Browse the corpus

Key messages What is the key question? A number of international centres have reported a clinically significant response to azithromycin therapy in open studies of lung transplant recipients with bronchiolitis obliterans syndrome (BOS) but there are also negative studies, where azithromycin therapy was not associated with gain in lung function. What is the bottom line? With no randomised placebo controlled studies performed or published there was a clear unmet need, met by this trial, which showed on average that azithromycin was superior to placebo treatment in our study population. Why read on? This study outlines strengthened evidence for the clinical practice of initiating azithromycin therapy for patients who develop BOS post lung transplantation. Introduction Lung transplantation can be the only life-sustaining intervention for end-stage lung disease.1 Good functional outcomes have been shown, with improved quality of life.2 Long-term survival remains limited by the development of the bronchiolitis obliterans syndrome (BOS) however.3 The histological lesion of BOS is obliterative bronchiolitis. This is characterised by epithelial alloimmune and non-alloimmune injury.4 5 Deterioration in allograft function is characterised by the development of progressive, small airway narrowing, fixed airflow limitation, progressive dyspnoea and premature death.6 International data shows in excess of 50% of patients surviving to 5 years after transplantation develop BOS,6 limiting 10-year survival to around 30%.7

in allograft function is characterised by the development of progressive, small airway narrowing, fixed airflow limitation, progressive dyspnoea and premature death.6 International data shows in excess of 50% of patients surviving to 5 years after transplantation develop BOS,6 limiting 10-year survival to around 30%.7 Therapeutic approaches have ranged from switching immunosuppression6 through to initiating cytolytic therapy. Such approaches, including the use of total lymphoid irradiation8 have, at best, reduced the rate of decline in graft function in BOS, with significant iatrogenic potential.8 In contrast, retrospective studies of macrolides, in particular low dose azithromycin, have indicated that up to 30% patients with BOS may gain lung function. A number of international centres have reported a clinical response to azithromycin therapy in around a third of patients,9–13 with better life expectancy.13 There have been negative studies, however, with no gain in lung function.14–16 The need for randomised controlled trial data has been highlighted,17 but these have not been performed. We tested the hypothesis that azithromycin therapy is superior to placebo in patients with BOS in a randomised double blind placebo controlled study. Some results have been presented as abstracts.18 19

In contrast, retrospective studies of macrolides, in particular low dose azithromycin, have indicated that up to 30% patients with BOS may gain lung function. A number of international centres have reported a clinical response to azithromycin therapy in around a third of patients,9–13 with better life expectancy.13 There have been negative studies, however, with no gain in lung function.14–16 The need for randomised controlled trial data has been highlighted,17 but these have not been performed. We tested the hypothesis that azithromycin therapy is superior to placebo in patients with BOS in a randomised double blind placebo controlled study. Some results have been presented as abstracts.18 19 Methods Study design Randomisation and masking This was a single-centre randomised double-blind placebo-controlled parallel group study comparing azithromycin (250 mg on alternate days) with placebo over 12 weeks in lung transplant recipients with BOS, with study drug taken in addition to existing medication. Patients were randomly assigned to a treatment arm in a 1:1 ratio using random permuted blocks within strata. Study medication was provided by Bilcare, (Bilcare GCS Europe, Powys, UK) a commercial clinical trial supplier, independent of the manufacturers of azithromycin. Patient population Patients were recruited between November 2006 and December 2010 from the Freeman Hospital.

Methods Study design Randomisation and masking This was a single-centre randomised double-blind placebo-controlled parallel group study comparing azithromycin (250 mg on alternate days) with placebo over 12 weeks in lung transplant recipients with BOS, with study drug taken in addition to existing medication. Patients were randomly assigned to a treatment arm in a 1:1 ratio using random permuted blocks within strata. Study medication was provided by Bilcare, (Bilcare GCS Europe, Powys, UK) a commercial clinical trial supplier, independent of the manufacturers of azithromycin. Patient population Patients were recruited between November 2006 and December 2010 from the Freeman Hospital. Withdrawal of patients from study Patients who had a rapid and severe deterioration in lung function were withdrawn from the study (for patient details see online supplementary appendix 2). This was defined a priori as a sustained 500 mL fall in FEV1 from baseline, before the full 12-week course of study treatment, thought to be due to BOS. Patients could also be withdrawn based on the clinical judgement of the responsible clinician. Following withdrawal patients were treated according to the usual centre and international practice, which included the use of open-label azithromycin. The randomised treatment allocation of withdrawn patients remained concealed. Assessments Spirometry FEV1 and FVC were measured at baseline, week 4, week 8 and week 12 in the Freeman Hospital.

Withdrawal of patients from study Patients who had a rapid and severe deterioration in lung function were withdrawn from the study (for patient details see online supplementary appendix 2). This was defined a priori as a sustained 500 mL fall in FEV1 from baseline, before the full 12-week course of study treatment, thought to be due to BOS. Patients could also be withdrawn based on the clinical judgement of the responsible clinician. Following withdrawal patients were treated according to the usual centre and international practice, which included the use of open-label azithromycin. The randomised treatment allocation of withdrawn patients remained concealed. Assessments Spirometry FEV1 and FVC were measured at baseline, week 4, week 8 and week 12 in the Freeman Hospital. Bronchoscopy, bronchoalveolar lavage and transbronchial biopsy Patients underwent bronchoscopy at baseline (prerandomisation) and at final visit (week 12) as previously described.20 Transbronchial biopsies were taken at each allograft bronchoscopy, fixed in 10% formalin, embedded in paraffin, and then stained with haematoxylin and eosin (H&E) to assess acute vascular and airway inflammation according to standard ISHLT criteria by a pathologist.21 Outcome measures The primary outcome measure was change in FEV1 from baseline to 12 weeks. Secondary outcome measures reported here are change in FVC from baseline and change in bronchoalveolar lavage (BAL) neutrophils.

Bronchoscopy, bronchoalveolar lavage and transbronchial biopsy Patients underwent bronchoscopy at baseline (prerandomisation) and at final visit (week 12) as previously described.20 Transbronchial biopsies were taken at each allograft bronchoscopy, fixed in 10% formalin, embedded in paraffin, and then stained with haematoxylin and eosin (H&E) to assess acute vascular and airway inflammation according to standard ISHLT criteria by a pathologist.21 Outcome measures The primary outcome measure was change in FEV1 from baseline to 12 weeks. Secondary outcome measures reported here are change in FVC from baseline and change in bronchoalveolar lavage (BAL) neutrophils. Study oversight Newcastle University Clinical Trials Unit monitored the study. An independent data monitoring committee was established to assess accumulating recruitment, safety and efficacy data and to oversee the trial conduct (Statistician (chair) and two consultant respiratory physicians). Sample size Estimates of SDs of differences in FEV1 from baseline to 12 weeks in patients with lung transplant with BOS were based on the Freeman Hospital data.41 A sample size of 64 patients, 32 patients per randomised group, allowed for 10% data attrition. A recruitment period of 30 months was estimated to be adequate to recruit 64 patients.

SDs of differences in FEV1 from baseline to 12 weeks in patients with lung transplant with BOS were based on the Freeman Hospital data.41 A sample size of 64 patients, 32 patients per randomised group, allowed for 10% data attrition. A recruitment period of 30 months was estimated to be adequate to recruit 64 patients. Statistical methods The mean difference in FEV1 between treatment groups was estimated using a multilevel random effects model assuming a normal error structure within and between patients.22 The models were fitted in MLwiN software (V.2.28).23 Random effects models allow appropriate estimation of the treatment effect (and associated SE) taking into account varying numbers of measurements within patients and varying time between measurements. Models were also adjusted for baseline FEV1 and the two randomisation stratification variables.24 Model assumptions were checked and analyses omitting possible outliers or influential observations were performed. The secondary outcome measure, FVC, was analysed in the same way. BAL neutrophil counts and their change from baseline to 12 weeks were summarised and presented as median and IQR. The intention-to-treat population (ITT) was defined as all randomised patients with BOS who received at least one dose of study drug. The per protocol population was defined as all randomised patients with BOS who followed the protocol and completed 12 weeks of study drug. Completers included all randomised patients with BOS who completed 12 weeks of study drug.

defined as all randomised patients with BOS who received at least one dose of study drug. The per protocol population was defined as all randomised patients with BOS who followed the protocol and completed 12 weeks of study drug. Completers included all randomised patients with BOS who completed 12 weeks of study drug. An ‘as treated’ analysis was also performed; this was a ‘post hoc’ analysis which had not been described in the statistical analysis plan and as such should be interpreted cautiously. Patients who withdrew or were withdrawn from study drug were treated with open-label azithromycin. The ‘as treated’ analysis provided an estimate of the treatment effect allowing for treatment to change over time. In this way a patient's measurement contributed to the treatment effect based on the treatment they were receiving at the time the measurement was taken and not the treatment as randomised (for extended details see online supplementary appendix 1). Results Study patients The CONSORT flow chart25 is presented in figure 1. Patient withdrawals are detailed in the online supplementary appendix 2 (for extended details see online supplementary appendix 1). Figure 1 CONSORT flow chart25 summarising the progress of patients through the trial. BOS, bronchiolitis obliterans syndrome; ITT, intention-to-treat. All 46 patients in the ITT analysis set had baseline and final visit FEV1 measured. Across all visits there were 177 FEV1 measurements: 2 patients had 5 FEV1 measurements, 36 patients had 4, 7 patients had 3, and 1 patient had 2 measurements.

Figure 1 CONSORT flow chart25 summarising the progress of patients through the trial. BOS, bronchiolitis obliterans syndrome; ITT, intention-to-treat. All 46 patients in the ITT analysis set had baseline and final visit FEV1 measured. Across all visits there were 177 FEV1 measurements: 2 patients had 5 FEV1 measurements, 36 patients had 4, 7 patients had 3, and 1 patient had 2 measurements. There were 33 patients in the Completers analysis set (16 azithromycin, 17 placebo). For the Completers analysis there were 124 FEV1 measures; 26 patients with 4 measurements, 6 patients with 3 and 1 patient with 2 measurements. The ‘as treated’ analysis used data on all 46 ITT patients and all their 177 FEV1 measurements. For the five placebo patients who were withdrawn and placed on open label azithromycin, their measurements post withdrawal (nine measurements in total across the five patients) contributed to the overall azithromycin treatment mean and not the placebo treatment mean.

46 ITT patients and all their 177 FEV1 measurements. For the five placebo patients who were withdrawn and placed on open label azithromycin, their measurements post withdrawal (nine measurements in total across the five patients) contributed to the overall azithromycin treatment mean and not the placebo treatment mean. Baseline characteristics of the study patients, for the ITT and Completer analysis sets, are summarised in table 1. All patients initially received standard immunosuppressant comprising ciclosporin, prednisolone and azathioprine. Patients with more than one episode of vascular rejection requiring augmented steroids in the 1st year post transplant and women with problematic hirsute were switched to tacrolimus treatment. In the ITT population 38 of 46 had switched to tacrolimus by enrolment; 18 in the azithromycin group and 20 in the placebo group. All switches were in the 1st year post transplant, well before study enrolment. All patients received proton pump inhibitors and statin therapy, throughout the study. There were no relevant differences in background immunosuppressant or other therapies between the two groups. Results of lavage microbiology did not lead to any change in baseline therapy and no patient enrolled was regarded as having a new infection. Table 1 Baseline characteristics for the intention-to-treat (ITT, n=46) and Completer (Comp, n=33) populations, by treatment allocation group

Baseline characteristics of the study patients, for the ITT and Completer analysis sets, are summarised in table 1. All patients initially received standard immunosuppressant comprising ciclosporin, prednisolone and azathioprine. Patients with more than one episode of vascular rejection requiring augmented steroids in the 1st year post transplant and women with problematic hirsute were switched to tacrolimus treatment. In the ITT population 38 of 46 had switched to tacrolimus by enrolment; 18 in the azithromycin group and 20 in the placebo group. All switches were in the 1st year post transplant, well before study enrolment. All patients received proton pump inhibitors and statin therapy, throughout the study. There were no relevant differences in background immunosuppressant or other therapies between the two groups. Results of lavage microbiology did not lead to any change in baseline therapy and no patient enrolled was regarded as having a new infection. Table 1 Baseline characteristics for the intention-to-treat (ITT, n=46) and Completer (Comp, n=33) populations, by treatment allocation group Baseline characteristic ITT azithromycin n=23 ITT placebo n=23 Comp azithromycin n=16 Comp placebo n=17 Sex Female 12 (52%) 8 (35%) 9 (56%) 7 (41%) Male 11 (48%) 15 (65%) 7 (44%) 10 (59%) Age (years), median (IQR) 51.0 (35–56) 51.0 (44–59) 53.5 (47.0–57.5) 54.0 (45.0–62.0) Pretransplant (Tx) disease Cystic fibrosis 7 (30%) 6 (26%) 2 (13%) 4 (24%) Emphysema 11 (48%) 8 (35%) 9 (56%) 6 (35%) Fibrosing alveolitis 2 (9%) 4 (17%) 2 (13%) 3 (18%) Other* 3 (13%) 5 (22%) 3 (19%) 4 (24%) Tx procedure Double lung 14 (61%) 13 (57%) 7 (44%) 9 (53%) Single lung 9 (39%) 9 (39%) 9 (56%) 7 (41%) Heart lung 0 1 (4%) 0 1 (6%) Years between Tx and BOS Median (IQR) 3.7 (1.3–7.4) 2.2 (1.3–5.0)† 4.2 (2.6–7.8) 2.0 (1.3–4.0) BOS stage 1 13 (57%) 17 (74%) 10 (63%) 13 (77%) 2 8 (35%) 4 (17%) 6 (38%) 3 (18%) 3 2 (9%) 2 (9%) 0 1 (6%) FEV1 (litres), median (IQR) 1.5 (1.2–2.4) 1.7 (1.5–2.5) 1.6 (1.2–2.4) 1.7 (1.5–2.2) FVC (L), median (IQR) 3.0 (2.3–3.6) 2.9 (2.2–3.6) 2.7 (2.1–3.6) 2.9 (2.1–3.5) TBB A and B scores‡ Missing 2 (9%) 4 (17%) 2 (13%) 3 (18%) Ax 5 (22%) 8 (35%) 3 (19%) 7 (41%) A0 12 (52%) 7 (30%) 7 (44%) 4 (24%) A1 4 (17%) 3 (13%) 4 (25%) 2 (12%) A2 0 1 (4%) 0 1 (6%) Missing 2 (9%) 4 (17%) 2 (13%) 3 (18%) Bx 5 (22%) 7 (30%) 5 (31%) 4 (24%) B0 5 (22%) 6 (26%) 2 (13%) 5 (29%) B1R 9 (39%) 6 (26%) 7 (44%) 5 (29%) B2R 2 (9%) 0 0 0 BAL microbiology Missing 2 (9%) 2 (9%) 2 (13%) 1 (6%) NPI 11 (48%) 14 (61%) 8 (50%) 11 (65%) ‘Any’ organism 10 (43%) 7 (30%) 6 (38%) 5 (29%) ‘Any’ includes: Pa 5 4 3 3 Asp Fum 2 0 1 0 Ca 4 2 2 1 Other 2 1 1 1 *Other Pre Tx disease: Obliterative Bronchiolitis, Sarcoid, Congenital heart disease, Histiocytosis X, Silicosis.

L microbiology Missing 2 (9%) 2 (9%) 2 (13%) 1 (6%) NPI 11 (48%) 14 (61%) 8 (50%) 11 (65%) ‘Any’ organism 10 (43%) 7 (30%) 6 (38%) 5 (29%) ‘Any’ includes: Pa 5 4 3 3 Asp Fum 2 0 1 0 Ca 4 2 2 1 Other 2 1 1 1 *Other Pre Tx disease: Obliterative Bronchiolitis, Sarcoid, Congenital heart disease, Histiocytosis X, Silicosis. †One patient randomised to the placebo arm >10 years post transplant (at 11.9 years). ‡ISHLT grades (ref 20) BAL Microbiology=Clinical microbiology. Other=Proteus mirabilis, Stenotrophomonas Maltophilia, Ralstonia Picketti, Candida species. Percentages for patients growing individual organisms are not given since some patients grew more than one organism. Asp Fum, Aspergillus fumigatus; BAL, bronchoalveolar lavage; BOS, bronchiolitis obliterans syndrome; Ca, Candida albicans; NPI, no pathogens identified; Pa, Pseudomanas aeruginosa; TBB, transbronchial biopsy.

