were performed during the same PICU stay, prior to the development of HAIs; these did not include surgery to stop bleeding or surgery for sternal closure. Effects of Dopamine To examine the effects of dopamine on the frequency of postoperative HAIs, we stratified patients into two groups: with dopamine and without dopamine, in which all the other risk factors (except for dopamine use) were adjusted through propensity score matching. Then, we compared the frequency of postoperative HAIs, with and without dopamine. In addition, we examined the dose-dependent effect of dopamine on postoperative HAIs. First, we stratified patients into three groups: patients who did not receive dopamine, patients who received dopamine for 3 days or less, and patients who received dopamine for more than 3 days. Then, we compared the frequency of postoperative HAIs among these groups. Second, we stratified patients depending on the total amount of dopamine (mg) used per kg of body weight within 3 days of surgery, into four groups: total amount of dopamine ≤ 5 mg/kg, > 5 and ≤ 10 mg/kg, > 10 and ≤ 15 mg/kg, and > 15 mg/kg. We compared the frequency of postoperative HAIs among these groups. Furthermore, to differentiate the dopamine effect from the inotrope effect, we added a maximum Vasoactive-Inotrope Score (17, 18) during the first 3 days following surgery into the additional multivariable analysis, for postoperative HAIs.

> 15 mg/kg. We compared the frequency of postoperative HAIs among these groups. Furthermore, to differentiate the dopamine effect from the inotrope effect, we added a maximum Vasoactive-Inotrope Score (17, 18) during the first 3 days following surgery into the additional multivariable analysis, for postoperative HAIs. Statistical Methods Categorical variables were evaluated using the chi-square test or Fisher exact test, as appropriate. Continuous variables were evaluated using the Wilcoxon rank-sum test. Statistical significance was defined as p value of less than or equal to 0.05. The individual potential risk factors for HAIs from the literature and our consideration were assessed via bivariate analysis. The significant risk factors (p < 0.1) in the bivariate analysis were included in multivariable logistic regression analysis. The risk factors highly correlated in the bivariate analysis (Cramer’s V > 0.7) were not included in the multivariable analysis together. Propensity score matching method was used to adjust the risk factors for dopamine use. The statistical analyses were conducted using JMP Version 10.0 (SAS Institute, Cary, NC) and Stata Version 14.0 (Stata Corp LLC, College Station, TX).

lysis (Cramer’s V > 0.7) were not included in the multivariable analysis together. Propensity score matching method was used to adjust the risk factors for dopamine use. The statistical analyses were conducted using JMP Version 10.0 (SAS Institute, Cary, NC) and Stata Version 14.0 (Stata Corp LLC, College Station, TX). RESULTS Study Surgical Procedures and Population During the study period, 571 cardiac surgeries were performed, and 526 surgeries (394 patients) were eligible for inclusion. A total of 45 surgeries were excluded, such as those performed on patients greater than 18 years old (n = 6) and those performed in the neonatal ICU (n = 6). Patients undergoing pacemaker implantation (n = 11), aortopexy (n = 1), tracheoinnominate artery fistula ligation (n = 1), and cardiac tumor excision (n = 1), as well as those with a history of repair of esophageal atresia (n = 1) were also excluded. Eleven, two, and one patient/s underwent two, three, and four surgeries, respectively, during a single PICU stay. We assessed only the first surgery during a single PICU stay, and hence, the second, third, and fourth surgeries were excluded (n = 18). None of the patients died within 48 hours following surgery. The number of patients in each RACHS-1 score category was as follows: 57 (11%) in category 1, 175 (33%) in category 2, 238 (45%) in category 3, 31 (6%) in category 4, none in category 5, and 25 (5%) in category 6. A total of 76.6% surgeries were performed with the patient under CPB (n = 403). Table 1 shows the patient characteristics and clinical outcomes, based on the presence of HAIs.

57 (11%) in category 1, 175 (33%) in category 2, 238 (45%) in category 3, 31 (6%) in category 4, none in category 5, and 25 (5%) in category 6. A total of 76.6% surgeries were performed with the patient under CPB (n = 403). Table 1 shows the patient characteristics and clinical outcomes, based on the presence of HAIs. TABLE 1. Patient Characteristics and Clinical Outcomes Based on the Presence of Healthcare-Associated Infections HAIs and Causative Pathogens In the cases of 526 surgeries, 81 postoperative HAIs were identified. Table 2 shows the number of postoperative HAIs, causative pathogens, and postoperative days on which HAIs were diagnosed. Of all the eligible surgeries, one surgery led to three postoperative HAIs during a single PICU stay, whereas eight surgeries led to two postoperative HAIs. Therefore, 71 surgeries had at least one postoperative HAI. The overall frequency of postoperative HAIs was 13.5 per 100 surgeries. TABLE 2. Number of Postoperative Healthcare-Associated Infections, Causative Pathogens, and Postoperative Day on Which the Healthcare-Associated Infection Was Diagnosed Seven HAIs, including UTIs (n = 4), BSIs (n = 2), and pneumonia (n = 1), presented before the surgical procedure. Two patients had both preoperative and postoperative HAIs; however, the causative pathogens and sites of infection were not the same between the preoperative and postoperative HAIs.

TABLE 2. Number of Postoperative Healthcare-Associated Infections, Causative Pathogens, and Postoperative Day on Which the Healthcare-Associated Infection Was Diagnosed Seven HAIs, including UTIs (n = 4), BSIs (n = 2), and pneumonia (n = 1), presented before the surgical procedure. Two patients had both preoperative and postoperative HAIs; however, the causative pathogens and sites of infection were not the same between the preoperative and postoperative HAIs. Risk Factors for Postoperative HAIs Table 3 shows the potential risk factors for postoperative HAIs in the bivariate analysis. The potential risk factors that were significant in the bivariate analysis (p < 0.1) were assessed for their correlation with each other, and none had a high correlation (Cramer’s V > 0.7) with another. Therefore, all the potential risk factors that were significant in the bivariate analysis were included in the multivariable analysis. The potential risk factors for postoperative HAIs included in the final multivariate analysis are shown in Table 3. Table 4 shows the risk factors for postoperative HAIs in the multivariable analysis. The significant risk factors for postoperative HAIs included mechanical ventilation greater than or equal to 3 days, dopamine use, genetic abnormality, and DSC. TABLE 3. Potential Risk Factors for Postoperative Healthcare-Associated Infections in Bivariate Analysis TABLE 4. Risk Factors for Postoperative Healthcare-Associated Infections in Multivariable Analysis

Risk Factors for Postoperative HAIs Table 3 shows the potential risk factors for postoperative HAIs in the bivariate analysis. The potential risk factors that were significant in the bivariate analysis (p < 0.1) were assessed for their correlation with each other, and none had a high correlation (Cramer’s V > 0.7) with another. Therefore, all the potential risk factors that were significant in the bivariate analysis were included in the multivariable analysis. The potential risk factors for postoperative HAIs included in the final multivariate analysis are shown in Table 3. Table 4 shows the risk factors for postoperative HAIs in the multivariable analysis. The significant risk factors for postoperative HAIs included mechanical ventilation greater than or equal to 3 days, dopamine use, genetic abnormality, and DSC. TABLE 3. Potential Risk Factors for Postoperative Healthcare-Associated Infections in Bivariate Analysis TABLE 4. Risk Factors for Postoperative Healthcare-Associated Infections in Multivariable Analysis Effects of Dopamine After adjusting the 23 potential risk factors included in Table 3 (except for the use of dopamine), through propensity score matching, 200 patients with dopamine and 200 controls without dopamine were compared, with regard to the occurrence rates of HAIs. The frequencies of postoperative HAIs, among patients with and without dopamine, were 57 of 200 (29%) and 12 of 200 (6%) (odds ratio, 6.24; 95% CI; 3.23–12.1; p < 0.001).

h propensity score matching, 200 patients with dopamine and 200 controls without dopamine were compared, with regard to the occurrence rates of HAIs. The frequencies of postoperative HAIs, among patients with and without dopamine, were 57 of 200 (29%) and 12 of 200 (6%) (odds ratio, 6.24; 95% CI; 3.23–12.1; p < 0.001). With regard to the dose-dependent effect, the frequencies of postoperative HAIs were 4.3% (14/326), 16.3% (13/80), and 36.7% (44/120), respectively, among patients who did not receive dopamine, patients who received dopamine for 3 days or less, and patients who received dopamine for more than 3 days (p < 0.001). In addition, the frequencies of postoperative HAIs were 15% (5/34), 17% (9/54), 38% (20/52), and 38% (23/60), respectively, among patients with a total amount of dopamine ≤ 5 mg/kg, > 5 and ≤ 10 mg/kg, > 10 and ≤ 15 mg/kg, and > 15 mg/kg (p = 0.007). With regard to the vasoactive inotropic effect, in the additional multivariable analysis which included a maximum Vasoactive-Inotrope Score during the first 3 days of surgery, dopamine use was a risk factor for postoperative HAIs (p= 0.005), whereas the use of inotrope was not (p = 0.76).

With regard to the dose-dependent effect, the frequencies of postoperative HAIs were 4.3% (14/326), 16.3% (13/80), and 36.7% (44/120), respectively, among patients who did not receive dopamine, patients who received dopamine for 3 days or less, and patients who received dopamine for more than 3 days (p < 0.001). In addition, the frequencies of postoperative HAIs were 15% (5/34), 17% (9/54), 38% (20/52), and 38% (23/60), respectively, among patients with a total amount of dopamine ≤ 5 mg/kg, > 5 and ≤ 10 mg/kg, > 10 and ≤ 15 mg/kg, and > 15 mg/kg (p = 0.007). With regard to the vasoactive inotropic effect, in the additional multivariable analysis which included a maximum Vasoactive-Inotrope Score during the first 3 days of surgery, dopamine use was a risk factor for postoperative HAIs (p= 0.005), whereas the use of inotrope was not (p = 0.76). DISCUSSION In this retrospective study, we demonstrated that mechanical ventilation greater than or equal to 3 days, dopamine use, genetic abnormality, and DSC were independently associated with the four common postoperative HAIs. To the best of our knowledge, this is the first study to examine the potential risk factors during the pre-, intra-, and postoperative periods, for these common HAIs, and to demonstrate the association between dopamine use and frequencies of HAI after pediatric cardiac surgery.

with the four common postoperative HAIs. To the best of our knowledge, this is the first study to examine the potential risk factors during the pre-, intra-, and postoperative periods, for these common HAIs, and to demonstrate the association between dopamine use and frequencies of HAI after pediatric cardiac surgery. The role of dopamine in critically ill patients has been questioned in various clinical settings (19, 20). Dopamine can modulate immune responses and may result in the inhibition of cytokine and chemokine production, inhibition of neutrophil chemotaxis, and disturbance of T-cell proliferation (21). In animal studies, bromocriptine—a dopamine receptor agonist that inhibits pituitary prolactin secretion—was found to suppress T-lymphocyte–dependent macrophage activation (22); similarly, it was also found to suppress the activity of T lymphocytes (23). Furthermore, dopamine suppressed prolactin levels and the T-cell response, in human studies (24, 25). Despite these theoretical disadvantages of immunosuppression, dopamine is still widely used in various clinical settings, including pediatric cardiac surgery (26–29). Recently, Ventura et al (12) compared the use of dopamine and epinephrine for pediatric septic shock in a randomized controlled trial and found that the rate of HAIs was higher in patients treated with dopamine than in those treated with epinephrine. However, no studies have investigated the association between dopamine use and postoperative HAIs in pediatric patients after cardiac surgery. In the present study, we identified dopamine use as an independent risk factor for HAIs after pediatric cardiac surgery; in addition, this factor is modifiable.

ith epinephrine. However, no studies have investigated the association between dopamine use and postoperative HAIs in pediatric patients after cardiac surgery. In the present study, we identified dopamine use as an independent risk factor for HAIs after pediatric cardiac surgery; in addition, this factor is modifiable. Dopamine was administered at the discretion of the treating providers; therefore, we adjusted for the factors associated with the use of dopamine through propensity score matching. Significant differences in the frequencies of postoperative HAIs between patients with and without dopamine remained, after adjusting the patient-level differences associated with dopamine use. In addition, we examined the dose-dependent effect of dopamine on postoperative HAIs. The duration of dopamine administration and the total amount of dopamine per kg of body weight were associated with postoperative HAIs in the bivariate analysis, and the presence of dose-dependent effects of dopamine on postoperative HAIs was noted. Furthermore, to differentiate the dopamine and inotrope effects, we added a maximum Vasoactive-Inotrope Score and found that while dopamine use remained a risk factor for postoperative HAIs, the use of inotrope did not.

te analysis, and the presence of dose-dependent effects of dopamine on postoperative HAIs was noted. Furthermore, to differentiate the dopamine and inotrope effects, we added a maximum Vasoactive-Inotrope Score and found that while dopamine use remained a risk factor for postoperative HAIs, the use of inotrope did not. The duration of mechanical ventilation, genetic abnormality, and DSC were reported as risk factors for postoperative HAIs in other studies, although they are not easily modifiable. In particular, mechanical ventilation was found to be a risk factor for postoperative HAIs (3), pneumonia (30), and BSI (31), after pediatric cardiac surgery. Invasive tracheal intubation leads to airway damage, and the tracheal tube serves as a reservoir for bacteria (32). The use of noninvasive ventilation, protocolized sedation, and a respiratory care program reduced the frequency of frequencies of postoperative HAIs (33, 34). Genetic abnormality is also a risk factor for postoperative HAIs (4, 5). The frequency of SSI was higher in children with 22q11.2 deletion syndrome (35), who were reported to have immunodeficiency (36). Although DSC has hemodynamic and respiratory benefits (37), it is a risk factor for postoperative HAIs (5, 9, 11), and prolonged DSC may increase the frequency of HAIs (38). Hence, minimizing the duration of DSC should be considered.

with 22q11.2 deletion syndrome (35), who were reported to have immunodeficiency (36). Although DSC has hemodynamic and respiratory benefits (37), it is a risk factor for postoperative HAIs (5, 9, 11), and prolonged DSC may increase the frequency of HAIs (38). Hence, minimizing the duration of DSC should be considered. The present study had certain limitations. First, it was conducted in a single center, which limits the generalizability of the results. Furthermore, the number of patients may not be sufficiently large to identify the other risk factors for HAIs. HAIs were detected in 13.5% of the cases of surgeries, which is similar to the data obtained from other studies (6.0–30.8%) (3, 5, 6, 9). The most common HAIs included BSIs and SSIs, and the common causative pathogens were Gram-positive bacteria for BSIs (14/30) and SSIs (17/30), and Gram-negative bacteria for UTIs (12/13) and pneumonia (5/8), as was noted in other studies (3, 6, 9, 10). Furthermore, our study population, the frequency and type of HAIs, and the causative pathogens were similar to those observed in other studies, reflecting the generalizability of our study. Second, the study was retrospective in nature. Hence, infection prevention and control strategies could not be manipulated. In our study, of the patients with nasal MRSA colonization, the frequencies of postoperative HAIs in the patients who received prophylactic vancomycin and those who did not were 16% and 3% (p = 0.06), respectively. In addition, the prophylactic antibiotic regimen for DSC was changed during the study period, making it difficult to analyze infections during the two-time frames, together. Nonetheless, there was no statistically significant difference in the frequency of HAIs between patients with DSC (65% during the vancomycin plus meropenem period and 79% during the cefazolin period) (p = 0.33). Similarly, the strategies for inotrope use, RBC transfusion, and postoperative steroid use could not be controlled and were based on the physician’s clinical decision. It is possible that the type of surgery affected the use of dopamine, as well as the frequency of postoperative HAIs. Although there are no documented guidelines for inotrope use, dobutamine was the most commonly used inotrope for postoperative low cardiac output syndrome, and milrinone was added when elevated systemic vascular resistance was observed. Dopamine was used initially, and epinephrine was the next choice when low systemic vascular resistance was observed.

guidelines for inotrope use, dobutamine was the most commonly used inotrope for postoperative low cardiac output syndrome, and milrinone was added when elevated systemic vascular resistance was observed. Dopamine was used initially, and epinephrine was the next choice when low systemic vascular resistance was observed. In addition, all the diagnoses of HAIs were retrospective. There was a possibility that the severity of illness affected the use of therapeutic antibiotics and the diagnosis of postoperative HAIs. Furthermore, some HAIs, especially cases of clinical sepsis which are diagnosed according to clinical manifestations, may have been overlooked. Nonetheless, during diagnosis, we prospectively reviewed the recorded documents on suspected infections, including the results of cultures, symptoms (i.e., fever, hypotension, and apnea), use of antibiotics, and results of blood tests. Third, we assessed only the first surgeries during a single PICU stay. Nevertheless, we included multiple surgeries during a single PICU stay as potential risk factors for postoperative HAIs. Interestingly, undergoing multiple surgeries during a single PICU stay was not an independent risk factor for HAIs. Fourth, although we employed the 2008 CDC/NHSN surveillance definition (13), according to the latest definition, only LCBSI is considered a BSI and is grouped into central line–associated BSI (CLABSI) and non-CLABSI. Therefore, we may have overdiagnosed BSI by including cases without positive blood cultures; however, we believe that those episodes may have been clinically important. Fifth, the U.S. guideline recommends the use of alcohol-containing preoperative skin preparatory agents (39), which were not available in the study period in Japan; therefore, we used 70% alcohol followed by povidone iodine for preoperative skin antisepsis.

elieve that those episodes may have been clinically important. Fifth, the U.S. guideline recommends the use of alcohol-containing preoperative skin preparatory agents (39), which were not available in the study period in Japan; therefore, we used 70% alcohol followed by povidone iodine for preoperative skin antisepsis. CONCLUSIONS This retrospective, single-center study identified the risk factors for BSI, SSI, pneumonia, and UTI, following pediatric cardiac surgery. The frequency of HAIs after pediatric cardiac surgery was 13.5%. The risk factors for HAIs after pediatric cardiac surgery include mechanical ventilation greater than or equal to 3 days, dopamine use, genetic abnormality, and DSC. Since the use of dopamine is an easily modifiable risk factor, and may serve as a potential target to reduce HAIs, further studies are needed to establish whether dopamine negatively impacts the development of HAIs. ACKNOWLEDGMENTS We thank the staff at our PICU for cooperating with us on this study, particularly Masayo Tsuda, Hideyuki Matsunaga, Nao Okuda, Takaaki Akamatsu, Noboru Matsumoto, and Atsushi Kawamura (PICU physicians). We would also like to thank Kenji Hirai (health information manager), Akihito Inoue (medical engineer), Makie Kinoshita (infection control nurse), and Futoshi Fujiwara (clinical microbiology laboratory personnel) for their assistance with data collection, as well as Kimiko Ueda and Yusuke Naito for their assistance with the statistical analyses. *See also p. 269. The authors have disclosed that they do not have any potential conflicts of interest.

d in U.S. populations (1–3). This finding has not been replicated in a lower mortality, resource-rich cohort nor has it been demonstrated to correlate with other important outcomes such as length of ventilation or ICU stay. Additionally, optimal timing for assessment of VIS in pediatric sepsis has not been established. Figure 1. Calculation of Vasoactive-Inotropic Score (VIS). This study aims to determine whether VIS is a valid surrogate outcome measure in pediatric sepsis by testing its association with important short-term outcomes. We also seek to determine the optimal timing for assessment of VIS. We hypothesize that VIS at early time points in sepsis (6, 12, 24, and 48 hr following arrival to the ICU) will be associated with key outcome measures including ICU LOS, ventilator days, and a composite outcome of cardiac arrest, need for extracorporeal membrane oxygenation (ECMO), and in-hospital mortality. VIS is easily measured and, if validated in pediatric sepsis, has the potential to standardize the quantification of hemodynamic support, identify, and stratify high-risk patient populations, and be used as a surrogate outcome measure for quality initiatives and research.

Early fluid resuscitation of critically ill pediatric patients has been become standard of care (1, 2). However, once a patient is stabilized, the focus must change to preventing or treating fluid overload as it is a risk factor for PICU mortality (3, 4) and has been associated with prolonged mechanical ventilation (5, 6). Typically, the first line of therapy for fluid overload in pediatric patients is intermittent furosemide, although some studies suggest that continuous furosemide administration is more beneficial than intermittent furosemide dosing due to more stable hemodynamics and improved urine output (UOP) (7). However, continuous infusions of furosemide are often prohibited in pediatric patients due to drug compatibility issues and the lack of a dedicated IV lumen. Furosemide is rarely used alone, and adjunctive diuretics such as chlorothiazide and spironolactone have been used for decades in pediatric patients (8).

Sepsis is a common critical illness leading to pediatric morbidity, mortality, and increased healthcare costs in the United States and worldwide (1–3). Nationally in 2005, there were more than 70,000 hospital admissions for pediatric severe sepsis, accounting for up to 7% of pediatric hospitalizations with an associated cost of nearly $5 billion (1). In recent U.S. database studies, pediatric severe sepsis mortality estimates ranged from 5% to 20% (1, 3). Pediatric sepsis is also associated with significant resource burden, with a median hospitalization cost of $77,000, median PICU length of stay (LOS) of 7 days, and long-term morbidity in survivors (2, 3). Pediatric septic shock is the end stage of severe sepsis, defined as those patients with evidence of cardiovascular dysfunction despite fluid resuscitation (4), and represents the patients at highest risk of mortality (5).

$77,000, median PICU length of stay (LOS) of 7 days, and long-term morbidity in survivors (2, 3). Pediatric septic shock is the end stage of severe sepsis, defined as those patients with evidence of cardiovascular dysfunction despite fluid resuscitation (4), and represents the patients at highest risk of mortality (5). Vasoactive and inotropic medications are standardly used to treat hypotension and cardiovascular dysfunction associated with pediatric septic shock (6). Currently, no uniform, validated measure or scoring system exists to describe the magnitude of hemodynamic support required in pediatric sepsis. In adult sepsis, clinical measures that both objectively describe illness severity and correlate with important outcomes such as mortality are being increasingly recognized as necessary to identify which patients are most at risk for poor outcomes (7). Low-mortality rates in pediatric sepsis limit the feasibility of using death as a primary endpoint, making validated surrogate measures critical for pediatric sepsis research. A validated score that accurately describes cardiovascular dysfunction and correlates with other clinically relevant outcomes such as duration of mechanical ventilation and ICU stay could be used to identify high-risk patients and as an outcome in research and quality improvement.

ritical for pediatric sepsis research. A validated score that accurately describes cardiovascular dysfunction and correlates with other clinically relevant outcomes such as duration of mechanical ventilation and ICU stay could be used to identify high-risk patients and as an outcome in research and quality improvement. One candidate scoring system was recently proposed by Gaies et al (8) for use in infant cardiac surgery. The Vasoactive-Inotropic Score (VIS), expanded from the previously described Inotropic Score (9), quantifies the amount of cardiovascular support required by infants postoperatively and includes dopamine, dobutamine, epinephrine, milrinone, vasopressin, and norepinephrine (Fig. 1). VIS has been shown to correlate with worse short-term clinical outcomes in infants following cardiac surgery (8, 10–12). Various studies have used VIS in pediatric sepsis to describe inotropic and vasopressor support in their population (13, 14). However, to our knowledge, only one previous study by Haque et al (15) has shown an association between VIS and mortality in pediatric sepsis. This study looked at a small cohort in India with a much higher mortality rate (42%) for pediatric sepsis than previously described in U.S. populations (1–3). This finding has not been replicated in a lower mortality, resource-rich cohort nor has it been demonstrated to correlate with other important outcomes such as length of ventilation or ICU stay. Additionally, optimal timing for assessment of VIS in pediatric sepsis has not been established.

oxygenation (ECMO), and in-hospital mortality. VIS is easily measured and, if validated in pediatric sepsis, has the potential to standardize the quantification of hemodynamic support, identify, and stratify high-risk patient populations, and be used as a surrogate outcome measure for quality initiatives and research. MATERIALS AND METHODS We conducted a secondary analysis of a single-center sepsis registry that includes children clinically identified with suspected sepsis in the emergency department (ED) from January 2012 to June 2015. This study was conducted at Children’s Hospital Colorado, a freestanding tertiary children’s hospital. Institutional ED protocols for sepsis include an “activation” process for suspected sepsis, which includes protocolized mobilization of additional personnel and equipment, paging, order sets and standardized procedures for rapid initiation of IV access, fluid resuscitation, antibiotics, and critical therapeutics. Bedside clinicians (nurses, advanced practice providers, or physicians) initiate sepsis activations when they clinically suspect infection in the presence of altered mental status or perfusion. Using these clinical indications for sepsis treatment follows international guidelines for recognition and treatment of pediatric septic shock (6). The sepsis registry was created to support sepsis clinical quality improvement work and approved by the Children’s Hospital Colorado Organizational Research Risk and Quality Improvement Review Panel. Extraction of data from the prospective registry into a limited dataset for analysis for generalizable research was approved by the Colorado Multiple Institution Review Board.

cal quality improvement work and approved by the Children’s Hospital Colorado Organizational Research Risk and Quality Improvement Review Panel. Extraction of data from the prospective registry into a limited dataset for analysis for generalizable research was approved by the Colorado Multiple Institution Review Board. ED patients were identified for inclusion in the sepsis registry through the presence of any of the following in the electronic health record (EHR): 1) use of a sepsis-specific order set at the time of ED treatment, 2) use of the sepsis activation paging system which is time-stamped in the EHR, and 3) missed cases. Missed cases were identified in the following manner: all patients placed in intensive care within 24 hours of ED care who received antibiotics or were hypotensive in the ED are identified through a monthly EHR query. Two trained reviewers reviewed these cases monthly, and cases considered missed sepsis were added to the database along with their clinical data. Missed case classification by the two trained reviewers was tested on an ongoing basis and a κ of greater than 0.8 was sustained.

are identified through a monthly EHR query. Two trained reviewers reviewed these cases monthly, and cases considered missed sepsis were added to the database along with their clinical data. Missed case classification by the two trained reviewers was tested on an ongoing basis and a κ of greater than 0.8 was sustained. Data were extracted from the EHR using extract, transfer, and load methodology. From the Epic Clarity database, reports generated using Crystal Reports 2008 were imported monthly into a Research Electronic Data Capture (REDCap) database (16). Additional monthly loads updated time-dependent fields that were incomplete at the original load, such as hospital LOS and 30-day mortality. Routine quality auditing identified specific variables as unreliably captured in the EHR; these underwent chart review and manual data entry monthly, including the less than 1% of encounters that included article (non-EHR) resuscitation charting.

complete at the original load, such as hospital LOS and 30-day mortality. Routine quality auditing identified specific variables as unreliably captured in the EHR; these underwent chart review and manual data entry monthly, including the less than 1% of encounters that included article (non-EHR) resuscitation charting. Additional exclusion criteria were applied to the larger registry to obtain the cohort for this study. Potential subjects were identified for inclusion in the cohort by identifying patients in the registry who had a vasoactive agent administered during hospitalization as indicated in the medical administration record. These cases subsequently underwent chart review to determine whether inclusion/exclusion criteria were met. Inclusion criteria were patients who received at least one vasoactive or inotropic agent in the first 48 hours after arrival to the ED or ICU. Exclusion criteria were admission to the neonatal ICU, age less than 60 days or greater than 18 years, need for an operating room (OR) procedure within the first 6 hours after admission, determination that the primary reason for requiring inotropic or vasoactive support was a definitive alternative diagnosis (Supplemental Table 1, Supplemental Digital Content 1, http://links.lww.com/PCC/A449), placement onto ECMO support prior to initiation of vasoactive/inotropic support, and initiation of vasoactive/inotropic support occurring only following an OR procedure.

ic or vasoactive support was a definitive alternative diagnosis (Supplemental Table 1, Supplemental Digital Content 1, http://links.lww.com/PCC/A449), placement onto ECMO support prior to initiation of vasoactive/inotropic support, and initiation of vasoactive/inotropic support occurring only following an OR procedure. Demographic and clinical information collected on all patients from the auto-extracted database included age at admission, Pediatric Index of Mortality 3 score (17), gender, death during hospitalization, vasopressor days, ICU and hospital LOS, presence of complex chronic condition as defined by Feudtner et al (18), initial lactate in 8 hours, amount of fluid boluses received in the first 24 hours, and time to antibiotic from initial vitals in the ED. Information that was manually chart reviewed included cardiac arrest, need for ECMO, underlying diagnosis, infectious source, need for and timing of OR procedures, need for intubation, and length of ventilation. VIS was the independent variable in this study. Maximal VIS in the first 48 hours and VIS at 6, 12, 24, and 48 hours following ED admission time and ICU admission time were recorded. VIS was manually calculated through investigator chart review on patients per Gaies et al (8) (Fig. 1). For a sample patient in the cohort on 9 µg/kg/min of dopamine, 0.12 µg/kg/min of norepinephrine, and 0.05 µg/kg/min of epinephrine at 48 hours after time of admission, VIS at 48 hours would be calculated as follows: 9 + (100 × 0.12) + (100 × 0.05) = 23.

nvestigator chart review on patients per Gaies et al (8) (Fig. 1). For a sample patient in the cohort on 9 µg/kg/min of dopamine, 0.12 µg/kg/min of norepinephrine, and 0.05 µg/kg/min of epinephrine at 48 hours after time of admission, VIS at 48 hours would be calculated as follows: 9 + (100 × 0.12) + (100 × 0.05) = 23. Patients who were in the OR in the first 48 hours but after 6 hours following admission were included with the VIS calculated only for those time points preceding the OR procedure in order to exclude patients who required intraoperative and postoperative vasoactive and inotropic support due to shock caused by conditions other than sepsis such as hemorrhage, anesthetic agents, or nonseptic systemic inflammatory response. For patients who died or were cannulated onto ECMO prior to the 48-hour time mark, VIS was calculated only for time points preceding death or ECMO. Primary outcomes were ICU LOS and ventilator-free days. The secondary outcome was a combined poor outcome variable defined as the following: cardiac arrest or need for cardiopulmonary resuscitation, need for ECMO, or death during hospitalization. For patients who had a cardiac arrest in the first 48 hours, subsequent VIS following the arrest event was excluded from analysis for the composite outcome but included in the analysis for ICU LOS, ventilator days, and need for intubation.

r cardiopulmonary resuscitation, need for ECMO, or death during hospitalization. For patients who had a cardiac arrest in the first 48 hours, subsequent VIS following the arrest event was excluded from analysis for the composite outcome but included in the analysis for ICU LOS, ventilator days, and need for intubation. Ventilator days were chart abstracted manually, and a ventilator day was defined as any amount of invasive mechanical ventilation on a given calendar day. Patients who died during hospitalization were assigned a maximum ventilator days value of 28 days. ICU LOS was auto-extracted. Patients who died were assigned the value of the maximum ICU LOS in the cohort, which was 964 hours or 48 days. Patients with tracheostomies who were baseline ventilated were defined as requiring intubation if they required escalation above their home settings, and total time ventilated was calculated as time from escalation to time of return to home ventilation settings. Patients who were intubated solely for an operative procedure were not counted as requiring intubation. Study data were collected and managed using REDCap, hosted at the University of Colorado. REDCap is a secure, web-based application designed to support data capture for research studies (16). A limited HIPAA-compliant dataset was exported from REDCap into SAS v. 9.4 (SAS Institute, Cary, NC) for analysis.