‡ISHLT grades (ref 20) BAL Microbiology=Clinical microbiology. Other=Proteus mirabilis, Stenotrophomonas Maltophilia, Ralstonia Picketti, Candida species. Percentages for patients growing individual organisms are not given since some patients grew more than one organism. Asp Fum, Aspergillus fumigatus; BAL, bronchoalveolar lavage; BOS, bronchiolitis obliterans syndrome; Ca, Candida albicans; NPI, no pathogens identified; Pa, Pseudomanas aeruginosa; TBB, transbronchial biopsy. Analysis of FEV1 data Figure 2A summarises FEV1 measurements as a two-panel spaghetti plot of FEV1 over time in the study. Figure 2B summarises FEV1 measurements in patients who were randomised to the placebo arm and withdrew or were withdrawn and then received open-label azithromycin. Results are summarised in table 2. For the ITT analysis (n=46, 177 observations), the estimated mean difference in FEV1 between treatment groups (azithromycin minus placebo) was 0.035 L, (on average higher in the azithromycin group) with a 95% CI for the mean difference of −0.112 L to 0.182 L (p=0.6). Nine out of 23 (39%) ITT patients in the azithromycin group had ≥10% gain in FEV1 from baseline. No patients in the placebo arm had ≥10% gain in FEV1 from baseline while on placebo (p<0.002, Fisher's exact test). Table 2 Mean difference in FEV1 between treatment groups for the intention-to-treat (ITT, n=46), ‘as treated’ (n=46) and Completer (n=33) populations

Analysis of FEV1 data Figure 2A summarises FEV1 measurements as a two-panel spaghetti plot of FEV1 over time in the study. Figure 2B summarises FEV1 measurements in patients who were randomised to the placebo arm and withdrew or were withdrawn and then received open-label azithromycin. Results are summarised in table 2. For the ITT analysis (n=46, 177 observations), the estimated mean difference in FEV1 between treatment groups (azithromycin minus placebo) was 0.035 L, (on average higher in the azithromycin group) with a 95% CI for the mean difference of −0.112 L to 0.182 L (p=0.6). Nine out of 23 (39%) ITT patients in the azithromycin group had ≥10% gain in FEV1 from baseline. No patients in the placebo arm had ≥10% gain in FEV1 from baseline while on placebo (p<0.002, Fisher's exact test). Table 2 Mean difference in FEV1 between treatment groups for the intention-to-treat (ITT, n=46), ‘as treated’ (n=46) and Completer (n=33) populations Outcome FEV1 (L) Mean difference in FEV1 (azithromycin minus placebo) 95% CI for population mean difference p Value ITT analysis 46 patients, 177 measurements Mean difference in FEV1 between treatment arms, adjusted for baseline FEV1, randomisation stratification variables (disease and transplant) and time since randomisation 0.035 −0.112 to 0.182 0.6 ‘As treated’ analysis 46 patients, 177 measurements Mean difference in FEV1 between treatment arms, adjusted for baseline FEV1, randomisation stratification variables (disease and transplant) and time since randomisation 0.306 0.181 to 0.431 <0.001 Completers analysis 33 patients, 124 measurements Mean difference in FEV1 between treatment arms, adjusted for baseline FEV1, randomisation stratification variables (disease and transplant) and time since randomisation 0.278 0.170 to 0.386 <0.001 Figure 2 (A) FEV1 measurements as a two-panel spaghetti plot of FEV1 over time in the study. The thickened lines denote FEV1 from the time a patient withdrew or was withdrawn from study medication. FEV1.0 versus days (placebo group; solid squares) and FEV1.0 versus days (azithromycin group; solid circles). (B) Descriptive plot of FEV1 data for patients treated with placebo who withdrew or were withdrawn from study medication. Symbols, different for each patient (key) denote FEV1 measurements. ID, anonymised patient identifier. ‘P’ indicates where a patient was being treated with placebo at the time FEV1 was measured. ‘A’ denotes where a patient was being treated with azithromycin at the time FEV1 was measured, after withdrawal from study. Patient 48 was withdrawn from study medication following stomach pains and did not receive azithromycin. Patient 51 withdrew consent and was treated with open-label azithromycin. The remaining four patients had ‘rapid fall’ in FEV1 and were withdrawn and treated with open label azithromycin.

fter withdrawal from study. Patient 48 was withdrawn from study medication following stomach pains and did not receive azithromycin. Patient 51 withdrew consent and was treated with open-label azithromycin. The remaining four patients had ‘rapid fall’ in FEV1 and were withdrawn and treated with open label azithromycin. For the ‘as treated’ analysis (n=46, 177 observations) the estimated mean difference in FEV1 between treatment groups (azithromycin minus placebo) was 0.306 L, with 95% CI for the mean difference: 0.181 L to 0.431 L (p=<0.001). For study Completers the estimated mean difference in FEV1 between treatment groups (azithromycin minus placebo) was 0.278 mL, with 95% CI for the mean difference: 0.170 L to 0.386 L (p=<0.001). Analysis of FVC data The results are summarised in table 3. For the ITT population, (n=46, 177 observations), the estimated mean difference in FVC between treatment groups (azithromycin minus placebo) was 0.099 L, with 95% CI for the mean difference: −0.026 L to 0.224 L (p=0.1). Table 3 Mean difference in FVC between treatment groups for the intention-to-treat (ITT, n=46), ‘as treated’ (n=46) and Completer (n=33) populations

Analysis of FVC data The results are summarised in table 3. For the ITT population, (n=46, 177 observations), the estimated mean difference in FVC between treatment groups (azithromycin minus placebo) was 0.099 L, with 95% CI for the mean difference: −0.026 L to 0.224 L (p=0.1). Table 3 Mean difference in FVC between treatment groups for the intention-to-treat (ITT, n=46), ‘as treated’ (n=46) and Completer (n=33) populations Outcome FVC (L) Mean difference in FVC (azithromycin minus placebo) 95% CI for population mean difference p Value ITT analysis 46 patients, 177 measurements Mean difference in FVC between treatment arms, adjusted for baseline FVC, randomisation stratification variables (disease and transplant) and time since randomisation 0.099 −0.026 to 0.224 0.1 ‘As treated’ analysis 46 patients, 177 measurements Mean difference in FVC between treatment arms, adjusted for baseline FVC, randomisation stratification variables (disease and transplant) and time since randomisation 0.272 0.158 to 0.386 <0.001 Completers analysis 33 patients, 124 measurements Mean difference in FVC between treatment arms, adjusted for baseline FVC, randomisation stratification variables (disease and transplant) and time since randomisation 0.248 0.115 to 0.381 <0.001 For the ‘as treated’ analysis (n=46, 177 observations) the estimated mean difference in FVC between treatment groups (azithromycin minus placebo) was 0.272 L, with 95% CI for the mean difference: 0.158 L to 0.386 L (p=<0.001).

n variables (disease and transplant) and time since randomisation 0.248 0.115 to 0.381 <0.001 For the ‘as treated’ analysis (n=46, 177 observations) the estimated mean difference in FVC between treatment groups (azithromycin minus placebo) was 0.272 L, with 95% CI for the mean difference: 0.158 L to 0.386 L (p=<0.001). For study Completers, the estimated mean difference in FVC between treatment arms (azithromycin minus placebo) was 0.248 L, with 95% CI for the mean difference: 0.115 L to 0.381 L (p=<0.001). BAL neutrophil data BAL data was not available from all patients due to a clinical decision that the sample was either not possible or prudent because of the clinical status of the patients during the bronchoscopy. At baseline BAL differential data were available for 39/46 (85%) of the ITT analysis set. The median per cent neutrophils in BAL was 25.8% (IQR 3.4–72.0%). BAL neutrophil data were available at baseline and final visit for 28/46 patients in the ITT analysis set (13/23 azithromycin, 15/23 placebo) and 25/33 in the Completers analysis set (12/16 azithromycin, 13/17 placebo). Summary statistics for baseline, final and change in BAL neutrophil percentage are given in table 4. There was no evidence of systematic changes in BAL neutrophil percentage associated with either azithromycin or placebo treatment for either the ITT (figure 3A) or study Completer populations (figure 3B). Table 4 Per cent neutrophils in bronchoalveolar lavage at baseline and final visit (week 12) for the intention-to-treat (ITT) (n=28/46) and Completer (n=25/33) populations, by treatment allocation

BAL neutrophil data were available at baseline and final visit for 28/46 patients in the ITT analysis set (13/23 azithromycin, 15/23 placebo) and 25/33 in the Completers analysis set (12/16 azithromycin, 13/17 placebo). Summary statistics for baseline, final and change in BAL neutrophil percentage are given in table 4. There was no evidence of systematic changes in BAL neutrophil percentage associated with either azithromycin or placebo treatment for either the ITT (figure 3A) or study Completer populations (figure 3B). Table 4 Per cent neutrophils in bronchoalveolar lavage at baseline and final visit (week 12) for the intention-to-treat (ITT) (n=28/46) and Completer (n=25/33) populations, by treatment allocation n Baseline Median (IQR) Final visit Median (IQR) Change from baseline Median (IQR) ITT azithromycin 13/23 16.6 (4.2 to 68.8) 32.0 (10.0 to 69.5) 9.8 (−10.4 to 17.7) ITT placebo 15/23 14.8 (2.4 to 56.0) 19.8 (2.0 to 52.2) −0.5 (−7.8 to 5.0) Completers azithromycin 12/16 16.1 (3.7 to 61.5) 31.5 (7.5 to 73.3)  11.9 (−7.7 to 18.9) Completers placebo 13/17 9.2 (2.0 to 52.5) 19.8 (1.5 to 46.9) −0.5 (−7.5 to 4.9) Figure 3 (A) The change in per cent neutrophils in bronchoalveolar lavage (BAL) from baseline to week 12 for the intention-to-treat (ITT) population, by treatment allocation group (n=28/46). Median change denoted by horizontal line. (B) The change in per cent neutrophils in BAL from baseline to week 12 for the Completer population, by treatment allocation group (n=25/33). Median change denoted by horizontal line. (C) The change in per cent neutrophils in baseline to week 12 for patients treated with azithromycin who had <10% gain in FEV1 (solid circles) and in patients treated with azithromycin who had a >10% gain in FEV1 (solid squares).

lation, by treatment allocation group (n=25/33). Median change denoted by horizontal line. (C) The change in per cent neutrophils in baseline to week 12 for patients treated with azithromycin who had <10% gain in FEV1 (solid circles) and in patients treated with azithromycin who had a >10% gain in FEV1 (solid squares). Transbronchial biopsy data Where paired data were available approximately half of the biopsies were graded as ‘Bx’ (ungradeable) for the B scores and a third were ‘Ax’ (ungradeable) for the A scores. Further analysis of the potential effect of azithromycin on biopsy scores was therefore not carried out. The data is summarised in online supplementary appendix 3. Safety Seven adverse events led to hospital admission or the prolonging of existing hospitalisation. They were therefore classified as serious, but not related to study medication. There were no other major safety issues to report with the trial. Discussion To our knowledge this is the first randomised controlled study of azithromycin therapy in BOS. Our trial showed that azithromycin improves FEV1 and FVC in a significant proportion of patients. As in previous open studies not all patients with BOS responded to azithromycin.

Safety Seven adverse events led to hospital admission or the prolonging of existing hospitalisation. They were therefore classified as serious, but not related to study medication. There were no other major safety issues to report with the trial. Discussion To our knowledge this is the first randomised controlled study of azithromycin therapy in BOS. Our trial showed that azithromycin improves FEV1 and FVC in a significant proportion of patients. As in previous open studies not all patients with BOS responded to azithromycin. We also noted that in patients treated with placebo who were withdrawn from the study due to rapid fall in lung function, treatment with open label azithromycin was associated with a significant gain in lung function leading to our decision to report our Completers and ‘as treated’ groups. We recognise that our ‘as treated’ analysis should be interpreted cautiously, as a post hoc analysis. Overall we conclude that azithromycin appears superior to placebo treatment in our study. An implication of the present study is that the definition of BOS, which currently includes the presence of irreversible airflow obstruction, should be revised or a new term introduced to describe a phenotype of patient who fulfils the definition of BOS, apart from showing a response to azithromycin. This has been suggested prominently by others.5

study is that the definition of BOS, which currently includes the presence of irreversible airflow obstruction, should be revised or a new term introduced to describe a phenotype of patient who fulfils the definition of BOS, apart from showing a response to azithromycin. This has been suggested prominently by others.5 A limitation of our study was randomisation of 48 subjects versus the target of 64. The incidence of BOS throughout the potential trial population was below that estimated. Consequently, our recruitment rate was on average one per month rather than the estimated rate of two per month. Our recruitment period ran for 48 months rather than the anticipated 30 months. This resulted in having data available from 46 patients for the ITT analysis compared with the target of 58. This shortfall would be expected to reduce the power of the study. However, the observed SD of the change in FEV1 from baseline was also much smaller than anticipated at design, a factor that would be expected to favour the study power. In line with CONSORT guidelines,25 we reported 95% CIs for the trial, allowing open interpretation of the findings to be considered along with study size.

ver, the observed SD of the change in FEV1 from baseline was also much smaller than anticipated at design, a factor that would be expected to favour the study power. In line with CONSORT guidelines,25 we reported 95% CIs for the trial, allowing open interpretation of the findings to be considered along with study size. Our study compliments a previous randomised trial in lung transplantation where azithromycin was used as prophylaxis against developing BOS. This had a primary end point of freedom from BOS and survival 2 years after lung transplantation.26 This found that BOS-free survival was better with azithromycin. Patients receiving azithromycin had better FEV1, and lower airway neutrophilia.26 When open-label azithromycin was initiated in patients with BOS, this was associated with an improvement of FEV1 in around half the patients treated.26 Our study was powered to detect a change in FEV1. FEV1 is the basis for the internationally recognised classification of BOS and has previously been reported in open studies of azithromycin therapy,9–13 and was reported in the sole previous randomised trial of azithromycin for BOS prophylaxis.26FVC may be sensitive to changes in the calibre of smaller, peripheral airways.27 28 Our data indicated that azithromycin was superior to placebo for FVC in the Completer population and ‘as treated’ analysis.

omycin therapy,9–13 and was reported in the sole previous randomised trial of azithromycin for BOS prophylaxis.26FVC may be sensitive to changes in the calibre of smaller, peripheral airways.27 28 Our data indicated that azithromycin was superior to placebo for FVC in the Completer population and ‘as treated’ analysis. Our current and previously reported data29 therefore suggest that FVC and other tests such as FEF25–75% may be useful end points for intervention studies in BOS. Previous open studies of azithromycin therapy, including work from our own centre have not reported FVC data9–13 or focused on physiological measurements which reflect smaller airway function. It would seem reasonable to report such measurements given the recognised small airway contribution to BOS pathophysiology, and that the measurements are readily made.