Ventilator days were chart abstracted manually, and a ventilator day was defined as any amount of invasive mechanical ventilation on a given calendar day. Patients who died during hospitalization were assigned a maximum ventilator days value of 28 days. ICU LOS was auto-extracted. Patients who died were assigned the value of the maximum ICU LOS in the cohort, which was 964 hours or 48 days. Patients with tracheostomies who were baseline ventilated were defined as requiring intubation if they required escalation above their home settings, and total time ventilated was calculated as time from escalation to time of return to home ventilation settings. Patients who were intubated solely for an operative procedure were not counted as requiring intubation. Study data were collected and managed using REDCap, hosted at the University of Colorado. REDCap is a secure, web-based application designed to support data capture for research studies (16). A limited HIPAA-compliant dataset was exported from REDCap into SAS v. 9.4 (SAS Institute, Cary, NC) for analysis. Patients’ demographics and baseline clinic characteristics were summarized using descriptive statistics. Median and interquartile range were reported for continuous variables, whereas frequency and percentage were reported for binary and categorical variables. The data distribution of continuous outcomes such as ICU LOS and ventilator days was examined to determine the models to be employed. As a result, ICU LOS was transformed in natural log scale and general linear model was fitted for this dependent variable. There were over 50% of subjects with a ventilator days value of zero, thus zero inflated Poisson model was fit for this dependent variable. Candidate covariates such as PIM3 score, complex chronic condition, initial lactate, fluid boluses within first 24 hours, and time to antibiotics and pneumonia were chosen based on clinical relevance and published sepsis literature (19–21). Spearman rank correlation test was performed to further narrow down the candidate covariates that were included in the multivariable models. Covariates included in the final model are shown in Supplemental Tables 2a–2e (Supplemental Digital Content 2, http://links.lww.com/PCC/A450). Multiple logistic regression models were fit for the binary outcomes of intubation and composite outcome. Best fit models were evaluated based on the statistical significance (p) and appropriate model fitting statistics (Akaike information criterion, R2 and Tjur’s R2). p value of less than 0.05 was considered statistically significant, and all analyses were performed using SAS V9.4.

mes of intubation and composite outcome. Best fit models were evaluated based on the statistical significance (p) and appropriate model fitting statistics (Akaike information criterion, R2 and Tjur’s R2). p value of less than 0.05 was considered statistically significant, and all analyses were performed using SAS V9.4. RESULTS Between January 2012 and June 2015, 215 patients were identified in the sepsis database as having received any vasoactive or inotropic support, of whom 138 patients met inclusion criteria and did not meet exclusion criteria (Fig. 2). Seven included patients had VIS calculations at limited time points, due to OR procedure or death in first 48 hours of ICU admission. For the composite outcome analysis only, two additional patients were excluded due to cardiac arrest occurring in the first 6 hours after admission, and one patient had VIS calculated at only the 6- and 12-hour time points due to cardiac arrest occurring between 12 and 24 hours. Figure 2. Exclusion diagram. NICU = neonatal ICU, OR = operating room, VIS = Vasoactive-Inotropic Score.

RESULTS Between January 2012 and June 2015, 215 patients were identified in the sepsis database as having received any vasoactive or inotropic support, of whom 138 patients met inclusion criteria and did not meet exclusion criteria (Fig. 2). Seven included patients had VIS calculations at limited time points, due to OR procedure or death in first 48 hours of ICU admission. For the composite outcome analysis only, two additional patients were excluded due to cardiac arrest occurring in the first 6 hours after admission, and one patient had VIS calculated at only the 6- and 12-hour time points due to cardiac arrest occurring between 12 and 24 hours. Figure 2. Exclusion diagram. NICU = neonatal ICU, OR = operating room, VIS = Vasoactive-Inotropic Score. Demographic and treatment characteristics are shown in Table 1. The most common underlying conditions were genetic/metabolic (21.6%) and oncologic (13.0%). The most common infectious sources were pneumonia (31.7%) and bacteremia (23.7%). Ninety-six percent of intubations occurred in the first 48 hours, and the absolute range of VIS at 48 hours was 0–40. Details of specific vasoactive and inotropic agent usage in study patients as well as the timing of intubations are described in Supplemental Table 3 (Supplemental Digital Content 3, http://links.lww.com/PCC/A451) and Supplemental Table 4 (Supplemental Digital Content 4, http://links.lww.com/PCC/A452), respectively. Change in VIS over time is demonstrated graphically in Supplemental Graph 1 (Supplemental Digital Content 5, http://links.lww.com/PCC/A453).

tal Table 3 (Supplemental Digital Content 3, http://links.lww.com/PCC/A451) and Supplemental Table 4 (Supplemental Digital Content 4, http://links.lww.com/PCC/A452), respectively. Change in VIS over time is demonstrated graphically in Supplemental Graph 1 (Supplemental Digital Content 5, http://links.lww.com/PCC/A453). TABLE 1. Demographic, Treatment, and Outcome Characteristics of Patient Population As a first step, VIS correlations with the primary outcomes of ICU LOS and ventilator days were calculated at prespecified time points after ED and ICU arrival on univariate analysis to identify the best VIS time point to use on multivariable analysis. VIS at time points calculated based on ICU arrival time (Table 2) performed best, therefore, arrival at ICU was used as time point zero for additional analyses (VIS for ED time points not shown). We then compared VIS at different time points after ICU admission to each other (Table 2). TABLE 2. Correlation of Vasoactive-Inotropic Score Calculated After ICU Arrival With ICU Length of Stay and Ventilator Days

As a first step, VIS correlations with the primary outcomes of ICU LOS and ventilator days were calculated at prespecified time points after ED and ICU arrival on univariate analysis to identify the best VIS time point to use on multivariable analysis. VIS at time points calculated based on ICU arrival time (Table 2) performed best, therefore, arrival at ICU was used as time point zero for additional analyses (VIS for ED time points not shown). We then compared VIS at different time points after ICU admission to each other (Table 2). TABLE 2. Correlation of Vasoactive-Inotropic Score Calculated After ICU Arrival With ICU Length of Stay and Ventilator Days Given that VIS at 48 hours had the strongest correlation with the primary outcomes (ICU LOS and ventilator days), VIS at 48 hours was chosen as the independent continuous variable in the multivariable analysis. VIS at 48 hours was independently associated with both ventilator days and ICU LOS, showing that for every unit increase in VIS, there was an 8% increase (p < 0.01) and 13% increase (p < 0.0001) in ventilator days and ICU LOS, respectively. We also tested the relationship of patients who required any vasoactive or inotropic support at 48 hours with both primary outcomes via multivariable analysis. VIS at 48 hours as a dichotomous variable did not have a statistically significant association with ventilator days on multivariable analysis. VIS at 48 hours as a dichotomous variable was associated with ICU LOS but had minimal effect on overall fit of the multivariable model compared with VIS at 48 hours as a continuous variable (change in R2 of 0.04).

iable did not have a statistically significant association with ventilator days on multivariable analysis. VIS at 48 hours as a dichotomous variable was associated with ICU LOS but had minimal effect on overall fit of the multivariable model compared with VIS at 48 hours as a continuous variable (change in R2 of 0.04). For the composite outcome, the model indicated a nonstatistically significant association with VIS at 48 hours, with an 8% increase in the odds have having the composite outcome with every unit increase in VIS (p = 0.24). This analysis was likely underpowered due to limited events; there were only 12 total composite outcome events in the cohort and five of those events happened prior to 48 hours and so were excluded from the analysis for VIS at 48 hours. Of the five excluded subjects from the VIS at 48-hour analysis, three patients suffered cardiac arrest or death between 12 and 24 hours (Supplemental Table 5, Supplemental Digital Content 6, http://links.lww.com/PCC/A454). We therefore performed a secondary analysis using VIS at 6 and 12 hours in the multivariable model in order to capture those events that occurred earlier in the ICU stay. We found that VIS at 12 hours had a strong independent association with cardiac arrest, ECMO, or death. For every unit increase in VIS at 12 hours, there was a 14% increase in the odds of subsequently experiencing the composite outcome (p < 0.01). Interestingly, in this model, for a fixed VIS at 12 hours, higher VIS at 6 hours was associated with a decreased risk of later having the composite outcome (OR = 0.87; 95% CI, 0.77–0.99). Additionally, when looking at the prearrest VIS trend in the five patients who experienced the composite outcome during the first 48 hours, four of the five demonstrated increasing VIS in the 8 hours prior to the event (Supplemental Table 5, Supplemental Digital Content 6, http://links.lww.com/PCC/A454; and Supplemental Graph 2, Supplemental Digital Content 7, http://links.lww.com/PCC/A455).

who experienced the composite outcome during the first 48 hours, four of the five demonstrated increasing VIS in the 8 hours prior to the event (Supplemental Table 5, Supplemental Digital Content 6, http://links.lww.com/PCC/A454; and Supplemental Graph 2, Supplemental Digital Content 7, http://links.lww.com/PCC/A455). DISCUSSION In pediatric sepsis, there is a need for validated surrogate outcome measures (7). Our study provides a critical first step in validating VIS for such purposes. Although VIS has been validated in pediatric cardiac surgery (8, 10–12), and has been used in previous studies of pediatric patients with severe sepsis to describe the severity of illness and as a measure of hemodynamic support (13, 14), only one prior study by Haque et al (15) has shown an association between maximal VIS and outcomes in pediatric sepsis. Our study complements the study by Haque et al (15) done in a resource-poor setting, by showing that in a resource-rich setting with a medically complex, low-mortality cohort of children with septic shock, VIS at 48 hours after ICU arrival is independently associated with short-term outcomes including ICU LOS and ventilator days.

omplements the study by Haque et al (15) done in a resource-poor setting, by showing that in a resource-rich setting with a medically complex, low-mortality cohort of children with septic shock, VIS at 48 hours after ICU arrival is independently associated with short-term outcomes including ICU LOS and ventilator days. Importantly, the association of VIS at 48 hours with these outcomes was independent of the validated PIM3 score. This finding is in part not surprising, as the PIM3 score was validated to predict mortality across the broad spectrum of PICU admissions and did not specifically target sepsis patients or outcomes other than mortality (17). However, it is also important to note that progressive cardiovascular instability plays a role in the pathophysiology of sepsis and the cardiovascular component of the PIM3 score is limited to systolic blood pressure and acid base status in the first hour. Our results indicate that VIS at 48 hours may capture a component of illness in pediatric sepsis that is not fully addressed by PIM3. VIS at 48 hours may therefore complement existing acuity scores as both a relatively early indicator of the likely duration of critical care support as well as a surrogate outcome that highlights cardiovascular instability.

ours may capture a component of illness in pediatric sepsis that is not fully addressed by PIM3. VIS at 48 hours may therefore complement existing acuity scores as both a relatively early indicator of the likely duration of critical care support as well as a surrogate outcome that highlights cardiovascular instability. Consistent with this concept, our study shows that persistent rather than early or maximal need for vasoactive and inotropic support in the first 48 hours is most strongly associated with duration of critical care support. Patients with a high VIS at 48 hours demonstrate ongoing cardiovascular dysfunction and are inherently at highest risk of a poor outcome. In contrast to previous studies of VIS in infant cardiac surgery (8, 11, 12) and the study by Haque et al (15) of VIS in pediatric sepsis (15), we did not find a strong correlation between maximal VIS and outcomes of interest. Early aggressive resuscitation is associated with improved outcomes and shock reversal in pediatric sepsis (6), and it is possible that early and maximal VIS may reflect attentive support of reversible pathophysiology compared with VIS at later time points.

rong correlation between maximal VIS and outcomes of interest. Early aggressive resuscitation is associated with improved outcomes and shock reversal in pediatric sepsis (6), and it is possible that early and maximal VIS may reflect attentive support of reversible pathophysiology compared with VIS at later time points. The relationship between high VIS and prolonged intubation and ICU LOS is intuitive. Patients requiring vasoactive or inotropic support due to vasoplegia or capillary leak syndrome are also those who are more likely to have acute respiratory distress syndrome or pulmonary edema requiring positive-pressure ventilation. Additionally, patients with myocardial dysfunction that require inotropic support are often intubated to reduce total body oxygen demand and improve left ventricular function. Conversely, prolonged ventilation can also directly lead to increased vasoactive and inotropic medication requirements due to increased intrathoracic pressure, decreased venous return, and increased right ventricular afterload. As the vast majority of patients requiring invasive ventilation were intubated prior to 48 hours (Supplemental Table 4, Supplemental Digital Content 5, http://links.lww.com/PCC/A452), VIS at 48 hours cannot be considered to be predictive of intubation. However, those patients with a persistent need for vasoactive or inotropic support 48 hours after admission may have prolonged multiple organ dysfunction increasing the duration and need for ICU care and ventilation. The association of VIS with ICU LOS and length of ventilation suggests that cardiovascular dysfunction, as represented by a high VIS, may be causal in both of these important outcomes. This may mean that interventions targeting a decrease in a patient’s VIS may ultimately affect these relevant outcomes as well.

lation. The association of VIS with ICU LOS and length of ventilation suggests that cardiovascular dysfunction, as represented by a high VIS, may be causal in both of these important outcomes. This may mean that interventions targeting a decrease in a patient’s VIS may ultimately affect these relevant outcomes as well. VIS at 48 hours after PICU arrival did not show a statistically significant association with the secondary composite outcome but low frequency of this outcome means that this analysis was likely underpowered. More important, many of these events occurred early in hospitalization and VIS at 48 hours is by definition not useful to accurately identify or predict early cardiovascular collapse. We therefore explored an earlier time point (VIS at 12 hr) in a secondary analysis and found that it was strongly and independently associated with subsequent cardiac arrest, ECMO, or death. Again, this finding was independent of the measured acuity score (PIM3) suggesting a complementary role for the quantification of cardiovascular support in this population. Of interest, for a fixed VIS at 12 hours, VIS at 6 hours was inversely associated with the composite outcome. Although it is possible that an early high VIS at 6 hours portends a better outcome for patients as it represents aggressive early resuscitation, this is unlikely as higher VIS at 6 hours is associated with worse outcomes when modeled separately from VIS at 12 hours. An alternative explanation for this relationship may be that direction or trend of VIS is important. Those patients who do not wean from support quickly or require increasing support between 6 and 12 hours are patients whose illness is worsening and are more likely to have a poor outcome. This concept is supported by our descriptive data showing that in four of five patients with early cardiovascular collapse, VIS was rising in the 4 hours immediately prior to the event, although the small number of cases precludes meaningful statistical analysis. Overall, although these data are suggestive that early and rising VIS may predict subsequent poor outcome, due to the small number of events, it can be considered exploratory only and will require validation in larger cohorts.

r to the event, although the small number of cases precludes meaningful statistical analysis. Overall, although these data are suggestive that early and rising VIS may predict subsequent poor outcome, due to the small number of events, it can be considered exploratory only and will require validation in larger cohorts. Overall, our study population is similar to pediatric severe sepsis populations presenting to the ED of tertiary centers with pediatric ICUs. The 6% mortality in our study of patients with septic shock is similar to that reported from ED-based pediatric sepsis registries with sepsis quality programs in place (19, 22, 23) but is lower than what has been reported in national database and ICU-based pediatric studies (1–3). Also, the median PIM3 score in our population was lower than in other ICU-based studies (2), suggesting that our population may have had a lower overall level of illness severity. Compared with previously published pediatric ED sepsis studies in the United States, our population had a similar proportion of patients with complex chronic conditions (3, 22, 23). A potentially unique aspect of our population was the median amount of bolus fluid (55 mL/kg) received by patients in the first 24 hours, which is lower than American College of Critical Care Medicine guidelines but is similar to what has been described in other pediatric tertiary hospital populations (23, 24). Despite this possible variation, VIS was still found to be independently associated with outcomes of interest when controlling for the amount of fluid resuscitation. Additionally, although vasoactive and inotropic usage may vary in other institutions in comparison to our ICU due to individual practice patterns and availability of specific drugs, such as those in resource-poor settings, VIS is calculated in a manner that incorporates all commonly used vasoactive and inotropic equally and thus can be used to compare across institutions despite different prescribing patterns.

rison to our ICU due to individual practice patterns and availability of specific drugs, such as those in resource-poor settings, VIS is calculated in a manner that incorporates all commonly used vasoactive and inotropic equally and thus can be used to compare across institutions despite different prescribing patterns. In summary, this study shows that VIS is a reliable marker of cardiovascular support that is independently associated with important outcomes in pediatric sepsis and may complement existing acuity scores as an early prognostic indicator of the duration of critical care support in this population. Low-mortality rates in pediatric sepsis mean that using death as a primary endpoint for many clinical investigations is often impractical. Our results indicate that VIS at 48 hours could serve as a surrogate cardiovascular outcome to help develop investigations with sufficient power to answer clinical questions in a timely and feasible manner. It can also be used to quantify hemodynamic support, allowing comparison of patient populations across studies and centers. Additionally, VIS is a simple score that can be easily calculated in real time at the bedside and requires only 48 hours of data collection. Finally, although only preliminary in nature, our study also suggests that trends in VIS prior to the 48-hour time point should be explored further in larger multiinstitutional studies as a potential early warning or risk-stratification method to clinically identify septic patients at increasing risk for sudden cardiovascular collapse.

only preliminary in nature, our study also suggests that trends in VIS prior to the 48-hour time point should be explored further in larger multiinstitutional studies as a potential early warning or risk-stratification method to clinically identify septic patients at increasing risk for sudden cardiovascular collapse. There are limitations to this study. Data were collected retrospectively from an EHR and quality improvement database primarily designed to assess early recognition and treatment of sepsis in the ED. It was also a single-center study of a tertiary children’s hospital center and may not be generalizable to all institutions that care for pediatric sepsis patients. Given that sepsis patients are very heterogeneous, determining time point zero of illness in sepsis is difficult, and it is possible that VIS values measured at discrete time points after ICU arrival in this study were affected by variations in the timing of a patient’s presentation to the ED and subsequent ICU admission. Although we excluded those patients diagnosed with dilated cardiomyopathy/myocarditis, all patients did not have an objective measure of cardiac function so we were unable to make comparisons or draw conclusions between groups of patients with and without myocardial dysfunction, nor did we have long-term follow-up on which patients in the cohort had persistent ventricular abnormalities. Additionally, these findings may not be generalizable to individual patients requiring early operative intervention or who suffer very early catastrophic events such as cardiac arrest, death, or cannulation onto ECMO, as these patients were excluded from this study. As is true in all observational studies, treatments were not dictated by the study, so individual clinician and institutional practices such as threshold for initiation of cardiovascular support may have affected the VIS. Additionally, when evaluating the VIS as a score itself, it is important to note that individual components of the score are weighted for convenience rather than considering the clinical relevance of each medication. Due to the lack of long-term follow-up in this registry-based study, we were unable to assess the association of VIS with longer term mortality after hospitalization or functional outcomes. Investigation of the relationship between VIS and functional outcomes is an important area for future prospective studies.

cation. Due to the lack of long-term follow-up in this registry-based study, we were unable to assess the association of VIS with longer term mortality after hospitalization or functional outcomes. Investigation of the relationship between VIS and functional outcomes is an important area for future prospective studies. CONCLUSIONS VIS at 48 hours after ICU admission in pediatric sepsis is an easily calculated clinical score that is independently associated with ICU LOS and length of ventilation. VIS at 12 hours is independently associated with risk of cardiac arrest, death, or need for ECMO. In pediatric sepsis, VIS is a reliable marker of cardiovascular support that may be used as a surrogate outcome for research studies and provide additive value to existing pediatric acuity scores in this population. ACKNOWLEDGMENTS We are grateful for the contributions of Kendra Kocher, Kathleen Grice, and Mimi Goodwin in the Section of Pediatric Emergency Medicine for their tireless commitment to the maintenance of the sepsis registry, managing data entry, quality control, and compliance. Supplementary Material *See also p. 803. This study was performed at Children’s Hospital Colorado, Aurora, CO. Drs. Davidson and Scott contributed equally as senior authors and mentors for the article. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (http://journals.lww.com/pccmjournal).

Drs. Davidson and Scott contributed equally as senior authors and mentors for the article. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (http://journals.lww.com/pccmjournal). Financial support for this study was provided by National Institutes of Health/Heart Lung and Blood Institute 1K23HL123634 (to Dr. Davidson) and the Brigid Hope Fund. Financial support for the database used in this study was provided by the Clinical and Operational Effectiveness and Patient Safety Grant of the University of Colorado (Grant Number 2300704). Dr. Davidson received support for article research from the National Institutes of Health (NIH), and his institution received funding from NIH/National Heart, Lung, and Blood Institute K23. The remaining authors have disclosed that they do not have any potential conflicts of interest.

OP) (7). However, continuous infusions of furosemide are often prohibited in pediatric patients due to drug compatibility issues and the lack of a dedicated IV lumen. Furosemide is rarely used alone, and adjunctive diuretics such as chlorothiazide and spironolactone have been used for decades in pediatric patients (8). The use of methylxanthine derivatives as diuretic adjuncts was first described in adults with congestive heart failure in the 1950s (aminophylline) (9), and as an adjunctive diuretic in 1978 (theophylline) (10), although their use in critically ill children was not reported until the late 1990s (11–13). Aminophylline is a methylxanthine derivative, a soluble compound of theophylline and ethylenediamine, that has bronchodilator properties through inhibition of phosphodiesterase type IV at high concentrations (> 10 μg/mL). However, it exhibits diuretic effects at lower concentration (2–3 μg/mL) via adenosine receptor antagonism (12), probably by acting as an afferent arteriole vasodilator and improving glomerular blood flow. Aminophylline is useful as prophylaxis against tacrolimus-induced nephrotoxicity (14), although it does not prevent acute kidney injury (AKI) in children following cardiopulmonary bypass (15). The pharmacokinetics of theophylline/aminophylline have recently been investigated in children following congenital heart surgery with cardiopulmonary bypass and revealed that lower dosages were needed in this population to target serum concentrations of 5–10 mg/L (16).

n children following cardiopulmonary bypass (15). The pharmacokinetics of theophylline/aminophylline have recently been investigated in children following congenital heart surgery with cardiopulmonary bypass and revealed that lower dosages were needed in this population to target serum concentrations of 5–10 mg/L (16). Our institution recently published a prospective, open-label study of the administration of aminophylline to 35 critically ill pediatric patients and examined its effects on UOP and inflammatory profiles (17). In this study, patients were administered an initial bolus of IV aminophylline, and dosing was adjusted to maintain theophylline levels 4–8 μg/mL. In this previous study, a large proportion (20%) had side effects including agitation, increased nasogastric output, cardiac ectopy, and tachydysrhythmias. Whether the maintenance of a minimum theophylline level is necessary for effective diuresis has never been examined. We hypothesized that low dose aminophylline therapy would be an effective diuretic regardless of theophylline trough level. We undertook a retrospective, single-center study of theophylline trough concentrations and UOP before and after the initiation of aminophylline diuretic therapy in a tertiary care PICU to determine whether theophylline trough levels would correlate with the change in UOP after the initiation of aminophylline.

h level. We undertook a retrospective, single-center study of theophylline trough concentrations and UOP before and after the initiation of aminophylline diuretic therapy in a tertiary care PICU to determine whether theophylline trough levels would correlate with the change in UOP after the initiation of aminophylline. MATERIALS AND METHODS Study Design and Participants This study is a single-center, retrospective chart review of patients who received aminophylline as an adjunctive diuretic during PICU admission between July 2010 and June 2015. To answer our primary research question of whether aminophylline/theophylline trough levels were correlated with diuresis, we screened patients less than 18 years old admitted to the PICU who were prescribed intermittent IV aminophylline for inclusion using the Pharmnet system by Cerner (Kansas City, MO). Patients were excluded if they 1) received continuous aminophylline infusion for bronchodilator purposes, 2) received aminophylline less than 48 hours after admission (so that adequate baseline laboratory values and UOP could be assessed), 3) received aminophylline for less than 48 hours, 4) received dialysis during admission, and 5) expired within 48 hours of admission. The Institutional Review Board of the Pennsylvania State University College of Medicine approved the study (STUDY00003422) prior to any patient screening or data collection.

e assessed), 3) received aminophylline for less than 48 hours, 4) received dialysis during admission, and 5) expired within 48 hours of admission. The Institutional Review Board of the Pennsylvania State University College of Medicine approved the study (STUDY00003422) prior to any patient screening or data collection. Standard Aminophylline Dosing and Monitoring The administration and dosing of aminophylline for diuresis were protocolized in our PICU prior to this study (17), and this Penn State Health Medical Center aminophylline dosing and monitoring guideline were built into our provider order entry system as an order set (Table S1, Supplemental Digital Content 1, http://links.lww.com/PCC/A667). However, all decisions regarding the initiation, dosing, frequency of dosing, and termination of aminophylline were made at the discretion of the clinical team. Typically, per the guideline, a loading dose of 3 mg/kg of aminophylline was administered IV, and maintenance dosing was administered based on the patient’s age. Because aminophylline is the 2:1 complex salt of theophylline and ethylenediamine, the laboratory measure of aminophylline levels is a serum theophylline concentration. Per the guideline, serum theophylline trough levels were drawn every morning, and dosing was adjusted to targeted trough concentrations of 4–8 μg/mL, although changes were made at the discretion of the attending physician and the clinical team.

tory measure of aminophylline levels is a serum theophylline concentration. Per the guideline, serum theophylline trough levels were drawn every morning, and dosing was adjusted to targeted trough concentrations of 4–8 μg/mL, although changes were made at the discretion of the attending physician and the clinical team. Patient Screening and Data Collection Patient data were collected from the Cerner Electronic Medical Record and transferred into a secure electronic database using REDCap software (Research Electronic Data Capture, Vanderbilt University). Data collected included admission date and time, sex, age, race, height (cm), weight (kg), diagnosis, details of aminophylline administration (including aminophylline loading dose and maintenance dosing, date and time), laboratory values including theophylline trough levels collected the morning after initiation, blood urea nitrogen (BUN) and serum creatinine (sCr) values, total fluid intake, and output from 48 hours before to 48 hours after the initiation of aminophylline therapy. For consistency, the 24 hours fluid intake, UOP, and fluid balance were recorded at 7 am each day, and output included urinary drainage tube, indwelling foley catheter, or diaper weight. The use of inotropes and vasopressors, as well as other diuretics, was collected.

ours after the initiation of aminophylline therapy. For consistency, the 24 hours fluid intake, UOP, and fluid balance were recorded at 7 am each day, and output included urinary drainage tube, indwelling foley catheter, or diaper weight. The use of inotropes and vasopressors, as well as other diuretics, was collected. Statistical Analysis Summary statistics including means and sds were computed for continuous variables (i.e., UOP) and frequencies and percentages were determined for categorical variables (i.e., ethnicity, diagnosis). For this and future analyses, clinical time intervals were defined relative to the initiation of aminophylline administration, for example, the period 24–48 hours prior to initiation was defined as “day –2,” the period 0–24 hours prior to initiation was defined as “day –1,” etc. All fluids, including intake, output, and balance, were recorded hourly, and hourly UOP was averaged over each 24 hours period and divided by the patients’ last recorded weight prior to the administration of aminophylline. The percent daily fluid balance was also calculated as fluid balance (L)/weight (kg) × 100 (18).

All fluids, including intake, output, and balance, were recorded hourly, and hourly UOP was averaged over each 24 hours period and divided by the patients’ last recorded weight prior to the administration of aminophylline. The percent daily fluid balance was also calculated as fluid balance (L)/weight (kg) × 100 (18). Correlation analysis between UOP and theophylline concentration was conducted based on Pearson’s correlation coefficient. The normality assumption was checked based on the Shapiro tests; thus, paired t tests or Wilcoxon signed-rank tests were used to assess the changes in primary (UOP) and secondary outcomes (e.g., sCr, BUN, fluid balance) before and after aminophylline initiation, as appropriate. Generalized estimating equation (GEE) regression models, a form of multivariate analysis used to estimate the variables of a generalized linear model with a possible unknown correlation between outcomes, were employed to investigate the marginal association between potential risk factors and UOP or fluid balance with daily repeated measures, where the variable estimates were obtained by Quasi-likelihood approach, and the p values for significance were based on Wald tests. The analyses were conducted in SAS 9.4 (SAS Institute Inc., Cary, NC), and the statistical significance threshold of p value of less than 0.05 was applied for all hypothesis testing.

res, where the variable estimates were obtained by Quasi-likelihood approach, and the p values for significance were based on Wald tests. The analyses were conducted in SAS 9.4 (SAS Institute Inc., Cary, NC), and the statistical significance threshold of p value of less than 0.05 was applied for all hypothesis testing. RESULTS Study Population and Characteristics A total of 452 patients admitted to our PICU during the 5-year period received aminophylline. Since our primary research question was whether theophylline levels were correlated with diuretic effect, we only included patients for whom 48 hours of baseline characteristics could be recorded, and then who received at least 48 hours of maintenance dosing of aminophylline. Therefore, 337 patients were excluded based upon criteria, leaving 115 patients for analysis. The patient population was predominantly young with 67% of the patient being less than 1 year old (Table 1) and overwhelmingly Caucasian (n = 85; 73.9%). The most common diagnosis was respiratory failure (n = 67; 58.3%), which included bronchiolitis, viral and bacterial pneumonia, and status asthmaticus (Table 1). The second most common diagnosis was surgical, congenital heart (n = 25; 21.7%). TABLE 1. Demographics of Patients Fulfilling Eligibility Criteria

RESULTS Study Population and Characteristics A total of 452 patients admitted to our PICU during the 5-year period received aminophylline. Since our primary research question was whether theophylline levels were correlated with diuretic effect, we only included patients for whom 48 hours of baseline characteristics could be recorded, and then who received at least 48 hours of maintenance dosing of aminophylline. Therefore, 337 patients were excluded based upon criteria, leaving 115 patients for analysis. The patient population was predominantly young with 67% of the patient being less than 1 year old (Table 1) and overwhelmingly Caucasian (n = 85; 73.9%). The most common diagnosis was respiratory failure (n = 67; 58.3%), which included bronchiolitis, viral and bacterial pneumonia, and status asthmaticus (Table 1). The second most common diagnosis was surgical, congenital heart (n = 25; 21.7%). TABLE 1. Demographics of Patients Fulfilling Eligibility Criteria Forty-four of the 115 patients (38.3%) received an aminophylline loading dose, although more frequently only aminophylline maintenance dosing was given (Table 2). Maintenance aminophylline dosing varied by age per protocol and physician discretion: 55 patients (48%) received 1.5 mg/kg/dose, 29 patients (25%) received 2 mg/kg/dose, and 13 patients (11%) received 3.3 mg/kg/dose of aminophylline. Only 13 of the patients had dose adjustments within 48 hours, three patients had dosing decreased due to high theophylline levels, and 10 patients had dosing increased due to low theophylline levels. A large proportion of our patients received furosemide, whereas chlorothiazide was less frequently used, and spironolactone was only administered on occasion (Table 2). It is a common practice in our PICU to use low dose inotropes (dobutamine) to increase cardiac output and indirectly increase renal perfusion. This fact, and our inclusion of surgical, congenital heart patients, accounts for our relatively frequent use of dobutamine, milrinone, and epinephrine in patients being administered diuretics (Table 2).

our PICU to use low dose inotropes (dobutamine) to increase cardiac output and indirectly increase renal perfusion. This fact, and our inclusion of surgical, congenital heart patients, accounts for our relatively frequent use of dobutamine, milrinone, and epinephrine in patients being administered diuretics (Table 2). TABLE 2. Frequency of Various Aminophylline Dosing Regimens, Other Diuretics Vasoactive Medications, and Other Medication Administration Diuretic Dosing The total daily dosing of aminophylline averaged 8.2 ± 2.9 mg/kg/d on day +1 and 7.0 ± 2.9 mg/kg/d on day +2, with the difference between days likely being due to bolus dosing being included in average daily dosing on day +1 (Fig. 1A). In our PICU, aminophylline is administered as an adjunctive diuretic; therefore, it is not surprising that furosemide (Fig. 1B) and chlorothiazide (Fig. 1C) dosing also increased significantly during the 4-day study period. Importantly, no patients received alternative diuretics such as mannitol or acetazolamide during the study period to enhance UOP. Interestingly, total fluid intake also increased significantly during the study period (Fig. 1D).

chlorothiazide (Fig. 1C) dosing also increased significantly during the 4-day study period. Importantly, no patients received alternative diuretics such as mannitol or acetazolamide during the study period to enhance UOP. Interestingly, total fluid intake also increased significantly during the study period (Fig. 1D). Figure 1. Diuretic dosing and fluid intake before/after aminophylline initiation. Scatter dot plot of daily aminophylline dosing of individual patients before (days –2, –1) and after (days +1, +2) aminophylline initiation (A). Furosemide dosing significantly increased during the study period (B), as did chlorothiazide dosing (C), and fluid intake (D). Statistical differences were calculated using nonparametric comparisons for each pair using the Wilcoxon method (*p < 0.05, **p < 0.005, ***p < 0.0005).

ys +1, +2) aminophylline initiation (A). Furosemide dosing significantly increased during the study period (B), as did chlorothiazide dosing (C), and fluid intake (D). Statistical differences were calculated using nonparametric comparisons for each pair using the Wilcoxon method (*p < 0.05, **p < 0.005, ***p < 0.0005). Aminophylline and UOP Average hourly UOP (mL/kg) increased significantly with the addition of aminophylline as compared with the day –2 and day –1 preaminophylline baseline (Fig. 2A). Theophylline troughs at 24 hours did not correlate with day +1 average hourly UOP (p = 0.78) (Fig. 2B), but the 48 hours theophylline levels did correlate with day +2 average hourly UOP (r2 = 0.080; p = 0.0022), although interestingly this relationship had a negative slope (Fig. 2C). Aminophylline therapy was assessed in two different ways in our statistical analyses: as a categorical variable (0, 1) where all patients were aminophylline “0” on days –2 and –1, then aminophylline “1” on days +1 and +2 and as a continuous variable using the total daily aminophylline dosage (mg/kg/d), where on days –2 and –1 this value was “0 mg/kg/d.” Univariate analysis demonstrated that the administration of aminophylline (categorical variable) was associated with increased UOP (p < 0.0001), whereas the amount of aminophylline administered or dosing (mg/kg/d, continuous variable) did not correlate with UOP (p = 0.59).

on days –2 and –1 this value was “0 mg/kg/d.” Univariate analysis demonstrated that the administration of aminophylline (categorical variable) was associated with increased UOP (p < 0.0001), whereas the amount of aminophylline administered or dosing (mg/kg/d, continuous variable) did not correlate with UOP (p = 0.59). Figure 2. Effect of aminophylline administration on urine output (UOP) and the association with theophylline levels. Average hourly UOP on days +1 and +2 (after aminophylline initiation) was significantly increased as compared with days –1 and –2 (prior to aminophylline) (A). Statistical differences were calculated using nonparametric comparisons for each pair using the Wilcoxon method (*p < 0.05, **p < 0.005, ***p < 0.0005). Linear regression analysis of average hourly UOP and theophylline levels at 24 hr (B) and 48 hr (C) on days +1 and +2, respectively. Theophylline levels at 48 hr were negatively correlated with UOP on day +2.

sing nonparametric comparisons for each pair using the Wilcoxon method (*p < 0.05, **p < 0.005, ***p < 0.0005). Linear regression analysis of average hourly UOP and theophylline levels at 24 hr (B) and 48 hr (C) on days +1 and +2, respectively. Theophylline levels at 48 hr were negatively correlated with UOP on day +2. We next investigated what factors, in addition to aminophylline administration, were associated with increased diuresis. Univariate analysis demonstrated that total daily furosemide dosing (p < 0.0001), chlorothiazide dosing (p < 0.0001), and fluid intake (p < 0.0001) were all significantly correlated with UOP over the 4-day time period, whereas inotrope and vasopressor use were not. A multivariate analysis using a GEE approach for repeated measures analysis was employed to investigate the association between UOP and the fluid intake and the administration of diuretics. Multivariate analysis demonstrated continued correlations with furosemide dosing (p = 0.0001), chlorothiazide dosing (p = 0.0168), fluid intake (p < 0.0001), and aminophylline administration when analyzed as a categorical variable (p < 0.0001) (Table 3). TABLE 3. Linear Regression Analysis of Factors Associated With Urine Output by Generalized Estimating Equation

We next investigated what factors, in addition to aminophylline administration, were associated with increased diuresis. Univariate analysis demonstrated that total daily furosemide dosing (p < 0.0001), chlorothiazide dosing (p < 0.0001), and fluid intake (p < 0.0001) were all significantly correlated with UOP over the 4-day time period, whereas inotrope and vasopressor use were not. A multivariate analysis using a GEE approach for repeated measures analysis was employed to investigate the association between UOP and the fluid intake and the administration of diuretics. Multivariate analysis demonstrated continued correlations with furosemide dosing (p = 0.0001), chlorothiazide dosing (p = 0.0168), fluid intake (p < 0.0001), and aminophylline administration when analyzed as a categorical variable (p < 0.0001) (Table 3). TABLE 3. Linear Regression Analysis of Factors Associated With Urine Output by Generalized Estimating Equation Aminophylline and Fluid Balance Given the ultimate importance of fluid balance in critically ill pediatric patients, the relationship between aminophylline administration and daily fluid balance (% volume/weight) was also examined (Fig. 3A). Fluid balance decreased significantly after the initiation of aminophylline on day +1 and day +2. The relationships between theophylline levels and fluid balance were also examined, and no correlations were found between theophylline levels at 24 hours or 48 hours and fluid balance on day +1 or +2 (Fig. 3, B and C). Univariate analysis demonstrated that fluid balance correlated with fluid intake (p < 0.0001), aminophylline administration when analyzed as a categorical variable (p < 0.0001), and chlorothiazide dosing (p = 0.025), but not aminophylline dosing (p = 0.59) or furosemide dosing (p = 0.5). Multivariate analysis using GEE regression analysis of factors associated with fluid balance continued to demonstrate a correlation between fluid balance and fluid intake (p < 0.0001), chlorothiazide dosing (p = 0.0002), and aminophylline administration as a categorical variable (p < 0.0001)(Table 4).

dosing (p = 0.5). Multivariate analysis using GEE regression analysis of factors associated with fluid balance continued to demonstrate a correlation between fluid balance and fluid intake (p < 0.0001), chlorothiazide dosing (p = 0.0002), and aminophylline administration as a categorical variable (p < 0.0001)(Table 4). TABLE 4. Linear Regression Analysis of Factors Associated With Fluid Balance by Generalized Estimating Equation Figure 3. Effect of aminophylline administration on fluid balance (% volume/weight) and the association with theophylline levels. Significantly lower fluid balance on days +1 and +2 (after aminophylline initiation) (A). Statistical differences were calculated using nonparametric comparisons for each pair using the Wilcoxon method (*p < 0.05, **p < 0.005, ***p < 0.0005). Linear regression analysis of daily fluid balance and theophylline levels at 24 hr (B) and 48 hr (C) on days +1 and +2, respectively.

er aminophylline initiation) (A). Statistical differences were calculated using nonparametric comparisons for each pair using the Wilcoxon method (*p < 0.05, **p < 0.005, ***p < 0.0005). Linear regression analysis of daily fluid balance and theophylline levels at 24 hr (B) and 48 hr (C) on days +1 and +2, respectively. Effect of Aminophylline on Laboratory Variables The effect aminophylline administration had on renal function as measured by sCr and BUN levels was also assessed. Although the sCr levels seemed to trend upwards slightly during the 4-day time interval, there was no statistically significant increase in sCr (Fig. S1, Supplemental Digital Content 2, http://links.lww.com/PCC/A668). In comparison, BUN levels rose significantly after aminophylline administration on day +2 when compared with day –2 (p < 0.001) and day –1 (p < 0.01) (Fig. S2, Supplemental Digital Content 3, http://links.lww.com/PCC/A669), presumably due to an increased prerenal state, although this could not be tested as fractional excretions of sodium were not available given the retrospective design. Importantly, no patients were known to have gastrointestinal bleeding, and thus, blood resorption was less likely to be the cause of elevated BUN over time.

sumably due to an increased prerenal state, although this could not be tested as fractional excretions of sodium were not available given the retrospective design. Importantly, no patients were known to have gastrointestinal bleeding, and thus, blood resorption was less likely to be the cause of elevated BUN over time. DISCUSSION Methylxanthines (caffeine) were first reported to increase urine production in patients with congestive heart failure and edema in 1864 (19). In more recent times, methylxanthines have been replaced by more powerful diuretics (20), although the use of aminophylline in the PICU as an adjunctive diuretic, albeit uncommon, has continued (13, 17). The peak serum concentration of aminophylline is reached within 30 minutes after IV dose, and its half-life varies with age: 20 hours for premature neonates, 3.5 hours for children, and 8.5 hours for adults. However, the clearance is quite variable, and it has been suggested that levels be closely monitored (21). Although there are no universal theophylline levels used among institutions for diuresis, our PICU standard guideline for aminophylline dosing targeted a goal theophylline trough of 4–8 μg/mL. However, it is possible that there is no need to target theophylline levels, and given the level-dependent negative side effect profile of aminophylline, we examined if theophylline trough levels correlated with UOP.

our PICU standard guideline for aminophylline dosing targeted a goal theophylline trough of 4–8 μg/mL. However, it is possible that there is no need to target theophylline levels, and given the level-dependent negative side effect profile of aminophylline, we examined if theophylline trough levels correlated with UOP. This study adds to the literature as it demonstrates that 24- and 48-hour theophylline trough levels are not predictive of UOP, and at 48 hours, in fact, are negatively correlated with UOP. Given that theophylline is partially renally cleared, these data suggest that theophylline levels are more dependent upon renal clearance, than renal clearance is dependent upon theophylline levels. Although intermittent measurement of theophylline levels may be warranted to guarantee the lack of toxicity, our data suggest that maintenance of theophylline trough levels above an arbitrary trough level is not necessary for the provision of the diuretic effect of aminophylline.

is dependent upon theophylline levels. Although intermittent measurement of theophylline levels may be warranted to guarantee the lack of toxicity, our data suggest that maintenance of theophylline trough levels above an arbitrary trough level is not necessary for the provision of the diuretic effect of aminophylline. By evaluating aminophylline therapy in our statistical analyses as both as a categorical and continuous variable, we were able to ask two separate questions regarding the relationship between aminophylline and diuresis: 1) does adding aminophylline increase UOP and improve fluid balance and 2) does a higher dose of aminophylline further increase UOP and improve fluid balance? Aminophylline administration, treated as a categorical variable, was correlated with increased UOP, but surprisingly, increased aminophylline dosing did not correlate with increased UOP. These data suggest that the addition of very low dose aminophylline is sufficient to generate improved diuresis and that further increases in aminophylline dosing may not enhance the diuretic effect. This result is in accordance with other studies that have found the diuretic benefit of aminophylline therapy (12, 13, 17, 22–25). Specifically, Tamburro et al (17) demonstrated an increase of UOP in 24 hours after aminophylline therapy from 35 patients, whereas BUN and creatinine concentration remained unchanged. Axelrod et al (24) demonstrated an improvement of renal function and UOP from aminophylline over a 7-day period among 31 cardiovascular ICU patients who had AKI. However, the same group performed a subsequent double-blinded, placebo-controlled, randomized control trial in 144 patients and did not demonstrate the same effect (15). In our retrospective study, aminophylline was used as an adjunctive diuretic, not to prevent or reverse AKI. To the best of our knowledge, this is the first article to investigate whether targeted theophylline levels are necessary for diuretic effect.

trial in 144 patients and did not demonstrate the same effect (15). In our retrospective study, aminophylline was used as an adjunctive diuretic, not to prevent or reverse AKI. To the best of our knowledge, this is the first article to investigate whether targeted theophylline levels are necessary for diuretic effect. The diuretic benefit of low dose aminophylline can be explained by its mechanism of adenosine receptor blockade at low dosage (theophylline concentration of 2–3 μg/mL) versus type IV phosphodiesterase inhibition at high levels (theophylline concentration > 10 μg/mL). Also, although low dose aminophylline augments diuresis, it likely does not confer bronchodilatory or anti-inflammatory properties from phosphodiesterase inhibition: an important caveat given that the majority of our patients suffered from respiratory failure. Elevated theophylline concentrations are associated with increased adverse effects including tachycardia, agitation, an increased risk for premature ventricular contractions, and arrhythmias. Due to our study design, inclusion only of patients that continued to receive aminophylline for 48 hours, we were not able to investigate the physiologic side effects of aminophylline/theophylline. However, given that our data indicate that diuretic effect is not dose-dependent, lower dosing may be more appropriate to optimize the risk-benefit ratio, and it may be prudent to avoid aminophylline boluses.

lline for 48 hours, we were not able to investigate the physiologic side effects of aminophylline/theophylline. However, given that our data indicate that diuretic effect is not dose-dependent, lower dosing may be more appropriate to optimize the risk-benefit ratio, and it may be prudent to avoid aminophylline boluses. Any conclusions drawn from this report regarding the correlation of aminophylline dosing and diuretic effect must be tempered by the obvious study limitations, most notably its retrospective design. Because our primary research question investigated the relationship between theophylline levels and UOP, we excluded many patients who received less than maintenance dosing aminophylline, or those that received aminophylline less than 48 hours after PICU admission. Hence, our study was not designed, nor should it be used, to investigate differential adverse effects of aminophylline dosing. Also, the demographics of our patient population are skewed toward a large Caucasian bias. However, no study has ever reported an ethnicity effect of aminophylline pharmacokinetics or pharmacodynamics. UOP was sometimes based on urine collection with diaper weights, which is inherently inaccurate although, we anticipate no systematic error from this practice. Our demographics were also skewed toward younger patients with more than 60% of our patient cohort between the ages of 6 weeks and 1 year old; however, this accurately reflects the general PICU/cardiac ICU patient population age demographics (26).

accurate although, we anticipate no systematic error from this practice. Our demographics were also skewed toward younger patients with more than 60% of our patient cohort between the ages of 6 weeks and 1 year old; however, this accurately reflects the general PICU/cardiac ICU patient population age demographics (26). Surprisingly, although furosemide dosing correlated with UOP, it did not correlate with fluid balance. This may be a power issue or due to a decrease in furosemide dosing because of improving fluid balance (Fig. 1B). Along this same line of thought, the diuretic effect seen with the initiation of aminophylline could be due to temporally related natural improvement of disease processes. We tried to account for this by establishing a 48 hours UOP and fluid balance baseline for each patient prior to the initiation of aminophylline, but we could not control for time in our analysis. To definitively determine how varying theophylline concentrations affect UOP would require a randomized control trial of several cohorts of patients, with each cohort having a theophylline trough goal and a protocol for attaining the goal. In the absence of such a trial, this study demonstrates that targeted theophylline levels are likely not needed for the maximal diuretic effect of aminophylline.

d require a randomized control trial of several cohorts of patients, with each cohort having a theophylline trough goal and a protocol for attaining the goal. In the absence of such a trial, this study demonstrates that targeted theophylline levels are likely not needed for the maximal diuretic effect of aminophylline. CONCLUSIONS In conclusion, aminophylline significantly improved UOP, and its diuretic effect was independent of both bolus dosing and the trough level. These results suggest that low dose aminophylline is a useful adjunctive diuretic and that no therapeutic trough level is required, although intermittent levels may be appropriate to assess for toxicity. Supplementary Material Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (http://journals.lww.com/pccmjournal). Drs. Tamburro and Halstead disclosed off-label product use of aminophylline as a diuretic. Dr. Tamburro’s institution received funding from the U.S. FDA Office of Orphan Product Development Grant Program, and he disclosed that Ony, LLC provided calfactant free of charge for the above listed U.S. FDA Office of Orphan Product Development Grant. The remaining authors have disclosed that they do not have any potential conflicts of interest.

The impact of medications with anticholinergic effects has long been recognized as a significant problem in adult medicine, such that multiple scales exist to measure the degree of “anticholinergic burden” (1–4). One reason for this concern is an association between delirium and medications with anticholinergic properties (1, 5–7), although studies of correlation in the adult ICU (AICU) have produced mixed results (7, 8). In the PICU, treatment with benzodiazepines—a drug class that has anticholinergic properties as well as potentiating the anticholinergic effect of other drugs—is associated with an increased risk of delirium. In one study, subjects who had received an anticholinergic had an odds ratio for developing delirium of 2.2 (p = 0.006) (9).

treatment with benzodiazepines—a drug class that has anticholinergic properties as well as potentiating the anticholinergic effect of other drugs—is associated with an increased risk of delirium. In one study, subjects who had received an anticholinergic had an odds ratio for developing delirium of 2.2 (p = 0.006) (9). The anticholinergic toxidrome includes either peripheral (i.e., tachycardia, hypertension, mydriasis, fever, constipation and urinary retention, myoclonus, tremor, and dry skin) or central (i.e., disorientation, confusion, hallucinations, and seizures) features, or both (10, 11). The “Anticholinergic Drug Scale” (ADS) rates individual medications using a variety of characteristics, including any known anticholinergic properties from pharmacologic receptor binding studies, known anticholinergic adverse events, or consensus expert opinion (2). The ADS scores range from 0 (none) to 1 (potential effect), to 2 (observed effect), or to 3 (effect almost always occurs) anticholinergic effects. For example, according to the ADS, diphenhydramine is given a value of 3, ranitidine a value of 2, and common PICU medications such as furosemide, midazolam, hydralazine, hydrocortisone, fentanyl, morphine, clindamycin, and gentamicin a score of 1 (2). Another tool is the Anticholinergic Cognitive Burden (ACB) scale, which is based on expert opinion and a systematic review of the literature (4). A value of 0–3 is assigned to 88 medications based on evidence of likely effect on cognitive function. Both the ADS and ACB scores have been used extensively in the adult population to sum up the total anticholinergic burden from multiple medications (5, 6, 8).

ert opinion and a systematic review of the literature (4). A value of 0–3 is assigned to 88 medications based on evidence of likely effect on cognitive function. Both the ADS and ACB scores have been used extensively in the adult population to sum up the total anticholinergic burden from multiple medications (5, 6, 8). In a recent systematic review, we were not able to identify any pediatric studies that assessed the anticholinergic drug burden in the PICU (12). Because many of the medications assigned a value on the ADS and ACB scores are commonly used in the PICU, we sought to quantify the extent of exposure to anticholinergic medications in a single center because it is a potentially modifiable risk factor for morbidity, including ICU delirium. METHODS The Boston Children’s Hospital (BCH) institutional review board (IRB) approved this retrospective cohort study (IRB-CR00020606-2). Eligible pediatric patients were identified via a computerized search system through the electronic medical record (EMR) programmed to identify subjects meeting the following criteria: age less than 18 years old; admitted to the PICU between January 2011 and December 2015; PICU length of stay of at least 15 days; and receiving ventilatory support (either invasive or noninvasive) during the first 15 days of admission. Exclusion criteria were newborns and patients with severe brain injury progressing to diagnosis of brain death during the admission. Patients with a PICU length of stay less than 15 days were excluded to reduce heterogeneity in the cohort.

support (either invasive or noninvasive) during the first 15 days of admission. Exclusion criteria were newborns and patients with severe brain injury progressing to diagnosis of brain death during the admission. Patients with a PICU length of stay less than 15 days were excluded to reduce heterogeneity in the cohort. All subjects received care in the same PICU, managed by a common group of PICU attendings and trainees, and subject to medication guidelines to the same degree. Our PICU manages sedative infusions using a nurse-implemented, goal-directed sedation protocol on which the Randomized Evaluation of Sedation Titration fOr REspiratory failure clinical trial was based (13). Data Collection The EMR data collection for each patient included patient demographics, medical history, clinical values from day of PICU admission, and throughout PICU stay. Severity of illness in the first 24 hours was measured using the Pediatric Risk of Mortality (PRISM)–III score (14). Medical history and reason for admission to PICU were collected from the PICU attending admission documentation. Medications administered during each of the first 15 days of the PICU stay were extracted from the medication administration record, including all changes to sedative infusion rates and intermittent bolus doses administered. Additional daily information included Withdrawal Assessment Tool (WAT)–1 scores (15), recorded by nursing on the patient flowsheet. Patients with any WAT-1 score over 3 at any point during data collection period were considered to have withdrawal.

es to sedative infusion rates and intermittent bolus doses administered. Additional daily information included Withdrawal Assessment Tool (WAT)–1 scores (15), recorded by nursing on the patient flowsheet. Patients with any WAT-1 score over 3 at any point during data collection period were considered to have withdrawal. Anticholinergic scores using the ADS (2) and ACB scale (4) were then calculated based on the value ascribed to each medication, using Microsoft Excel (Microsoft Office Standard 2010, Version 14.0.7190.5000). Dose adjustment for the daily dose of medications has been proposed for the ADS (2); however, based on the methods of Wolters et al (8), who studied ACB and ADS in AICU patients, we did not dose-adjust level 1 medications (which includes all sedative and analgesic medications). These data were managed using the Research Electronic Data Capture database hosted at BCH (16). Each patient chart was reviewed for diagnosis of “Delirium” by diagnostic codes or referral to the clinical psychology/psychiatry service at BCH. Psychiatry notes were reviewed for diagnosis of delirium and, along with administration of an antipsychotic medication (not administered before PICU admission), were all considered diagnostic for delirium. Widespread screening using bedside delirium assessment tools was not carried out as routine practice during the time period studied, for reasons explained in our previous review, for example, the overlap in case definition with benzodiazepine withdrawal and the anticholinergic toxidrome (12).

d diagnostic for delirium. Widespread screening using bedside delirium assessment tools was not carried out as routine practice during the time period studied, for reasons explained in our previous review, for example, the overlap in case definition with benzodiazepine withdrawal and the anticholinergic toxidrome (12). Statistical Analyses Descriptive data are presented as median (interquartile range [IQR]) because of presumed nonnormal distribution. Nonparametric comparison between groups was performed using the Kruskal-Wallis test. All statistical computations were carried out with IBM SPSS Statistics for Windows, Version 23 (IBM, Armonk, NY). Maximum, minimum, mean, median, and IQR of ADS and ACB scores for each patient over the 15 PICU days was determined using SPSS (IBM SPSS Statistics for Windows, Version 23 [IBM]). As the median ADS score reported in an AICU study was 2 (8), we defined an ADS score greater than 3 as “high,” and likewise as the 75th percentile of ADS scores in our cohort was 7, and we further defined a score greater than 7 as “very high.” We also calculated the total sum of all ADS scores per patient over the first 15 days of PICU stay, so-called “cumulative anticholinergic exposure,” to gain a measure of the consistency of high anticholinergic exposure, as we hypothesized that this would be a better measure of the consistency of high anticholinergic exposure throughout the 15 days.

um of all ADS scores per patient over the first 15 days of PICU stay, so-called “cumulative anticholinergic exposure,” to gain a measure of the consistency of high anticholinergic exposure, as we hypothesized that this would be a better measure of the consistency of high anticholinergic exposure throughout the 15 days. We hypothesized that ADS score would be associated with the primary benzodiazepine exposure, which was midazolam in this cohort, so maximum dose was calculated for each patient in mg/kg, and the subjects were divided into tertiles based on distribution (lower tertile, no midazolam; middle tertile, midazolam < 2.2 mg/kg/d; upper tertile, midazolam ≥ 2.2 mg/kg/d). ADS median score by patient was then compared based on tertile of midazolam dose using the Kruskal-Wallis test. We also hypothesized that anticholinergic burden would be associated with critical illness severity, so we analyzed the PRISM-III score (14). Since the median PRISM-III for the cohort was 9.5 for the cohort (IQR, 3–13.75), ADS score medians were analyzed for each quartile of the PRISM-III score. Withdrawal was defined per subject as a WAT-1 score of 3 or greater on any of the 15 days of data collection.

ness severity, so we analyzed the PRISM-III score (14). Since the median PRISM-III for the cohort was 9.5 for the cohort (IQR, 3–13.75), ADS score medians were analyzed for each quartile of the PRISM-III score. Withdrawal was defined per subject as a WAT-1 score of 3 or greater on any of the 15 days of data collection. RESULTS Eighty-eight PICU patients met the inclusion criteria for study (on average 18 cases/yr). The median age was 7.9 years (IQR, 1.4–14.3 yr) (Table 1). The cohort consisted of subjects with a high rate of comorbidities, with only 10 (11%; 95% CI, 6–20%) that were previously well, and only 23 (26%; 17–37%) not receiving any prescription medications before PICU admission. The median raw PRISM-III score was 9 (IQR, 3–13), and 53 subjects (60%; 49–71%) were on vasopressors on the day of PICU admission. During the course of the admission, 76 subjects (86%; 77–93%) were on continuous infusion of sedative agents. TABLE 1. Analysis of Patient Characteristics on PICU Admission

RESULTS Eighty-eight PICU patients met the inclusion criteria for study (on average 18 cases/yr). The median age was 7.9 years (IQR, 1.4–14.3 yr) (Table 1). The cohort consisted of subjects with a high rate of comorbidities, with only 10 (11%; 95% CI, 6–20%) that were previously well, and only 23 (26%; 17–37%) not receiving any prescription medications before PICU admission. The median raw PRISM-III score was 9 (IQR, 3–13), and 53 subjects (60%; 49–71%) were on vasopressors on the day of PICU admission. During the course of the admission, 76 subjects (86%; 77–93%) were on continuous infusion of sedative agents. TABLE 1. Analysis of Patient Characteristics on PICU Admission ADS Scores The daily ADS scores are shown in Table 2. The median ADS score across all 1,320 scores for 88 subjects was 5 (IQR, 3–7). When considering individual data for each of the 88 subjects, the median of individual maximum ADS score was 8 (IQR, 6–10), and the median number of days spent with high ADS score (above 3) was 11 (IQR, 8–14) out of the first 15 PICU days. There was no association between age, sex, comorbidities, or admission severity (PRISM-III) and either the median ADS scores for the population or individual maximum ADS scores. None of the PICU outcomes (i.e., mortality, length of stay, duration of mechanical ventilation, and diagnosis of delirium) was associated with ADS scores, with the exception of sedation withdrawal, which was correlated with maximum ADS score over the first 15 days (p = 0.042) (Table 3). Last, on examining the median ADS score for each of the 88 subjects, and then separating the population into quartiles, we found no differences in patient characteristics or PICU outcomes (Table 4).

ception of sedation withdrawal, which was correlated with maximum ADS score over the first 15 days (p = 0.042) (Table 3). Last, on examining the median ADS score for each of the 88 subjects, and then separating the population into quartiles, we found no differences in patient characteristics or PICU outcomes (Table 4). TABLE 2. Anticholinergic Exposure in PICU by Several Measures TABLE 3. Analysis of PICU Patient Admission Characteristics and Outcomes by Anticholinergic Drug Scale Scores TABLE 4. Admission Characteristics and PICU Outcomes by Quartile of Cumulative Anticholinergic Exposure, As Measured by Anticholinergic Drug Scale Score ACB Scores The ACB scores for the cohort were similar to ADS scores, with the notable difference that midazolam (ADS = 1) carries a score of 0, and ranitidine (ADS = 1) carries a score of 1 on the ACB. The median ACB score was 2 (IQR, 1–4) overall 1,320 PICU days measured. As the results of further analyses conducted using the ACB scores were very similar to the ADS score results, we elected to focus on ADS scores.

hat midazolam (ADS = 1) carries a score of 0, and ranitidine (ADS = 1) carries a score of 1 on the ACB. The median ACB score was 2 (IQR, 1–4) overall 1,320 PICU days measured. As the results of further analyses conducted using the ACB scores were very similar to the ADS score results, we elected to focus on ADS scores. Medications and ADS Scores The four medications that most commonly contributed to the ADS scores were all low-level (i.e., ADS score = 1) anticholinergic drugs (Fig. 1). Midazolam was the most commonly used medication in subjects who had a daily ADS score of 8 or greater (classified as “very high”), received by 94% of subjects on those days. The next most commonly administered medications were morphine, vancomycin, and steroids (including hydrocortisone, prednisone, prednisolone, methylprednisolone, and dexamethasone). Ranitidine (ADS score 2) was the fifth most common medication, contributing to 56% of daily scores greater than or equal to 8, and diphenhydramine (ADS score 3) contributed to 40% of all daily scores greater than or equal to 8. Figure 1. Medications contributing to high Anticholinergic Drug Scale (ADS) score. The medications (x-axis) contributing to the percentage of ADS scores above 8 (black bar) and 4–7 (gray bar). Individual contributions from named medication represented in parentheses on x-axis legend.