erapy, including work from our own centre have not reported FVC data9–13 or focused on physiological measurements which reflect smaller airway function. It would seem reasonable to report such measurements given the recognised small airway contribution to BOS pathophysiology, and that the measurements are readily made. As a secondary end point we analysed the effects of azithromycin therapy on BAL neutrophils, reflecting our long-standing interest30 31 BOS has a neutrophilic pathophysiology and it has been suggested that the clinical heterogeneity of BOS may be clarified by considering a distinct patient subset with neutrophilic reversible allograft dysfunction.5 This proposed dichotomy may have important therapeutic implications in predicting patients who might respond to azithromycin.5 In a significant number of our patients no BAL data were available. The samples that were available confirmed our previous finding,30 and others’32 33 that BOS is accompanied by an elevated BAL neutrophil count. However azithromycin therapy was not associated with a change in neutrophils. Our findings are therefore different to the Leuven Centre which has published data indicating that a fall in neutrophilic inflammation occurs in patients with a clinical response to azithromycin. We feel that the limited data available for analysis in our study precludes firm conclusions being drawn. We would therefore recommend that further research is performed to clarify the relationship between BAL neutrophil levels and the clinical effectiveness of azithromycin treatment.

clinical response to azithromycin. We feel that the limited data available for analysis in our study precludes firm conclusions being drawn. We would therefore recommend that further research is performed to clarify the relationship between BAL neutrophil levels and the clinical effectiveness of azithromycin treatment. Apart from potential anti-inflammatory benefits, which may include effects on neutrophil numbers and function, other effects of macrolide therapy have been reviewed elsewhere and warrant further study. These include immunomodulatory mechanisms, interference in the formation of infective biofilms and alleviation of extraoesophageal reflux and microaspiration, with promotion of gastric motility.34 It is increasingly recognised that reflux and aspiration may be an important injury in lung allografts,35 and we would recommend that characterisation of reflux disease is considered in future intervention trials in BOS.

d alleviation of extraoesophageal reflux and microaspiration, with promotion of gastric motility.34 It is increasingly recognised that reflux and aspiration may be an important injury in lung allografts,35 and we would recommend that characterisation of reflux disease is considered in future intervention trials in BOS. In our study azithromycin treatment was not associated with significant adverse events. There were no deaths or graft losses during the study. It has been suggested that prolonged treatment with azithromycin may have adverse effects which may be of potential relevance in lung transplantation. These include gastrointestinal effects, loss of hearing36 and the development of macrolide resistant organisms.34 It is also suggested that azithromycin could predispose patients to the development of non-TB mycobacterial infection.37 Debate has also been generated by a study of azithromycin use in a group of patients with pre-existing heart disease. Here azithromycin treatment was associated with an increased rate of cardiovascular related mortality.38 It has also been shown that the macrolide antibiotics may cause cholestatic hepatitis at an estimated rate of 3.6 per 100 000.39

by a study of azithromycin use in a group of patients with pre-existing heart disease. Here azithromycin treatment was associated with an increased rate of cardiovascular related mortality.38 It has also been shown that the macrolide antibiotics may cause cholestatic hepatitis at an estimated rate of 3.6 per 100 000.39 While azithromycin treatment in BOS is generally safe, and provides a therapeutic opportunity in a pathophysiology causing significant morbidity and mortality,9 10 12–16 26 40 it remains a research priority to elucidate which patients benefit from azithromycin, what the optimum timing of treatment is and to provide long-term follow-up data. This might lessen the possibility of iatrogenic consequences of therapy, although these are also the subject of investigation and debate, with the possibility that risks have been overestimated generally, and may not be especially relevant in the specialised setting of lung transplantation. We consider the potential benefits of alternate day low dose azithromycin 250 mg outweigh the potential risks in lung transplantation. Ideally the results of this trial should be replicated. We conclude that this study provides strengthened evidence for the clinical practice of initiating azithromycin therapy for patients who develop BOS post lung transplantation.

While azithromycin treatment in BOS is generally safe, and provides a therapeutic opportunity in a pathophysiology causing significant morbidity and mortality,9 10 12–16 26 40 it remains a research priority to elucidate which patients benefit from azithromycin, what the optimum timing of treatment is and to provide long-term follow-up data. This might lessen the possibility of iatrogenic consequences of therapy, although these are also the subject of investigation and debate, with the possibility that risks have been overestimated generally, and may not be especially relevant in the specialised setting of lung transplantation. We consider the potential benefits of alternate day low dose azithromycin 250 mg outweigh the potential risks in lung transplantation. Ideally the results of this trial should be replicated. We conclude that this study provides strengthened evidence for the clinical practice of initiating azithromycin therapy for patients who develop BOS post lung transplantation. Supplementary Material Web supplement The authors thank all staff in the Freeman Hospital transplant unit. The authors also thank the volunteer data monitoring and ethics committee for the trial; Professor Martin Bland (Professor of Health Statistics at the University of York), Dr Chris Stenton (Consultant Respiratory Physician, Royal Victoria Infirmary, Newcastle upon Tyne) and Dr Graham Burns (Consultant Respiratory Physician, Royal Victoria Infirmary, Newcastle upon Tyne). This research was only possible due to the courage of the participant lung transplant recipients and their families.

, Dr Chris Stenton (Consultant Respiratory Physician, Royal Victoria Infirmary, Newcastle upon Tyne) and Dr Graham Burns (Consultant Respiratory Physician, Royal Victoria Infirmary, Newcastle upon Tyne). This research was only possible due to the courage of the participant lung transplant recipients and their families. Contributors: PAC was clinical science lead for this study. He co-wrote the manuscript; VAR was trial statistician. She analysed the data, wrote the statistical sections and helped to write the manuscript; TS was trial manager and reviewed the manuscript; JL, AJF, GM were clinical co-investigators, who recruited patients and reviewed the manuscript; GJ was lab scientist responsible for all patient BAL sample results and reviewed the manuscript; CW was science lead and principal investigator for this study. He wrote the manuscript drafts. Funding: Funded by Medical Research Council Project grant G0500705 and a British Lung Foundation Trevor Clay Award. Competing interests: None. Patient consent: Obtained. Ethics approval: Local ethics committee and UK Medicines and Healthcare products Regulatory Agency (MHRA). Provenance and peer review: Not commissioned; externally peer reviewed.

Key messages What is the key question? Does paracetamol exposure during pregnancy or lactation lead to the development of allergic airways disease in early life? What is the bottom line? Maternal exposure to paracetamol, either during pregnancy, or lactation, or both, does not affect development of house dust mite induced allergic airways disease in neonatal mice. Why read on? This is the first mechanistic study to investigate a causal link between prenatal and early life paracetamol exposure and the subsequent development of allergic airways disease.

What is the bottom line? Maternal exposure to paracetamol, either during pregnancy, or lactation, or both, does not affect development of house dust mite induced allergic airways disease in neonatal mice. Why read on? This is the first mechanistic study to investigate a causal link between prenatal and early life paracetamol exposure and the subsequent development of allergic airways disease. Introduction An association between paracetamol use in pregnancy and increased risk of early childhood wheezing was first reported in a large population-based birth cohort study several years ago.1 Paracetamol exposure in late gestation was subsequently associated with an increased risk of doctor-diagnosed asthma and elevated total IgE at 7 years, but not allergen skin test positivity, lung function or bronchial responsiveness; early gestational exposure was associated with an increased risk of asthma but not raised IgE.2 3 It has been argued that confounding by unmeasured behavioural factors linked to paracetamol use is an unlikely explanation for these findings.4 Some evidence has recently been reported for interactions between maternal antioxidant gene polymorphisms and prenatal paracetamol exposure on childhood asthma risk, thus strengthening the likelihood of causality.3 5 Since the original report, a body of epidemiological evidence from other birth cohort studies has accumulated,6 with most,5 7 8 though not all,9 confirming a link between prenatal exposure and subsequent asthma or wheezing.

aracetamol exposure on childhood asthma risk, thus strengthening the likelihood of causality.3 5 Since the original report, a body of epidemiological evidence from other birth cohort studies has accumulated,6 with most,5 7 8 though not all,9 confirming a link between prenatal exposure and subsequent asthma or wheezing. Infant paracetamol exposure has also been linked longitudinally to later wheezing and asthma,10–12 independently of the effect of prenatal exposure3 and also to atopy.3 13 While the most likely explanation for the association with asthma is confounding by early respiratory infection,12–14 we cannot rule out the possibility that paracetamol exposure, perhaps in synergy with viral infection, might promote persistence of wheezing.

y of the effect of prenatal exposure3 and also to atopy.3 13 While the most likely explanation for the association with asthma is confounding by early respiratory infection,12–14 we cannot rule out the possibility that paracetamol exposure, perhaps in synergy with viral infection, might promote persistence of wheezing. While definitive evidence of a causal link can only come from trials in humans, primary prevention trials in pregnancy and infancy pose considerable challenges; one immediate way forward is to carry out experimental studies in animal models.6 In an adult mouse model it was demonstrated that paracetamol, in the equivalent of human therapeutic doses could, via its reactive metabolite, activate the transient receptor potential ankyrin-1 channel, leading to neurogenic airway inflammation, thus offering an alternative potential mechanism by which paracetamol might influence asthma pathogenesis.15 Using an established and characterised neonatal model of house dust mite (HDM) induced allergic airways disease (AAD)16 we have investigated whether maternal exposure to paracetamol during pregnancy and lactation promotes the development of AAD in the offspring.

which paracetamol might influence asthma pathogenesis.15 Using an established and characterised neonatal model of house dust mite (HDM) induced allergic airways disease (AAD)16 we have investigated whether maternal exposure to paracetamol during pregnancy and lactation promotes the development of AAD in the offspring. Materials and methods Animals and paracetamol administration BALB/c mice were housed at Imperial College animal facility and used at 8–16 weeks of age. Female mice were administered 100 µL of liquid paracetamol (Calpol) (120 mg/5mLs paracetamol) or phosphate-buffered saline (PBS) by oral gavage either during pregnancy or lactation alone (figure 1), or both during pregnancy and lactation (figure 2). The dose of paracetamol was based upon the maximum recommended dose of 60 mg/kg/day for an average human. The potential effects of oral gavage in inducing stress during pregnancy were assessed by including a group of naïve pregnant females that underwent no treatment. Since there was no significant effect of oral gavage alone, data for PBS treated and naïve mothers were combined. In order to mimic findings from the epidemiological data which only showed effects with frequent and regular maternal paracetamol use, we administered paracetamol on 5 days every week. Litters were housed with their mothers until weaned at 3 weeks. All mice and litters were maintained in specific pathogen-free conditions and given food and water ad libitum. UK Home Office guidelines for animal welfare based on the Animals Scientific Procedures act 1986 were observed.

paracetamol on 5 days every week. Litters were housed with their mothers until weaned at 3 weeks. All mice and litters were maintained in specific pathogen-free conditions and given food and water ad libitum. UK Home Office guidelines for animal welfare based on the Animals Scientific Procedures act 1986 were observed. Figure 1 Similar airway hyper-responsiveness (AHR) following paracetamol exposure during either pregnancy or lactation. Experimental protocol for paracetamol exposure during pregnancy (A). Pregnant females were treated with 100 µL of 1.2 mg/mL liquid paracetamol by oral gavage 5 days a week, with a break on day 4 and day 7, during pregnancy (↑). Neonates were challenged with either house dust mite (HDM) or phosphate-buffered saline (PBS) intranasally on day 3 of life, three times a week for 3 weeks (↓). AHR measured as (B and E) lung resistance and (C and F) dynamic compliance to increasing doses of methacholine was determined 4 h after the last HDM challenge. Experimental protocol for paracetamol exposure during lactation (D). Female mice were mated and left for the duration of the pregnancy. On the day of birth, mothers were treated with 100 µL of 1.2 mg/mL liquid paracetamol by oral gavage 5 days a week, with a break on day 4 and day 7, during lactation, neonates were challenged with PBS or HDM from day 3 of life. Combined data of at least two experiments (n=10 for control mice and n=16–24 for HDM-exposed mice). No significant differences were found.

100 µL of 1.2 mg/mL liquid paracetamol by oral gavage 5 days a week, with a break on day 4 and day 7, during lactation, neonates were challenged with PBS or HDM from day 3 of life. Combined data of at least two experiments (n=10 for control mice and n=16–24 for HDM-exposed mice). No significant differences were found. Figure 2 Experimental protocol for paracetamol exposure during pregnancy and lactation. Female BALB/c mice aged 6–8 weeks were treated with 100 µL of 1.2 mg/mL liquid paracetamol by oral gavage 5 days a week, with a break on day 4 and day 7, during pregnancy and lactation (↑). Neonatal mice were treated with either house dust mite (HDM) or phosphate-buffered saline (PBS) intranasally from day 3 of life, three times a week for either 3 weeks or 6 weeks (↓). Induction of AAD Pups were exposed to 10 µg (10 µL of 1 mg/mL protein weight solution in PBS) of purified HDM extract (Greer Laboratories, Lenoir, North Carolina, USA) or PBS intranasally from day 3 of life for 3 days/week for 2 weeks, followed by 15 μg for up to 3 weeks or 6 weeks.16 Measurement of AHR Airway hyper-responsiveness (AHR) was measured as a terminal procedure 4 h after last allergen challenge in response to increasing doses of methacholine (3–100 mg/mL, Sigma, Poole, UK) in tracheostomised, anaesthetised mice using a Flexivent system (Scireq, Montreal, Canada) as described previously.16

Measurement of AHR Airway hyper-responsiveness (AHR) was measured as a terminal procedure 4 h after last allergen challenge in response to increasing doses of methacholine (3–100 mg/mL, Sigma, Poole, UK) in tracheostomised, anaesthetised mice using a Flexivent system (Scireq, Montreal, Canada) as described previously.16 Cell recovery Airway Lumen Bronchoalveolar lavage (BAL) was performed using three aliquots of 0.3 mL PBS for 3-week-old mice and 0.4 mL PBS for 6-week-old mice, via a tracheal cannula. BAL fluid was centrifuged (200×g, 5 min at 4°C), supernatants were stored at −80°C for cytokine analysis and cells were resuspended in 0.5 mL complete media (Roswell Park Memorial Institute medium (RPMI)+10% fetal calf serum (FCS), 2 mM L-Glutamine, 100 U/mL Penicillin/Streptomycin). Lung parenchyma One lobe of lung tissue was mechanically chopped and incubated at 37°C for 1 h in complete media containing 0.15 mg/mL collagenase (Type D, Roche Diagnostics, Lewes, UK) and 25 μg/mL DNase (Type 1, Roche Diagnostics). Cells were recovered by filtration through a 70 μm nylon sieve, washed twice and resuspended in 1 mL complete media. Serum Blood was collected by cardiac puncture and transferred into a microtainer with serum separation gel (BD, Fischer scientific). Tubes were centrifuged at 14 000 RPM for 5 min to separate serum. Tubes were then stored at −20°C prior to analysis of immunoglobulins.

Lung parenchyma One lobe of lung tissue was mechanically chopped and incubated at 37°C for 1 h in complete media containing 0.15 mg/mL collagenase (Type D, Roche Diagnostics, Lewes, UK) and 25 μg/mL DNase (Type 1, Roche Diagnostics). Cells were recovered by filtration through a 70 μm nylon sieve, washed twice and resuspended in 1 mL complete media. Serum Blood was collected by cardiac puncture and transferred into a microtainer with serum separation gel (BD, Fischer scientific). Tubes were centrifuged at 14 000 RPM for 5 min to separate serum. Tubes were then stored at −20°C prior to analysis of immunoglobulins. Cytocentrifuge preparation and differential counts of Wright-Giemsa stained BAL and lung cells Lung and BAL cells were applied to glass slides by centrifugation and stained with Wright-Giemsa (Thermo Fisher Scientific, Waltham, Massachusetts, USA). Percentages of macrophages, lymphocytes/mononuclear cells, eosinophils and neutrophils were determined under 40× magnification by counting cells in randomly selected fields and dividing this number by the total number of cells (400) counted. To obtain absolute numbers, this percentage was multiplied by the total number of cells recovered in lavage fluid and lung digest suspension. All cell counts were performed blind by the same observer.

tion by counting cells in randomly selected fields and dividing this number by the total number of cells (400) counted. To obtain absolute numbers, this percentage was multiplied by the total number of cells recovered in lavage fluid and lung digest suspension. All cell counts were performed blind by the same observer. Analysis of cytokines Cytokines were analysed in BAL samples and lung tissue homogenate supernatants. Lung tissue was homogenised at 50 mg/mL in Hank's balanced salt solution (HBSS) containing protease inhibitor tablets (Roche Diagnostics), centrifuged (800×g, 10 min) and the supernatant collected. BAL and lung homogenate cytokine levels were measured using paired antibodies for murine interleukin (IL) 4, IL-5, IL-25, IL-33 and interferon (IFN)-γ (BD Bioscience UK and R&D Systems, Abingdon, UK) in standardised sandwich ELISAs according to the manufacturer's protocol. Kits to measure IL-13 were purchased from R&D Systems. Paired antibodies for IgE (R&D systems) were used to measure serum IgE levels. Total lung collagen levels Recently synthesised acid-soluble collagen was measured in the lung by biochemical assay (Sircol collagen assay; Biocolor, Belfast, UK) according to manufacturer's instructions and normalised for tissue weight.