Medications and ADS Scores The four medications that most commonly contributed to the ADS scores were all low-level (i.e., ADS score = 1) anticholinergic drugs (Fig. 1). Midazolam was the most commonly used medication in subjects who had a daily ADS score of 8 or greater (classified as “very high”), received by 94% of subjects on those days. The next most commonly administered medications were morphine, vancomycin, and steroids (including hydrocortisone, prednisone, prednisolone, methylprednisolone, and dexamethasone). Ranitidine (ADS score 2) was the fifth most common medication, contributing to 56% of daily scores greater than or equal to 8, and diphenhydramine (ADS score 3) contributed to 40% of all daily scores greater than or equal to 8. Figure 1. Medications contributing to high Anticholinergic Drug Scale (ADS) score. The medications (x-axis) contributing to the percentage of ADS scores above 8 (black bar) and 4–7 (gray bar). Individual contributions from named medication represented in parentheses on x-axis legend. Midazolam Midazolam was administered to 73 of 88 (83; 73–90%) of the cohort at some point during the PICU stay. ADS scores were also computed after excluding all benzodiazepines, and the median score over the whole 15-day period for all subjects was 4 (IQR, 2–5). After excluding all sedative medications, the median ADS score for the whole population data dropped to 3 (IQR, 1–4). When midazolam dose per day in the whole population was divided into tertiles (lower tertile, no midazolam; middle tertile, midazolam < 2.2 mg/kg/d; upper tertile, midazolam ≥ 2.2 mg/kg/d), there were 460 of 1,320 days (34.8%) without midazolam exposure. The median ADS score was higher in the upper tertile compared with middle and lower tertiles (7 [IQR, 6–8] vs 4.5 [IQR, 4–6] vs 4 [IQR, 2–5]; p < 0.001) (Fig. 2).

midazolam; middle tertile, midazolam < 2.2 mg/kg/d; upper tertile, midazolam ≥ 2.2 mg/kg/d), there were 460 of 1,320 days (34.8%) without midazolam exposure. The median ADS score was higher in the upper tertile compared with middle and lower tertiles (7 [IQR, 6–8] vs 4.5 [IQR, 4–6] vs 4 [IQR, 2–5]; p < 0.001) (Fig. 2). Figure 2. The Anticholinergic Drug Scale (ADS) score based on dose of midazolam administration. Group 1 received no midazolam (median, 4; interquartile range [IQR], 2–5), group 2 less than 2.2 mg/kg/d (median, 4.5; IQR, 4–6), and group 3 greater than or equal to 2.2 mg/kg/d (median, 7; IQR, 6–8). p values represent Kruskal-Wallis comparison between groups.

le (ADS) score based on dose of midazolam administration. Group 1 received no midazolam (median, 4; interquartile range [IQR], 2–5), group 2 less than 2.2 mg/kg/d (median, 4.5; IQR, 4–6), and group 3 greater than or equal to 2.2 mg/kg/d (median, 7; IQR, 6–8). p values represent Kruskal-Wallis comparison between groups. DISCUSSION In this report, we have demonstrated a high level of exposure to anticholinergic medications in patients with respiratory failure requiring prolonged mechanical ventilation and sedation in the PICU. Compared with an AICU population, in which the median anticholinergic score was 2 (8), the median of 5 in our PICU study is cause for concern and further study. Additionally, in this cohort of children, a median of 8 of the first 15 days of admission was characterized by a high burden of anticholinergic medications. The medications associated with these elevated scores are commonly used in PICU. In fact, the two most prevalent—midazolam, followed by morphine—are mainstays of PICU sedation and analgesic practice (13). With the exception of ranitidine (ADS = 2) and diphenhydramine (ADS = 3), high ADS scores were attributed to medications with low anticholinergic toxicity (ADS = 1); however, toxicity is likely additive (2).

e two most prevalent—midazolam, followed by morphine—are mainstays of PICU sedation and analgesic practice (13). With the exception of ranitidine (ADS = 2) and diphenhydramine (ADS = 3), high ADS scores were attributed to medications with low anticholinergic toxicity (ADS = 1); however, toxicity is likely additive (2). In a recent review of the literature, we did not identify any studies that had evaluated and characterized anticholinergic burden in the pediatric population (12). However, studies have examined the incidence of delirium in the PICU as it relates to anticholinergic medication as a dichotomous variable, and reported exposure of 68–74% of subjects to this medication class (9, 17). In a recent study examining medications administered to AICU patients, the median number of “low potency” anticholinergic medications was 1.5 in those patients who were never delirious, versus 2.4 in those patients developing delirium (< 0.0001) (7). Another AICU study demonstrated a median daily ADS score of 2 (IQR, 1–3), with 90% of the scores being accounted for by lowest level mediations (8). Hence, in comparison, it appears that the burden of anticholinergic medications in our PICU cohort is considerably higher. It is certainly feasible that the impact of medications with anticholinergic properties is manifested differently in younger age groups due to varying parasympathetic activity; however, the evidence linking anticholinergic medications to delirium in children (9) warrants further study of this relationship.

ably higher. It is certainly feasible that the impact of medications with anticholinergic properties is manifested differently in younger age groups due to varying parasympathetic activity; however, the evidence linking anticholinergic medications to delirium in children (9) warrants further study of this relationship. Benzodiazepines have also long been associated with delirium in critically ill AICU (7, 18–20) and PICU patients (9, 17, 21). The biologic mechanisms underlying this association are multifactorial. Benzodiazepines activate γ-aminobutyric acid receptors in the CNS and alter levels of neurotransmitters including dopamine, serotonin, acetylcholine, norepinephrine, and glutamate (18, 22, 23). The medication most commonly associated with high anticholinergic burden in our cohort was midazolam, one of several benzodiazepines judged to have anticholinergic properties by the ADS (2, 24). The properties of benzodiazepines have been long discussed in the context of anticholinergic activity (2, 24). A study of adult palliative care patients showed that increased levels of midazolam are associated with decreased levels of serum cholinesterase activity (25), which may be another mechanism accounting for this association. We found that midazolam not only contributed to the total anticholinergic score but was also associated with higher exposure, greater than its individual contribution (Fig. 2). It follows that midazolam exposure would track with other common medications administered in sedated patients, including morphine and fentanyl, which also contribute to the ADS, so this association is not unexpected.

score but was also associated with higher exposure, greater than its individual contribution (Fig. 2). It follows that midazolam exposure would track with other common medications administered in sedated patients, including morphine and fentanyl, which also contribute to the ADS, so this association is not unexpected. Aside from the relationship between benzodiazepines and delirium, anticholinergic medication burden has been associated with delirium in several studies, although the results are mixed. A large AICU study did not find an association between increasing ADS score and onset of delirium (8). However, another large AICU study did find both low and high potency anticholinergic medications to be associated with delirium (7). One PICU study has also found an association between anticholinergic medication and delirium occurrence (9). Our retrospective cohort was not powered to detect a difference in rate of psychiatrist-diagnosed delirium, based on anticholinergic exposure. Our goal was to demonstrate the degree of exposure to these medications in PICU practice and raise awareness about the possibility of contributing to an anticholinergic toxidrome or delirium.

ive cohort was not powered to detect a difference in rate of psychiatrist-diagnosed delirium, based on anticholinergic exposure. Our goal was to demonstrate the degree of exposure to these medications in PICU practice and raise awareness about the possibility of contributing to an anticholinergic toxidrome or delirium. One likely impact of increased awareness and investigation of the role of anticholinergic medication burden in the PICU would be discussion of alternative medication choices to those with anticholinergic effects. Alternatives to ranitidine (ADS 2), diphenhydramine (ADS 3), and sedative agents with low-level anticholinergic effects (midazolam, morphine, and fentanyl) exist. Several studies have demonstrated a lower incidence of delirium with dexmedetomidine infusion compared with other common agents (26, 27). This study is limited by the lack of systematic screening for delirium using available scales, limiting our ability to draw any conclusion about an association with delirium prevalence (4). The relatively small sample size analyzed limits our power to rule out statistical association; however, the sample was limited to enhance homogeneity of the cohort with a prolonged PICU stay. The retrospective design also does not allow us to determine a cause and effect relationship for anticholinergic medication exposure and scale scores. However, we have established a high rate of anticholinergic medication exposure in these patients, which is likely to be a potentially modifiable contributor to PICU morbidity.

rospective design also does not allow us to determine a cause and effect relationship for anticholinergic medication exposure and scale scores. However, we have established a high rate of anticholinergic medication exposure in these patients, which is likely to be a potentially modifiable contributor to PICU morbidity. CONCLUSIONS PICU patients receive a large number of medications during sedation for mechanical ventilation, many of which also have low-level anticholinergic effects, which may be additive. Whether we are seeing discrete syndromes of anticholinergic toxicity, sedative effects and withdrawal, or delirium remains to be further explored (4). This exposure represents a modifiable risk factor that is, in our experience, not frequently considered by PICU clinicians. Alternative medications should be considered in patients who require a high number of medications with anticholinergic burden. Therefore, we propose further, prospective study of anticholinergic exposure using these validated scales, and its relationship with clinical signs and symptoms of anticholinergic toxicity, delirium and agitation, and other PICU morbidities. The authors acknowledge that a nonspecific funding source has been used for this article. The authors have disclosed that they do not have any potential conflicts of interest.

Mortality prognostic scores are valuable tools for assessing the quality of care provided to critically ill patients (1). These are mathematical models, based on the presumption that there is a predictable relationship among severity of illness, evidenced by certain physiologic alterations, patient characteristics (diagnoses or complex chronic conditions [CCCs]), and risk of death (2, 3). It is assumed that the probability of death calculated before the initiation of intensive care treatment is independent of the quality of care received in the PICU. Therefore, these scores might be employed for assessing the results of each institution when compared with other institutions, directly (by comparing adjusted mortality) or indirectly (by comparing the number of actual deaths with the number of deaths predicted by the model) (4). In the field of intensive care, the standardized mortality ratio (SMR), an indicator that compares the observed number of deaths with the predicted number of deaths in a specific period, is typically employed to assess the performance of PICUs (5). An SMR of 1 indicates a perfect agreement between observed and estimated mortality. If the performance of a PICU is higher or lower than expected, then the ratio will be below or above 1, respectively. However, for this interpretation to be valid, an adequate mortality prediction model is needed. At the same time, the use of prognostic mortality scores is essential for the assessment of quality of care in the PICU (6).

of a PICU is higher or lower than expected, then the ratio will be below or above 1, respectively. However, for this interpretation to be valid, an adequate mortality prediction model is needed. At the same time, the use of prognostic mortality scores is essential for the assessment of quality of care in the PICU (6). The Ministry of Health of Argentina recommends using the Pediatric Index of Mortality (PIM) 2 as a model for predicting the risk of death in PICUs (7). Similarly, this score is used for adjusting mortality by the Quality Benchmarking Program of the Argentine Society of Intensive Care (SATI-Q) (8, 9). The performance of PIM2 was assessed in our country and Latin America and showed an adequate capacity for discriminating between nonsurvivors and survivors (10, 11). However, from the moment the model was developed to this day, the quality of intensive care provided around the world has changed dramatically due to advancements in technology and treatments. In the last few years, modifications in the prediction capacity of the score have been detected in the countries where it was developed, showing an actual mortality below the PIM2 prediction (12, 13). In order to correct these changes observed in calibration, the authors published an updated version called PIM3 in 2013 (14). The performance of this new model is yet to be assessed in Argentina. Similarly, literature related to validations of the score in populations different from the ones it was generated in is scarce (15–17).

correct these changes observed in calibration, the authors published an updated version called PIM3 in 2013 (14). The performance of this new model is yet to be assessed in Argentina. Similarly, literature related to validations of the score in populations different from the ones it was generated in is scarce (15–17). As using updated tools for assessing results from PICUs is of vital importance, we designed this research whose main objective was to evaluate the performance of the PIM3 score in a sample of patients admitted to PICUs in Argentina. Employing the PIM2 score as an instrument for predicting mortality, locally validated but developed more than 20 years ago, may condition an overvalued estimation of the results obtained in the ICUs in a country. For this reason, determining if this new version of the model can be used to adjust the risk of death of critically ill children in our country is essential. If so, an updated instrument that allows establishing a comparison between local intensive care and international care will be available. MATERIALS AND METHODS We designed a multicenter, observational, prospective, cross-sectional study. PICUs of Argentina were invited to participate through the scientific societies that group pediatric intensivists of the country.

As using updated tools for assessing results from PICUs is of vital importance, we designed this research whose main objective was to evaluate the performance of the PIM3 score in a sample of patients admitted to PICUs in Argentina. Employing the PIM2 score as an instrument for predicting mortality, locally validated but developed more than 20 years ago, may condition an overvalued estimation of the results obtained in the ICUs in a country. For this reason, determining if this new version of the model can be used to adjust the risk of death of critically ill children in our country is essential. If so, an updated instrument that allows establishing a comparison between local intensive care and international care will be available. MATERIALS AND METHODS We designed a multicenter, observational, prospective, cross-sectional study. PICUs of Argentina were invited to participate through the scientific societies that group pediatric intensivists of the country. All patients requiring intensive care between 1 month and 16 years old, admitted consecutively in participating PICUs between May 15, 2016, and February 15, 2017, were included. Newborns were excluded because they are usually treated in neonatal ICUs in Argentina. Patients admitted from other PICUs and those referred to other units for continuation of treatment or still hospitalized by March 1, 2017, were excluded from the analyses.

s between May 15, 2016, and February 15, 2017, were included. Newborns were excluded because they are usually treated in neonatal ICUs in Argentina. Patients admitted from other PICUs and those referred to other units for continuation of treatment or still hospitalized by March 1, 2017, were excluded from the analyses. The following data were recorded during each admission: admission diagnosis, age, gender, length of stay in PICU, days of mechanical ventilation, outcome (survival or death) at discharge from PICU, and variables necessary for calculating the PIM3 score. The coefficients for each variable included in the model, the odds ratio, and their respective 95% CI are shown in Supplemental Table 1 (Supplemental Digital Content 1, http://links.lww.com/PCC/A772) (14). The presence of any CCC, as defined by the Feudtner classification (18), at the time of admission was also recorded. For describing participating units, number of beds, type of institution (general, pediatric), and type of management (private, public, or social security) were also recorded. The instrument employed for data collection was the SATI-Q software (Hardineros Sistemas, Buenos Aires, Argentina), a computing tool provided free of charge to Argentine PICUs that voluntarily participate in the SATI-Q Program. This program is an initiative sponsored by SATI with the purpose of collecting data related to quality standards in intensive care (9). For this study, the software was updated to include the PIM3 score calculation as a new function.

free of charge to Argentine PICUs that voluntarily participate in the SATI-Q Program. This program is an initiative sponsored by SATI with the purpose of collecting data related to quality standards in intensive care (9). For this study, the software was updated to include the PIM3 score calculation as a new function. To guarantee the quality of the data, each participating PICU designated a contact in charge of data recording and supervision. For standardization of the PIM3 calculation method, a standard operating procedure manual was created and delivered to each PICU. For training purposes, each person in charge in every unit was asked to calculate the score for five example cases. The results were sent via e-mail and subsequently discussed with the main research team. Each PICU sent the database within the first 2 months of data collection. At that time, the understanding of the protocol and the guidelines for the construction of the score were assessed, and the necessary modifications were implemented. An e-mail address for direct queries was made available for clarifying any doubts that arose during the study period. Once the data collection period concluded, each participating unit sent the database to the coordination center for analysis. The data were sent encrypted and anonymized, for ensuring the safety of the data and the protection of personal information. The database is registered in the Argentine National Directorate for the Protection of Personal Data.

concluded, each participating unit sent the database to the coordination center for analysis. The data were sent encrypted and anonymized, for ensuring the safety of the data and the protection of personal information. The database is registered in the Argentine National Directorate for the Protection of Personal Data. Ethical Considerations The ethical and scientific aspects of the protocol were assessed and approved by the Research Ethics Committees of the participating units. In all cases, the need for informed consent for participation was waived because the data collected were considered routine practice in every PICU, the data protection requirements were met, and due to the observational characteristics of the study. Statistical Analysis Continuous quantitative variables were expressed as median and interquartile range (IQR) according to their distribution. Continuous discrete variables were expressed as median and range, and categorical variables were expressed as frequencies and percentages. The performance of the PIM3 score was assessed by analyzing its discrimination and calibration in the general population and different subgroups (age, diagnoses at admission, and presence of CCC). Simultaneously, SMR and its 95% CI were analyzed.

and categorical variables were expressed as frequencies and percentages. The performance of the PIM3 score was assessed by analyzing its discrimination and calibration in the general population and different subgroups (age, diagnoses at admission, and presence of CCC). Simultaneously, SMR and its 95% CI were analyzed. The discrimination or ability of the model to differentiate between survivors and nonsurvivors was assessed by calculating the area under the receiver operating characteristic curve (AUC-ROC) and its 95% CI. Calibration or degree of agreement between the number of predicted and observed events was calculated by using the Hosmer-Lemeshow goodness-of-fit test in the general population and stratified by deciles of risk. In order to analyze the performance of PIM3 in the different subgroups, age groups were classified as follows: 1–11 months/12–59 months/60–119 months/120–191 months. The analyzed diagnostic groups were as follows: 1) cardiac (including postoperative), 2) injury, 3) neurologic, 4) postoperative (noncardiac), 5) respiratory, and 6) miscellaneous (19). The AUC-ROC and SMR with their corresponding 95% CI were calculated in each age subgroup, diagnostic category, and according to the presence of any CCC.

In order to analyze the performance of PIM3 in the different subgroups, age groups were classified as follows: 1–11 months/12–59 months/60–119 months/120–191 months. The analyzed diagnostic groups were as follows: 1) cardiac (including postoperative), 2) injury, 3) neurologic, 4) postoperative (noncardiac), 5) respiratory, and 6) miscellaneous (19). The AUC-ROC and SMR with their corresponding 95% CI were calculated in each age subgroup, diagnostic category, and according to the presence of any CCC. Sample size was calculated based on the formula N = 10 k/p, where N is the minimum number of cases to be included, “k” is the number of independent variables included in the PIM3 logistic regression model, and “p” is the smallest proportion of expected positive cases in the population (deaths) (20). As PIM3 comprises 13 independent variables and expected proportion of deaths was calculated at 8% according to the latest general pediatric reports of the SATI-Q program (21), a sample size of 1,625 patients was calculated for obtaining a minimum of 130 events. All statistical analyses were performed using Excel 2010 (Microsoft, Redmond, WA), Access 2010 (Microsoft, Redmond, WA), MedCalc version 16.4 (MedCalc Software bvba, Ostend, Belgium), and STATA IC/11 (StataCorp LLC, College Station, TX).

Sample size was calculated based on the formula N = 10 k/p, where N is the minimum number of cases to be included, “k” is the number of independent variables included in the PIM3 logistic regression model, and “p” is the smallest proportion of expected positive cases in the population (deaths) (20). As PIM3 comprises 13 independent variables and expected proportion of deaths was calculated at 8% according to the latest general pediatric reports of the SATI-Q program (21), a sample size of 1,625 patients was calculated for obtaining a minimum of 130 events. All statistical analyses were performed using Excel 2010 (Microsoft, Redmond, WA), Access 2010 (Microsoft, Redmond, WA), MedCalc version 16.4 (MedCalc Software bvba, Ostend, Belgium), and STATA IC/11 (StataCorp LLC, College Station, TX). RESULTS Fifty-two PICUs agreed to participate in the study. After the data collection period concluded, 49 units sent their records. PICUs characteristics and distribution of admitted patients according to the setting are described in Table 1 (22). The median number of beds per PICU was 10 (range of 3–26). The median number of admitted patients per unit during the study period was 108 (IQR, 53–182); 38 PICUs (77, 5%) had less than or equal to 200 admissions. TABLE 1. Characteristics of PICUs and Volume of Admissions

RESULTS Fifty-two PICUs agreed to participate in the study. After the data collection period concluded, 49 units sent their records. PICUs characteristics and distribution of admitted patients according to the setting are described in Table 1 (22). The median number of beds per PICU was 10 (range of 3–26). The median number of admitted patients per unit during the study period was 108 (IQR, 53–182); 38 PICUs (77, 5%) had less than or equal to 200 admissions. TABLE 1. Characteristics of PICUs and Volume of Admissions During the study period, 7,075 admissions were registered in the participating PICUs; 473 (6.68%) were excluded from the analysis because of exclusion criteria, for being still hospitalized at the end of the study period or due to the existence of missing data necessary for the calculation of PIM3 (Fig. 1). A total of 6,602 records were analyzed. Their characteristics are detailed in Table 2. Respiratory conditions were the main reason for admission in the PICU, and patients less than 1 year old were the predominant age group. Although the sample showed high prevalence of CCC, variability according to reason for admission was evident. Only 4% of patients admitted due to injuries had some CCC versus 85.8% of admissions due to cardiac conditions. TABLE 2. Characteristics of the Patient's Population (n = 6,602) Figure 1. Flow chart of the study population.

During the study period, 7,075 admissions were registered in the participating PICUs; 473 (6.68%) were excluded from the analysis because of exclusion criteria, for being still hospitalized at the end of the study period or due to the existence of missing data necessary for the calculation of PIM3 (Fig. 1). A total of 6,602 records were analyzed. Their characteristics are detailed in Table 2. Respiratory conditions were the main reason for admission in the PICU, and patients less than 1 year old were the predominant age group. Although the sample showed high prevalence of CCC, variability according to reason for admission was evident. Only 4% of patients admitted due to injuries had some CCC versus 85.8% of admissions due to cardiac conditions. TABLE 2. Characteristics of the Patient's Population (n = 6,602) Figure 1. Flow chart of the study population. Observed mortality in the sample was 8.04% (531/6,602), whereas mortality predicted by PIM3 was 6.17% (407 deaths). SMR was 1.3 (95% CI, 1.2–1.42). The difference between the number of deaths observed and PIM3 predicted deaths was statistically significant (p < 0.001). The AUC-ROC for the entire cohort was 0.83 (95% CI, 0.82–0.85), showing an adequate performance of the score for discriminating between nonsurvivors and survivors.

Observed mortality in the sample was 8.04% (531/6,602), whereas mortality predicted by PIM3 was 6.17% (407 deaths). SMR was 1.3 (95% CI, 1.2–1.42). The difference between the number of deaths observed and PIM3 predicted deaths was statistically significant (p < 0.001). The AUC-ROC for the entire cohort was 0.83 (95% CI, 0.82–0.85), showing an adequate performance of the score for discriminating between nonsurvivors and survivors. Table 3 shows the observed mortality and expected mortality in the different risk deciles, according to the goodness-of-fit test (Hosmer-Lemeshow). In all cases, the observed mortality was higher than PIM3 predicted mortality. The difference was statistically significant in the general population and for most deciles of mortality risk (χ2, 135.63; 8 degrees of freedom; p < 0.001). However, the difference between observed and expected mortality in the highest predicted mortality deciles (> 6.48%) was not statistically significant. TABLE 3. Hosmer-Lemeshow Goodness-of-Fit Test for Deciles of Mortality Risk: χ2: 135.63; 8 degree of freedom; p < 0.001 An analysis of discrimination and calibration of the model according to the volume of PICU number of admissions showed that for units with less than or equal to 200 admissions, the AUC-ROC was 0.84 (95% CI, 0.82–0.87) versus 0.82 (95% CI, 0.80–0.85) in units with more than 200 admissions. The SMR was 1.37 (95% CI, 1.21–1.55) and 1.25 (95% CI, 1.1–1.4), respectively.

the model according to the volume of PICU number of admissions showed that for units with less than or equal to 200 admissions, the AUC-ROC was 0.84 (95% CI, 0.82–0.87) versus 0.82 (95% CI, 0.80–0.85) in units with more than 200 admissions. The SMR was 1.37 (95% CI, 1.21–1.55) and 1.25 (95% CI, 1.1–1.4), respectively. Analysis by Age Group The discrimination ability of PIM3 was adequate in all age groups. The observed mortality was higher than the mortality predicted by the score in all groups, especially in patients more than 120 months old. The difference between observed mortality and PIM3 predicted mortality was statistically significant (Table 4), except in children less than 1 year old. TABLE 4. Model Fit and Discrimination by Age Groups Analysis by Diagnostic Group PIM3 showed adequate discrimination ability in all diagnostic groups. The lowest discrimination ability was observed in patients admitted for respiratory disease, showing an AUC-ROC of 0.70, with a 95% CI lower limit of 0.66. Regarding the score calibration, the observed mortality was higher than the predicted mortality in all diagnostic groups, except in patients admitted for injuries. The difference between the observed mortality and PIM3 predicted mortality was statistically significant in all diagnostic groups, except in the neurologic category, injury category, and, although borderline, in postoperative admissions (Table 5). TABLE 5. Model Fit and Discrimination by Admission Diagnostic Groups

Regarding the score calibration, the observed mortality was higher than the predicted mortality in all diagnostic groups, except in patients admitted for injuries. The difference between the observed mortality and PIM3 predicted mortality was statistically significant in all diagnostic groups, except in the neurologic category, injury category, and, although borderline, in postoperative admissions (Table 5). TABLE 5. Model Fit and Discrimination by Admission Diagnostic Groups Analysis According to the Presence of CCC PIM3 showed adequate discrimination ability in patients with some CCC when admitted to the PICU and in previously healthy patients. In this group, the observed mortality was higher than the PIM3 predicted mortality; the difference was statistically significant (276 observed deaths vs 179.3 predicted deaths). Although the observed mortality in previously healthy patients was higher than the expected mortality, this difference was not statistically significant (Table 6). TABLE 6. Model Fit and Discrimination According to the Presence of Chronic Complex Conditions DISCUSSION Mortality risk prediction models in PICUs are usually built in developed countries, based on a population with particular characteristics according to its case mix, available resources, and health system organization. Before being implemented by a country as tools for measuring intensive care quality and individual performance of PICUs, they must be validated in a locally representative sample of patients.

on a population with particular characteristics according to its case mix, available resources, and health system organization. Before being implemented by a country as tools for measuring intensive care quality and individual performance of PICUs, they must be validated in a locally representative sample of patients. This study was carried out to assess the performance of the PIM3 score in a population of patients admitted to PICUs in Argentina, a medium-to-high income country (according to the World Bank classification), and with health system characteristics different from the countries where the score was developed (23). The obtained results indicate that the score has an adequate capacity for discriminating between survivors and nonsurvivors, in the general population and all the different age and diagnostic subgroups. However, the observed mortality exceeds the mortality predicted by the score. In general terms, its performance is comparable to PIM2 according to the validation study carried out in Argentina in 2009 and Latin America in 2013, both in terms of discrimination and calibration capacity (10, 11). This study yielded a PIM3 AUC-ROC of 0.83 in the general population: 83% of nonsurvivors showed a higher PIM3 predicted death probability than survivors, compared with 88% of patients in the population in which the score was developed (14). Similarly, previous PIM2 validation studies carried out in Argentina and Latin America showed an adequate discrimination ability, with AUC-ROCs of 0.84 and 0.82, respectively (10, 11).

PIM3 predicted death probability than survivors, compared with 88% of patients in the population in which the score was developed (14). Similarly, previous PIM2 validation studies carried out in Argentina and Latin America showed an adequate discrimination ability, with AUC-ROCs of 0.84 and 0.82, respectively (10, 11). Instead of 407 predicted deaths, 531 deaths were observed in the sample. The observed mortality was higher than the expected mortality in all intervals of risk probability and in the majority of the analyzed subgroups (age, diagnostic, presence of CCC). Similarly, during the PIM2 validation study carried out in Argentina in 2009, 297 deaths were observed versus 246 predicted deaths. This tends to happen when mortality risk prediction scores are used in populations other than those they were developed in, especially when said populations have different characteristics, in terms of admission pathologies in PICUs, comorbidities, and health system (fragmentation, accessibility, available resources, and social-sanitary conditions). Regarding the characteristics of the admitted population, we can mention that PIM3 predicted mortality in our population was 6.17%, higher than the mortality predicted in the regions in which the model was developed (4.1% in the United Kingdom/Ireland and 2.8% in Australasia). Only 18.8% of patients in our sample had an elective admission versus 41% in the original population. Likewise, 16% of patients in our population were admitted for postsurgery recovery versus 39.7% of patients in the original sample.

the model was developed (4.1% in the United Kingdom/Ireland and 2.8% in Australasia). Only 18.8% of patients in our sample had an elective admission versus 41% in the original population. Likewise, 16% of patients in our population were admitted for postsurgery recovery versus 39.7% of patients in the original sample. The profile of children admitted in PICU in our population, more severely ill and non elective admissions, may reflect differences in admission criteria in argentine PICUs, or less accessibility to these units. These differences might not be adequately expressed by the model, affecting its performance when used in our area. But, at the same time, the excess of deaths observed might be interpreted as differences in the quality of care provided in our PICUs compared with the units in which PIM3 was developed. So far, few studies assessing the performance of the model in external populations have been published. In a retrospective study, Wofler et al (15) reported an adequate performance of the score in a sample of Italian PICUs. In this population, AUC-ROC was 0.88 and SMR was 0.98. There were no statistically significant differences between predicted and observed mortality. On the contrary, Lee et al (16) reported an AUC-ROC of 0.77 and SMR of 1.29 in a sample of 1,656 patients from one Korean PICU. The PIM3 score showed higher performance in Italy, possibly as it has a population and health system of characteristics similar to the United Kingdom and Australasia.

d and observed mortality. On the contrary, Lee et al (16) reported an AUC-ROC of 0.77 and SMR of 1.29 in a sample of 1,656 patients from one Korean PICU. The PIM3 score showed higher performance in Italy, possibly as it has a population and health system of characteristics similar to the United Kingdom and Australasia. The analysis by age group indicated that teenagers showed the greatest difference between observed mortality and model-predicted mortality, reporting an SMR of 1.89. Similarly, Wofler et al (15) reported an SMR of 1.4 for this population in Italian PICUs (15). These groups will likely show characteristics affecting the score performance, such as different CCCs that were not considered as an adjustment variable by PIM3. This challenge in mortality risk prediction for teenage patients is also evidenced by other scores built with different statistical techniques such as the one introduced by Arzeno et al (2), who suggest the need to develop specific scores for this population.

that were not considered as an adjustment variable by PIM3. This challenge in mortality risk prediction for teenage patients is also evidenced by other scores built with different statistical techniques such as the one introduced by Arzeno et al (2), who suggest the need to develop specific scores for this population. In the analysis performed according to the diagnosis at admission, observed mortality was higher than mortality predicted by PIM3 in all groups, except in patients admitted for injury. This is similar to the results observed in the PIM2 validation carried out in Latin America, in which 22 Argentinian PICUs participated (11). This finding is possibly related to a higher score performance in patients without previous comorbidity because only 25 patients (4%) admitted for injuries had some CCC in our sample. Calibration capacity was inadequate, showing statistically significant differences between expected and observed mortality in all groups, except in patients admitted for neurologic problems and injuries. Paradoxically, calibration in the original sample failed in patients admitted for neurologic problems.

n our sample. Calibration capacity was inadequate, showing statistically significant differences between expected and observed mortality in all groups, except in patients admitted for neurologic problems and injuries. Paradoxically, calibration in the original sample failed in patients admitted for neurologic problems. The analysis of the population according to the presence of CCC showed that PIM3 performance in patients with previous comorbidities was inadequate in terms of calibration. Although mortality in patients without CCC was higher than expected (6.99% vs 6.25%), the difference was not statistically significant and SMR was 1.12. In contrast, the SMR for patients with CCC was 1.54. Currently, no studies on the assessment of score performance in this particular group have been published. The PIM3 model considers leukemia, postinduction lymphoma, neurodegenerative diseases, or bone marrow transplant, as adjustment variables of high and very high risk of death, but excludes as risk factors conditions such as HIV or post-liver transplant admissions, which are still associated with higher mortality rates in our country. According to reports from the World Health Organization, the HIV-AIDS mortality rate reached 0.9 per 100,000 inhabitants in Australia versus 8.9 per 100,000 in Argentina (24) by 2012. Similarly, other oncologic or immune-hematologic pathologies, which are not considered in the score as risk factors, may be associated with a higher mortality rate in PICUs in our region. Decreased availability of palliative care and anticipated decisions regarding end-of-life care may influence access to ICUs for patients with low recovery capacity in Argentina. The above mentioned factors might partially explain a better performance of the model in children admitted without comorbidities.

our region. Decreased availability of palliative care and anticipated decisions regarding end-of-life care may influence access to ICUs for patients with low recovery capacity in Argentina. The above mentioned factors might partially explain a better performance of the model in children admitted without comorbidities. Other conditions that might explain the excess of deaths observed in our population are the characteristics of the Argentinean health system, highly fragmented and not centralized, existing high number of PICUs that treat a small volume of patients (25). In our study, 77% of the units that participated had a low volume of admissions (200 or less). In these units, the SMR was higher than in the units with larger volume of admissions. It is possible that reductions in mortality could be achieved if critical patients were admitted to large PICUs, as proposed by other authors (26).

udy, 77% of the units that participated had a low volume of admissions (200 or less). In these units, the SMR was higher than in the units with larger volume of admissions. It is possible that reductions in mortality could be achieved if critical patients were admitted to large PICUs, as proposed by other authors (26). As a limitation of the study, we can mention that neonatal patients were not included even though the original PIM3 model has included this in their assessments. This is because neonatal (<28 d old) admissions are managed in neonatal ICUs in our country. Another limitation is that not all PICUs in the country were included, as no entity groups them in a mandatory manner. However, units in public and private hospitals, and in general and pediatric hospitals, are represented in the sample. Furthermore, PICUs from all five regions of the country considered by the Ministry of Health took part in the study although the central region provinces showed a clear predominance, which reflects the concentration of the Argentine population in that region (27).

and pediatric hospitals, are represented in the sample. Furthermore, PICUs from all five regions of the country considered by the Ministry of Health took part in the study although the central region provinces showed a clear predominance, which reflects the concentration of the Argentine population in that region (27). In contrast, the study results could be generalized for PICUs that are members of the SATI-Q pediatric program, as 30 of the 33 units participating in the registry in 2015 were also included in this research (28). This program, sponsored by the SATI since 2005, has the voluntary participation of units located in different provinces of the country. It represents a source of free and publicly available data, which provides information on quality indicators in Argentinean PICUs for benchmarking purposes. A general report on predefined quality indicators, like the SMR resulting from the analysis of the total number of admissions in the participating units, is prepared and published annually.

free and publicly available data, which provides information on quality indicators in Argentinean PICUs for benchmarking purposes. A general report on predefined quality indicators, like the SMR resulting from the analysis of the total number of admissions in the participating units, is prepared and published annually. Understanding the performance of PIM3 in a local representative sample allows us to use the score as a mortality prediction tool for the construction of SMR in each individual PICU and at a national level. The results of this study show that using PIM3 to predict mortality, the actual SMR for PICUs participating in the SATI-Q program is 1.3, instead of the values observed in recent years using PIM2 (21). The performance of each participating PICU can be assessed by comparing their obtained SMRs with this value. This analysis conducted on an annual basis, can detect changes in population characteristics and the score performance, and will allow the comparison of each PICU against a local SMR and the comparison of SMR over time, as it has been performed with the use of PIM2 to date.

omparing their obtained SMRs with this value. This analysis conducted on an annual basis, can detect changes in population characteristics and the score performance, and will allow the comparison of each PICU against a local SMR and the comparison of SMR over time, as it has been performed with the use of PIM2 to date. We suggest that it would be optimal to switch from PIM2 to PIM3 as the score to predict mortality in Argentine PICUs given that using a nonupdated model, like PIM2, might result in the misconception that care in our units is better than it actually is. On the other hand, using an up-to-date tool to compare care provided in local PICUs with international care for benchmarking purposes is necessary to highlight characteristics in the local care model that could have an impact on our patients’ outcomes. An objective measurement of the results is necessary to evaluate the impact of the measures aimed at improving the care of the critical pediatric patient, either by improving the detection, the quality of the initial care, the accessibility to the PICUs, or the human and technological resources available in them.

We suggest that it would be optimal to switch from PIM2 to PIM3 as the score to predict mortality in Argentine PICUs given that using a nonupdated model, like PIM2, might result in the misconception that care in our units is better than it actually is. On the other hand, using an up-to-date tool to compare care provided in local PICUs with international care for benchmarking purposes is necessary to highlight characteristics in the local care model that could have an impact on our patients’ outcomes. An objective measurement of the results is necessary to evaluate the impact of the measures aimed at improving the care of the critical pediatric patient, either by improving the detection, the quality of the initial care, the accessibility to the PICUs, or the human and technological resources available in them. CONCLUSIONS This study assessed the performance of the PIM3 score in a large sample of patients admitted to PICUs in Argentina. The score showed an adequate ability to discriminate between the population of patients who survive and those who die. Instead, observed mortality was higher than predicted mortality in the general population and the population stratified by age, diagnosis or presence of CCC. The use of an updated instrument such as PIM3 will allow an actual comparison between pediatric intensive care provided in the country and care provided internationally. This might also allow future planning of pediatric intensive care services in Argentina.

e population stratified by age, diagnosis or presence of CCC. The use of an updated instrument such as PIM3 will allow an actual comparison between pediatric intensive care provided in the country and care provided internationally. This might also allow future planning of pediatric intensive care services in Argentina. ACKNOWLEDGMENTS We express our most sincere gratitude to all PICU members who participated in the study and for their commitment to the project. We also thank the Argentine Society of Intensive Care who provided the SATI-Q software as a data collection tool for this study and has continuously supported research projects proposed by its members. Supplementary Material Members of VALIDARPIM3 Argentine Group are listed in Appendix 1.

ACKNOWLEDGMENTS We express our most sincere gratitude to all PICU members who participated in the study and for their commitment to the project. We also thank the Argentine Society of Intensive Care who provided the SATI-Q software as a data collection tool for this study and has continuously supported research projects proposed by its members. Supplementary Material Members of VALIDARPIM3 Argentine Group are listed in Appendix 1. This work was performed at the PICUs of the following institutions: Hospital de Niños Ricardo Gutiérrez; Hospital de Niños Sor María Ludovica; Hospital Universitario Austral; Hospital General de Agudos “Carlos G. Durand”; Hospital de Pediatría “Prof. Dr. Juan P. Garrahan” PICU 44 y PICU 72; Hospital Zonal Ramón Carrillo; Hospital Interzonal General de Agudos “Dr. Abraham Piñeyro”; Hospital Del Niño Jesús de Tucumán; Hospital Dr. Humberto Notti; Hospital Pediátrico Juan Pablo II; Hospital Zonal General de Agudos Dr. Lucio Melendez; Sanatorio Anchorena; Complejo Medico Policial “Churruca-Visca”; Hospital De Niños De La Santísima Trinidad; Sanatorio De La Trinidad Mitre; Hospital General de Niños Pedro de Elizalde; Hospital Provincia de Rosario; Hospital Español de Rosario; Hospital El Cruce, Dr. Néstor Carlos Kirchner. Alta Complejidad en Red; Clinica del Niño de Quilmes; Hospital Guillermo Rawson; Hospital Publico Materno Infantil de Salta; Hospital de Niños Zona Norte; Hospital de San Luis; Clinica Modelo de Moron; Hospital Dr. H Notti Cardiovascular PICU; Hospital Penna; Hospital Isola Puerto Madryn; Hospital Regional de Rio Gallegos, Santa Cruz; Sanatorio de Niños de Rosario; Hospital Regional Castro Rendón; Hospital Zonal de Caleta Olivia; Hospital Avelino Castelán; Hospital de Niños V J Vilela; Hospital Interzonal Especializado Materno Infantil de Mar del Plata; Hospital de Niños Sor María Ludovica, Cardiovascular PICU; Hospital de Clínicas José de San Martin; Hospital Materno Infantil San Roque; Clinica Velez Sarfield; Sanatorio Sagrado Corazón; Hospital Pediátrico del Niño Jesús de Córdoba; Hospital Municipal de Trauma y Emergencias Dr. Federico Abete; Hospital de Niños Dr. Héctor Quintana; Corporación Medica de General San Martin; Hospital de Quemados; Hospital de Niños de San Justo; Sanatorio Trinidad Ramos Mejía; Hospital de Niños Dr. O Allassia; and Sociedad Argentina de Terapia Intensiva.

de Córdoba; Hospital Municipal de Trauma y Emergencias Dr. Federico Abete; Hospital de Niños Dr. Héctor Quintana; Corporación Medica de General San Martin; Hospital de Quemados; Hospital de Niños de San Justo; Sanatorio Trinidad Ramos Mejía; Hospital de Niños Dr. O Allassia; and Sociedad Argentina de Terapia Intensiva. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (http://journals.lww.com/pccmjournal). Supported, in part, by Healthcare Research Grants “Dr. Abraam Sonis,” individual category, granted by the National Ministry of Health, through the Health Research Directorate. The authors have disclosed that they do not have any potential conflicts of interest.

Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (http://journals.lww.com/pccmjournal). Supported, in part, by Healthcare Research Grants “Dr. Abraam Sonis,” individual category, granted by the National Ministry of Health, through the Health Research Directorate. The authors have disclosed that they do not have any potential conflicts of interest. APPENDIX 1. MEMBERS OF VALIDARPIM3 ARGENTINE GROUP Luis Aramayo (Hospital Zonal Ramón Carrillo, Rio Negro); Pedro Portero (Hospital Interzonal General de Agudos “Dr. Abraham Piñeyro,” Buenos Aires); Priscilla Botta (Hospital Del Niño Jesús, Tucumán); Marta Mosciaro (Hospital Dr. Humberto Notti, Mendoza); Segundo Español (Hospital pediátrico Juan Pablo II, Corrientes); Walter Lorenz (Hospital Zonal General de Agudos Dr. Lucio Melendez, Buenos Aires); Alberto Hernández (Hospital de Pediatría” Prof. Dr. Juan P. Garrahan” Unidad 72 Ciudad Autónoma de Buenos Aires); Rosana Poterala (Sanatorio Anchorena, Ciudad Autónoma de Buenos Aires); Gustavo Gonzalez (Complejo Medico Policial “Churruca-Visca,” Ciudad Autónoma de Buenos Aires); Ramon Pogonza (Hospital De Niños De La Santísima Trinidad, Córdoba); Facundo Jorro (Sanatorio De La Trinidad Mitre, Ciudad Autónoma de Buenos Aires); Carolina Sabatini (Hospital General de Niños Pedro de Elizalde, Ciudad Autónoma de Buenos Aires); Marta De Barelli (Hospital Provincial, Rosario, Hospital Español, Rosario); Karina Cinquegranni (Hospital El Cruce Dr. Néstor Carlos Kirchner, Alta Complejidad en Red, Buenos Aires); Sergio Suarez (Clinica del Niño, Quilmes, Buenos Aires); Javier Ponce (Hospital Guillermo Rawson, San Juan); Sandra Chuchuy (Hospital Publico Materno Infantil de Salta, Salta); Gustavo Sciolla (Hospital de Niños Zona Norte, Santa Fe); Maria Eugenia Passini (Hospital de San Luis, San Luis); Rose Marie Deheza (Clínica Modelo, Morón, Buenos Aires); Maria Mackern (Hospital Dr. H Notti Cardiovascular, Mendoza); Juan Fabris (Hospital Penna, Bahia Blanca, Buenos Aires); Ana Rodriguez Calvo (Hospital Isola Puerto Madryn, Chubut); Claudia Benaroya (Hospital Regional de Rio Gallegos, Santa Cruz); Maria A.

n Luis); Rose Marie Deheza (Clínica Modelo, Morón, Buenos Aires); Maria Mackern (Hospital Dr. H Notti Cardiovascular, Mendoza); Juan Fabris (Hospital Penna, Bahia Blanca, Buenos Aires); Ana Rodriguez Calvo (Hospital Isola Puerto Madryn, Chubut); Claudia Benaroya (Hospital Regional de Rio Gallegos, Santa Cruz); Maria A. Boretto (Sanatorio de Niños, Rosario, Santa Fe); German Kaltenbach (Hospital Regional Castro Rendón, Neuquén); Carlos Rodriguez (Hospital Zonal de Caleta Olivia, Santa Cruz); Marisol Ramos (Hospital Avelino Castelán, Chaco); Silvia Lanatti (Hospital de Niños VJ Vilela, Santa Fe); Paula Medici (Hospital Interzonal Especializado Materno Infantil de Mar del Plata, Buenos Aires); Claudia Pedraza (Hospital de Niños Sor María Ludovica, Unidad Cardiovascular, La Plata, Buenos Aires); Juan Varón Redondo (Hospital de Clínicas José de San Martin, Ciudad Autónoma de Buenos Aires); Marcelo Itharte (Hospital Materno Infantil San Roque, Entre Ríos); Gabriel Boggio (Clinica Velez Sarfield, Cordoba); Sebastián De Giuseppe (UCIP Sagrado Corazón, Ciudad Autónoma de Buenos Aires); Marlene Velazquez (Hospital Pediátrico del Niño Jesús, Córdoba); Yanina Fortini (Hospital Municipal de Trauma y Emergencias Dr. Federico Abete, Buenos Aires); Alejandra Ribonetto (Hospital de Niños Dr. Héctor Quintana, Jujuy); Gaston Morales (Corporación Medica de General San Martin, Buenos Aires); Jorge Cavagna (Hospital de Quemados, Ciudad Autónoma de Buenos Aires); Matias Penazzi (Hospital de Niños de San Justo, Buenos Aires); Daniel Capra (Sanatorio Trinidad Ramos Mejía, Buenos Aires); and Ariel Albano (Hospital de Niños Dr O Allassia, Santa Fe).

In critical illness, liver test abnormalities are associated with a poor outcome independently of other organ dysfunctions. Plasma concentrations of total bilirubin, gamma-glutamyl transpeptidase (γGT), alkaline phosphatase (ALP), alanine aminotransferase (ALT), and aspartate aminotransferase (AST) are often above the normal range in critically ill adults, the severity of which has been associated with a more complicated course of critical illness, a longer duration of ICU stay and a higher risk of death (1–6). However, despite its common occurrence and its association with poor outcome in critically ill adults, the prevalence and prognostic value of liver test abnormalities in pediatric critical illness remain largely unexplored.

ated course of critical illness, a longer duration of ICU stay and a higher risk of death (1–6). However, despite its common occurrence and its association with poor outcome in critically ill adults, the prevalence and prognostic value of liver test abnormalities in pediatric critical illness remain largely unexplored. Timely identification of true liver dysfunction may lead to prevention or attenuation of its consequences (7). Grossly, deranged liver test results often indicate severe liver injury as a result of hepatic ischemia and/or a true cholestatic liver dysfunction (8). However, the precise etiology and clinical relevance of milder cholestatic abnormalities in response to critical illness are not completely understood. It has been suggested that the presence of mild hyperbilirubinemia in response to critical illness may merely reflect a biochemical epiphenomenon or could be part of an adaptive and beneficial stress response (7, 9). Such a possibility was recently corroborated by findings from a large randomized controlled trial (RCT) in adult ICU patients (the Early Parenteral Nutrition Completing Enteral Nutrition in Adult Critically Ill Patients [EPaNIC] trial) (10). The EPaNIC trial investigated the impact of delaying initiation of parenteral nutrition (PN) to beyond the first week of critical illness (late PN) and found fewer infections, less organ failure, and faster recovery as compared with early PN (10, 11). Remarkably, total bilirubin plasma concentrations were significantly higher in adult patients who did not get early PN, whereas biochemical markers of hepatocyte lysis and cholestasis did not rise to the same extent as compared with patients who had received early PN (12). Furthermore, withholding PN during the first week of adult critical illness lowered the occurrence rate of biliary sludge (12). Also in critically ill children, withholding PN in the first week in the PICU showed to be clinically superior to providing early PN with fewer infections, less organ failure and faster recovery (13). During this first week in the PICU, late PN was also associated with lower peak plasma concentrations of γGT and ALP, whereas plasma total bilirubin peaked higher (13).

n the first week in the PICU showed to be clinically superior to providing early PN with fewer infections, less organ failure and faster recovery (13). During this first week in the PICU, late PN was also associated with lower peak plasma concentrations of γGT and ALP, whereas plasma total bilirubin peaked higher (13). As a preplanned secondary analysis of the Early versus Late Parenteral Nutrition in the Pediatric ICU (PEPaNIC) RCT (14), we here investigate the prevalence and prognostic value of abnormal liver test results in critically ill children treated in the PICU and the impact hereon of late PN versus early PN. Neonates less than 28 days old (n = 209) were excluded from this analysis to avoid confounding induced by neonatal jaundice (15). In the 1,231 PEPaNIC patients between 28 days and 17 years old, plasma concentrations of total bilirubin and of the liver enzymes ALT, AST, γGT, and ALP were quantified systematically. With use of currently accepted criteria, we documented the prevalence of cholestasis (plasma total bilirubin > 2 mg/dL [34.2 μmol/L] [16–21]) and of hypoxic hepatitis (≥ 20-fold the upper limit of normality for plasma ALT and AST concentrations [4, 6, 17, 22]) and assessed the change in liver test results during the first week in PICU. In addition, we assessed the predictive value for mortality of liver test results on the first day in PICU.

μmol/L] [16–21]) and of hypoxic hepatitis (≥ 20-fold the upper limit of normality for plasma ALT and AST concentrations [4, 6, 17, 22]) and assessed the change in liver test results during the first week in PICU. In addition, we assessed the predictive value for mortality of liver test results on the first day in PICU. MATERIALS AND METHODS Patients This study is a preplanned secondary analysis of the previously published PEPaNIC RCT, performed in three PICUs located in Belgium, the Netherlands, and Canada, of which the detailed study protocol and primary results have been published elsewhere (13, 14).

μmol/L] [16–21]) and of hypoxic hepatitis (≥ 20-fold the upper limit of normality for plasma ALT and AST concentrations [4, 6, 17, 22]) and assessed the change in liver test results during the first week in PICU. In addition, we assessed the predictive value for mortality of liver test results on the first day in PICU. MATERIALS AND METHODS Patients This study is a preplanned secondary analysis of the previously published PEPaNIC RCT, performed in three PICUs located in Belgium, the Netherlands, and Canada, of which the detailed study protocol and primary results have been published elsewhere (13, 14). In brief, 1,440 patients (from term newborns to children 17 yr old) were randomized either to receive early supplementation of insufficient enteral nutrition with PN to reach the estimated caloric target (early PN), or to withhold supplemental PN in the first 7 days in the PICU, even if enteral intake was well below estimated caloric targets (late PN). Clinicians in Edmonton used measured or estimated resting energy expenditure to define individual caloric targets whereas in Leuven and Rotterdam, caloric and macronutrient targets were based on weight and age, according to published guidelines. For a further detailed description of feeding strategies that were used during the study, we refer to the original publications (13, 14). Withholding PN for up to 7 days in pediatric critical illness showed to be beneficial, with fewer new infections, an earlier weaning from mechanical ventilation, and a shorter PICU-stay. Written informed consent had been obtained from next of kin. The study protocol and consent forms were approved by the institutional ethical review boards. Of the 1,440 PEPaNIC-patients, neonates less than 28 days old (n = 209) were excluded from the current analysis to avoid confounding induced by neonatal jaundice (Supplementary Fig. 1, Supplemental Digital Content 1, http://links.lww.com/PCC/A770).

consent forms were approved by the institutional ethical review boards. Of the 1,440 PEPaNIC-patients, neonates less than 28 days old (n = 209) were excluded from the current analysis to avoid confounding induced by neonatal jaundice (Supplementary Fig. 1, Supplemental Digital Content 1, http://links.lww.com/PCC/A770). Plasma Concentrations of Bilirubin, C-Reactive Protein, and Liver Enzymes For all patients, plasma concentrations of total bilirubin, γGT, ALP, ALT, AST, and C-reactive protein (CRP) had been measured systematically at the participating PICUs as part of the routine performed laboratory tests, using standard routine automated laboratory assays (Roche/Hitachi cobas 8000 c702 modular analyzer; Roche, Vilvoorde, Belgium). Blood samples were collected in vacutainers by experienced ICU nurses via an arterial catheter, or, in those rare cases were the arterial catheter was no longer in place, through a venous catheter, limiting risk of hemolysis. Plasma total bilirubin concentrations were quantified by the colorimetric diazo-method (lower detection limit 0.1 mg/dL). Liver enzymes ALT and AST were quantified using kinetic colorimetric ALT and AST assays without pyridoxal phosphate activation according to the International Federation of Clinical Chemistry (IFCC) recommendations (lower detection limit ALT and AST 5 U/L). Plasma γGT concentrations were quantified with the Szasz kinetic colorimetric assay (lower detection limit 3 U/L) and plasma ALP concentrations using the adenosine monophosphate kinetic colorimetric assay (lower detection limit 5 U/L), both according to the IFCC recommendations. However, as admission plasma total bilirubin had not been measured during the trial period, as reference, we also quantified plasma total bilirubin concentrations with use of the colorimetric total bilirubin assay (Thermo Scientific, Waltham, MA) in stored admission samples from 100 randomly selected PEPaNIC-patients. This random selection of 50 late PN patients and 50 early PN patients were comparable for demographics, Severity of Illness scores, and admission features.

rations with use of the colorimetric total bilirubin assay (Thermo Scientific, Waltham, MA) in stored admission samples from 100 randomly selected PEPaNIC-patients. This random selection of 50 late PN patients and 50 early PN patients were comparable for demographics, Severity of Illness scores, and admission features. No standardized definitions for critical illness-induced cholestasis or hypoxic hepatitis are available for the pediatric critically ill population. We therefore defined critical illness-induced cholestasis in our study population as a plasma total bilirubin concentration higher than 2 mg/dL (34 μmol/L), which is the commonly used cutoff in critically ill adults and has been identified as a cutoff for increased morbidity in children with biliary atresia (18, 20, 21). Hypoxic hepatitis was defined as a plasma concentration of ALT and AST of greater than or equal to 20-fold higher than the upper limit of normality, again, based on the commonly used adult and pediatric cutoffs (4, 6, 17, 22).

identified as a cutoff for increased morbidity in children with biliary atresia (18, 20, 21). Hypoxic hepatitis was defined as a plasma concentration of ALT and AST of greater than or equal to 20-fold higher than the upper limit of normality, again, based on the commonly used adult and pediatric cutoffs (4, 6, 17, 22). Statistical Analyses As liver tests were performed as part of routine laboratory measurements in the participating centers, liver test data were available for 60–70% of all patients on each day in the PICU. To detect, with 95% certainty and at least 80% power, a difference of about 20% in bilirubin plasma concentrations (12), 400 patients suffice. Furthermore, as withholding early PN has previously shown to shorten PICU-stay (13), fewer late PN patients were still in the PICU at later time points than was the case for early PN patients, and these were likely to have a higher disease severity upon admission. This could induce bias when analyzing the effect of late PN versus early PN on liver test results at later time points. To correct for missing liver test results and in order to avoid bias evoked by shortened PICU-stay in late PN patients, all analyses of liver tests were adjusted for baseline demographics and severity of illness characteristics. On each study day, patients with available liver test results in the early PN and late PN groups were matched with use of propensity scores obtained by logistic regression (one-to-one nearest neighbor matching without replacement and with a caliper of 0.2) for demographics (gender, age, weight), Severity of Illness scores (Pediatric Index of Mortality 2 probability of death at admission, Pediatric Logistic Organ Dysfunction [PELOD] score at first PICU day), admission features (diagnostic group, elective/emergency admission, presence of infection at admission, center of inclusion), and the nutritional risk score (Screening Tool for Risk on Nutritional Status and Growth kids score quantified at PICU admission) ( Supplementary Fig. 2, Supplemental Digital Content 1, http://links.lww.com/PCC/A770; and Supplementary Table 1A-N, Supplemental Digital Content 1, http://links.lww.com/PCC/A770, for baseline characteristics of all propensity matched subsets). Comparisons between groups were made with chi-square tests, Student’s t test, and Mann-Whitney U tests, as appropriate.

Content 1, http://links.lww.com/PCC/A770; and Supplementary Table 1A-N, Supplemental Digital Content 1, http://links.lww.com/PCC/A770, for baseline characteristics of all propensity matched subsets). Comparisons between groups were made with chi-square tests, Student’s t test, and Mann-Whitney U tests, as appropriate. The predictive value of day 1 PICU liver test results on PICU mortality was analyzed with use of a multivariate logistic regression analysis, adjusting for demographics, Severity of Illness scores, admission features, and study randomization (early/late PN). In brief, in an initial visual exploratory univariable analysis, we compared deciles of plasma total bilirubin, ALP, γGT, ALT, and AST concentrations and PICU mortality and identified potential thresholds for further evaluation in multivariable analyses (23). With a bootstrap approach consisting of 1,000 samples (24), we evaluated the range of 10% available concentration values centered at the visually identified cutoffs for each marker. Concentration values were dichotomized as greater than or less than or equal to the threshold, and entered as a covariate in a multivariable regression analysis, adjusting for demographics, Severity of Illness scores, admission features, nutritional risk score, and randomization to early or late PN as described above. The threshold with the minimum p value was identified in the multivariable model for each bootstrap sample, and the most frequently occurring in the 1,000 samples was selected as the final threshold for that marker (25)

dmission features, nutritional risk score, and randomization to early or late PN as described above. The threshold with the minimum p value was identified in the multivariable model for each bootstrap sample, and the most frequently occurring in the 1,000 samples was selected as the final threshold for that marker (25) Data are presented as numbers with percentages, mean ± sem, or median (interquartile range), as is also indicated in the figure legends. All statistical analyses were performed with SPSS (IBM, North Castle, Armonk, NY) including the R based plugin for propensity score matching, JMP (SAS Institute, Cary, NC) and MATLAB 2014b (The MathWorks, Natick, MA). For all comparisons, a p value of less than or equal to 0.05 was considered significant.

gure legends. All statistical analyses were performed with SPSS (IBM, North Castle, Armonk, NY) including the R based plugin for propensity score matching, JMP (SAS Institute, Cary, NC) and MATLAB 2014b (The MathWorks, Natick, MA). For all comparisons, a p value of less than or equal to 0.05 was considered significant. RESULTS The Prevalence of Cholestasis and Hypoxic Hepatitis in Pediatric Critical Illness and the Effect of Nutritional Strategy Total daily caloric intake during the first 10 days in PICU and PICU baseline characteristics of the 1,231 patients between 28 days and 17 years old are shown in Figure 1 and Table 1, respectively. In this subset of patients, excluding neonates less than 28 days old, patients randomized to late PN received significantly less calories during their first week in the PICU (Fig. 1). As was the case for the total population, randomization to late PN resulted in a better clinical outcome also in this subset (Table 1). Indeed, patients randomized to late PN acquired fewer new infections and required a shorter PICU- and hospital stay (Table 1). Although there were fewer new infections with late PN, the peak plasma concentrations of CRP were higher with late PN during the first 7 days in ICU (Table 1). TABLE 1. Baseline and Outcome Characteristics of 1,231 Early Parenteral Nutrition and Late Parenteral Nutrition Patients

RESULTS The Prevalence of Cholestasis and Hypoxic Hepatitis in Pediatric Critical Illness and the Effect of Nutritional Strategy Total daily caloric intake during the first 10 days in PICU and PICU baseline characteristics of the 1,231 patients between 28 days and 17 years old are shown in Figure 1 and Table 1, respectively. In this subset of patients, excluding neonates less than 28 days old, patients randomized to late PN received significantly less calories during their first week in the PICU (Fig. 1). As was the case for the total population, randomization to late PN resulted in a better clinical outcome also in this subset (Table 1). Indeed, patients randomized to late PN acquired fewer new infections and required a shorter PICU- and hospital stay (Table 1). Although there were fewer new infections with late PN, the peak plasma concentrations of CRP were higher with late PN during the first 7 days in ICU (Table 1). TABLE 1. Baseline and Outcome Characteristics of 1,231 Early Parenteral Nutrition and Late Parenteral Nutrition Patients Figure 1. Daily caloric intake. A, Present daily total caloric intake in kilocalories per day per kg bodyweight (BW) provided by the enteral route, the parenteral route, or both (total). B, Present daily macronutrient intake split by carbohydrates, proteins, and lipids. White bars represent patients with early parenteral nutrition (PN) supplementation, black bars present values of late PN patients. Number of patients still in ICU per day are presented below the panels. Data are presented as mean ± sem.

B, Present daily macronutrient intake split by carbohydrates, proteins, and lipids. White bars represent patients with early parenteral nutrition (PN) supplementation, black bars present values of late PN patients. Number of patients still in ICU per day are presented below the panels. Data are presented as mean ± sem. The daily prevalence of cholestasis increased with time in the PICU, with the proportion of patients fulfilling the predefined criterion during the first 7 days in the PICU ranging between 3.8% and 4.9%. The daily prevalence of hypoxic hepatitis during the first week in the PICU ranged between 0.8% and 2.2%. The prevalence of both cholestasis and hypoxic hepatitis was not different for patients in the late PN and the early PN group (Table 2). Also, daily PELOD scores remained comparable throughout the first 7 days in ICU (Table 3). TABLE 2. Daily Prevalence of Critical Illness-Induced Cholestasis and Hypoxic Hepatitis in PICU TABLE 3. Daily Pediatric Logistic Organ Dysfunction Scores of Early and Late Parenteral Nutrition Patients

The daily prevalence of cholestasis increased with time in the PICU, with the proportion of patients fulfilling the predefined criterion during the first 7 days in the PICU ranging between 3.8% and 4.9%. The daily prevalence of hypoxic hepatitis during the first week in the PICU ranged between 0.8% and 2.2%. The prevalence of both cholestasis and hypoxic hepatitis was not different for patients in the late PN and the early PN group (Table 2). Also, daily PELOD scores remained comparable throughout the first 7 days in ICU (Table 3). TABLE 2. Daily Prevalence of Critical Illness-Induced Cholestasis and Hypoxic Hepatitis in PICU TABLE 3. Daily Pediatric Logistic Organ Dysfunction Scores of Early and Late Parenteral Nutrition Patients The Effect of Withholding PN for 1 Week in PICU on the Daily Profile of Plasma Total Bilirubin Concentrations and Other Liver Tests From the first day after randomization onward, withholding PN in the first week of critical illness elevated plasma total bilirubin concentrations throughout the 7-day study intervention window as compared with early PN administration (Fig. 2; and Supplementary Table 1A-G, Supplemental Digital Content 1, http://links.lww.com/PCC/A770). However, from day 8 onwards, as soon as the total caloric intake became comparable for both groups, plasma total bilirubin concentrations were no longer different between late PN and early PN patients (Fig. 2; and Supplementary Table 1A-G, Supplemental Digital Content 1, http://links.lww.com/PCC/A770). Also, peak plasma total bilirubin concentrations during the first week was higher in late PN patients as compared with early PN patients (Table 1).

ations were no longer different between late PN and early PN patients (Fig. 2; and Supplementary Table 1A-G, Supplemental Digital Content 1, http://links.lww.com/PCC/A770). Also, peak plasma total bilirubin concentrations during the first week was higher in late PN patients as compared with early PN patients (Table 1). Figure 2. Daily plasma total bilirubin concentrations of all patients still in the PICU. Black dots present values of late parenteral nutrition (PN) patients, whereas gray dots represent patients with early PN supplementation. Number of patients still in ICU and of which plasma bilirubin concentrations were available at each day are presented below the panels. Data are presented as mean ± sem. *p ≤ 0.05 early vs late PN patients, corrected for baseline characteristics including severity of illness by performing the analyses in propensity score matched subsets per day (Supplementary Table 1A-G, Supplemental Digital Content 1, http://links.lww.com/PCC/A770). aAdmission values represent the mean ± sem of 50 randomly selected patients per intervention arm. To convert plasma total bilirubin concentrations in mg/dL to μmol/L, multiply by 17.1.

in propensity score matched subsets per day (Supplementary Table 1A-G, Supplemental Digital Content 1, http://links.lww.com/PCC/A770). aAdmission values represent the mean ± sem of 50 randomly selected patients per intervention arm. To convert plasma total bilirubin concentrations in mg/dL to μmol/L, multiply by 17.1. Although peak plasma concentrations of ALP and γGT during the first week of critical illness were significantly lower in children receiving late PN compared with those receiving early PN (Table 1), there were no significant differences when analyzed on a daily basis (Supplementary Table 1H-N, Supplemental Digital Content 1, http://links.lww.com/PCC/A770). Also, daily plasma concentrations of ALT and AST did not differ significantly between the two treatment groups (Supplementary Table 1H-N, Supplemental Digital Content 1, http://links.lww.com/PCC/A770).

n analyzed on a daily basis (Supplementary Table 1H-N, Supplemental Digital Content 1, http://links.lww.com/PCC/A770). Also, daily plasma concentrations of ALT and AST did not differ significantly between the two treatment groups (Supplementary Table 1H-N, Supplemental Digital Content 1, http://links.lww.com/PCC/A770). The Predictive Value of Day 1 Liver Test Results for PICU Mortality Plasma total bilirubin concentrations on the first PICU day displayed a U-shaped relationship with PICU mortality (Fig. 3). Using a multivariable bootstrap approach, we identified 0.20 mg/dL (3.42 μmol/L) and 0.76 mg/dL (13 μmol/L) as cutoffs in this U-shaped relationship. Indeed, when plasma total bilirubin concentrations were categorized into three categories (< 0.20 mg/dL, between 0.20 and 0.76 mg/dL, and > 0.76 mg/dL [< 3.42 μmol/L, 3.42–13 μmol/L, and > 13 μmol/L]), the group with plasma total bilirubin concentrations between 0.20 and 0.76 mg/dL (3.42–13 μmol/L) had a significantly lower mortality as compared with the patients in the lower category (T2 vs T1: odds ratio [OR] 0.25 [0.10–0.62]; p = 0.003), and the highest category (T2 vs T3: 0.34 [0.15–0.79]; p = 0.01). The latter remained significant after adjusting for baseline risk factors, severity of illness, and randomization to late PN versus early PN (T2 vs T1: 0.17 [0.04–0.67]; p = 0.01 and T2 vs T3: 0.15 [0.04–0.55]; p = 0.003).