Analysis of cytokines Cytokines were analysed in BAL samples and lung tissue homogenate supernatants. Lung tissue was homogenised at 50 mg/mL in Hank's balanced salt solution (HBSS) containing protease inhibitor tablets (Roche Diagnostics), centrifuged (800×g, 10 min) and the supernatant collected. BAL and lung homogenate cytokine levels were measured using paired antibodies for murine interleukin (IL) 4, IL-5, IL-25, IL-33 and interferon (IFN)-γ (BD Bioscience UK and R&D Systems, Abingdon, UK) in standardised sandwich ELISAs according to the manufacturer's protocol. Kits to measure IL-13 were purchased from R&D Systems. Paired antibodies for IgE (R&D systems) were used to measure serum IgE levels. Total lung collagen levels Recently synthesised acid-soluble collagen was measured in the lung by biochemical assay (Sircol collagen assay; Biocolor, Belfast, UK) according to manufacturer's instructions and normalised for tissue weight. RNA extraction and qPCR Total RNA was extracted from 50 mg to 100 mg of lung tissue by using a Qiagen RNeasy Mini Kit. Total RNA (1 μg) was reverse transcribed into cDNA using a High Capacity cDNA Reverse Transcription Kit (Life Technologies) as per the manufacturer's instructions. Real-time PCR reactions were performed using fast-qPCR mastermix (Life technologies) on a Viaa-7 (Life Technologies) instrument with TaqMan primer sets for murine amphiregulin, matrix metallopeptidase 2 (MMP-2), fibronectin, Found In Inflammatory Zone (FIZZ)1, vimentin, collagen-1α1 and Hypoxanthine-guanine phosphoribosyltransferase (HPRT) (Life Technologies) and gene expression was analysed using the change-in-threshold ΔΔCt- method.

nstrument with TaqMan primer sets for murine amphiregulin, matrix metallopeptidase 2 (MMP-2), fibronectin, Found In Inflammatory Zone (FIZZ)1, vimentin, collagen-1α1 and Hypoxanthine-guanine phosphoribosyltransferase (HPRT) (Life Technologies) and gene expression was analysed using the change-in-threshold ΔΔCt- method. Statistical analysis Data are expressed as median or mean±SEM. Group comparisons were performed using a non-parametric analysis of variance (ANOVA) test (Kruskal-Wallis), followed by a Dunn’s post test for multiple comparisons between groups. Graph generation and statistical analyses were performed using GraphPad Prism software (V.5.00; GraphPad). Results Maternal paracetamol exposure during pregnancy alone did not affect AHR in neonatal mice In order to determine whether prenatal exposure to paracetamol during pregnancy altered lung function in neonatal mice exposed to HDM, pregnant females were treated with paracetamol on the 1st day of mating and throughout pregnancy. Paracetamol treatment was stopped on the day they gave birth. Offspring were exposed to intranasal PBS or HDM from day 3 of life, for 3 weeks (figure 1A). In utero paracetamol exposure alone did not result in worse lung function in HDM-exposed neonatal mice at weaning (3 weeks) (figure 1B, C).

and throughout pregnancy. Paracetamol treatment was stopped on the day they gave birth. Offspring were exposed to intranasal PBS or HDM from day 3 of life, for 3 weeks (figure 1A). In utero paracetamol exposure alone did not result in worse lung function in HDM-exposed neonatal mice at weaning (3 weeks) (figure 1B, C). Maternal paracetamol exposure during lactation alone did not affect AHR in neonatal mice Therefore, exposure during lactation was investigated to determine the influence of oral ingestion in the neonate via breast milk.17 Female mice were mated and left for the duration of the pregnancy. On the day of birth, mothers received the first dose of oral paracetamol which continued for 3 weeks. Pups were exposed to either intranasal PBS or HDM (figure 1D). HDM-exposed pups from paracetamol-treated mothers had similar AHR compared with those from PBS-treated mothers (figure 1E, F). There was no effect of paracetamol on inflammation in the BAL or lung (see online supplementary figure S1A,B), levels of IL-13 or IL-33 in the lungs (see online supplementary figure S1C,D), or total IgE and HDM-specific IgE (see online supplementary figure S1E, F).

ed with those from PBS-treated mothers (figure 1E, F). There was no effect of paracetamol on inflammation in the BAL or lung (see online supplementary figure S1A,B), levels of IL-13 or IL-33 in the lungs (see online supplementary figure S1C,D), or total IgE and HDM-specific IgE (see online supplementary figure S1E, F). Maternal paracetamol exposure during pregnancy and lactation does not affect AAD in offspring As paracetamol exposure during pregnancy or lactation alone had no effect on neonatal AAD, additive effects of perinatal exposure during pregnancy and lactation were determined in early life (3 weeks) and in young adulthood (6 weeks) (figure 2). HDM-exposed pups from mothers treated with paracetamol during pregnancy and lactation had similar airway resistance and compliance at 3 weeks and 6 weeks compared with those from naïve or PBS-treated mothers (figure 3A–D). The total number of inflammatory cells recruited to the lung and BAL and eosinophils in the lung were unaffected by maternal paracetamol at 3 weeks and 6 weeks (figure 3E–J).

actation had similar airway resistance and compliance at 3 weeks and 6 weeks compared with those from naïve or PBS-treated mothers (figure 3A–D). The total number of inflammatory cells recruited to the lung and BAL and eosinophils in the lung were unaffected by maternal paracetamol at 3 weeks and 6 weeks (figure 3E–J). Figure 3 Allergic airways disease characterisation in house dust mite (HDM)-exposed neonatal mice from paracetamol-treated mothers. Airway hyper-responsiveness (AHR) in response to increasing doses of methacholine measured as lung resistance (A and B) and compliance (C and D) following 3 weeks or 6 weeks of HDM or phosphate-buffered saline (PBS) exposure, was determined in pups from either paracetamol-treated or non-paracetamol-treated mothers. Total inflammatory cells in the lung (E and F), total number of lung eosinophils (G and H) and total inflammatory cells in the bronchoalveolar lavage (BAL, I and J) were assessed at 3 weeks and 6 weeks. Combined data from two experiments (○ naïve females and ▪ PBS treated females) or paracetamol-treated mothers (◊ and ▾); (n=10–12 for control mice and n=11–15 for HDM-exposed mice). Horizontal bars represent median. There were no statistically significant differences between neonates exposed to paracetamol and those without paracetamol exposure at any time point.

emales and ▪ PBS treated females) or paracetamol-treated mothers (◊ and ▾); (n=10–12 for control mice and n=11–15 for HDM-exposed mice). Horizontal bars represent median. There were no statistically significant differences between neonates exposed to paracetamol and those without paracetamol exposure at any time point. Maternal paracetamol exposure during pregnancy and lactation does not alter neonatal inflammatory cytokines or IgE The Th2 cytokines IL-5 and IL-13 were measured in BAL and lung and were significantly increased in HDM-exposed neonatal mice compared with PBS controls at 3 weeks, but maternal paracetamol did not alter levels of BAL (data not shown) or lung IL-5 or IL-13 (figure 4A, B). The same was true for the innate cytokine IL-33 (figure 4C). At 6 weeks levels of IL-5, IL-13 and IL-33 were still increased in HDM-treated groups, compared with controls (figure 4D–F), but there was no additional effect of maternal paracetamol. At 3 weeks and 6 weeks serum total IgE levels were significantly higher in HDM-exposed neonatal mice compared with controls, but there was no impact of maternal paracetamol exposure on total IgE levels (figure 5A). The same was also true for HDM specific IgE levels following 3 and 6 weeks of HDM exposure (figure 5B).

. At 3 weeks and 6 weeks serum total IgE levels were significantly higher in HDM-exposed neonatal mice compared with controls, but there was no impact of maternal paracetamol exposure on total IgE levels (figure 5A). The same was also true for HDM specific IgE levels following 3 and 6 weeks of HDM exposure (figure 5B). Figure 4 Levels of Th2 and innate cytokines in the lung. The levels of interleukin (IL) 5 (A and D), IL-13 (B and E) and IL-33 (C and F) in neonatal lung homogenate supernatant from non-paracetamol-treated (○ naïve females and ▪ phosphate-buffered saline (PBS) treated females) or paracetamol-treated mothers (◊ and ▾) were assessed by ELISA at 3 weeks and 6 weeks. Combined data from two experiments (n=10–12 for control mice and n=11–15 for house dust mite (HDM)-exposed mice). Significant differences between HDM-exposed and PBS-exposed neonates from non-paracetamol-treated or paracetamol-treated mothers are shown as **p<0.01 and ***p<0.001.

sessed by ELISA at 3 weeks and 6 weeks. Combined data from two experiments (n=10–12 for control mice and n=11–15 for house dust mite (HDM)-exposed mice). Significant differences between HDM-exposed and PBS-exposed neonates from non-paracetamol-treated or paracetamol-treated mothers are shown as **p<0.01 and ***p<0.001. Figure 5 Total and house dust mite (HDM)-specific IgE levels. Serum from HDM-treated or phosphate-buffered saline (PBS)-treated neonatal mice from either non-paracetamol or paracetamol treated mothers was assessed by ELISA for total IgE (A) or HDM-specific IgE (B) following 3 weeks or 6 weeks of HDM or PBS exposure. Combined data from two experiments (○ naïve females and ▪ PBS-treated females) or paracetamol-treated mothers (◊ and ▾) (n=10–12 for control mice and n=11–15 for HDM-exposed mice). Significant differences between HDM-exposed and PBS-exposed neonates from non-paracetamol-treated or paracetamol-treated mothers are shown as **p<0.01 and ***p<0.001. ns, not significant.

les and ▪ PBS-treated females) or paracetamol-treated mothers (◊ and ▾) (n=10–12 for control mice and n=11–15 for HDM-exposed mice). Significant differences between HDM-exposed and PBS-exposed neonates from non-paracetamol-treated or paracetamol-treated mothers are shown as **p<0.01 and ***p<0.001. ns, not significant. Maternal paracetamol did not affect airway remodelling in HDM-exposed neonatal mice Increased subepithelial reticular membrane thickness is an early feature of airway remodelling in preschool wheeze18 and childhood asthma,19 we therefore analysed mRNA levels of genes associated with remodelling including FIZZ1,20 fibronectin,21 vimentin,22 MMP-2,23 amphiregulin24 and collagen-1α1,25 from neonatal mice exposed to paracetamol during pregnancy alone (figure 1A), or lactation alone (figure 1D). Paracetamol treatment during pregnancy or lactation led to an upregulation of FIZZ1 expression in 3 week old HDM-treated neonates, in comparison to PBS-treated controls (figure 6A). Minimal changes were observed in the expression of vimentin and fibronectin, following 3 weeks of HDM in mice exposed to paracetamol in utero, in comparison to PBS exposed controls (figure 6B,C). However, these alterations did not lead to lung function changes (figure 1 and see online supplementary figure S2A–D). Importantly, the effect of paracetamol exposure was transient and was not observed following 6 weeks of HDM exposure. Expression levels of amphiregulin, MMP-2 and collagen-1α1 were unaffected by paracetamol exposure following 3 weeks or 6 weeks of HDM treatment (figure 6D–F). In order to determine whether differences in protein were present, total lung collagen was quantified by biochemical assay. Maternal paracetamol treatment during pregnancy, lactation, or both, had no impact on collagen levels in neonatal mice exposed to HDM for 3 weeks (figure 7A–C) or in young adulthood at week 6 (figure 7D).

to determine whether differences in protein were present, total lung collagen was quantified by biochemical assay. Maternal paracetamol treatment during pregnancy, lactation, or both, had no impact on collagen levels in neonatal mice exposed to HDM for 3 weeks (figure 7A–C) or in young adulthood at week 6 (figure 7D). Figure 6 Increased levels of airway remodelling gene expression in house dust mite (HDM) neonatal mice from paracetamol-treated mothers. RNA was extracted from lungs from 3-week-old and 6-week-old neonatal mice exposed to HDM or phosphate-buffered saline (PBS), from mothers treated with paracetamol or PBS during pregnancy alone or lactation alone. mRNA levels for Found In Inflammatory Zone (FIZZ)1 (A), vimentin (B), fibronectin (C), amphiregulin (D), matrix metallopeptidase 2 (MMP)-2 (E) and collagen-1α1 (F) are shown as fold induction normalised to the control group, PBS-exposed neonates from PBS-treated mothers. Data represents three individual experiments (3 week data) or one representative experiment (6 week data). Significant differences between HDM exposed neonates from control or paracetamol-treated mothers and between PBS control mice are shown as *p<0.05 and **p<0.01.

roup, PBS-exposed neonates from PBS-treated mothers. Data represents three individual experiments (3 week data) or one representative experiment (6 week data). Significant differences between HDM exposed neonates from control or paracetamol-treated mothers and between PBS control mice are shown as *p<0.05 and **p<0.01. Figure 7 Similar levels of lung collagen in all house dust mite (HDM)-exposed neonatal mice, regardless of paracetamol exposure. Total lung collagen was measured in lung supernatants from neonatal mice treated with HDM or phosphate-buffered saline (PBS) for 3 weeks or 6 weeks, from mothers exposed to paracetamol during pregnancy or lactation only, and during both pregnancy and lactation using the Sircol biochemical assay. Combined data from at least two experiments (n=10 for control mice and n=16–24 for HDM-exposed mice). Horizontal bars represent median, *p<0.05 and **p<0.01.

ks or 6 weeks, from mothers exposed to paracetamol during pregnancy or lactation only, and during both pregnancy and lactation using the Sircol biochemical assay. Combined data from at least two experiments (n=10 for control mice and n=16–24 for HDM-exposed mice). Horizontal bars represent median, *p<0.05 and **p<0.01. Discussion To our knowledge, this is the first mechanistic study to investigate a causal link between maternal paracetamol intake and asthma in early life. We have provided direct in vivo evidence that maternal paracetamol exposure either during pregnancy, lactation, or during both, does not affect AHR in offspring exposed to inhaled HDM. There was also no enhanced inflammation, Th2 cytokines or increased levels of serum total or HDM-specific IgE. Although expression of early remodelling genes FIZZ1 and fibronectin were increased at weaning (3 weeks) following in utero paracetamol exposure, there was no associated change in protein (total lung collagen) or lung function, and the differences were not maintained into later life (6 weeks).