R] 0.25 [0.10–0.62]; p = 0.003), and the highest category (T2 vs T3: 0.34 [0.15–0.79]; p = 0.01). The latter remained significant after adjusting for baseline risk factors, severity of illness, and randomization to late PN versus early PN (T2 vs T1: 0.17 [0.04–0.67]; p = 0.01 and T2 vs T3: 0.15 [0.04–0.55]; p = 0.003). Figure 3. U-shaped relationship between plasma total bilirubin concentrations on day 1 and PICU mortality. Each interval represents 10% of all patients with available data on day 1 (n = 913). To convert plasma total bilirubin concentrations in mg/dL to μmol/L, multiply by 17.1.

R] 0.25 [0.10–0.62]; p = 0.003), and the highest category (T2 vs T3: 0.34 [0.15–0.79]; p = 0.01). The latter remained significant after adjusting for baseline risk factors, severity of illness, and randomization to late PN versus early PN (T2 vs T1: 0.17 [0.04–0.67]; p = 0.01 and T2 vs T3: 0.15 [0.04–0.55]; p = 0.003). Figure 3. U-shaped relationship between plasma total bilirubin concentrations on day 1 and PICU mortality. Each interval represents 10% of all patients with available data on day 1 (n = 913). To convert plasma total bilirubin concentrations in mg/dL to μmol/L, multiply by 17.1. Similarly, plasma ALP concentrations on the first PICU day revealed a U-shaped relation with PICU mortality (Supplementary Fig. 3, Supplemental Digital Content 1, http://links.lww.com/PCC/A770). Using a multivariable bootstrap approach, we identified 94 IU/L and 200 IU/L as cutoffs in this U-shaped relationship. When plasma ALP concentrations were categorized into three categories (< 95 IU/L, between 95 and 200 IU/L, and > 200 IU/L), patients with plasma ALP concentrations between 95 and 200 IU/L had a significantly lower mortality as compared with the patients in the lower category (T2 vs T1: OR, 0.38 [0.17–0.90]; p = 0.02), and the highest category (T2 vs T3: 0.28 [0.12–0.68]; p = 0.004). However, when adjusted for demographics, Severity of Illness scores, admission features, and randomization, plasma ALP was no longer independently associated with PICU mortality. Plasma γGT and PICU mortality displayed a biphasic relation, using a multivariable bootstrap approach, we identified 11 IU/L as cutoff (Supplementary Fig. 4, Supplemental Digital Content 1, http://links.lww.com/PCC/A770). Patients with plasma γGT greater than 11 IU/L had a significantly higher mortality as compared with patients with plasma γGT less than or equal to 11 IU/L (OR, 10.17 [3.64–42.41]; p < 0.001), also after adjustment for baseline characteristics including severity of illness (OR, 4.92 [1.40–240.12]; p = 0.01).

/PCC/A770). Patients with plasma γGT greater than 11 IU/L had a significantly higher mortality as compared with patients with plasma γGT less than or equal to 11 IU/L (OR, 10.17 [3.64–42.41]; p < 0.001), also after adjustment for baseline characteristics including severity of illness (OR, 4.92 [1.40–240.12]; p = 0.01). The parenchymal lysis enzymes ALT and AST concentrations also displayed a biphasic relation with PICU mortality, using a multivariable bootstrap approach, we identified 27 IU/L for ALT and 99 IU/L for AST as cutoffs (Supplementary Figs. 5 and 6, Supplemental Digital Content 1, http://links.lww.com/PCC/A770). PICU mortality was significantly higher in patients with plasma ALT greater than 27 IU/L (OR, 21.95 [9.33–64.38]; p < 0.001) and plasma AST greater than 99 IU/L (OR, 9.632 [4.74–21.65]; p < 0.001) as compared with patients with plasma concentrations below these thresholds. This remained significant after adjustment for baseline characteristics including severity of illness (ALT: OR 8.22 [2.96–26.97]; p < 0.001) and (AST: OR 4.80 [1.78–13.86]; p = 0.001).

T greater than 99 IU/L (OR, 9.632 [4.74–21.65]; p < 0.001) as compared with patients with plasma concentrations below these thresholds. This remained significant after adjustment for baseline characteristics including severity of illness (ALT: OR 8.22 [2.96–26.97]; p < 0.001) and (AST: OR 4.80 [1.78–13.86]; p = 0.001). DISCUSSION Overt cholestasis, defined as plasma concentrations of bilirubin greater than 2 mg/dL, and hypoxic hepatitis, defined as plasma concentrations of ALT and AST of greater than or equal to 20-fold higher than the upper limit of normality, were found to be rare in this PICU population and were unrelated to nutritional strategy. However, withholding PN during week 1 in the PICU increased plasma total bilirubin concentrations. A mild elevation in plasma total bilirubin on the first PICU day was associated with a lower risk of death.

per limit of normality, were found to be rare in this PICU population and were unrelated to nutritional strategy. However, withholding PN during week 1 in the PICU increased plasma total bilirubin concentrations. A mild elevation in plasma total bilirubin on the first PICU day was associated with a lower risk of death. The prevalence of cholestasis was found to be less than half the prevalence in adult critically ill populations and remained less than 5% throughout the first week in the PICU (1–3, 26). Altered liver physiology, such as reduced metabolic function, a decrease in liver blood flow, and a reduction of the hepatic regenerative capacity in the adult or aging patient may increase susceptibility to liver injury in response to severe illness and could explain the increased risk of illness-induced cholestasis in older subjects (27, 28). Alternatively, the threshold values used for the definition of cholestasis in adults may not be transferable to the pediatric population. Indeed, in our pediatric study population, plasma total bilirubin concentrations above 0.76 mg/dL (13 μmol/L) on the first day in PICU were already independently associated with higher mortality, which is also much below the cutoffs used in the development of previously used pediatric organ dysfunction and mortality scores (3.5 mg/dL [60 μmol/L] in the PRISM score and 5 mg/dL [85 μmol/L] in the PELOD score [16, 29, 30]). During the first 7 days in the PICU, the prevalence of hypoxic hepatitis also remained low (< 2.2% of patients), but this low number was overall comparable to that reported for adult ICU patients (4, 6, 26, 31). Of note, we did not assess liver oxygenation to ascertain presence of hypoxia, nor did we exclude presence of potential confounding factors such as medication, blood transfusions, and infections which might have affected liver test results (7).

overall comparable to that reported for adult ICU patients (4, 6, 26, 31). Of note, we did not assess liver oxygenation to ascertain presence of hypoxia, nor did we exclude presence of potential confounding factors such as medication, blood transfusions, and infections which might have affected liver test results (7). Both the prevalence of cholestasis and of hypoxic hepatitis were not affected by randomization to either early PN or late PN. Avoiding early administration of PN during the first week of pediatric critical illness did result in higher daily and peak levels of plasma total bilirubin, whereas daily levels of plasma γGT, ALP, ALT, or AST were unaffected, and peak levels of plasma γGT and ALP were lowered. In a previous study in critically ill adults, avoiding early PN administration also increased levels of plasma total bilirubin, but lowered other markers of cholestasis and the prevalence of biliary sludge (10). When PN is administered chronically, in doses that exceed energy expenditure or containing high doses of lipids, this can cause cholestasis outside the context of critical illness (32). This may potentially also be an additional toxic threat to hepatocytes during critical illness. Indeed, PN induced cholestasis can exert toxic accumulation of endogenous toxins and drugs by inhibition of the hepatobiliary excretion (32, 33). It has indeed been postulated, based on available age- and weight-adjusted guideline recommendations, that early PN might have been clinically inferior to late PN because of glucose and/or lipid overfeeding and protein underfeeding (34–36). However, a postrandomization observational study, testing the independent associations between average doses of administered glucose, lipids, and amino acids and the likelihood of worse clinical outcomes, demonstrated that not the doses of glucose or lipids, but instead the protein doses explained harm caused by early PN (36). In contrast, the observation that early administration of PN actually reduced plasma bilirubin levels argues against nutrition-induced liver injury. Importantly, short-term caloric restriction, both in health and during critical illness, has shown to quickly increase bilirubin and/or bile acids, similar to our current findings (37–41). However, withholding PN also increased CRP values during the first week in ICU, which might be indicative of more inflammation (13).

mportantly, short-term caloric restriction, both in health and during critical illness, has shown to quickly increase bilirubin and/or bile acids, similar to our current findings (37–41). However, withholding PN also increased CRP values during the first week in ICU, which might be indicative of more inflammation (13). Inflammation is a known contributor to critical-illness induced cholestasis and might as such have contributed to the observed increase in plasma bilirubin in late PN patients (7, 9).

mportantly, short-term caloric restriction, both in health and during critical illness, has shown to quickly increase bilirubin and/or bile acids, similar to our current findings (37–41). However, withholding PN also increased CRP values during the first week in ICU, which might be indicative of more inflammation (13). Inflammation is a known contributor to critical-illness induced cholestasis and might as such have contributed to the observed increase in plasma bilirubin in late PN patients (7, 9). Plasma total bilirubin concentrations on the first PICU day displayed a U-shaped relationship with PICU mortality. That a mild elevation in plasma total bilirubin on the first PICU day was associated with a lower risk of death, may question the role of bilirubin as a valid marker of liver dysfunction. Possibly, a mild rise in plasma bilirubin may be an adaptive response to critical illness (9, 38). Indeed, antioxidant and antiinflammatory effects have been associated with elevations of plasma bilirubin after challenge with oxidative stressors and lipopolysaccharide (LPS) (42–44). Furthermore, administration of bilirubin has shown to improve survival and attenuated liver injury in response to LPS infusion (45). Additionally, knockout mice that lack heme oxygenase, which is the rate-limiting enzyme in bilirubin formation, were shown to be more likely to die and to suffer significantly more liver and renal dysfunction after LPS challenge, as compared with wild-type mice (46). Other potential protective effects of heme oxygenase induction are a direct modulation of antioxidant genes transcription, an increased production of the tissue mediator carbon monoxide and a decrease in cellular prooxidant heme (42). Bilirubin might also reduce cellular damage through biliverdin reductase action, which has recently been documented to be increased in septic patients (47). Indeed, when bilirubin reacts with reactive oxygen species, toxicity is neutralized and bilirubin is oxidized to biliverdin. Biliverdin can be converted back to bilirubin through biliverdin reductase, and by repeating this cycle, the antioxidative and cytoprotective effect of small amounts of bilirubin is greatly amplified (44, 47). In our study, however, total plasma bilirubin concentrations on the first day of critical illness were associated with PICU mortality in a “U-shape” relationship, in which both low and high levels of bilirubin were associated with higher risk of death. Also, other studies have described a nonlinear relationship between plasma bilirubin and mortality, both in critically illness (12) and other diseases (48).

illness were associated with PICU mortality in a “U-shape” relationship, in which both low and high levels of bilirubin were associated with higher risk of death. Also, other studies have described a nonlinear relationship between plasma bilirubin and mortality, both in critically illness (12) and other diseases (48). As expected, high plasma concentrations of the hepatocyte lysis enzymes ALT and AST were, above a certain threshold, independently associated with increased PICU mortality and likely reflect the severity of shock-induced liver injury (4).

illness were associated with PICU mortality in a “U-shape” relationship, in which both low and high levels of bilirubin were associated with higher risk of death. Also, other studies have described a nonlinear relationship between plasma bilirubin and mortality, both in critically illness (12) and other diseases (48). As expected, high plasma concentrations of the hepatocyte lysis enzymes ALT and AST were, above a certain threshold, independently associated with increased PICU mortality and likely reflect the severity of shock-induced liver injury (4). Our study has several limitations to take into account. First, admission total bilirubin and other liver test results were not routinely assessed, and the bilirubin values on the first day in PICU were already affected by randomization, which precluded the comparison of changes over time within the same patient in the two intervention groups. Second, because liver tests were measured as part of routine laboratory tests, and not daily, approximately 30% of patient liver test data was missing per day in the PICU. In order to adjust for missing data and to avoid bias as much as possible, propensity score matched subsets were used to analyze liver test data. Third, the low risk of death in this pediatric study population may have limited the relevance of the observed association between admission plasma liver tests and mortality. Fourth, liver ultrasounds were not performed routinely, which would have strengthened the information on prevalence of abnormal liver tests. In addition, the predictive cutoff values that were identified with a multivariable bootstrap approach, are merely illustrative for the association with worse outcome. These cutoffs cannot be used in clinical practice without validation in a properly designed study including information on biliary inflammation, cholestasis, and sludge.

predictive cutoff values that were identified with a multivariable bootstrap approach, are merely illustrative for the association with worse outcome. These cutoffs cannot be used in clinical practice without validation in a properly designed study including information on biliary inflammation, cholestasis, and sludge. CONCLUSIONS The overall prevalence of critical illness-induced cholestasis and hypoxic hepatitis was low in the studied critically ill children and was not affected by nutritional strategy. However, withholding PN during the first 7 days of critical illness mildly elevated plasma total bilirubin concentrations. Furthermore, plasma total bilirubin concentration on the first day of critical illness was associated with PICU mortality in a “U-shape” relationship, in which plasma total bilirubin concentrations between 0.20 and 0.76 mg/dL (3.42–13 μmol/L) were associated with lower mortality than either lower or high plasma bilirubin. Whether this may indicate a beneficial role for mildly elevated bilirubin levels as part of the stress response to critical illnesses, requires further research. ACKNOWLEDGMENTS We thank the patients and their family or guardians for their willingness to participate in the PEPaNIC randomized controlled trial. We also thank the original PEPaNIC authors and the clinical research staff for the recruitment and data entry and Pieter Wouters for his excellent data management and the database exports. Supplementary Material *See also p. 1169. Drs. Langouche and Van den Berghe had equal contribution.

ACKNOWLEDGMENTS We thank the patients and their family or guardians for their willingness to participate in the PEPaNIC randomized controlled trial. We also thank the original PEPaNIC authors and the clinical research staff for the recruitment and data entry and Pieter Wouters for his excellent data management and the database exports. Supplementary Material *See also p. 1169. Drs. Langouche and Van den Berghe had equal contribution. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (http://journals.lww.com/pccmjournal). Supported, in part, by the Methusalem Program of the Flemish Government (Drs. Langouche and Van den Berghe) via the KU Leuven University (METH/08/07); by an ERC Advanced grant (AdvG-2012–321670) to Dr. Van den Berghe from the Ideas Program of the European Union 7th framework program. Dr. Van den Berghe’s institution received funding from Flemish Government (METHUSALEM) and European Research Council. The remaining authors have disclosed that they do not have any potential conflicts of interest.

Catecholaminergic polymorphic ventricular tachycardia (CPVT) is a genetic condition with an estimated prevalence of 1 in 10,000 (1, 2), characterized by epinephrine (adrenaline)-induced polymorphic ventricular tachycardia (VT) leading to syncope and sudden death in children and adolescents. Genetic tests in those who died from sudden unexplained death (autopsy negative) under 35 years old reveal this condition almost as commonly as long QT syndrome (3, 4). Common triggers include exercise or emotion (1, 5–8). Outpatient diagnosis depends on exercise-induced ventricular ectopy, bidirectional, or polymorphic VT (2, 9). Epinephrine challenges have been used to demonstrate provocation of polymorphic or bidirectional VT (9–11). Genetic testing reveals mutations in the cardiac ryanodine receptor gene (RyR2) in 50–60% of those with CPVT (1, 2, 6–9, 12). Long-term treatments for CPVT are β-blockers, flecainide, and left cardiac sympathetic denervation (2, 7, 13–17) with implantable cardiac defibrillator (ICD) indicated in some patients (1, 2, 9, 15, 18).

esting reveals mutations in the cardiac ryanodine receptor gene (RyR2) in 50–60% of those with CPVT (1, 2, 6–9, 12). Long-term treatments for CPVT are β-blockers, flecainide, and left cardiac sympathetic denervation (2, 7, 13–17) with implantable cardiac defibrillator (ICD) indicated in some patients (1, 2, 9, 15, 18). Epinephrine is recommended as part of current advanced cardiopulmonary resuscitation (CPR) guidelines (19–22). Here we describe three children who presented following cardiac arrest, ultimately proven to be due to CPVT, where the clinical course was either made worse through the administration of epinephrine and/or improved by the use of opiate analgesia and/or general anesthesia with avoidance of epinephrine. In two of these cases, we also describe how epinephrine testing while on extracorporeal membrane oxygenation (ECMO) was used to diagnose CPVT, and in one where flecainide stopped the VT. Each subject is enrolled in the consent-based and ethically approved Cardiac Inherited Disease Registry New Zealand, which permits deidentified publication of clinical and genetic information. Because some potentially identifying features are presented, signed informed consent was obtained from each of the parents after they reviewed the article.

the consent-based and ethically approved Cardiac Inherited Disease Registry New Zealand, which permits deidentified publication of clinical and genetic information. Because some potentially identifying features are presented, signed informed consent was obtained from each of the parents after they reviewed the article. CASES Case A A 4-year-old boy (approximately 20 kg) collapsed while playing outside with another young child, unwitnessed by any adults. An estimated 2–4 minutes passed before CPR was started by an adult family member. Ambulance staff arrived 5 minutes after initiation of CPR and found the child in asystole. Return of spontaneous circulation occurred 10 minutes after initiation of chest compressions and bag-mask ventilation only. On route to hospital at 23 minutes after CPR was started, he had another cardiac arrest, described as pulseless electrical activity (PEA). At this time, the first bolus of IV epinephrine (0.2 mg~0.01 mg/kg) was given, after which he went into ventricular fibrillation (VF) and was shocked with 100 J into sinus bradycardia at 50–60 beats/min.

t 23 minutes after CPR was started, he had another cardiac arrest, described as pulseless electrical activity (PEA). At this time, the first bolus of IV epinephrine (0.2 mg~0.01 mg/kg) was given, after which he went into ventricular fibrillation (VF) and was shocked with 100 J into sinus bradycardia at 50–60 beats/min. In the emergency department, at 34 minutes, he was breathing spontaneously with a sinus bradycardia of 40 beats/min and poor peripheral perfusion, he received two further boluses of 0.2 mg IV epinephrine, 3 and 7 minutes later. There were frequent multimorphic ventricular extrasystoles (VEs). At 46 minutes, he had another VF arrest, at which point CPR was recommenced then two 100 J shocks, IV saline (10 mL/kg), and calcium (1 mL/kg 10% calcium gluconate) were given. After this, he was intubated and ventilated. He was still in VF after the intubation, and with CPR continuing, an additional 0.2 mg epinephrine bolus was given at 56 minutes with another direct current (DC) shock and amiodarone (10 µg/kg), after which he reverted to sinus rhythm at 110 beats/min for around 6 minutes, but with intermittent ventricular ectopy.

ted. He was still in VF after the intubation, and with CPR continuing, an additional 0.2 mg epinephrine bolus was given at 56 minutes with another direct current (DC) shock and amiodarone (10 µg/kg), after which he reverted to sinus rhythm at 110 beats/min for around 6 minutes, but with intermittent ventricular ectopy. At 64 minutes, he had another low output arrest with sinus rhythm but very weak/absent central pulses and interspersed runs of VT. A fifth epinephrine bolus was given followed by an epinephrine infusion was started at 0.05 µg/kg/min. In addition, he received IV magnesium, lignocaine, and saline boluses and had a further 0.2 mg epinephrine bolus 10 minutes later, after which the epinephrine infusion was increased to 0.5 µg/kg/min. The rhythm alternated between sinus bradycardia and rapid polymorphic VT. At 78 minutes on the advice of a pediatric electrophysiologist, the epinephrine infusion was stopped and 200 µg (10 µg/kg) of fentanyl was given IV. This dramatically altered the course of events. Within minutes, CPR was ceased as he had a return of spontaneous circulation with a sinus bradycardia of 50 beats/min and mean blood pressure of 75–85 mm Hg. Following the resuscitation, he was put on ECMO because of severe metabolic acidosis (pH, 6.9; lactate, 6.1 mmol/L), hypoxemia, and poor myocardial function (ejection fraction, 18%). In total, he received six epinephrine boluses, an epinephrine infusion, and 4 DC shocks for VF.

At 78 minutes on the advice of a pediatric electrophysiologist, the epinephrine infusion was stopped and 200 µg (10 µg/kg) of fentanyl was given IV. This dramatically altered the course of events. Within minutes, CPR was ceased as he had a return of spontaneous circulation with a sinus bradycardia of 50 beats/min and mean blood pressure of 75–85 mm Hg. Following the resuscitation, he was put on ECMO because of severe metabolic acidosis (pH, 6.9; lactate, 6.1 mmol/L), hypoxemia, and poor myocardial function (ejection fraction, 18%). In total, he received six epinephrine boluses, an epinephrine infusion, and 4 DC shocks for VF. The 12 lead electrocardiogram was normal and the left ventricular function returned rapidly to normal while on ECMO. There were no features to suggest myocarditis, and there was no family history of sudden death or syncope. After 24 hours on ECMO, with stable hemodynamic and metabolic parameters, and good ejection on echocardiography, an epinephrine challenge was done to confirm the diagnosis of CPVT. An infusion was started at 0.2 µg/kg/min which induced polymorphic VT. A flecainide bolus of 40 mg over a few minutes led to gradual resolution of the tachycardia, confirming the diagnosis (Figs. 1–3). Figure 1. Series of 12 lead electrocardiograms. All recorded at 25 mm/s. Initiation of ventricular tachycardia with epinephrine challenge. Figure 2. Series of 12 lead electrocardiograms. All recorded at 25 mm/s. More organized, bidirectional ventricular tachycardia after the commencement of the flecainide infusion.

After 24 hours on ECMO, with stable hemodynamic and metabolic parameters, and good ejection on echocardiography, an epinephrine challenge was done to confirm the diagnosis of CPVT. An infusion was started at 0.2 µg/kg/min which induced polymorphic VT. A flecainide bolus of 40 mg over a few minutes led to gradual resolution of the tachycardia, confirming the diagnosis (Figs. 1–3). Figure 1. Series of 12 lead electrocardiograms. All recorded at 25 mm/s. Initiation of ventricular tachycardia with epinephrine challenge. Figure 2. Series of 12 lead electrocardiograms. All recorded at 25 mm/s. More organized, bidirectional ventricular tachycardia after the commencement of the flecainide infusion. Figure 3. Series of 12 lead electrocardiograms. All recorded at 25 mm/s. Termination of ventricular tachycardia during flecainide administration, resolving here to frequent ventricular ectopy and after this into sinus rhythm. An ICD was inserted, and regular atenolol and flecainide were started. Genetic testing confirmed a de novo RyR2 mutation (p.Ser2246Leu), absent in both parents. His ICD has not fired in 12 months of follow-up.

Figure 3. Series of 12 lead electrocardiograms. All recorded at 25 mm/s. Termination of ventricular tachycardia during flecainide administration, resolving here to frequent ventricular ectopy and after this into sinus rhythm. An ICD was inserted, and regular atenolol and flecainide were started. Genetic testing confirmed a de novo RyR2 mutation (p.Ser2246Leu), absent in both parents. His ICD has not fired in 12 months of follow-up. Case B Case B, a 10-year-old girl (weight 30 kg), had a background of presumed epileptic seizures associated with exercise. While running, she was witnessed to collapse, followed by a brief seizure. She was found to be pulseless and not breathing. CPR was initiated by bystanders for 8 minutes before ambulance staff arrived. DC shock (150 J) was delivered for VF, and 0.8 mg (~0.03 mg/kg) IV epinephrine was given. Return of spontaneous circulation was recorded after 20 minutes, and she was transported to the local hospital. A second VF arrest occurred at 38 minutes, on arrival at the hospital, with delivery of a 70 J DC shock and return of spontaneous circulation. IV magnesium and amiodarone were given, and a propofol infusion was started. She had another VF arrest and received seven recorded DC shocks, but no epinephrine. She returned to sinus rhythm at 81 beats/min, with occasional ventricular ectopy and remained stable for over an hour on a propofol infusion.

ous circulation. IV magnesium and amiodarone were given, and a propofol infusion was started. She had another VF arrest and received seven recorded DC shocks, but no epinephrine. She returned to sinus rhythm at 81 beats/min, with occasional ventricular ectopy and remained stable for over an hour on a propofol infusion. At 2 and 3 hours, she had further VF arrests receiving five and four DC shocks, respectively. She was given IV magnesium, amiodarone, and after her cardiac rhythm stabilized, she was started on an amiodarone infusion. She received no epinephrine during these resuscitations. She was transferred to the PICU at a tertiary children’s hospital following the third VF arrest and proceed to have three further arrests in the next 8 hours with frequent multimorphic ventricular ectopy between arrests. Another arrest occurred at 9 hours, and she received 5 DC shocks and an IV epinephrine bolus, IV magnesium, and atropine. She was started on an epinephrine infusion at 0.05 µg/kg/min, up titrated to 0.2 µg/kg/min. After the infusion was started, she received two further epinephrine boluses and four further DC shocks over 10 minutes. Specialist pediatric electrophysiology advice was sought, CPVT was suspected, and at 20 minutes into the arrest, IV fentanyl (10 µg/kg) was given, continuous IV epinephrine was stopped, and the VT/VF storm settled over 2 minutes. A final VF arrest occurred 40 minutes later, which resolved with a DC shock, IV lignocaine, magnesium, and further fentanyl (10 µg/kg).

was sought, CPVT was suspected, and at 20 minutes into the arrest, IV fentanyl (10 µg/kg) was given, continuous IV epinephrine was stopped, and the VT/VF storm settled over 2 minutes. A final VF arrest occurred 40 minutes later, which resolved with a DC shock, IV lignocaine, magnesium, and further fentanyl (10 µg/kg). She was started on ECMO as she had very poor ventricular function, poor cardiac output, and bradycardia. Family history was negative, the resting electrocardiogram was normal, the heart was structurally normal on echocardiography, and there was nothing to suggest myocarditis. To confirm the suspicion of CPVT, while stable on ECMO, she received a 0.2 mg epinephrine bolus, then a 0.2 µg/kg/min infusion of epinephrine which caused a short run of ventricular bigeminy and finally a 0.5 µg/kg/min infusion that caused self-reverting VT/VF (Figs. 4 and 5). Figure 4. Series of 12 lead electrocardiograms. All recorded at 25 mm/s. Ventricular bigeminy following epinephrine infusion at 0.2 μg/min. Figure 5. Ventricular fibrillation following epinephrine infusion at 0.5 μg/min. She underwent a left cardiac sympathectomy, had an ICD implanted, and was started on nadolol and flecainide; genetic testing confirmed a de novo RyR2 mutation (p.Ser2246Leu), absent in both parents. She had no further episodes of cardiac arrest, and her ICD has not fired in 74 months of follow-up.

Figure 5. Ventricular fibrillation following epinephrine infusion at 0.5 μg/min. She underwent a left cardiac sympathectomy, had an ICD implanted, and was started on nadolol and flecainide; genetic testing confirmed a de novo RyR2 mutation (p.Ser2246Leu), absent in both parents. She had no further episodes of cardiac arrest, and her ICD has not fired in 74 months of follow-up. Case C Case C, a 5-year-old girl, weighing 17 kg, was climbing down into the water from the back of a boat to the point where her legs were submerged in water when she became suddenly unresponsive, and fell backward into the water. She was pulled back onto the boat, and CPR was started by a family member. A nurse who was nearby continued CPR “more aggressively.” An estimated 10 minutes passed before a nearby automatic external defibrillator was located and applied. A shockable rhythm was identified, and a shock delivered. CPR was continued until emergency services arrived by helicopter approximately 15 minutes following the arrest, she was breathing spontaneously, and CPR was stopped. She was taken to the local hospital with a Glasgow Coma Score of 7, where an electrocardiogram showed polymorphic broad complexes at 200 beats/min. The electrocardiogram was faxed to specialists at the tertiary children’s hospital and reviewed by the electrophysiology specialist. CPVT was suspected, and the local team was advised to avoid epinephrine and anesthetise her and deliver opiates. This suspicion was based on key features in the history—particularly that the child was previously well and that loss of consciousness occurred in water, along with the presence of multimorphic ventricular ectopy on the electrocardiogram (Figs. 1–5). She was treated with fentanyl (10 µg/kg) followed by rocuronium, intubation, and ventilation at which point her rhythm stabilized and her tachycardia resolved. A fentanyl infusion (3 µg/kg/hr) was started. She was transferred to the ICU at the tertiary children’s hospital. She had no further arrests, nor any ventricular ectopy. She underwent cerebral cooling but did not require ECMO. Family history was negative, the resting electrocardiogram was normal, and the heart was structurally normal on echocardiography.

started. She was transferred to the ICU at the tertiary children’s hospital. She had no further arrests, nor any ventricular ectopy. She underwent cerebral cooling but did not require ECMO. Family history was negative, the resting electrocardiogram was normal, and the heart was structurally normal on echocardiography. Case C had a hybrid ICD inserted and was started on nadolol and flecainide; genetic testing revealed a denovo RyR2 mutation (p.Gly4140Glu). She had no further episodes of arrest, and her ICD has not fired in 43 months of follow-up. DISCUSSION The first two cases presented here show that the repeated use of epinephrine was ineffective and likely contributed to VT/VF storms over many hours. A comparison with the third case, where epinephrine was avoided, and fentanyl delivered early, led to a dramatically better clinical course.