otal or HDM-specific IgE. Although expression of early remodelling genes FIZZ1 and fibronectin were increased at weaning (3 weeks) following in utero paracetamol exposure, there was no associated change in protein (total lung collagen) or lung function, and the differences were not maintained into later life (6 weeks). An epidemiological association between paracetamol exposure during pregnancy and wheezing in early childhood was first demonstrated over a decade ago in the Avon Longitudinal Study of Parents and Children (ALSPAC) birth cohort.1 Frequent paracetamol use, particularly in the latter half of pregnancy, was associated with persistent wheeze in children at age 3.5 years and asthma at age 7 years;1 2 the relation with asthma was independent of the association between infant paracetamol exposure and asthma.3 Many other birth cohort studies have since shown a link between maternal paracetamol use during pregnancy and the development of wheezing or asthma in children.6

en at age 3.5 years and asthma at age 7 years;1 2 the relation with asthma was independent of the association between infant paracetamol exposure and asthma.3 Many other birth cohort studies have since shown a link between maternal paracetamol use during pregnancy and the development of wheezing or asthma in children.6 We first tried to mimic the effects of in utero exposure alone by administering paracetamol to female mice only during pregnancy, but in contrast to the epidemiological data showing a positive association between in utero paracetamol exposure and elevated serum IgE at 7 years,2 we saw no difference in either total or allergen-specific IgE levels. In addition, we found no effect of prenatal paracetamol exposure on any other features of AAD in our animal model. One explanation may be that the epidemiological association with maternal use in pregnancy is explained by residual or unmeasured confounding factors. For example, few cohorts had information on indication for use. A recent, large prospective study designed to look at the effects of maternal and infant paracetamol ingestion and subsequent asthma has only shown an association between prenatal paracetamol and asthma in early childhood, the effect was not sustained until mid-childhood.26 In order to see whether even a short-term increase in susceptibility to AAD was present following maternal paracetamol exposure, we assessed pups at 3 weeks and 6 weeks of age. In keeping with the findings from Sordillo et al, we saw a transient increase in expression of remodelling genes only at 3 weeks. However, there was no enhanced effect on lung function, inflammation or sensitisation apparent from maternal paracetamol at either 3 weeks or 6 weeks.

sed pups at 3 weeks and 6 weeks of age. In keeping with the findings from Sordillo et al, we saw a transient increase in expression of remodelling genes only at 3 weeks. However, there was no enhanced effect on lung function, inflammation or sensitisation apparent from maternal paracetamol at either 3 weeks or 6 weeks. Another issue is whether infant ingestion may impact subsequent asthma development. One epidemiological study demonstrated that paracetamol ingestion by infants in the 1st year of life was associated with troublesome lower lung symptoms independently of lower respiratory tract infections in the first 3 years. There was no association between paracetamol ingestion during the 1st year and asthma at 7 years of age, or between maternal paracetamol ingestion during pregnancy and asthma or troublesome lower lung symptoms.27 Similarly, a much larger prospective study has shown adjustment for respiratory infections in early life substantially reduces the association between infant paracetamol intake and subsequent asthma.26 However, evidence from other epidemiological studies on whether the link between infant use and asthma is confounded by respiratory infection is conflicting.6 12 Paracetamol can be rapidly secreted into breast milk and is detectable at higher concentrations in the milk compared with maternal plasma.17 Furthermore, paracetamol metabolites detected in the urine from babies from breastfeeding mothers, differ significantly from those detected in the urine of healthy adults who have ingested paracetamol. Neonates have significantly higher levels of paracetamol in their urine compared with adults, and adults have greater levels of urinary paracetamol sulfate,17 confirming that the pharmacokinetics and pharmacodynamics of paracetamol differ substantially in murine28 and human neonates and infants compared with children and adults.29 As it is technically very difficult to perform oral gavage in newborn mice, the importance of oral paracetamol ingestion in early life on subsequent AAD was indirectly determined in our study by administering paracetamol to mothers only during lactation and assessing AAD just prior to weaning at 3 weeks. However, we did not see any effect of oral ingestion via breast milk on the development of AAD in 3 week-old mice. There was also no long-term effect after weaning in mice exposed to HDM for 6 weeks. Paracetamol exposure did not alter inflammation or Th2 cytokines and levels of IgE were also unaffected.

ing at 3 weeks. However, we did not see any effect of oral ingestion via breast milk on the development of AAD in 3 week-old mice. There was also no long-term effect after weaning in mice exposed to HDM for 6 weeks. Paracetamol exposure did not alter inflammation or Th2 cytokines and levels of IgE were also unaffected. Various genes have now been identified as important regulators of bronchial subepithelial basement membrane thickening and as such we analysed the mRNA expression levels of several fibrotic and early remodelling genes. Interestingly, we found there was a very small (less than twofold), but statistically significant increase in the mRNA levels of fibronectin and vimentin in allergen-exposed mice from mothers treated with paracetamol during pregnancy, but not during lactation; whereas FIZZ1 was upregulated after paracetamol exposure during lactation or pregnancy. FIZZ1 is induced in the lung following ovalbumin or Alternaria challenge,20 30 while fibronectin has been shown to be essential for the development of ovalbumin-induced airway fibrosis and AHR.20 Vimentin and fibronectin are also associated with airway collagen deposition and airway remodelling.21 22 Despite the transient increase in the expression of these genes we found no change in protein levels of lung collagen in HDM-treated neonatal mice exposed to maternal paracetamol.

induced airway fibrosis and AHR.20 Vimentin and fibronectin are also associated with airway collagen deposition and airway remodelling.21 22 Despite the transient increase in the expression of these genes we found no change in protein levels of lung collagen in HDM-treated neonatal mice exposed to maternal paracetamol. We acknowledge that the findings from a murine model may not be directly applicable to the human situation. However, murine models have been used extensively to study paracetamol toxicity.31 Furthermore, the transient changes with paracetamol in FIZZ1, fibronectin and vimentin suggest that paracetamol crosses the placenta. There is no literature on whether paracetamol crosses into breast milk of lactating mice, but conversely, there is no evidence to suggest that breast milk metabolism in mice is significantly different from humans. To our knowledge, this is the first mechanistic investigation of a direct effect of paracetamol exposure either in utero, or orally via breast milk, or at both times, on neonatal HDM-induced AAD. The strengths include; (1) use of an age-appropriate model with inhaled allergen challenge, (2) administration of maternal paracetamol with a similar frequency and dosage, equivalent to those in epidemiological studies shown to have effects and (3) absence of any confounding from respiratory infections. With these best possible conditions, there was no effect of maternal paracetamol on the development of any parameters of AAD in offspring either in early life (3 weeks of age) or later at 6 weeks. These data may help to demonstrate why a recent systematic review of studies looking at the effect of paracetamol in pregnancy and early life and subsequent asthma has concluded (A) the association between exposure during pregnancy and asthma in childhood is highly variable between studies and is not robust, and (B) the association between paracetamol ingestion in infancy and subsequent asthma is confounded by respiratory infections.32

egnancy and early life and subsequent asthma has concluded (A) the association between exposure during pregnancy and asthma in childhood is highly variable between studies and is not robust, and (B) the association between paracetamol ingestion in infancy and subsequent asthma is confounded by respiratory infections.32 Supplementary Material Web figures Contributors: DCPL designed and conducted the majority of the experiments, analysed the data and wrote the first draft of the manuscript; SAW, AJB, LGG and JB performed and analysed the revised experimental work; AB and SOS reviewed and edited the manuscript; SS and CML designed the study, supervised the project and edited the manuscript. Funding: This work was supported by Asthma UK (Grant ID: 11/051) and the Wellcome Trust (Grant ID: 087618/Z/08/Z). AB was supported by the NIHR Respiratory Disease Biomedical Research Unit at the Royal Brompton and Harefield NHS Foundation Trust and Imperial College London. Competing interests: AJB, SOS, CML and SS are members of the MRC Asthma UK Centre for Allergic Mechanisms of Disease. CML is a Wellcome Senior Research Fellow in Basic Biomedical Science. Provenance and peer review: Not commissioned; internally peer reviewed.

Introduction Inhaled corticosteroids and β-2 agonists have been the first-line treatment for asthma since the 1960s, despite almost half a century of targeted research.1 These control the symptoms of the majority of patients, who experience mild-to-moderate forms of the disease. However, 5%–10% of the asthma population exhibit severe symptoms that are not adequately controlled by conventional inhaled therapy.2 While this represents a small proportion of asthma sufferers, they account for almost half of healthcare costs related to the disease and the majority of asthma-related deaths.3 There is a great unmet need to discover new treatments in severe asthma and also for the different sub-phenotypes that are progressively recognised to represent distinct disease mechanisms.4 Many promising drugs that perform well in preclinical animal studies fail in humans due to lack of safety and/or efficacy, suggesting the current preclinical testing strategy, focusing largely on murine in vivo models, which do not recapitulate the complexity of the human disease, is not sophisticated enough to meet today's respiratory drug development needs.5 A new approach to drug discovery and development in this area is necessary.

acy, suggesting the current preclinical testing strategy, focusing largely on murine in vivo models, which do not recapitulate the complexity of the human disease, is not sophisticated enough to meet today's respiratory drug development needs.5 A new approach to drug discovery and development in this area is necessary. Although some facets of the asthmatic phenotype can be modelled in certain animal species, no animals that are commonly used to study the condition (including mice, rats, guinea pigs and rabbits) develop an asthma-like syndrome which completely correlates to the human disease. As asthma is a disease unique to humans, the development and application of human tissue-based approaches with which to study the disease should be considered a priority. Methods The NC3Rs, working with Asthma UK, the UK Respiratory Research Collaborative and the Human Tissue Authority, recently surveyed the UK asthma research community to better understand the extent of human tissue use in this area. The survey was distributed as an online questionnaire through the partner networks and others, including the British Thoracic Society, the British Lung Foundation and the British Association for Lung Research. The survey was completed by 59 individuals from academia, pharmaceutical companies, small and medium enterprises and the National Health Service. The survey was divided into sections to capture information on: The way human lung tissue is currently used in asthma research; The level of knowledge surrounding the regulatory requirements and guidance on human tissue use;

The survey was distributed as an online questionnaire through the partner networks and others, including the British Thoracic Society, the British Lung Foundation and the British Association for Lung Research. The survey was completed by 59 individuals from academia, pharmaceutical companies, small and medium enterprises and the National Health Service. The survey was divided into sections to capture information on: The way human lung tissue is currently used in asthma research; The level of knowledge surrounding the regulatory requirements and guidance on human tissue use; The perceived barriers to wider uptake of human tissue-based approaches in asthma research. Results Only 14% of people surveyed do not currently use human tissue in their asthma research programmes. Of those who do (86%), 16% use it exclusively, while the majority use it in conjunction with immortalised cells or cryopreserved material (47%) or alongside animal studies (37%). The type of human tissue used is predominantly limited to that which is easy to access, with 86% of respondents using either normal or diseased primary cells, 47% using sputum and 53% using biofluids. In contrast, the percentage using whole lung is much lower at 24%.

Results Only 14% of people surveyed do not currently use human tissue in their asthma research programmes. Of those who do (86%), 16% use it exclusively, while the majority use it in conjunction with immortalised cells or cryopreserved material (47%) or alongside animal studies (37%). The type of human tissue used is predominantly limited to that which is easy to access, with 86% of respondents using either normal or diseased primary cells, 47% using sputum and 53% using biofluids. In contrast, the percentage using whole lung is much lower at 24%. The type of tissue currently used in asthma research contrasts strongly with the type of tissue that researchers would like to use if they could access it, with only 13% wanting to use primary cells, while 44% expressed a desire to use whole lungs. The reasons given for not being able to use these types of tissue focused on the difficulty in accessing them, with normal tissue being in such high demand that there is not enough available to support studies. Respondents also indicated a desire to use larger tissue samples, such as wedge resections or whole lobes for functional experiments on isolated airways, blood vessels and other tissue components.

ulty in accessing them, with normal tissue being in such high demand that there is not enough available to support studies. Respondents also indicated a desire to use larger tissue samples, such as wedge resections or whole lobes for functional experiments on isolated airways, blood vessels and other tissue components. The survey highlighted a number of barriers to using fresh human tissue in asthma research (figure 1A). The majority of those questioned (71%) were concerned about the availability of either normal or diseased fresh human tissue. Another widely considered barrier is the practical issues related to the acquisition and storage of tissue (59%). Respondents considered that increasing access to a reliable supply of normal and diseased human tissue (71%), increased specific funding for human tissue research programmes (56%) and evidence that human tissue-based methods are more predictive than current animal models (46%) would enable greater adoption of human tissue methods as part of asthma research programmes (figure 1B). Other opportunities for increasing adoption of human tissue approaches are considered subsequently.

programmes (56%) and evidence that human tissue-based methods are more predictive than current animal models (46%) would enable greater adoption of human tissue methods as part of asthma research programmes (figure 1B). Other opportunities for increasing adoption of human tissue approaches are considered subsequently. Figure 1 (A) The top five barriers to the use of human tissue in asthma research. (B) Top five changes that would enable researchers to use more human tissue in their asthma research programmes. (C) Extent to which respondents agree or disagree that better access to, and wider use of, fresh human tissue would (i) reduce the number of animals used for asthma research, (ii) inform the development of more predictive animal models, (iii) speed up the development of efficacious new therapies and (iv) improve our understanding of the pathobiology of asthma. The impacts of wider adoption of human tissue for asthma research were far reaching, with respondents agreeing that this would reduce animal use, result in more predictive models, lead to new therapies and improve our understanding of the human disease (figure 1C).

Figure 1 (A) The top five barriers to the use of human tissue in asthma research. (B) Top five changes that would enable researchers to use more human tissue in their asthma research programmes. (C) Extent to which respondents agree or disagree that better access to, and wider use of, fresh human tissue would (i) reduce the number of animals used for asthma research, (ii) inform the development of more predictive animal models, (iii) speed up the development of efficacious new therapies and (iv) improve our understanding of the pathobiology of asthma. The impacts of wider adoption of human tissue for asthma research were far reaching, with respondents agreeing that this would reduce animal use, result in more predictive models, lead to new therapies and improve our understanding of the human disease (figure 1C). Discussion The survey suggests that there is already widespread use of human tissue in asthma research, with over 86% of respondents reporting that they use human tissue in some capacity. The survey also highlights that human tissue is used in a wide variety of research areas, including immunology, physiology, pharmacology, genetics and epigenetics, and compound evaluation. However, it is clear that most of this human tissue use is limited to isolation of individual cells rather than studies of airways or other components of the lung allowing for experiments in complex intact tissue.

areas, including immunology, physiology, pharmacology, genetics and epigenetics, and compound evaluation. However, it is clear that most of this human tissue use is limited to isolation of individual cells rather than studies of airways or other components of the lung allowing for experiments in complex intact tissue. The consensus among the researchers questioned is that there are clear benefits that can be realised by increasing the use of human tissue in asthma research. These include a greater understanding of the pathobiology of asthma, quicker development of efficacious new therapies to treat the disease, the development of more predictive animal models of asthma and a reduction in the overall number of animals used in asthma research programmes.

of human tissue in asthma research. These include a greater understanding of the pathobiology of asthma, quicker development of efficacious new therapies to treat the disease, the development of more predictive animal models of asthma and a reduction in the overall number of animals used in asthma research programmes. So what is preventing more human tissue use? The most common barrier is access, particularly to normal human tissue and also to diseased tissue. The major obstacle to an effective supply of human tissues is the logistical hurdles that need to be overcome. A lot of human lung tissue is being disposed of in hospitals all over the country. There is a need for greater cooperation between pathologists, transplant surgeons and researchers, and special research funding for resources to manage the collection, processing and transport of tissues for research without compromising the requirements of the caregivers. Owing to these hurdles, 28% of those using human tissue use a commercial supplier, either in the UK or internationally, to ensure a regular supply of lung tissue. This suggests there is a need for a more widely accessible UK-based lung tissue bio-resource or tissue network to obtain fresh human tissue, which 83% of respondents indicate they would be likely to access.

tissue use a commercial supplier, either in the UK or internationally, to ensure a regular supply of lung tissue. This suggests there is a need for a more widely accessible UK-based lung tissue bio-resource or tissue network to obtain fresh human tissue, which 83% of respondents indicate they would be likely to access. The regulatory requirements for the removal, storage and use of human tissues are seen as a barrier to increased human tissue use by 44% of respondents. However, the survey results suggest that this is at least partly a perceived barrier, as there was a lack of knowledge surrounding some aspects of the regulations. The licensing requirements concerning the storage and use of human tissue for research, and the requirements for consent to use and store human tissue for research purposes were considered particularly unclear. As many as 42% of respondents felt there was an insufficient level of information available on the regulatory framework surrounding the collection and use of human tissue. A large percentage (68%) of respondents felt that the regulatory requirements hold up research unnecessarily, with a similar proportion reporting that requirements are complicated and difficult to understand (70%), and that they are not streamlined (75%). This information is easily accessible online and represents one area where better dissemination and improved awareness of existing information could potentially have a significant impact on the uptake of human-tissue use in asthma research.

ted and difficult to understand (70%), and that they are not streamlined (75%). This information is easily accessible online and represents one area where better dissemination and improved awareness of existing information could potentially have a significant impact on the uptake of human-tissue use in asthma research. Journals also have a role to play in enabling more widespread human tissue use, according to 51% of survey respondents. A greater willingness on the part of journals to publish human tissue research without accompanying animal model data would increase the evidence base to support the use of these models and provide confidence to encourage their wider uptake.

ling more widespread human tissue use, according to 51% of survey respondents. A greater willingness on the part of journals to publish human tissue research without accompanying animal model data would increase the evidence base to support the use of these models and provide confidence to encourage their wider uptake. The current survey illustrates the views of a representative proportion of the asthma research community. It provides important evidence that a more concerted effort is needed to support asthma researchers in adopting human tissue-based approaches. Many barriers exist to this, and some of these may be more easily overcome than others, but the potential benefits to the science base, drug development and the 3Rs (the replacement, reduction and refinement of animals in research) are many. The asthma research community needs to come together to share information on how these barriers can be overcome, and it is important that funders, journal editors and regulators all play their part in supporting this. Additionally, there needs to be a meeting of minds between caregivers and researchers to ensure that as much as possible of the human tissue being collected in hospitals, which is surplus to healthcare needs, is made available to the research community. Recognising this need, the NC3Rs is developing tools to help researchers access relevant information to expedite the wider adoption of human tissue-based models in basic research and drug development. In conclusion, we believe it is important that the major funding bodies and other key stakeholders give higher priority to allocate the resources required to enhance and facilitate respiratory research using human tissues.

expedite the wider adoption of human tissue-based models in basic research and drug development. In conclusion, we believe it is important that the major funding bodies and other key stakeholders give higher priority to allocate the resources required to enhance and facilitate respiratory research using human tissues. The authors would like to acknowledge the contribution made by the members of the NC3Rs Asthma Advisory Group and the HTA in developing the survey, the partner organisations for disseminating the survey and the respondents who took the time to complete the survey. Correction notice: This article have been corrected since it was published Online First. Changes has been made in the 'Discussion' section and 'Acknowledgments' has been updated. Competing interests: None. Provenance and peer review: Not commissioned; internally peer reviewed.

Key messages What is the key question? Is vitamin D deficiency a risk factor for the development of acute respiratory distress syndrome (ARDS)? What is the bottom line? Patients with and at risk of ARDS are highly likely to be deficient, and severity of vitamin D deficiency relates to increased epithelial damage, the development of ARDS and survival. Why read on? We present evidence that an easily treatable vitamin deficiency may increase the risk of ARDS in patients at risk. Introduction Acute respiratory distress syndrome (ARDS) occurs due to either direct or indirect proinflammatory insults. However, only a proportion of at-risk patients develop ARDS, with research suggesting that genetic, age, social and other factors play a role in determining who develops ARDS.1 More than 1 billion people worldwide are believed to have vitamin D deficiency.2 Vitamin D has important functions besides bone and calcium homeostasis3 with cells of the innate and adaptive immune system responding to vitamin D. Vitamin D deficiency may therefore increase the risk of bacterial and viral infection.

Introduction Acute respiratory distress syndrome (ARDS) occurs due to either direct or indirect proinflammatory insults. However, only a proportion of at-risk patients develop ARDS, with research suggesting that genetic, age, social and other factors play a role in determining who develops ARDS.1 More than 1 billion people worldwide are believed to have vitamin D deficiency.2 Vitamin D has important functions besides bone and calcium homeostasis3 with cells of the innate and adaptive immune system responding to vitamin D. Vitamin D deficiency may therefore increase the risk of bacterial and viral infection. Vitamin D deficiency is associated with an increased risk of intensive care admission and mortality in patients with pneumonia.4 Deficiency is common in critically ill patients and associated with adverse outcome.3 Gram-positive bacteria, invasive pneumococcal disease and meningococcal disease are more common when 25(OH)D3 levels are low.5 Recent data from an Austrian study in critically ill deficient patients suggests that when treatment with vitamin D is successful in raising levels >75 nmol/L there is a mortality benefit.6 Vitamin D may improve outcomes by reducing both local and systemic inflammatory responses as a result of modulating cytokine responses.7 In a mouse model of lethal endotoxaemia, survival post intravenous lipopolysaccharide (LPS) was significantly poorer in the vitamin D receptor knockout mice.8

Vitamin D deficiency is associated with an increased risk of intensive care admission and mortality in patients with pneumonia.4 Deficiency is common in critically ill patients and associated with adverse outcome.3 Gram-positive bacteria, invasive pneumococcal disease and meningococcal disease are more common when 25(OH)D3 levels are low.5 Recent data from an Austrian study in critically ill deficient patients suggests that when treatment with vitamin D is successful in raising levels >75 nmol/L there is a mortality benefit.6 Vitamin D may improve outcomes by reducing both local and systemic inflammatory responses as a result of modulating cytokine responses.7 In a mouse model of lethal endotoxaemia, survival post intravenous lipopolysaccharide (LPS) was significantly poorer in the vitamin D receptor knockout mice.8 Our aim was to define the prevalence and severity of vitamin D deficiency in patients with ARDS and to establish whether vitamin D deficiency is a risk factor for and/or a driver of the exaggerated and persistent inflammation that is a hallmark of ARDS. To achieve these aims, we employed translational clinical studies and in vitro primary cell work and murine models. Methods Patient cohorts Patient cohort details are outlined in the online supplement but consisted of 52 patients with ARDS, 57 patients undergoing oesophagectomy (at risk of ARDS) and 8 patients undergoing oesophagectomy who had high-dose vitamin D supplementation prior to surgery.

Our aim was to define the prevalence and severity of vitamin D deficiency in patients with ARDS and to establish whether vitamin D deficiency is a risk factor for and/or a driver of the exaggerated and persistent inflammation that is a hallmark of ARDS. To achieve these aims, we employed translational clinical studies and in vitro primary cell work and murine models. Methods Patient cohorts Patient cohort details are outlined in the online supplement but consisted of 52 patients with ARDS, 57 patients undergoing oesophagectomy (at risk of ARDS) and 8 patients undergoing oesophagectomy who had high-dose vitamin D supplementation prior to surgery. Patients with ARDS: 52 patients who were recruited into the first beta agonist lung injury trial (BALTI-1) study9 and the translational sub-study of BALTI-2.10 There was no difference in age, sex, pre-enrolment Lung Injury score or acute physiology and chronic health evaluation (APACHE) II in these two groups of patients. Vitamin D levels were determined from ARDS patient plasma collected on the day of enrolment. In the oesophagectomy cohort, blood was collected on the day of the operation—pre any intervention. Aetiology of ARDS is outlined in table 1. As these cohorts of patients were diagnosed prior to the Berlin criteria, throughout the paper we have used ARDS to indicate patients meeting criteria for acute lung injury or ARDS according to the definition of the American European Consensus Conference.11 Table 1 Comparison of demographics between ARDS and at-risk patients who were undergoing oesophagectomy

Aetiology of ARDS is outlined in table 1. As these cohorts of patients were diagnosed prior to the Berlin criteria, throughout the paper we have used ARDS to indicate patients meeting criteria for acute lung injury or ARDS according to the definition of the American European Consensus Conference.11 Table 1 Comparison of demographics between ARDS and at-risk patients who were undergoing oesophagectomy ARDS (n=52) At risk (n=65) p Value Male, n (%) 30 (57) 57 (87.6) <0.001 Age (years), mean (SD) 61.3 (16.7) 62.8 (10.8) 0.560 Predisposing condition, n (%) Pneumonia 18 (35) Other sepsis 24 (46) Aortic aneurysm repair 3 (6) Chest trauma 2 (3.8) Pancreatitis 1 (1.9) Transfusion-related lung injury 1 (1.9) Other 3 (6) Oesophagectomy 65 (100) n/a LIS, median (IQR) 2.75 (2.50–3.19) 1.5 (1.0–2.0) <0.001 APACHE II median (IQR) 24 (19–28) 12 (8–14) <0.001 Worst P/F ratio during admission, mean (SD) 14.9 (5.2) 31.5 (9.9) <0.001 Hospital survival, n (%) 16 (30.8) 62 (95.4) <0.001 Length of hospital stay for survivors (days), median (IQR) 35 (16–49) 17 (10–28) 0.025 Statistical tests used are χ2 t test where data is normally distributed and Kruskal–Wallis for non-parametric data. APACHE, acute physiology and chronic health evaluation; ratio of arterial oxygen tension to the fraction of inspired oxygen (Pao2:Fio2), arterial oxygen tension: fractional inspired oxygen; ARDS, acute respiratory distress syndrome; LIS, lung injury score.

ARDS (n=52) At risk (n=65) p Value Male, n (%) 30 (57) 57 (87.6) <0.001 Age (years), mean (SD) 61.3 (16.7) 62.8 (10.8) 0.560 Predisposing condition, n (%) Pneumonia 18 (35) Other sepsis 24 (46) Aortic aneurysm repair 3 (6) Chest trauma 2 (3.8) Pancreatitis 1 (1.9) Transfusion-related lung injury 1 (1.9) Other 3 (6) Oesophagectomy 65 (100) n/a LIS, median (IQR) 2.75 (2.50–3.19) 1.5 (1.0–2.0) <0.001 APACHE II median (IQR) 24 (19–28) 12 (8–14) <0.001 Worst P/F ratio during admission, mean (SD) 14.9 (5.2) 31.5 (9.9) <0.001 Hospital survival, n (%) 16 (30.8) 62 (95.4) <0.001 Length of hospital stay for survivors (days), median (IQR) 35 (16–49) 17 (10–28) 0.025 Statistical tests used are χ2 t test where data is normally distributed and Kruskal–Wallis for non-parametric data. APACHE, acute physiology and chronic health evaluation; ratio of arterial oxygen tension to the fraction of inspired oxygen (Pao2:Fio2), arterial oxygen tension: fractional inspired oxygen; ARDS, acute respiratory distress syndrome; LIS, lung injury score. Pulse Contour Cardiac Output Monitoring (PiCCO) Extravascular lung water (EVLW) was measured using the single-indicator transpulmonary thermodilution system (PiCCO-II; Pulsion) as described previously.12 In our study, the coefficient of variance for this system was <7% for all parameters.

APACHE, acute physiology and chronic health evaluation; ratio of arterial oxygen tension to the fraction of inspired oxygen (Pao2:Fio2), arterial oxygen tension: fractional inspired oxygen; ARDS, acute respiratory distress syndrome; LIS, lung injury score. Pulse Contour Cardiac Output Monitoring (PiCCO) Extravascular lung water (EVLW) was measured using the single-indicator transpulmonary thermodilution system (PiCCO-II; Pulsion) as described previously.12 In our study, the coefficient of variance for this system was <7% for all parameters. Vitamin D status: 25(OH)D3 was measured by tandem mass spectroscopy using appropriate Vitamin D External Quality Assessment Scheme control. 1,25(OH)2D and vitamin D binding protein (VDBP) were measured by ELISA. Definition of vitamin D status is controversial, with different figures used throughout the literature, but for this study we have considered plasma 25-OH vitamin D3 levels <50 nmol/L as deficient and levels <20 nmol/L as severe deficiency.13 14 In addition, patients in the at-risk cohort with 25(OH) vitamin D3 <20 nmol/L had significantly lower 1,25(OH)2 vitamin D than patients with higher levels (<20 nmol/L 74 pmol/L vs >20 nmol/L 90 pmol/L, p=0.029). Murine cytokines were measured by multiplex array (R&D, UK) or ELISA as per the manufacturer's instruction. ATII cell isolation and culture ATII cells were extracted from lung resection specimens according to the methods described previously15 (see online supplementary material). Microarray analysis is outlined in the online supplementary material.

Murine cytokines were measured by multiplex array (R&D, UK) or ELISA as per the manufacturer's instruction. ATII cell isolation and culture ATII cells were extracted from lung resection specimens according to the methods described previously15 (see online supplementary material). Microarray analysis is outlined in the online supplementary material. Wound repair, proliferation and cell death assays were performed as described previously.16

ATII cell isolation and culture ATII cells were extracted from lung resection specimens according to the methods described previously15 (see online supplementary material). Microarray analysis is outlined in the online supplementary material. Wound repair, proliferation and cell death assays were performed as described previously.16 Mouse methods Wild-type (WT) C57Bl/6 mice were obtained from Harlan UK, Oxford, UK, and maintained at BMSU, Birmingham University, UK. Once weaned, vitamin D deficiency was induced in WT pups by feeding them a vitamin D-deficient chow (Harlan, USA) for 6 weeks pre-intra-tracheal (IT) LPS. 25(OH)-vitamin D was assessed by direct ELISA (ImmunDiagnostik, Germany). The LPS challenge model was performed as described previously.17 Briefly, mice are anaesthetised and 50 µg LPS (Sigma, UK) instilled by IT route as a model of direct lung injury. Mice were sacrificed at neutrophilic peak, 48 h post-LPS instillation, and bronchoalveolar lavage (BAL) performed with two washes of 0.6 mL phosphate buffered saline (PBS)/EDTA (200 nM) installations to determine the local effects on inflammation. Untreated controls were also analysed to determine lung parameters in vitamin D-deficient mice. BAL fluid (BALF) was assessed for cellular inflammation by cell count, neutrophilia and neutrophil apoptosis (flow cytometry), markers of epithelial damage BALF receptor for glycosylated endpoints (BALF RAGE), protein permeability index (ratio of BALF protein:plasma protein) as well as cytokines by luminex array (R&D systems, UK). Results represent mice from three separate experiments with at least four replicates per group. Oxygen saturations were measured at 48 h post-LPS and compared with WT mice given PBS by MouseOx II Plus (n=8 for each condition). All experiments were performed in accordance with UK laws with approval of local animal ethics committee.

esent mice from three separate experiments with at least four replicates per group. Oxygen saturations were measured at 48 h post-LPS and compared with WT mice given PBS by MouseOx II Plus (n=8 for each condition). All experiments were performed in accordance with UK laws with approval of local animal ethics committee. Statistics Data were analysed using SPSS for Windows 16.0 (SPSS, Chicago, Illinois, USA). Data were tested for normality and analysed by unpaired t tests or Mann–Whitney U test. Data are expressed as mean (SD) unless otherwise indicated. A χ2 or Fisher’s exact test was used to compare proportions.

esent mice from three separate experiments with at least four replicates per group. Oxygen saturations were measured at 48 h post-LPS and compared with WT mice given PBS by MouseOx II Plus (n=8 for each condition). All experiments were performed in accordance with UK laws with approval of local animal ethics committee. Statistics Data were analysed using SPSS for Windows 16.0 (SPSS, Chicago, Illinois, USA). Data were tested for normality and analysed by unpaired t tests or Mann–Whitney U test. Data are expressed as mean (SD) unless otherwise indicated. A χ2 or Fisher’s exact test was used to compare proportions. To test the hypothesis that low 25(OH)D levels are associated with the development of ARDS in the at-risk oesophagectomy cohort (N=65), we performed multivariable logistic regression with the exposure of interest being 25(OH)D3 level <20 nmol/L and ARDS as the outcome. Adjusted ORs were estimated by multivariable logistic regression models with inclusion of covariate terms chosen, a priori, thought to be plausibly associated with both 25(OH)D3 level and ARDS in the oesophagectomy patient cohort. We sought to build a parsimonious model that did not unnecessarily adjust for covariates that did not affect bias or the causal relation between exposure and outcome. Model calibration was assessed using the Hosmer–Lemeshow (HL) χ2 goodness-of-fit test and the accompanying p value. Bayesian information criterion and Akaike information criterion were also used to determine global model fit. Covariates included in the logistic regression model were age, gender, diagnosis, staging and pack-years smoked. The discriminatory ability for ARDS was quantified using the c-statistic. In all analyses, p values are two-tailed and values below 0.05 were considered statistically significant.