Case C had a hybrid ICD inserted and was started on nadolol and flecainide; genetic testing revealed a denovo RyR2 mutation (p.Gly4140Glu). She had no further episodes of arrest, and her ICD has not fired in 43 months of follow-up. DISCUSSION The first two cases presented here show that the repeated use of epinephrine was ineffective and likely contributed to VT/VF storms over many hours. A comparison with the third case, where epinephrine was avoided, and fentanyl delivered early, led to a dramatically better clinical course. CPVT thus presents a unique challenge to resuscitators who work with children and adolescents because it must be managed differently from other types of arrhythmias. Most cases are children between 4 and 14 years old (although presentation at age >40 yr and in infants is recognized), and occur with some level of activity, especially swimming, or with fright or excitement (23). There is usually frequent ventricular ectopy between arrests. There is a male predominance. Aside from frequent ectopy, these characteristics are also common among children who suffer cardiac arrest secondary to long QT syndrome type 1 (6, 11, 24), a condition in which epinephrine is not contraindicated during resuscitation attempts, even though exercise, swimming, and excitement are known triggers for long QT type 1 in particular. The finding of bidirectional VT, as demonstrated in Figure 2, is practically pathognomonic for CPVT (1, 5, 9, 24, 25), although it has been associated with Andersen-Tawil syndrome (25) and digoxin overdose (26).

ttempts, even though exercise, swimming, and excitement are known triggers for long QT type 1 in particular. The finding of bidirectional VT, as demonstrated in Figure 2, is practically pathognomonic for CPVT (1, 5, 9, 24, 25), although it has been associated with Andersen-Tawil syndrome (25) and digoxin overdose (26). Once recognized, the optimum therapies available are IV opiates and general anesthetics (which are useful treatments in any VT/VF storm anyway). In CPVT, these work to reduce the catecholaminergic stimulation to break the cycle that triggers the VT. The dose of IV fentanyl used in our cases, 10 μg/kg, is a high dose which may not be familiar to emergency physicians who commonly use a tenth of the dose for analgesia. However, this high dose is required to reduce the catecholaminergic stimulation. Hypotension and low cardiac output can be difficult to manage because catecholamines should be avoided, and ECMO may have to be considered to support cardiac output and blood pressure. Adjunct medications for rhythm control include flecainide and β-blockers (27). Long-term management includes long-term β-blockade, flecainide, left cervical sympathetic denervation, and ICDs (1, 14, 16–18, 24). Flecainide has been shown to have a specific effect in CPVT in a mouse model in reducing sarcoplasmic calcium release (16). In CPVT, the intracellular calcium levels typically cascade out of control with an adrenergic stimulus making cardiomyocytes highly excitable and vulnerable to dysrhythmia. The therapeutic effect of flecainide in the mouse has been replicated in humans (with or without the typical RyR2 gene mutations), so much so that in combination with β-blockers an ICD may sometimes be withheld even after a cardiac arrest (14–17).

us making cardiomyocytes highly excitable and vulnerable to dysrhythmia. The therapeutic effect of flecainide in the mouse has been replicated in humans (with or without the typical RyR2 gene mutations), so much so that in combination with β-blockers an ICD may sometimes be withheld even after a cardiac arrest (14–17). An awareness that epinephrine may sometimes be proarrhythmic in other acute cardiac deterioration is important; examples may include myocarditis or acute myocardial ischemia (such as due to anomalous coronary arterial supply). Epinephrine may also induce paradoxical hypotension during the acute management of cardiac arrest due to obstructive hypertrophic cardiomyopathy (28). In such cases, β1-adrenergic effects increase contractility and thus the outflow obstruction, whereas β2-adrenergic effects reduce systemic vascular resistance. Similarly, in quetiapine poisoning, the drug’s α-adrenergic blockade allows epinephrine’s potent β2-adrenergic effects on the peripheral vasculature to dominate, worsening vasodilation (29). It is worthy of note that an initial rhythm of asystole, sometime after a cardiac arrest, does not exclude a primary VF or polymorphic VT event. We have previously reported a case of CPVT where syncope correlated with polymorphic VT, recorded by an implanted digital loop recorder; the device only activated when prolonged asystole was detected after this (30). The finding of PEA in the first case in this series is unusual and presumably reflected an acutely hypoxic myocardium. Thus neither asystole nor PEA at presentation in a young person with a sudden unexpected cardiac arrest excludes CPVT.

r; the device only activated when prolonged asystole was detected after this (30). The finding of PEA in the first case in this series is unusual and presumably reflected an acutely hypoxic myocardium. Thus neither asystole nor PEA at presentation in a young person with a sudden unexpected cardiac arrest excludes CPVT. CPVT is an important differential diagnosis of any young person (typically 3 to 40 yr old) who has suffered an unexplained sudden cardiac arrest. Features which should raise suspicion include the following: 1) victim was previously well; 2) cardiac arrest occurred during a physical activity (especially in water) or with excitement; 3) electrocardiogram shows frequent VEs (usually but not always multimorphic); 4) VEs become more frequent (or join to form VT) with epinephrine, and become less frequent with opiates and anesthesia; and 5) bidirectional VT (where ventricular complex QRS axis alternate by 180°) is virtually pathognomonic when seen but is not needed for the diagnosis (Fig. 3). An algorithm has been proposed for the investigation of such patients in the ICU (11). Confirmation of the diagnosis, even if the patient has suffered brain death and treatment may be withdrawn, will guide further investigations and prognosis for surviving family (10, 11). During epinephrine or isoprenaline infusions, flecainide should be immediately available. In the event of good neurologic recovery, an exercise test is diagnostic (2, 9, 11).

e patient has suffered brain death and treatment may be withdrawn, will guide further investigations and prognosis for surviving family (10, 11). During epinephrine or isoprenaline infusions, flecainide should be immediately available. In the event of good neurologic recovery, an exercise test is diagnostic (2, 9, 11). CONCLUSIONS A high suspicion for CPVT is important for all clinicians responsible for the emergency resuscitation of children and young adults (24). In the context of unremitting ventricular arrhythmia, not responding appropriately to standard resuscitation measures, IV opiates and general anesthesia, potentially flecainide, and the avoidance of epinephrine, can be life-saving. ACKNOWLEDGMENTS We gratefully acknowledge Charlene Nell, Department of Cardiology, Green Lane Cardiovascular Services, Auckland City Hospital, New Zealand, for assistance with article preparation. *See also p. 297. Dr. Skinner receives salary support from Cure Kids. The authors have disclosed that they do not have any potential conflicts of interest. Each child is enrolled in the Cardiac Inherited Disease Registry, part of which gives consent for publication of deidentified data. In addition, although there are no identifying features (name, year of birth, etc.) and no photos, each family has reviewed this article and consent has been obtained for publication from the guardians of each of the three children.

Outborn infants, especially extremely preterm (EPT, birth at < 28 wk of gestation) infants, are at increased risk of mortality and morbidity in comparison to inborn infants because of their prominent vulnerability (1–3). There has been no clear progress in terms of the prevention of EPT births until recently, and the frequency of preterm birth is currently not decreasing (4). Short- and long-term outcomes of EPT infants have recently improved (5–8). However, long-term neurodevelopmental impairments (NDIs) remain a major concern, and there are few reports concerning the long-term outcomes of outborn EPT infants (5–8). EPT infants are at the highest risk and have the most specialized needs such as subspecialized neonatologists, highly skilled nursing staff, pediatric surgical specialists, physiologic monitoring equipment, and laboratory facilities. An appropriate timing must be selected for maternal transport to avoid the delivery of EPT infants in a facility without the necessary capabilities, where it is extremely difficult to stabilize the infants at birth and transport them to a neonatal ICU (NICU) of a tertiary center. The American Academy of Pediatrics and the American College of Obstetrics and Gynecologists Guidelines for Perinatal Care recommend that hospital-based services should be organized within geographical regions. Maternity hospitals should be located based on functional capabilities to provide adequate care for pregnant women and neonates at increased risk of EPT births (8, 9). The regionalization of perinatal care has been in place since around 1990 in Japan. However, Japan is unique in that approximately 50% of deliveries are managed at birth centers or level I hospitals for maternal care (10). Transporting women at risk of very preterm births to tertiary centers has significantly decreased the neonatal mortality and morbidity rate (11, 12). Considering the organization of parturient facilities in Japan and the difficulty of predicting EPT deliveries, it is possible that EPT outborn infants may be born at birth centers or level I maternity hospitals.

births to tertiary centers has significantly decreased the neonatal mortality and morbidity rate (11, 12). Considering the organization of parturient facilities in Japan and the difficulty of predicting EPT deliveries, it is possible that EPT outborn infants may be born at birth centers or level I maternity hospitals. In most cases, there is insufficient transportation time because of the rapid progression of labor. Thus, solving the present perinatal problems (e.g., appropriate resuscitation and transportation of vulnerable EPT infants) at birth centers or level I maternity hospitals remains a crucial issue. This study aimed to evaluate whether the birth status (outborn or inborn) of EPT infants influenced their short- and long-term outcomes based on information from the Neonatal Research Network of Japan (NRNJ) database.

In most cases, there is insufficient transportation time because of the rapid progression of labor. Thus, solving the present perinatal problems (e.g., appropriate resuscitation and transportation of vulnerable EPT infants) at birth centers or level I maternity hospitals remains a crucial issue. This study aimed to evaluate whether the birth status (outborn or inborn) of EPT infants influenced their short- and long-term outcomes based on information from the Neonatal Research Network of Japan (NRNJ) database. MATERIALS AND METHODS The protocol of this study was approved by the central internal review board at Tokyo Women’s Medical University, where all data were collected and stored. Written informed consent was obtained from the parents or guardians of all infants in the NRNJ. This was a retrospective analysis of 12,164 EPT infants, defined as those with a gestational age between 22 + 0 and 27 + 6 weeks, among very-low-birth-weight (VLBW) singleton and twin infants. Clinical data between 2003 and 2011 were obtained from the NRNJ database, which was created with a grant from the Ministry of Health, Labor, and Welfare of Japan. There was a systematic registration of VLBW infants so that almost all cases that were handled in the study period would be present in the medical records. This database collected data on greater than 50% of EPT infants born in Japan during the study period, and 186 facilities (including 84 tertiary centers) were registered for the NRNJ in 2011. This database contains information on morbidity, mortality, and follow-up data of EPT infants in participating facilities. Exclusion criteria were as follows: multiple births (more than triplets), any major congenital malformations, admission after greater than 1 day after birth, transported to other hospitals at birth, and those without records of mortality status and information on the mode of delivery. Data for 81 infants who were born alive but died in the delivery room were also excluded (Fig. 1). Gestational age was calculated based on ultrasound examination during the first trimester and the data of the last menstrual period. The presence of pathologic chorioamnionitis was examined based on Blanc’s criteria. Premature rupture of membranes (PROM) was defined as the rupture of membranes before the onset of labor. Antenatal corticosteroid (ACS) administration was defined as at least one dose of ACS to the mother at any time before delivery to accelerate fetal lung maturity. Small for gestational age (SGA) was defined as a birth weight below the 10th percentile of the standard birth weight for gestational age published by the Japan Pediatric Society (13).

ACS) administration was defined as at least one dose of ACS to the mother at any time before delivery to accelerate fetal lung maturity. Small for gestational age (SGA) was defined as a birth weight below the 10th percentile of the standard birth weight for gestational age published by the Japan Pediatric Society (13). Respiratory distress syndrome (RDS) was diagnosed on the basis of clinical presentation and characteristic radiographic appearance. Chronic lung disease (CLD) was defined as a persistent requirement for supplemental oxygen 28 days after birth or at 36 weeks postmenstrual age. Patent ductus arteriosus (PDA) was diagnosed based on both echocardiographic and clinical findings. Severe intraventricular hemorrhage (IVH) was defined as grade III to IV, according to the Papile criteria. Cystic periventricular leukomalacia (PVL) was defined as any cyst formation observed at any time in the periventricular white matter on cranial ultrasound or head MRI. The presence of sepsis was determined by positive blood culture results. Necrotizing enterocolitis (NEC) was defined according to Bell’s classification stage II or greater. Retinopathy of prematurity (ROP) was considered to be severe if the worst stage of ROP was III (intermediate) or greater according to the criteria proposed by the task force of the Ministry of Health, Labor and Welfare of Japan, which was equivalent to stage III or greater in the International Classification of ROP, and if treatment was required with laser coagulation, cryocoagulation therapy, or both. Neonatal mortality was defined as death of an infant before hospital discharge. The follow-up protocol consisted of routine physical and neurologic evaluations and developmental assessments at three years (36–42 mo) of chronological age for surviving EPT infants at each participating center (14). Neurologic evaluation at 3 years old included signs and symptoms of cerebral palsy (CP) and sensory abnormality. CP was defined as a nonprogressive CNS disorder characterized by abnormal muscle tone in at least one extremity and abnormal control of movement and posture. Visual impairment was defined as unilateral or bilateral blindness diagnosed by ophthalmologists. Severe hearing impairment was defined as when amplification was required. The assessment of cognitive function was performed using the Kyoto Scale of Psychologic Development (KSPD) test (15).

ntrol of movement and posture. Visual impairment was defined as unilateral or bilateral blindness diagnosed by ophthalmologists. Severe hearing impairment was defined as when amplification was required. The assessment of cognitive function was performed using the Kyoto Scale of Psychologic Development (KSPD) test (15). This test was administered by experienced testers who were certified psychologists blinded to the perinatal details at each participating center. All information about the infants was collected anonymously, and the stored data were unlinked from individual data. When the development quotient (DQ) was less than 70, the infant was judged as having a cognitive impairment according to the protocol of the Society for Follow-up Study of High-risk Infants. If the KSPD assessment was not available, the pediatrician estimated the child’s developmental level as delayed or not delayed. In cases judged as delayed, the developmental level was assumed as equivalent to a DQ score less than 50. Infants with CP, visual impairment, severe hearing impairment, and cognitive impairment were designated as having NDI. We analyzed the effect of place of birth (outborn vs inborn) on short- and long-term outcomes. Short-term outcomes were RDS, CLD, surgery for PDA, severe IVH, cystic PVL, sepsis, NEC or intestinal perforation, severe ROP, and neonatal mortality. The chosen long-term outcomes were death after hospital discharge up to 3 years, CP, visual impairment, severe hearing impairment, cognitive impairment, and NDI.

tcomes. Short-term outcomes were RDS, CLD, surgery for PDA, severe IVH, cystic PVL, sepsis, NEC or intestinal perforation, severe ROP, and neonatal mortality. The chosen long-term outcomes were death after hospital discharge up to 3 years, CP, visual impairment, severe hearing impairment, cognitive impairment, and NDI. Figure 1. Flow chart of study population. Analyzed data (n = 12,164) were obtained from the Neonatal Research Network of Japan database (2003–2011). GA = gestational age, NICU = neonatal ICU. Statistical analyses were performed using SPSS Version 24.0 (SPSS, Chicago, IL). The significance level was set at 0.05. Continuous data are presented as means ± sd. Absolute standardized differences (ASDs) were calculated to evaluate the statistical difference at baseline between those who dropped out in the evaluation of long-term prognosis and those who completed follow-up, with an ASD above 10% representing a meaningful imbalance. Differences in perinatal baseline characteristics between the outborn and inborn infants were tested using a chi-square test and t test, as appropriate, to find confounders in the relationship analysis between being outborn and outcomes (short- and long-term outcome). The relationship between being outborn and outcomes was examined by logistic regression analysis. With each outcome (short-term and long-term outcome) as a dependent variable, the odds ratio (OR) and 95% CI of being outborn (vs inborn) were calculated.

between being outborn and outcomes (short- and long-term outcome). The relationship between being outborn and outcomes was examined by logistic regression analysis. With each outcome (short-term and long-term outcome) as a dependent variable, the odds ratio (OR) and 95% CI of being outborn (vs inborn) were calculated. RESULTS The study group is illustrated in Figure 1. A total of 35,057 VLBW infants were registered in the database between 2003 and 2011. In total, 2,354 infants were excluded because of major congenital malformation, 1,300 infants were excluded because of multiple births (more than triplets), and 21,631 infants were excluded because they were born at gestational ages greater than or equal to 28 + 0 weeks or less than 22 + 0 weeks. After exclusion of inappropriate cases (infants without records of mortality status), data of 12,245 singleton and twin infants born at gestational ages between 22 + 0 and 27 + 6 weeks were available. Of these, 81 infants died in the delivery room and were excluded. A final number of 12,164 EPT infants were enrolled as the study population. Of the 12,164 subjects, 785 (6.5%) were outborn EPT infants. Of the 785 outborn and 11,379 inborn infants, 109 (13.8%) and 1,591 (14.0%), respectively, died before discharge from the hospital. Of the surviving 676 outborn and 9,788 inborn infants, follow-up data at 3 years old were collected from 319 (47.2%) and 4,573 (46.7%), respectively. Maternal and delivery characteristics for the outborn and inborn groups are listed in Table 1. There were no differences between the groups in terms of the distribution of infants from 22 to 27 gestational weeks. The outborn group had significantly lower rates of twin birth (p < 0.01), pregnancy-induced hypertension (PIH) (p < 0.01), pathologic chorioamnionitis (p < 0.01), PROM (p < 0.01), administration of ACS (p < 0.01), cesarean delivery (p < 0.01), and SGA (p < 0.01) and higher rate of Apgar (5 min) less than or equal to 3 (p < 0.01). The outborn group also had lower maternal age (p < 0.01) and higher birth weight (p < 0.01). The effects on short-term outcomes are listed in Table 2.

PROM (p < 0.01), administration of ACS (p < 0.01), cesarean delivery (p < 0.01), and SGA (p < 0.01) and higher rate of Apgar (5 min) less than or equal to 3 (p < 0.01). The outborn group also had lower maternal age (p < 0.01) and higher birth weight (p < 0.01). The effects on short-term outcomes are listed in Table 2. The multivariate logistic regression analysis performed after adjusting for confounders (twin birth, PIH, pathologic chorioamnionitis, PROM, ACS administration, cesarean delivery, SGA, Apgar [5 min] ≤ 3, maternal age, and birth weight) showed that outborn infants had higher odds of severe IVH (adjusted OR, 1.49; 95% CI, 1.11–2.00; p < 0.01) and NEC or FIP (adjusted OR, 1.58; 95% CI, 1.09–2.30; p = 0.015) for short-term outcomes. The effects on long-term outcomes are listed in Table 3. There were no significant differences between the outborn and inborn groups, except for the rate of cognitive impairment (adjusted OR, 1.49; 95% CI, 1.01–2.20; p = 0.04) in the multivariate logistic regression analysis. ASD values were calculated to evaluate statistical difference in baseline characteristics between those who dropped out of the evaluation of long-term prognosis and those who completed follow-up. All ASD values were under 10% (Table 4). TABLE 1. Demographics and Baseline Characteristics (n = 12,164) TABLE 2. The Multivariate Logistic Regression Analysis of Short-Term Outcomes TABLE 3. The Multivariate Logistic Regression Analysis of Long-Term Outcomes TABLE 4. Baseline Characteristics of Missing Data Not Evaluated at 3 Years Old

The multivariate logistic regression analysis performed after adjusting for confounders (twin birth, PIH, pathologic chorioamnionitis, PROM, ACS administration, cesarean delivery, SGA, Apgar [5 min] ≤ 3, maternal age, and birth weight) showed that outborn infants had higher odds of severe IVH (adjusted OR, 1.49; 95% CI, 1.11–2.00; p < 0.01) and NEC or FIP (adjusted OR, 1.58; 95% CI, 1.09–2.30; p = 0.015) for short-term outcomes. The effects on long-term outcomes are listed in Table 3. There were no significant differences between the outborn and inborn groups, except for the rate of cognitive impairment (adjusted OR, 1.49; 95% CI, 1.01–2.20; p = 0.04) in the multivariate logistic regression analysis. ASD values were calculated to evaluate statistical difference in baseline characteristics between those who dropped out of the evaluation of long-term prognosis and those who completed follow-up. All ASD values were under 10% (Table 4). TABLE 1. Demographics and Baseline Characteristics (n = 12,164) TABLE 2. The Multivariate Logistic Regression Analysis of Short-Term Outcomes TABLE 3. The Multivariate Logistic Regression Analysis of Long-Term Outcomes TABLE 4. Baseline Characteristics of Missing Data Not Evaluated at 3 Years Old DISCUSSION In this study, we found that the outborn group included more cases at small risk of EPT and had a significantly higher frequency of severe IVH, NEC or FIP, and cognitive impairment. We could not find studies about the long-term outcomes of outborn EPT infants, although there were some reports on short-term outcomes. To the best of our knowledge, this is the first large cohort study to explore short- and long-term outcomes of EPT inborn and outborn infants. Higher rates of mortality and morbidity for outborn infants have been reported (1–3).

m outcomes of outborn EPT infants, although there were some reports on short-term outcomes. To the best of our knowledge, this is the first large cohort study to explore short- and long-term outcomes of EPT inborn and outborn infants. Higher rates of mortality and morbidity for outborn infants have been reported (1–3). Significant differences were observed between the inborn and outborn groups in demographic and clinical factors. The outborn group may include more cases at smaller risk of EPT when compared with the inborn group, since the inborn group had significantly higher rates of twin birth, PIH, pathologic chorioamnionitis, PROM, SGA, and lower birth weight. If only background factors are considered, a good outcome for outborn infants can be expected, with the exceptions of ACS treatment and higher rate of a low Apgar score at 5 minutes. We found no difference in mortality risk between outborn and inborn infants. However, a significant difference in severe IVH and NEC or FIP was found, even after adjusting for perinatal risk factors. Nevertheless, the rates of all these morbidities remain low compared with other studies (7). There were no significant differences in long-term outcomes between the two groups, except for the rate of cognitive impairment. The high rate of cognitive impairment may be due to the high rate of severe IVH and NEC or FIP seen in the short-term outcomes.

l these morbidities remain low compared with other studies (7). There were no significant differences in long-term outcomes between the two groups, except for the rate of cognitive impairment. The high rate of cognitive impairment may be due to the high rate of severe IVH and NEC or FIP seen in the short-term outcomes. Our data did not include stillbirths and deaths in the delivery room at birth hospitals other than the participating tertiary centers. Data about resuscitation of outborn livebirths were only available for those referred to tertiary centers and not for outborn infants who died at the birth hospital without being referred. Thus, it was not possible to obtain information on decision-making or other aspects of care for outborn livebirths that were not referred to tertiary centers. Insufficient time to administer ACS, suboptimal resuscitation at birth, and lack of medical equipment and staff expertise have been identified as risk factors contributing to mortality in EPT infants (16). Previous studies emphasize the value of high-volume level III NICU care in minimizing neonatal mortality and morbidity among VLBW infants (17, 18).

CS, suboptimal resuscitation at birth, and lack of medical equipment and staff expertise have been identified as risk factors contributing to mortality in EPT infants (16). Previous studies emphasize the value of high-volume level III NICU care in minimizing neonatal mortality and morbidity among VLBW infants (17, 18). Unfortunately, there is no nationwide standard for the transportation of EPT infants (including use of transport incubators and presence of a neonatologist and neonatal nursing staff) in Japan at present. Improvements in the regional perinatal system, such as optimized transportation to tertiary centers, can improve the outcomes of EPT infants, as several factors at birth or during transport play a role in the health risks of outborn EPT infants (2, 19). The risk of severe IVH and NEC or FIP in outborn infants may reflect the influence of inappropriate circumstances such as insufficient resuscitation at the time of birth and unstable environment such as hypo- or hyperthermia in an ambulance during the transportation period. The association between admission temperature and mortality and morbidity in EPT infants has been reported (19). It seems unlikely that there will be a dramatic improvement in this situation in the future, as small regional birth centers and level I maternity care hospitals handle approximately 50% of deliveries in Japan. In other words, EPT outborn births will continue to occur for a long period in Japan. Part of the solution to improve neonatal retrieval services of the births in birth centers and level I maternity care hospitals could be to increase efforts to transport the pregnancies at risk of EPT in utero. Retrieval services means not only emergency transport services but also for subspecialized neonatologists to be present at births of the EPT infants in nontertiary hospitals to resuscitate and to stabilize the infants whenever possible. The regional maternal and neonatal transport system has been improving gradually, but the neonatal retrieval services for the regional perinatal system have remained largely unimproved in the past 20 years. Improved quality of resuscitation was reported when dedicated neonatal retrieval teams resuscitated outborn infants in the delivery room (16).

natal transport system has been improving gradually, but the neonatal retrieval services for the regional perinatal system have remained largely unimproved in the past 20 years. Improved quality of resuscitation was reported when dedicated neonatal retrieval teams resuscitated outborn infants in the delivery room (16). Retrieval services may make it possible to provide outborn infants with intensive care or interventions such as intubation, surfactant administration, cardiorespiratory monitoring, ventilation support, and accurate drug administration at the birth hospital and during transport, thus increasing the possibility that outborn infants will receive intensive care approximately equal to that of a level III NICU at birth centers or level I maternity care hospitals. It was reported that the mortality before NICU admission for EPT outborn infants was significantly higher when compared with that for inborn infants (20). We may need to equip the regional perinatal system to send the neonatal retrieval team to birth hospitals in case of emergency as much as possible; however, there are several issues to address, such as securing neonatologists, sharing information between the birth hospital and tertiary center, and arranging the ambulance for exclusive use for newborn babies. Naturally, obstetricians at birth hospitals should acquire enough knowledge and skills to resuscitate EPT infants and to stabilize their general condition until the arrival of the retrieval team or transport to a tertiary hospital.

l and tertiary center, and arranging the ambulance for exclusive use for newborn babies. Naturally, obstetricians at birth hospitals should acquire enough knowledge and skills to resuscitate EPT infants and to stabilize their general condition until the arrival of the retrieval team or transport to a tertiary hospital. Our study has some limitations. The first was the amount of missing data (53.2%) related to information on NDI at 3 years. We made a comparison for maternal and delivery characteristics between infants evaluated and not evaluated for NDI at 3 years. All ASD values to evaluate statistical difference were under 10%, and the low follow-up rate might not have a clinically important influence on short- and long-term outcomes between EPT infants who had data available for NDI and survivors who were lost to follow-up. The second was that we used the KSPD scale for evaluating developmental outcomes instead of the Bayley Scales of Infant Development III (Bayley III), which is widely used throughout the world. However, one study reported that the KSPD scale correlated well with Bayley III (15). Furthermore, our study design was a case-control comparison using the same developmental scale. Thus, we could not expect any discrepancy in our results. The third was that stillbirths and deaths of neonates who could not be transported to tertiary centers were not included in our data. These data may influence the result of our analysis for the short- and long-term outcomes to some extent. Recently, the unification of the nationwide neonatal and obstetrical data was discussed at the meeting of the Japan Society of Perinatal and Neonatal Medicine. We should acquire the information for outborn livebirths that were not referred to tertiary centers to more accurately determine the current provision of perinatal care.

the unification of the nationwide neonatal and obstetrical data was discussed at the meeting of the Japan Society of Perinatal and Neonatal Medicine. We should acquire the information for outborn livebirths that were not referred to tertiary centers to more accurately determine the current provision of perinatal care. CONCLUSIONS Our study pointed out that there was a significantly higher frequency of severe IVH, NEC or FIP, and cognitive impairment in outborn infants compared with the inborn infants. Thus, outborn/inborn birth status may influence short- and long-term outcomes of EPT infants. For the improvement of the current perinatal care standard, we should obtain more precise data of outborn EPT infants, including livebirths that were not referred to tertiary centers. Based on this analysis, more effort is needed to improve our perinatal environment, such as developing a neonatal retrieval system for outborn EPT infants. ACKNOWLEDGMENTS We thank all patients for their generous contributions. We also thank the data center at Tokyo Women’s Medical University (Tokyo, Japan) for providing organizational support to the Neonatal Research Network of Japan.