determine global model fit. Covariates included in the logistic regression model were age, gender, diagnosis, staging and pack-years smoked. The discriminatory ability for ARDS was quantified using the c-statistic. In all analyses, p values are two-tailed and values below 0.05 were considered statistically significant. Results Plasma vitamin D status in patients with or at risk of ARDS Patients with ARDS (100%) were vitamin D-deficient (plasma 25(OH)D3 <50 nmol/L). In total, 55 (96%) out of 57 unsupplemented oesophagectomy patients at risk of ARDS were deficient preoperatively but levels were higher than in patients with ARDS. Both patients at risk and patients with ARDS had significantly lower levels of 25(OH)D3 than normal controls (see figure 1). Demographics of patients groups based on vitamin D levels are illustrated in table 2. Table 2 Comparison of demographics between patients with severe deficiency, moderate deficiency and vitamin D supplemented at-risk patients undergoing oesophagectomy

Results Plasma vitamin D status in patients with or at risk of ARDS Patients with ARDS (100%) were vitamin D-deficient (plasma 25(OH)D3 <50 nmol/L). In total, 55 (96%) out of 57 unsupplemented oesophagectomy patients at risk of ARDS were deficient preoperatively but levels were higher than in patients with ARDS. Both patients at risk and patients with ARDS had significantly lower levels of 25(OH)D3 than normal controls (see figure 1). Demographics of patients groups based on vitamin D levels are illustrated in table 2. Table 2 Comparison of demographics between patients with severe deficiency, moderate deficiency and vitamin D supplemented at-risk patients undergoing oesophagectomy Patients with severe 25-OH vitamin D3 deficiency (n=25) Patients with moderate 25-OH vitamin D3 deficiency (n=32) Patients who received vitamin D supplementation (n=8) p Value Male, n (%) 21 (84) 28 (87.5) 8 (100) 0.706 Age, years median (IQR) 60.0 (52.0–68.5) 65.5 (54.5–72.0) 68.0 (63.8–71.3) 0.122 BMI median (IQR) 24.3 (20.6–27.9) 25.4 (21.7–28.3) 24.1 (22.0–26.6) 0.698 ASA median (IQR) 2.0 (2.0–2.0) 2.0 (2.0–2.0) 2.0 (2.0–2.75) 0.950 Postoperative P/F ratio 41.0 (34.3–53.3) 39.8 (32.0–52.0) 47.8 (39.4–50.8) 0.476 Preoperative plasma 25-OH vitamin D3 level (nmol/L) median (IQR) 13.7 (10.9–16.7) 27.6 (22.5–34.9) 66.9 (42.5–92.6) <0.001 All p values shown are Kruskal–Wallis tests from SPSS. ASA, American Society of Anaesthesiologists physical status classification system; BMI, body mass index.

Patients with severe 25-OH vitamin D3 deficiency (n=25) Patients with moderate 25-OH vitamin D3 deficiency (n=32) Patients who received vitamin D supplementation (n=8) p Value Male, n (%) 21 (84) 28 (87.5) 8 (100) 0.706 Age, years median (IQR) 60.0 (52.0–68.5) 65.5 (54.5–72.0) 68.0 (63.8–71.3) 0.122 BMI median (IQR) 24.3 (20.6–27.9) 25.4 (21.7–28.3) 24.1 (22.0–26.6) 0.698 ASA median (IQR) 2.0 (2.0–2.0) 2.0 (2.0–2.0) 2.0 (2.0–2.75) 0.950 Postoperative P/F ratio 41.0 (34.3–53.3) 39.8 (32.0–52.0) 47.8 (39.4–50.8) 0.476 Preoperative plasma 25-OH vitamin D3 level (nmol/L) median (IQR) 13.7 (10.9–16.7) 27.6 (22.5–34.9) 66.9 (42.5–92.6) <0.001 All p values shown are Kruskal–Wallis tests from SPSS. ASA, American Society of Anaesthesiologists physical status classification system; BMI, body mass index. Figure 1 Plasma 25(OH)D3 levels in acute respiratory distress syndrome (ARDS) versus at risk and normal controls. The horizontal bar represents the median, and the boxes represent IQRs. Vertical lines show minimum–maximum range. Fifty-two patients with ARDS, 57 at-risk patients undergoing oesophagectomy, 18 healthy controls. In at-risk patients, preoperative median plasma levels of 25(OH)D3 were significantly lower in those patients who were ventilated with ARDS postoperatively (ARDS 16.97 nmol/L vs no ARDS 25.46 nmol/L, p=0.014). Oesophagectomy patients with severe vitamin D deficiency (plasma 25(OH)D3 <20 nmol/L) had a 37.5% risk of postoperative lung injury as opposed to a 15% risk with vitamin D levels >20 nmol/L (figure 2).

in those patients who were ventilated with ARDS postoperatively (ARDS 16.97 nmol/L vs no ARDS 25.46 nmol/L, p=0.014). Oesophagectomy patients with severe vitamin D deficiency (plasma 25(OH)D3 <20 nmol/L) had a 37.5% risk of postoperative lung injury as opposed to a 15% risk with vitamin D levels >20 nmol/L (figure 2). Figure 2 Risk of postoperative acute respiratory distress syndrome in severe 25(OH)D3 deficiency versus less severe deficiency. Severe deficiency (n=25), less severe (n=32). In the at-risk oesophagectomy cohort, preoperative vitamin D status was the only measure to have a significant difference in oesophagectomy patients who develop lung injury postoperatively (see table 3). Table 3 Univariate analysis of predictors of postoperative ARDS in patients undergoing oesophagectomy

Figure 2 Risk of postoperative acute respiratory distress syndrome in severe 25(OH)D3 deficiency versus less severe deficiency. Severe deficiency (n=25), less severe (n=32). In the at-risk oesophagectomy cohort, preoperative vitamin D status was the only measure to have a significant difference in oesophagectomy patients who develop lung injury postoperatively (see table 3). Table 3 Univariate analysis of predictors of postoperative ARDS in patients undergoing oesophagectomy Patients with ARDS (n=15) Patients without ARDS (n=50) p Value Male, n (%) 14 (93) 43 (86) 0.448 Age (years), median (IQR) 61 (53–66) 66 (56–71) 0.304 BMI (kg/cm2), median (IQR) 25.2 (23.9–29.1) 24.8 (21.6–28.1) 0.460 FEV1 (L), median (IQR) 2.69 (2.28–3.50) 2.85 (2.38–3.3) 0.901 FVC (L), median (IQR) 4.3 (3.4–5.2) 4.1 (3.5–4.7) 0.450 Tumour type=adenocarcinoma, n (%) 12 (80) 35 (70) 0.511 Tumour stage, n (%)* T2 4 (27) 12 (24) 0.865 T3 11 (73) 37 (76) N0 3 (20) 12 (24) 0.719 N1–2 12 (80) 37 (76) Smoker, n (%) Current 6 (40) 11 (22) 0.304 Former 7 (47) 34 (68) Never 2 (13) 5 (10) Pack-years, median (IQR) 30 (20–40) 30 (15–45) 0.740 Plasma 25-OH vitamin D3 (nmol/L), Median (IQR) 16.97 (12.98–22.46) 25.46 (17.35–39.77) 0.014 Plasma 1,25(OH)2 vitamin D (pmol/L), median (IQR) 68 (47–91) 89 (76–109) 0.007 Only preoperative 25(OH)D3 and 1,25(OH)2D vitamin D were significantly different in univariate analysis. *No lung function available for seven patients, pack-years not available for two patients. One patient (without lung injury) had a benign tumour—not included in staging data.

Patients with ARDS (n=15) Patients without ARDS (n=50) p Value Male, n (%) 14 (93) 43 (86) 0.448 Age (years), median (IQR) 61 (53–66) 66 (56–71) 0.304 BMI (kg/cm2), median (IQR) 25.2 (23.9–29.1) 24.8 (21.6–28.1) 0.460 FEV1 (L), median (IQR) 2.69 (2.28–3.50) 2.85 (2.38–3.3) 0.901 FVC (L), median (IQR) 4.3 (3.4–5.2) 4.1 (3.5–4.7) 0.450 Tumour type=adenocarcinoma, n (%) 12 (80) 35 (70) 0.511 Tumour stage, n (%)* T2 4 (27) 12 (24) 0.865 T3 11 (73) 37 (76) N0 3 (20) 12 (24) 0.719 N1–2 12 (80) 37 (76) Smoker, n (%) Current 6 (40) 11 (22) 0.304 Former 7 (47) 34 (68) Never 2 (13) 5 (10) Pack-years, median (IQR) 30 (20–40) 30 (15–45) 0.740 Plasma 25-OH vitamin D3 (nmol/L), Median (IQR) 16.97 (12.98–22.46) 25.46 (17.35–39.77) 0.014 Plasma 1,25(OH)2 vitamin D (pmol/L), median (IQR) 68 (47–91) 89 (76–109) 0.007 Only preoperative 25(OH)D3 and 1,25(OH)2D vitamin D were significantly different in univariate analysis. *No lung function available for seven patients, pack-years not available for two patients. One patient (without lung injury) had a benign tumour—not included in staging data. ARDS, acute respiratory distress syndrome; BMI, body mass index.

Patients with ARDS (n=15) Patients without ARDS (n=50) p Value Male, n (%) 14 (93) 43 (86) 0.448 Age (years), median (IQR) 61 (53–66) 66 (56–71) 0.304 BMI (kg/cm2), median (IQR) 25.2 (23.9–29.1) 24.8 (21.6–28.1) 0.460 FEV1 (L), median (IQR) 2.69 (2.28–3.50) 2.85 (2.38–3.3) 0.901 FVC (L), median (IQR) 4.3 (3.4–5.2) 4.1 (3.5–4.7) 0.450 Tumour type=adenocarcinoma, n (%) 12 (80) 35 (70) 0.511 Tumour stage, n (%)* T2 4 (27) 12 (24) 0.865 T3 11 (73) 37 (76) N0 3 (20) 12 (24) 0.719 N1–2 12 (80) 37 (76) Smoker, n (%) Current 6 (40) 11 (22) 0.304 Former 7 (47) 34 (68) Never 2 (13) 5 (10) Pack-years, median (IQR) 30 (20–40) 30 (15–45) 0.740 Plasma 25-OH vitamin D3 (nmol/L), Median (IQR) 16.97 (12.98–22.46) 25.46 (17.35–39.77) 0.014 Plasma 1,25(OH)2 vitamin D (pmol/L), median (IQR) 68 (47–91) 89 (76–109) 0.007 Only preoperative 25(OH)D3 and 1,25(OH)2D vitamin D were significantly different in univariate analysis. *No lung function available for seven patients, pack-years not available for two patients. One patient (without lung injury) had a benign tumour—not included in staging data. ARDS, acute respiratory distress syndrome; BMI, body mass index. The odds of ARDS in patients with 25(OH)D3 <20 nmol/L was 3.5-fold that of patients with 25(OH)D3 ≥20 nmol/L (OR=3.5 (95% CI 1.06 to 11.6; p=0.040)). Following adjustment for gender, age, diagnosis, staging data, and pack-years, patients with 25(OH)D3 <20 nmol/L had a 4.2-fold higher odds of ARDS than patients with 25(OH)D <20 nmol/L (OR=4.2 (95% CI 1.13 to 15.9; p=0.032)). The adjusted model showed good calibration (HL χ2 11.10, p=0.20) and discrimination for ARDS (area under the curve 0.73). When 25(OH)D was analysed with logistic regression as a continuous exposure in 1 nmol/L increments, the odds of ARDS decreases by 17% for every 1 nmol/L decrease in 25(OH)D (OR 0.83 (95% CI 0.69 to 0.98; p=0.033)), adjusted for age, gender, diagnosis, staging and pack-years smoked.

for ARDS (area under the curve 0.73). When 25(OH)D was analysed with logistic regression as a continuous exposure in 1 nmol/L increments, the odds of ARDS decreases by 17% for every 1 nmol/L decrease in 25(OH)D (OR 0.83 (95% CI 0.69 to 0.98; p=0.033)), adjusted for age, gender, diagnosis, staging and pack-years smoked. Plasma levels of 1,25(OH)2D are lowest in ITU non-survivors Median plasma 1,25(OH)2D levels were also significantly lower in patients with ARDS (35.5 pmol/L) than in at-risk patients (85 pmol/L, p=0.0001). Plasma 1,25(OH)2D was lower at admission to intensive therapy unit (ITU) in patients who died than survivors (figure 3). Plasma levels of 1,25(OH)2D were lower in oesophagectomy patients at risk of ARDS who subsequently went on to be ventilated for ARDS (68 pmol/L (IQR 47–91)) than those who did not get postoperative ARDS (89 pmol/L (IQR 76–109), p=0.007). Figure 3 Plasma 1,25(OH)2D was significantly higher in patients with acute respiratory distress syndrome who survived at least 28 days following admission than those who died. The horizontal bar represents the median, and the boxes represent IQRs. Vertical lines show minimum–maximum range. Died (n=32), survived (n=20). Plasma VDBP levels are lower in patients with ARDS. 25(OH)D3 circulates tightly bound to the VDBP (also known as Gc-actin).18 VDBP levels were 40 mg/dL in normal controls, 19 mg/dL in ARDS and 28.7 mg/mL in the at-risk patients at the beginning of oesophagectomy (figure 4).