CONCLUSIONS Our study pointed out that there was a significantly higher frequency of severe IVH, NEC or FIP, and cognitive impairment in outborn infants compared with the inborn infants. Thus, outborn/inborn birth status may influence short- and long-term outcomes of EPT infants. For the improvement of the current perinatal care standard, we should obtain more precise data of outborn EPT infants, including livebirths that were not referred to tertiary centers. Based on this analysis, more effort is needed to improve our perinatal environment, such as developing a neonatal retrieval system for outborn EPT infants. ACKNOWLEDGMENTS We thank all patients for their generous contributions. We also thank the data center at Tokyo Women’s Medical University (Tokyo, Japan) for providing organizational support to the Neonatal Research Network of Japan. List of participating units: Sapporo City Hospital, Asahikawa Kosei Hospital, Kushiro Red Cross Hospital, Obihiro Kosei Hospital, Tenshi Hospital, NTT East Sapporo Hospital, Nikko Kinen Hospital, Sapporo Prefecture Medical University, Asahikawa Medical University, Aomori Prefecture Central Hospital, Iwate Medical University, Iwate Prefecture Ohfunato Hospital, Prefecture Kuji Hospital, Iwate Prefecture Ninohe Hospital, Sendai Red Cross Hospital, Akita Red Cross Hospital, Akita University, Tsuruoka City Shonai Hospital, Yamagata, Yamagata Prefecture Central Hospital, Fukusima Prefecture Medical University, Takeda General Hospital, National Fukushima Hospital, Tsukuba University, Tsuchiura Kyodo Hospital, Ibaraki Children’s Hospital, Dokkyo Medical University, Jichi Medical University, Ashikaga Red Cross Hospital, Gunma Prefecture Children’s Hospital, Kiryu Kosei General Hospital, Gunma University, Saitama Prefecture Children’s Hospital, National Nishisaitama Central Hospital, Saitama Medical University Medical Center, Kawaguchi City Medical Center, Jichi Medical University Saitama Medical Center, Asahi Central Hospital, Chiba City Kaihin Hospital, Kameda General Hospital, Tokyo Women’s Medical University Yachiyo Medical Center, Juntendo University Urayasu Hospital, Tokyo Metropolitan Children’s Medical Center, Tokyo Women’s Medical University, Aiiku Hospital, Nihon University, National International Medical Center, Teikyo University, Showa University, Japan Red Cross Hospital, National Center for Child Health and Development, Tokyo Metropolitan Otsuka Hospital, Toho University, Tokyo Metropolitan Bokuto Hospital, Tokyo Jikei Medical University, Saint Luku Hospital, Juntendo University, Sanikukai Hospital, Katsushika Red Cross Hospital, Yokohama Rosai Hospital, Yokohama City University Medical Center, Marianna Medical University, Kanagawa Children’s Medical Center, Tokai University, Kitazato University, Odawara City Hospital, Nippon Medical School Musashi Kosugi Hospital, Saiseikai Eastern Yokohama Hospital, Yokohama Medical Center, Yamanashi Prefecture Central Hospital, Nagano Children’s Hospital, Shinshu University, Iida City Hospital, National Shinshu Ueda Medi

Medical Center, Tokai University, Kitazato University, Odawara City Hospital, Nippon Medical School Musashi Kosugi Hospital, Saiseikai Eastern Yokohama Hospital, Yokohama Medical Center, Yamanashi Prefecture Central Hospital, Nagano Children’s Hospital, Shinshu University, Iida City Hospital, National Shinshu Ueda Medi cal Center, Saku General Hospital, Niigata University, Niigata Central Hospital, Niigata City Hospital, Nagaoka Red Cross Hospital, Koseiren Takaoka Hospital, Toyama Prefectural Central Hospital, Toyama University, Ishikawa Prefectural Central Hospital, Kanazawa, Medical University, Fukui Prefectural Hospital, Fukui University, Gifu Prefectural Medical Center, Takayama Red Cross Hospital, Seirei Hamamatsu Hospital, Shizuoka Saiseikai Hospital, Shizuoka Children’s Hospital, Hamamatsu Medical University, Yaizu City Hospital, Fujieda City Hospital, Nagoya Red Cross Daini Hospital, Nagoya University, Nagoya Red Cross Daiici Hospital, Anjokosei Hospital, Koritsu Tosei Hospital, Komaki City Hospital, Toyota Memorial Hospital, Okazaki City Hospital, Konankosei Hospital, National Mie Central Medical Center, Ise Red Cross Hospital, Yokkaichi City Hospital, Otsu Red Cross Hospital, Shiga Medical University, Nagahama Red Cross Hospital, Uji Tokushukai Hospital, Japan Baptist Hospital, Kyoto University, Kyoto Red Cross Daiichi Hospital, National Maizuru Medical Center, Fukuchiyama City Hospital, Kyoto Prefecture Medical University, Kyoto City Hospital, Yodogawa Christian Hospital, Osaka Medical Center and Research Institute for Maternal and Child Health, Osaka University, Takatsuki General Hospital, Kansai Medical University, Osaka City General Hospital, Osaka City Sumiyoshi Hospital, Aizenbashi Hospital, Toyonaka City Hospital, National Cerebral and Cardiovascular Center, Kitano Hospital, Saiseikai Suita Hospital, Chifune Hospital, Bell Land General Hospital, Rinku General Hospital, Yao City Hospital, Osaka City University, Kobe Children’s Hospital, Kobe University, Saiseikai Hyogo Hospital, Kobe City Medical Center Central Hospital, Hyogo Medical University, Himeji Red Cross Hospital, Toyooka General Hospital, Hyogo Prefectural Awaji Hospital, Nara Prefecture Medical University, Wakayama Prefecture Medical University, Tottori Prefectural Central Hospital, Tottori University, Shimane Prefectural Central Hospital, Matsue Red Cross Hospital, Kurashiki Central Hospital, Tsuyama Central Hospital, Kawasaki Medical University, National Okayama Medical Center, Okayama Red Cross H

e Medical University, Wakayama Prefecture Medical University, Tottori Prefectural Central Hospital, Tottori University, Shimane Prefectural Central Hospital, Matsue Red Cross Hospital, Kurashiki Central Hospital, Tsuyama Central Hospital, Kawasaki Medical University, National Okayama Medical Center, Okayama Red Cross H ospital, Hiroshima City Central Hospital, Hiroshima Prefectural Hospital, Tsuchiya General Hospital, National Kure Medical Center, Yamaguchi Prefecture Medical Center, Tokushima University, Kagawa University, Shikoku Medical Center for Children and Adults, Matsuyama Red Cross Hospital, Ehime Prefectural Central Hospital, Kochi Health Science Center, Saint Maria Hospital, National Kyushu Medical Center, Kurume University, Kitakyushu City Hospital, University of Occupational and Environmental Health Japan, Fukuoka University, Kyushu University, Iizuka Hospital, National Kokura Medical Center, National Saga Hospital, National Nagasaki Medical Center, Kumamoto City Hospital, Kumamoto University, Oita Prefectural Hospital, Almeida Memorial Hospital, Nakatsu City Hospital, Miyazaki University, Kagoshima City Hospital, Imakyure General Hospital, Okinara Prefectural Central Hospital, Naha City Hospital, and Okinawa Red Cross Hospital. *See also p. 994. The authors have disclosed that they do not have any potential conflicts of interest.

Severe sepsis and septic shock are a significant cause of morbidity and mortality in patients admitted to ICUs (1–3). Adjunctive systemic corticosteroids can be used in the presence of shock with cardiovascular failure (CV Failure) in an attempt to improve hemodynamics, but there is still controversy regarding its efficacy and indications (4–6), and even when the pediatric septic shock population was stratified by mortality risk, the analysis did not show benefit from systemic steroids administration (7). Two large, randomized, controlled trials of cortisol replacement therapy in patients with septic shock have shown opposite results regarding survival benefit (8, 9). Currently, the recommendation from the Surviving Sepsis Campaign 2012 is to administer stress doses of hydrocortisone for patients with catecholamine-resistant shock or patients with suspected or proven absolute adrenal insufficiency (10). One of the possible causes for the conflicting results found in these trials may lie in the fact that the individual patient response to steroids is variable, and there is no consensus regarding who would benefit from its use and how to properly identify these patients (11).

Severe sepsis and septic shock are a significant cause of morbidity and mortality in patients admitted to ICUs (1–3). Adjunctive systemic corticosteroids can be used in the presence of shock with cardiovascular failure (CV Failure) in an attempt to improve hemodynamics, but there is still controversy regarding its efficacy and indications (4–6), and even when the pediatric septic shock population was stratified by mortality risk, the analysis did not show benefit from systemic steroids administration (7). Two large, randomized, controlled trials of cortisol replacement therapy in patients with septic shock have shown opposite results regarding survival benefit (8, 9). Currently, the recommendation from the Surviving Sepsis Campaign 2012 is to administer stress doses of hydrocortisone for patients with catecholamine-resistant shock or patients with suspected or proven absolute adrenal insufficiency (10). One of the possible causes for the conflicting results found in these trials may lie in the fact that the individual patient response to steroids is variable, and there is no consensus regarding who would benefit from its use and how to properly identify these patients (11). Over a decade ago, Marik et al (12) introduced the concept of critical illness–related corticosteroid insufficiency (CIRCI) as a condition in which the level of endogenous cortisol is thought to be low relative to the degree of illness severity. It has been described in association with a broad spectrum of critical illnesses, including septic shock (8), acute respiratory distress syndrome (13), traumatic brain injury (14), liver failure (15), burns (16), pancreatitis (17), and following cardiopulmonary bypass (18). With the new CIRCI definition, the field of critical care medicine aimed to enter an era of personalized medicine through the use of a simple and quick method, the adrenocorticotropic hormone (ACTH) stimulation test, to identify individuals who would specifically have a better response to the use of adjunctive steroid therapy (8, 19), but subsequent studies failed to demonstrate its efficacy at identifying a subpopulation that would clearly benefit from systemic steroids culminating in the most recent surviving sepsis campaign guidelines not recommending routine use of ACTH stimulation tests (9, 10).

to the use of adjunctive steroid therapy (8, 19), but subsequent studies failed to demonstrate its efficacy at identifying a subpopulation that would clearly benefit from systemic steroids culminating in the most recent surviving sepsis campaign guidelines not recommending routine use of ACTH stimulation tests (9, 10). Therefore, investigators have more recently sought alternative mechanisms that may account for interpatient corticosteroid response variability and perhaps develop new diagnostic tools by evaluating peripheral steroid resistance and cortisol metabolism (20–23). In order to act, circulating cortisol has to diffuse across the cell membrane and bind to the intracellular cytosolic glucocorticoid receptor (GCR)-α. The cortisol-GCR complex then migrates to the nucleus where it inhibits the transcription of inflammatory genes by nuclear factor κ-light-chain-enhancer of activated B cells or activator protein-1, thus inhibiting the production of inflammatory cytokines and intracellular adhesion molecule-1 (24, 25). This could imply that the extent of its effect is proportional to the GCR expression, subtype, and affinity in a determined target cell (26).

actor κ-light-chain-enhancer of activated B cells or activator protein-1, thus inhibiting the production of inflammatory cytokines and intracellular adhesion molecule-1 (24, 25). This could imply that the extent of its effect is proportional to the GCR expression, subtype, and affinity in a determined target cell (26). Using genome-wide expression profiling, we recently reported a subclass of children with septic shock characterized by decreased expression of a group of genes corresponding to the GCR signaling pathway (27–29). This subclass of patients had a higher level of illness severity and a higher mortality rate compared with two other identified gene expression–based subclasses. We hypothesized that a subset of critically ill patients with cardiovascular dysfunction is characterized by decreased expression of the GCR and that this group has worse severity of illness measured by Pediatric Risk of Mortality (PRISM) III (30) and multiple organ failure (OF) burden (31). We performed a prospective, observational cohort study to characterize GCR expression in peripheral WBCs of critically ill pediatric patients.

sed expression of the GCR and that this group has worse severity of illness measured by Pediatric Risk of Mortality (PRISM) III (30) and multiple organ failure (OF) burden (31). We performed a prospective, observational cohort study to characterize GCR expression in peripheral WBCs of critically ill pediatric patients. METHODS Patients and Data Collection The study protocol was approved by Cincinnati Children’s Hospital Medical Center (CCHMC) Institutional Review Board and written informed consent was obtained from a parent or legal guardian for each enrolled patient. Subjects were eligible if they were admitted to the PICU at CCHMC and had an indwelling catheter (central venous catheter or an arterial catheter) from which blood samples could be obtained. Patients were excluded from the study if informed consent was not obtained or if the attending physician did not approve enrollment.

jects were eligible if they were admitted to the PICU at CCHMC and had an indwelling catheter (central venous catheter or an arterial catheter) from which blood samples could be obtained. Patients were excluded from the study if informed consent was not obtained or if the attending physician did not approve enrollment. After initial enrollment, blood samples were obtained in the first 24 hours of admission to the PICU for GCR flow cytometry analysis and random serum cortisol levels. Clinical and laboratory data were prospectively collected daily until discharge from PICU using a standardized paper based collection form. The following variables were evaluated initially: CV Failure (defined as the need for vasoactive or inotropic drug support at the admission day), use of steroids (defined as administration of any dose of steroids during the present admission prior to sample collection either in the emergency department or PICU), and chronic steroids use (above 14 d of steroids preceding admission). Severity of illness (PRISM III) was evaluated at admission, maximum number of OF was followed up daily from admission until the seventh day of PICU stay, and 28-day mortality was evaluated.

ple collection either in the emergency department or PICU), and chronic steroids use (above 14 d of steroids preceding admission). Severity of illness (PRISM III) was evaluated at admission, maximum number of OF was followed up daily from admission until the seventh day of PICU stay, and 28-day mortality was evaluated. Laboratory Procedure A single random total cortisol level was collected simultaneously with the GCR expression samples. Serum was sent to the main laboratory at CCHMC for cortisol level analysis using a chemiluminiscent microparticle immunoassay and the Architect i200 SR Analyzer (Abbot Laboratories, Abbot Park, IL). GCR receptor expression was measured using flow cytometry using surface antibodies to determine cell type (Pacific Blue Mouse Anti-Human CD4—BD Pharmingen (Franklin Lakes, NJ) for CD4 lymphocytes; Alexa Fluor 700 Mouse Anti-Human CD8—BD Pharmingen for CD8 lymphocytes; Phycoerytrin Mouse Anti-Human CD14—BD Pharmingen for monocytes; Alexa Flour 647 Mouse Anti-human CD66b—BD Pharmingen for neutrophils). After surface staining, cells were permeabilized to detect intracellular GCR receptors (anti-glucocorticoid receptor [FITC] Mouse, clone 5E4 MA1-81793—Thermo Scientific, Waltham, MA). Isotype (Mouse IgG1 [HyblgG1] [fluorescein isothiocyanate]—Abcam, Cambridge Science Park, Cambridge, UK) and fluorescence minus one (FMO) controls were used. Fluorescence samples were fixed in paraformaldehyde and read within a maximum of 5 days on a BD LSR II machine (BD biosciences, Franklin Lakes, NJ). The results were then analyzed using FACSDiva software (BD biosciences). The lymphocyte population (P1) and monocyte and neutrophil populations (P2) were discriminated, and gates were generated for cell type (x-axis) and GCR expression (y-axis). The results were gated in four areas, and mean fluorescence was measured for 10,000 events on the area expressing both the surface antigen and the GCR antigen. FMO sample values were subtracted from GCR values resulting in a mean fluorescence intensity (MFI), and those values were used for statistical analysis. Isotype controls were used for experimental control (to check if blocking was appropriate).

events on the area expressing both the surface antigen and the GCR antigen. FMO sample values were subtracted from GCR values resulting in a mean fluorescence intensity (MFI), and those values were used for statistical analysis. Isotype controls were used for experimental control (to check if blocking was appropriate). Statistical Analysis Data were analyzed using SigmaStat Software (Systat Software, San Jose, CA). For the primary analysis, the patient population was initially divided into patients with CV Failure and hemodynamically stable (No CV Failure) and both groups had their GCR expression compared. For the secondary analysis, the patient population was first divided in two groups according to PRISM III scores less than or equal to 7 (the lower 50th percentile) and PRISM III scores greater than or equal to 8 (the upper 50th percentile), and secondarily, the population was divided into two groups according to maximum number of OF (admission to day 7 using Goldstein et al [31] criteria): no OF or one system (the lower 50th percentile) and two or more systems (the higher 50th percentile). For nonnormally distributed variables, Mann-Whitney rank-sum test was performed, Fisher exact test was used for categorical data, t test was used to compare the groups for primary and secondary analysis, and linear regression was used to evaluate the relationship between MFI values (dependent variable) and cortisol levels (independent variable). RESULTS A cohort of 52 critically ill children were studied with 28 subjects in the No CV Failure group and 24 in the CV Failure group.

For nonnormally distributed variables, Mann-Whitney rank-sum test was performed, Fisher exact test was used for categorical data, t test was used to compare the groups for primary and secondary analysis, and linear regression was used to evaluate the relationship between MFI values (dependent variable) and cortisol levels (independent variable). RESULTS A cohort of 52 critically ill children were studied with 28 subjects in the No CV Failure group and 24 in the CV Failure group. The demographic characteristics of the subjects enrolled are shown in Table 1. The CV Failure group had a higher proportion of patients with sepsis, higher cortisol levels, higher PRISM III scores, and a higher number of OFs compared with the no CV Failure group. No other differences were noted. Forty-eight percent of the patients received steroids for various indications, including oncologic therapy, management of cerebral edema after neurosurgery, active immunosuppression, and airway edema (Supplementary Table 1, Supplemental Digital Content 1, http://links.lww.com/PCC/A154). TABLE 1. Demographic Characteristics and Comparison Between Cardiovascular Failure and Non-Cardiovascular Failure Groups GCR expression showed a large range of variability between different individuals (CD4 from 12 to 1,784 MFI, CD8 from 671 to 2,305 MFI, CD14 from 260 to 4,212 MFI, and CD66b from 126 to 8,853 MFI).

The demographic characteristics of the subjects enrolled are shown in Table 1. The CV Failure group had a higher proportion of patients with sepsis, higher cortisol levels, higher PRISM III scores, and a higher number of OFs compared with the no CV Failure group. No other differences were noted. Forty-eight percent of the patients received steroids for various indications, including oncologic therapy, management of cerebral edema after neurosurgery, active immunosuppression, and airway edema (Supplementary Table 1, Supplemental Digital Content 1, http://links.lww.com/PCC/A154). TABLE 1. Demographic Characteristics and Comparison Between Cardiovascular Failure and Non-Cardiovascular Failure Groups GCR expression showed a large range of variability between different individuals (CD4 from 12 to 1,784 MFI, CD8 from 671 to 2,305 MFI, CD14 from 260 to 4,212 MFI, and CD66b from 126 to 8,853 MFI). Primary Analysis When comparing the CV Failure and No CV Failure groups, the subjects with shock had a significantly lower GCR expression both in CD4 lymphocytes (p = 0.036) and CD8 lymphocytes (p = 0.019) (Fig. 1). GCR expression in monocytes and neutrophils showed a trend to be lower in the subjects with CV Failure, but this association did not reach statistical significance.

lure groups, the subjects with shock had a significantly lower GCR expression both in CD4 lymphocytes (p = 0.036) and CD8 lymphocytes (p = 0.019) (Fig. 1). GCR expression in monocytes and neutrophils showed a trend to be lower in the subjects with CV Failure, but this association did not reach statistical significance. Figure 1. Glucocorticoid receptor (GCR) expression in CD4 lymphocytes, CD8 lymphocytes, monocytes (CD14), and neutrophils (CD66b) in patients with cardiovascular (CV) failure and without CV failure. GCR expression was lower in patients with CV failure when compared with patients without CV failure, p < 0.05 by rank-sum test. MFI = mean fluorescence intensity. Secondary Analysis A secondary analysis was performed to further determine if decreased GCR expression is associated with greater illness severity. Accordingly, the study cohort was divided into two groups, representing the lower and upper 50th percentile of PRISM III scores or maximum number of OF. When PRISM III score groups were compared, subjects within the upper 50th percentile of PRISM III values had a significantly lower GCR expression in CD4 lymphocytes (p = 0.008) and CD8 lymphocytes (p = 0.010) when compared with patients within the lower 50th percentile of PRISM III values. Subjects in the upper 50th percentile and subjects in the lower 50th percentile of PRISM values did not show a different level of GCR expression in monocytes (p = 0.098) or neutrophils (p = 0.124) (Fig. 2).

and CD8 lymphocytes (p = 0.010) when compared with patients within the lower 50th percentile of PRISM III values. Subjects in the upper 50th percentile and subjects in the lower 50th percentile of PRISM values did not show a different level of GCR expression in monocytes (p = 0.098) or neutrophils (p = 0.124) (Fig. 2). Figure 2. Glucocorticoid receptor (GCR) expression in CD4 lymphocytes, CD8 lymphocytes, monocytes (CD14), and neutrophils (CD66b) in patients with admission Pediatric Risk of Mortality (PRISM) III score ≤ 7 and admission PRISM III score ≥ 8. *GCR expression was lower in patients with PRISM III score ≥ 8 when compared with patients with PRISM III score ≤ 7, p < 0.05 by rank-sum test. When OF score groups were compared, subjects within the higher 50th percentile of maximum number of OF had a significantly lower GCR expression in CD4 lymphocytes (p = 0.010) and CD8 lymphocytes (p = 0.009) when compared with patients within the lower 50th percentile of maximum OF. Patients with the higher 50th percentile of maximum number of OF showed a trend to have lower GCR expression in monocytes (p = 0.053) than patients in the lower maximum OF number group. Subjects within the upper 50th percentile of maximum number of OF did not show a statistically different GCR expression in neutrophils (p = 0.132) when compared with patients within the lower 50th percentile number of maximum OF (Fig. 3).

sion in monocytes (p = 0.053) than patients in the lower maximum OF number group. Subjects within the upper 50th percentile of maximum number of OF did not show a statistically different GCR expression in neutrophils (p = 0.132) when compared with patients within the lower 50th percentile number of maximum OF (Fig. 3). Figure 3. Glucocorticoid receptor (GCR) expression in CD4 lymphocytes, CD8 lymphocytes, monocytes (CD14), and neutrophils (CD66b) in patients with organ failure (OF) ≤ 1 and OF ≥ 2. *GCR expression lower in patients with OF ≥ 2 when compared with patients with OF ≤ 1, p < 0.05 by rank-sum test. A linear regression analysis was performed using random cortisol levels as the independent variable and GCR expression as the dependent variable. This analysis showed no linear correlation between the variables in all cell types (CD4 Rsqr = 0.06, CD8 Rsqr = 0.06, CD14 Rsqr = 0.03, and CD66b Rsqr = 0.06) (Fig. 4). When a nonparametric statistical test was applied, there was a weak, but statistically significant, inverse correlation between cortisol levels and GCR expression in CD4 (Spearman rank correlation coefficient, –0.277; p = 0.047) and CD8 (Spearman rank correlation coefficient, –0.288; p = 0.038) lymphocytes. Figure 4. Linear regression analysis of cortisol levels and glucocorticoid receptor (GCR) expression correlation. MFI = mean fluorescence intensity.

A linear regression analysis was performed using random cortisol levels as the independent variable and GCR expression as the dependent variable. This analysis showed no linear correlation between the variables in all cell types (CD4 Rsqr = 0.06, CD8 Rsqr = 0.06, CD14 Rsqr = 0.03, and CD66b Rsqr = 0.06) (Fig. 4). When a nonparametric statistical test was applied, there was a weak, but statistically significant, inverse correlation between cortisol levels and GCR expression in CD4 (Spearman rank correlation coefficient, –0.277; p = 0.047) and CD8 (Spearman rank correlation coefficient, –0.288; p = 0.038) lymphocytes. Figure 4. Linear regression analysis of cortisol levels and glucocorticoid receptor (GCR) expression correlation. MFI = mean fluorescence intensity. DISCUSSION Our primary observation is that GCR expression is decreased in CD4 and CD8 lymphocytes of critically ill children with cardiovascular dysfunction in the first 24 hours of admission. Our secondary observation is that GCR expression is decreased in patients with greater severity of illness measured by PRISM III scores as well as in patients with greater OF burden. Both results are consistent with our gene expression studies that showed repression of genes associated with the GCR signaling pathway in the cohort of children with septic shock who had the worst outcome (27). These findings suggest the need to evaluate PICU patients at a pharmacogenomic level for GCR expression in lymphocytes as a risk factor associated with worse severity of illness and the presence of CV Failure.

ted with the GCR signaling pathway in the cohort of children with septic shock who had the worst outcome (27). These findings suggest the need to evaluate PICU patients at a pharmacogenomic level for GCR expression in lymphocytes as a risk factor associated with worse severity of illness and the presence of CV Failure. In general, within the same patient, GCR expression was decreased in all cell types (Supplementary Table 1, Supplemental Digital Content 1, http://links.lww.com/PCC/A154). There was a significant overlap between the CV failure group and the higher 50th percentile of PRISM III and OF levels (Fig. 5), which could account for the similar statistical findings between the three groups. Figure 5. Venn diagram representing the patient distribution in cardiovascular (CV) failure group, organ failure (OF) > 1 group, and Pediatric Risk of Mortality (PRISM) III > 7 group.

There was a significant overlap between the CV failure group and the higher 50th percentile of PRISM III and OF levels (Fig. 5), which could account for the similar statistical findings between the three groups. Figure 5. Venn diagram representing the patient distribution in cardiovascular (CV) failure group, organ failure (OF) > 1 group, and Pediatric Risk of Mortality (PRISM) III > 7 group. It is known that GCR-α, when not bound to a ligand, is located in the cytoplasm and that binding of cortisol to GCR-α leads to receptor translocation to the nucleus to affect gene expression (32, 33). GCR-β, on the other hand, is predominantly found in the nucleus, and it inhibits the GCR-α-mediated gene activation (32). Other splice variants, such as GCRγ (GCRP), GCRτ1, and GCRτ2, have been reported in association with glucocorticoid resistance (32–34). The antibody and methodology used in our study allowed us to measure total GCR levels without distinction of subtypes and separation for location of expression, which constitutes a limitation of the current study. It is not possible to speculate if the lower total GCR levels presented here correlate with lower nuclear GCR-α levels and decreased anti-inflammatory effect, but GCR-β has been reportedly expressed in lower concentrations than GCR-α (33).

for location of expression, which constitutes a limitation of the current study. It is not possible to speculate if the lower total GCR levels presented here correlate with lower nuclear GCR-α levels and decreased anti-inflammatory effect, but GCR-β has been reportedly expressed in lower concentrations than GCR-α (33). Previous studies have reported total and cytoplasmic decrease in GCR levels in peripheral blood mononuclear cells from critically ill children (35), decreased GCR-receptor binding in cytosolic extracts of peripheral WBCs from critically ill adult patients (36), and down-regulation of GCR-α in adults with sepsis or septic shock, findings that are supportive of our results. van den Akker et al (37) also showed that GCR-α and GCRP messenger RNA expression was transiently decreased in neutrophils of children with septic shock compared with the same patients 3 months after the episode. In agreement with the fact that a lower expression of GCR was associated with shock and worse severity of illness, in vitro studies have previously shown that an increased GCR expression is protective for sepsis (38) and sepsis-induced acute lung injury (39). van den Akker et al (37) reported different findings with GCR levels not showing correlation with PRISM III scores or the presence of shock. The smaller number of patients enrolled in their study may account for a difference in power between both studies that could explain the difference between our results and their findings.

Akker et al (37) reported different findings with GCR levels not showing correlation with PRISM III scores or the presence of shock. The smaller number of patients enrolled in their study may account for a difference in power between both studies that could explain the difference between our results and their findings. Our findings suggest that patients who have lower GCR expression have higher severity of illness scores. Both in vitro studies (40, 41) and in vivo studies (42–44) suggest the existence of a phenomenon characterized by peripheral resistance mechanism of cortisol insufficiency that could be happening in the lower GCR expression individuals. In a healthy individual, cortisol is secreted in a diurnal pattern under the influence of corticotropin with a circadian rhythm throughout the day and 90% of the circulating cortisol is bound to corticosteroid-binding globulin with only less than 10% in the bioavailable free form that can be measured in saliva and urine (45, 46). This diurnal variation is lost in severe illness, and the percentage of circulating free cortisol increases due to a decrease in cortisol-binding globulin levels, and inflammatory cytokines can change cortisol metabolism, increasing cortisol levels (45). All these issues illustrate the challenges associated with the interpretation of a single random total cortisol level and may account for the finding that cortisol levels, albeit collected at the same time as GCR samples, were not correlated to them.

ines can change cortisol metabolism, increasing cortisol levels (45). All these issues illustrate the challenges associated with the interpretation of a single random total cortisol level and may account for the finding that cortisol levels, albeit collected at the same time as GCR samples, were not correlated to them. The lack of a linear correlation between random total cortisol level and GCR expression illustrates that it would not be possible to infer if the individual GCR expression is increased or decreased based solely on circulating random cortisol level. These findings emphasize the importance of measuring GCR expression. Similar findings were previously described by Imamura et al (47) in cord blood from term newborns where GCR-α expression did not correlate with cortisol level or GCR-β expression and by Indyk et al (35) who showed that nuclear GCR levels reflecting ongoing cortisol activity did not correlate to total, free, salivary, and urinary-free cortisol levels. There were a large amount of patients with a cortisol level lower than 10 μg/dL (Fig. 4), the majority of these patients were postoperative patients, but some CV failure patients also showed a low cortisol level. These findings are compatible with data reported by Menon et al (48) (Supplementary Table 1, Supplemental Digital Content 1, http://links.lww.com/PCC/A154).

isol level lower than 10 μg/dL (Fig. 4), the majority of these patients were postoperative patients, but some CV failure patients also showed a low cortisol level. These findings are compatible with data reported by Menon et al (48) (Supplementary Table 1, Supplemental Digital Content 1, http://links.lww.com/PCC/A154). Our findings do not suggest that the evaluation of the hypothalamic-pituitary-adrenal axis should be regarded as not useful. Rather, we suggest that the mechanisms of action of glucocorticoids in critical illness are complex and the evaluation of GCR expression could add information to the assessment of the presence of peripheral resistance to corticosteroids. Future studies could focus on further investigating if a population with a decreased GCR expression has a good response to cortisol therapy or if they may not benefit from its use but could still suffer from side effects of hyperglycemia, myopathy, and immune suppression (49–51). For future protocols studying the effectiveness and indications of the use of steroids in critically ill patients, our study suggests that solely analyzing cortisol levels and ACTH stimulation response may not completely predict response to steroid therapy and that the individual GCR expression has to be taken into account when designing these studies. Further relevant investigation could also evaluate if there are drugs previously studied in vitro (41, 52) that could modulate the GCR expression in vivo and if that timely modulation could impact on the inflammation severity, morbidity, and mortality of study subjects.

taken into account when designing these studies. Further relevant investigation could also evaluate if there are drugs previously studied in vitro (41, 52) that could modulate the GCR expression in vivo and if that timely modulation could impact on the inflammation severity, morbidity, and mortality of study subjects. It would be most optimal to analyze tissue/end-organ GCR expression as this is probably where its level will reliably be clinically relevant, but we were restricted to analyzing peripheral blood cells’ GCR expression since it was easily and less invasively available. It is not known if GCR expression in WBCs correlates with GCR expression in tissues and end organs. Our protocol did not allow us to separate nuclear and cytoplasmatic GCR expression levels; instead, we measured total GCR expression. We could not separately evaluate the different GCR isoforms because the antibody used did not allow us to do that. We measured random cortisol but not salivary, urinary, free cortisol, or cortisol-binding globulin levels. The current laboratory procedure we followed is a technology- and hardware-dependent, time-consuming process that would be difficult to be done at the bedside. Lastly, we divided our patient sample by their median PRISM III scores and OF values for convenience to be able to have two groups with the same number of individuals to run the statistical analysis.

lowed is a technology- and hardware-dependent, time-consuming process that would be difficult to be done at the bedside. Lastly, we divided our patient sample by their median PRISM III scores and OF values for convenience to be able to have two groups with the same number of individuals to run the statistical analysis. CONCLUSIONS This study suggests that patients with shock and increased illness severity have lower GCR expression in both CD4 and CD8 lymphocytes, consistent with gene expression studies (27–29). We speculate that this difference may have conceptual implications, identifying a subset population of critically ill children that may present a peripheral resistance form of critical illness–related cortisol insufficiency. GCR expression does not seem to correlate with random serum cortisol concentrations. Future studies could focus on studying GCR expression variability and isoform distribution in the pediatric critically ill population as well as on different strategies to optimize glucocorticoid response. ACKNOWLEDGMENTS Flow cytometry analysis was performed at Shriners Hospitals for Children, Cincinnati, and supported by a grant from the Shriners of North America (SSF 84070). Supplementary Material *See also p. 489. This study was performed at the Cincinnati Children’s Hospital Medical Center and Research Foundation, University of Cincinnati College of Medicine, Cincinnati, OH.

ACKNOWLEDGMENTS Flow cytometry analysis was performed at Shriners Hospitals for Children, Cincinnati, and supported by a grant from the Shriners of North America (SSF 84070). Supplementary Material *See also p. 489. This study was performed at the Cincinnati Children’s Hospital Medical Center and Research Foundation, University of Cincinnati College of Medicine, Cincinnati, OH. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (http://journals.lww.com/pccmjournal). Supported, in part, by the Nurturing Children’s Development Program, Procter & Gamble, and National Institute of Health Grants (R01GM064619 and R01GM099773). Dr. Shibata received grant support from Procter & Gamble (Nurturing Children’s Development Program), is employed by the Hospital Israelita Albert Einstein, and received support for article research from the National Institutes of Health (Dr. Wong's grants). Dr. Shibata and her institution received grant support from Procter & Gamble (grant for investigation of glucocorticoid receptor expression in severe sepsis and septic shock patients). Her institution received grant support from the NIH (NIH grants R01GM64619 and R01GM099773 for Dr. Wong’s laboratory). Dr. Troster is employed by the Hospital Israelita Albert Einstein. Dr. Wong has patents with the U.S. Patent Office and received support for article research from the NIH. His institution received grant support from the NIH.