Figure 3 Plasma 1,25(OH)2D was significantly higher in patients with acute respiratory distress syndrome who survived at least 28 days following admission than those who died. The horizontal bar represents the median, and the boxes represent IQRs. Vertical lines show minimum–maximum range. Died (n=32), survived (n=20). Plasma VDBP levels are lower in patients with ARDS. 25(OH)D3 circulates tightly bound to the VDBP (also known as Gc-actin).18 VDBP levels were 40 mg/dL in normal controls, 19 mg/dL in ARDS and 28.7 mg/mL in the at-risk patients at the beginning of oesophagectomy (figure 4). Figure 4 Plasma vitamin D binding protein measured by ELISA in acute respiratory distress syndrome (ARDS) versus at risk and normal controls. Fifty-two patients with ARDS, 57 at-risk patients undergoing oesophagectomy, 18 healthy controls. Vitamin D levels and perioperative changes in epithelial integrity in patients at risk of ARDS We measured perioperative changes in an in vivo measure of the integrity of the alveolar–capillary barrier, namely EVLW accumulation (extravascular lung water index (EVLWI)) and pulmonary vascular permeability index (PVPI) using a PiCCO2 catheter19–22 and related this to the patient's vitamin D status. Severe vitamin D deficiency (25-(OH)D3 <20 nmol/L) was associated with an increased accumulation of EVLW as assessed by PiCCO EVLWI and evidence of increases of PVPI, a marker of alveolar capillary permeability. Patients supplemented with vitamin D prior to oesophagectomy had significantly reduced changes in PiCCO EVLWI and PVPI than unsupplemented patients (figures 5 and 6).

ociated with an increased accumulation of EVLW as assessed by PiCCO EVLWI and evidence of increases of PVPI, a marker of alveolar capillary permeability. Patients supplemented with vitamin D prior to oesophagectomy had significantly reduced changes in PiCCO EVLWI and PVPI than unsupplemented patients (figures 5 and 6). Figure 5 Changes in extravascular lung water index (EVLWI) at the end of oesophagectomy and on the morning of postoperative day 1. EVLWI was measured using Pulse Contour Cardiac Output Monitoring II catheter at the end of the operation and on the morning after the operation (day 1). Severe deficient (n=25), moderate (n=32) and supplemented (n=8). Figure 6 Changes in Pulse Contour Cardiac Output Monitoring pulmonary vascular permeability index (PVPI) at the end of oesophagectomy and the morning of postoperative day 1. Severe deficient (n=25), moderate (n=32) and supplemented (n=8). Vitamin D deficiency is a determinant of inflammation and epithelial injury in the intratracheal LPS murine model of ALI/ARDS We studied the response to 50 µg of IT LPS in WT or mice made vitamin D deficient by dietary manipulation. Deficient mice were fed a vitamin D-free diet for 6 weeks and had median plasma vitamin 25(OH)D3 levels of 8 nmol/L (SEM 1.15 nmol/L) vs 42 nmol/L (SEM 2.17 nmol/L) in WT (p=0.001). Untreated vitamin D-deficient mice had no observed lung damage or inflammation (figure 7, and data not shown).

nt by dietary manipulation. Deficient mice were fed a vitamin D-free diet for 6 weeks and had median plasma vitamin 25(OH)D3 levels of 8 nmol/L (SEM 1.15 nmol/L) vs 42 nmol/L (SEM 2.17 nmol/L) in WT (p=0.001). Untreated vitamin D-deficient mice had no observed lung damage or inflammation (figure 7, and data not shown). Figure 7 Lung injury and inflammation was significantly higher in vitamin D-deficient mice compared with wild-type (WT) following intra-tracheal (IT)-lipopolysaccharide (LPS). Levels of tumour necrosis factor-α and CXCL1/KC in UTCs were below the detection threshold of the assays performed. UTC, untreated control; N.D., not detected. Following LPS challenge, vitamin D-deficient mice had increased evidence of alveolar epithelial damage as measured by BALF RAGE and BALF permeability index (figure 7A). Cellular inflammation and neutrophil apoptosis in BALF were also elevated in vitamin D-deficient mice, along with release of proinflammatory cytokines tumour necrosis factor-α, CXCL1/KC and vascular endothelial growth factor (figure 7B and C, respectively). These changes resulted in significantly lower oxygen saturation as measured by pulse oximetry (figure 7C), which we have previously demonstrated as a physiological measure of murine lung function.15

tory cytokines tumour necrosis factor-α, CXCL1/KC and vascular endothelial growth factor (figure 7B and C, respectively). These changes resulted in significantly lower oxygen saturation as measured by pulse oximetry (figure 7C), which we have previously demonstrated as a physiological measure of murine lung function.15 Vitamin D is trophic for alveolar epithelial cells in vitro ATII cells were treated with 100 nmol/L of 25(OH)D3 for 24 h. Microarray analysis revealed that vitamin D treatment caused a sustained activation or inhibition of 660 genes that included pathways involved in vitamin metabolism as well as regulators of cell growth, differentiation and response to wounding (GEOSET record GSE46749). The online supplementary table SA and SB outline the top 25 genes up-regulated and down-regulated by vitamin D and a heat map illustrating up-regulated and down-regulated genes. Table 4 outlines the biological processes and molecular functions modified by vitamin D treatment. Several of the identified pathways had significant relevance to proliferation, wound repair and apoptosis, so we tested the functional effects of vitamin D upon these important repair/protective processes. Table 4 List of 30 statistically significant gene ontology (GO) terms implicated by differential expression of genes in day 3 epithelial (type II like) cells treated with vitamin D3 100 nM relative to untreated cells

Vitamin D is trophic for alveolar epithelial cells in vitro ATII cells were treated with 100 nmol/L of 25(OH)D3 for 24 h. Microarray analysis revealed that vitamin D treatment caused a sustained activation or inhibition of 660 genes that included pathways involved in vitamin metabolism as well as regulators of cell growth, differentiation and response to wounding (GEOSET record GSE46749). The online supplementary table SA and SB outline the top 25 genes up-regulated and down-regulated by vitamin D and a heat map illustrating up-regulated and down-regulated genes. Table 4 outlines the biological processes and molecular functions modified by vitamin D treatment. Several of the identified pathways had significant relevance to proliferation, wound repair and apoptosis, so we tested the functional effects of vitamin D upon these important repair/protective processes. Table 4 List of 30 statistically significant gene ontology (GO) terms implicated by differential expression of genes in day 3 epithelial (type II like) cells treated with vitamin D3 100 nM relative to untreated cells GO Annotated genes Total p Value 587 2230 Immune response 48 76 0.00000 Immune system process 58 103 0.00000 Cytokine activity 25 36 0.00001 Extracellular process 45 86 0.00003 Signal transducer activity 75 173 0.00008 Molecular transducer activity 75 173 0.00008 Plasma membrane 133 356 0.00013 DNA replication 20 29 0.00015 Receptor activity 57 124 0.00015 Defence response 42 83 0.00015 Monoxygenase activity 12 13 0.00018 ATPase activity, coupled to transmembrane movement of substances 6 107 (0.00018) Primary active transmembrane transporter activity 6 107 (0.00018) Hydrolase activity, acting on acid anhydrides, catalysing transmembrane movement of substances 6 107 (0.00018) P-P-bond-hydrolysis-driven transmembrane transporter activity 6 107 (0.00018) ATPase activity, coupled to movement of substances 6 107 (0.00018) Cell surface receptor linked signal transduction 69 162 0.00025 Chemotaxis 14 17 0.00027 Taxis 14 17 0.00027 Response to external stimulus 50 108 0.00030 Response to wounding 38 76 0.00043 Heme binding 12 14 0.00060 Tetrapyrrole binding 12 14 0.00060 Cellular biosynthetic process 15 145 (0.00105) Biosynthetic process 29 214 (0.00121) Extracellular region part 58 137 0.00168 Nucleotide biosynthetic process 1 61 (0.00168) Cell cycle process 44 97 0.00208 Nucleobase-containing small molecule metabolic process 3 69 (0.00412) Nucleotide metabolic process 3 68 (0.00466) p values of underrepresented GO terms are denoted in parentheses.

ss 29 214 (0.00121) Extracellular region part 58 137 0.00168 Nucleotide biosynthetic process 1 61 (0.00168) Cell cycle process 44 97 0.00208 Nucleobase-containing small molecule metabolic process 3 69 (0.00412) Nucleotide metabolic process 3 68 (0.00466) p values of underrepresented GO terms are denoted in parentheses. Effect of physiologically relevant doses of 25(OH)D3 upon primary human alveolar type II cells 25(OH)D3 at physiologically relevant concentrations stimulated scratch wound repair, cell proliferation and attenuated soluble Fas ligand (sFasL)-mediated cell death (see figures 8–10). Figure 8 Scratch wound repair response of primary human alveolar type II cells to 25(OH)D3. Wound area after 24 h was compared with baseline and expressed as fold change in wound area. Data represents experiments using cells from six separate lung resection specimens. Analysis of variance p=0.001. Figure 9 Proliferation of primary human ATII cells in response to physiological doses of 25(OH)D3 by bromodeoxyuridine incorporation. Experiments were performed using cells from four donors. Analysis of variance p=0.001. Figure 10 Cellular response to soluble Fas ligand (sFasL) 10 ng/mL induced cell death. Experiments were performed using ATII cells from four donors. 100 nmol/L 25(OH)D3 was added at the time of addition of sFasL.

Figure 9 Proliferation of primary human ATII cells in response to physiological doses of 25(OH)D3 by bromodeoxyuridine incorporation. Experiments were performed using cells from four donors. Analysis of variance p=0.001. Figure 10 Cellular response to soluble Fas ligand (sFasL) 10 ng/mL induced cell death. Experiments were performed using ATII cells from four donors. 100 nmol/L 25(OH)D3 was added at the time of addition of sFasL. Discussion We have assessed the vitamin D status of a large cohort of patients with ARDS and a well-characterised group of patients at risk of ARDS, namely patients undergoing oesophagectomy. In ARDS cases, vitamin D deficiency was ubiquitous. Survivors of ARDS had significantly higher levels of vitamin D than non-survivors. Our finding of a 30% reduction in VDBP in patients with ARDS supports a role for either reduced production or increased losses as an explanation for some of the degree of deficiency seen. Equally the low observed levels of circulating 1,25(OH)2D in patients with ARDS suggests a problem with renal metabolism as this is probably the major source of circulating 1,25(OH)2D.23

h ARDS supports a role for either reduced production or increased losses as an explanation for some of the degree of deficiency seen. Equally the low observed levels of circulating 1,25(OH)2D in patients with ARDS suggests a problem with renal metabolism as this is probably the major source of circulating 1,25(OH)2D.23 Several studies have suggested that vitamin D deficiency may be a risk factor for adverse outcome in pneumonia24 and lower respiratory tract infections in neonates.25 Other studies have suggested patients with sepsis have significant vitamin D deficiency.26 27 Our data suggests in the high-risk oesophagectomy group that vitamin D status is also a pre-existing risk factor for ARDS—especially when deficiency is severe. Patients undergoing oesophagectomy with severe preoperative vitamin D deficiency had greater risk of postoperative ARDS and increases in PiCCO measures of alveolar permeability than those with less severe deficiency. In our animal model of LPS-induced lung injury, vitamin D deficiency was associated with greater BALF cellular inflammation and cytokine release at 48 h. Increased epithelial damage and accumulation of apoptotic neutrophils was also evident. Mice that were vitamin D deficient became more hypoxic, suggesting physiologically worse lung injury.

LPS-induced lung injury, vitamin D deficiency was associated with greater BALF cellular inflammation and cytokine release at 48 h. Increased epithelial damage and accumulation of apoptotic neutrophils was also evident. Mice that were vitamin D deficient became more hypoxic, suggesting physiologically worse lung injury. Our animal data is in keeping with recently published data in hamsters treated with LPS.28 In contrast, Klaff et al found reduced neutrophil chemotactic potential to the chemokine KC ex vivo in mice deficient in vitamin D but no differences in LPS-induced BALF neutrophilia. They used a 72 h time point and a much lower dose of LPS (2.5 µg), which we suggest accounts for the differences with our study.

LPS.28 In contrast, Klaff et al found reduced neutrophil chemotactic potential to the chemokine KC ex vivo in mice deficient in vitamin D but no differences in LPS-induced BALF neutrophilia. They used a 72 h time point and a much lower dose of LPS (2.5 µg), which we suggest accounts for the differences with our study. In both our human at-risk patients and the murine model, we have demonstrated evidence of increased permeability of the alveolar capillary barrier in response to one lung ventilation (EVLWI and PVPI) and LPS challenge respectively (PPI) when severe deficiency is present, suggesting that vitamin D might have protective effects on the alveolar epithelium as well as being anti-inflammatory. 1,25(OH)2D has been shown to induce DNA incorporation in human alveolar type II cells,29 but the effects of physiologically relevant doses of 25(OH)D have not been addressed upon type II cells previously. To address whether 25(OH)D3 has effects on ATII cells, we demonstrated considerable functional activity of a physiological dose of 25(OH)D3 by microarray analysis. Physiologically relevant doses of 25(OH)D3 stimulated wound repair, cellular proliferation and reduced sFasL-induced cell death. These in vitro experiments suggest that 25(OH)D3 may play a trophic role on adult human alveolar epithelial cells.

ctional activity of a physiological dose of 25(OH)D3 by microarray analysis. Physiologically relevant doses of 25(OH)D3 stimulated wound repair, cellular proliferation and reduced sFasL-induced cell death. These in vitro experiments suggest that 25(OH)D3 may play a trophic role on adult human alveolar epithelial cells. This study has limitations. First, although we obtained blood from patients with ARDS as soon as possible following admission to ITU, we are unable to be sure that levels of 25(OH)D3 were low prior to the development of ARDS in that cohort or whether levels fall because of the development of ARDS. Second, our at-risk data from oesophagectomy patients has to be divided into severe deficiency and moderate deficiency because of the severity of vitamin D deficiency observed in that patient group. Third, our data in oesophagectomy patients’ needs validating in an additional significant cohort as well as in other patient groups at risk of ARDS. Finally, our comparison of EVLWI between our patients in BALTI prevention cohort versus our oesophagectomy patients in the open label vitamin D replacement study is a potential limitation. However, the translational protocol of assessments and the two centres in which those assessments were performed was the same between the two studies. We are currently conducting a randomised placebo controlled trial to confirm these results due to finish recruitment in mid-2015, which should further address this question.30

r, the translational protocol of assessments and the two centres in which those assessments were performed was the same between the two studies. We are currently conducting a randomised placebo controlled trial to confirm these results due to finish recruitment in mid-2015, which should further address this question.30 Taken together, these data suggest that vitamin D deficiency is ubiquitous in patients with ARDS and relates to adverse outcome. In patients undergoing oesophagectomy, severe preoperative deficiency is associated with evidence of increased alveolar epithelial damage and EVLW accumulation as well as an increased risk of postoperative lung injury. Novel in vitro data further suggest a trophic and antiapoptotic role of physiologically relevant doses of 25(OH)D3 upon primary adult human alveolar epithelial cells. Finally, preoperative restoration of vitamin D levels in patients with oesophageal cancer who are at risk of ARDS resulted in significantly less accumulation of EVLW than in unsupplemented patients postoperatively. In conclusion, we suggest that clinical strategies should be developed to replete vitamin D levels in patients at risk of ARDS and this approach might also have value as a treatment for established ARDS. Supplementary Material Web supplement We would like to thank the staff and patients of the Queen Elizabeth Hospital Birmingham and Heart of England NHS trust for their help in recruiting to and taking part in these studies. We would like to thank Teresa Melody and Amy Bradley for trial nurse support for the study.

In conclusion, we suggest that clinical strategies should be developed to replete vitamin D levels in patients at risk of ARDS and this approach might also have value as a treatment for established ARDS. Supplementary Material Web supplement We would like to thank the staff and patients of the Queen Elizabeth Hospital Birmingham and Heart of England NHS trust for their help in recruiting to and taking part in these studies. We would like to thank Teresa Melody and Amy Bradley for trial nurse support for the study. Correction notice: This article has been corrected since it was published Online First. The author's name William M Fraser was incorrect and should be William D Fraser. Contributors: RCAD and DP are joint first authors. DRT, FG, ARM, PMS, MSC, WMF and GDP designed the study. SL, VD, SZ, CRB, DP, BN, RM, AMT and ES recruited the patients and undertook lab analysis/animal work. WW, PMS, WMF, KBC and MSC provided expert advice in their specialist areas. All authors contributed to the writing of the paper. DT is the guarantor of the data. Funding: These studies were funded by the Wellcome trust (DRT, PMS, MSC, SL), QEHB charities (RCAD, VD, DRT), the Medical Research Council UK (DRT, DP, RCAD) and the NIHR (DP, DRT, GDP). DB was funded by an ERS long-term training fellowship and a Marie Curie Intra-European Fellowship. Competing interests: None declared. Patient consent: Obtained Ethics approval: NHS LREC West midlands and all clinical investigations were conducted according to Declaration of Helsinki principles. Provenance and peer review: Not commissioned; externally peer reviewed.

Funding: These studies were funded by the Wellcome trust (DRT, PMS, MSC, SL), QEHB charities (RCAD, VD, DRT), the Medical Research Council UK (DRT, DP, RCAD) and the NIHR (DP, DRT, GDP). DB was funded by an ERS long-term training fellowship and a Marie Curie Intra-European Fellowship. Competing interests: None declared. Patient consent: Obtained Ethics approval: NHS LREC West midlands and all clinical investigations were conducted according to Declaration of Helsinki principles. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: The microarray dataset outlined in this paper is freely available to search on PubMed GEOSET.

Key messages What is the key question? What underlies impaired interferon responses to rhinovirus in asthmatics. What is the bottom line? Suppression of Toll-like receptor (TLR)7 by interleukin-5-induced lung eosinophilia impairs interferon responses to rhinovirus, leading to exaggerated inflammatory responses and exacerbation. Why read on? This provides a mechanism whereby therapies modulating TLR7 or targeting eosinophils may ameliorate virus-induced asthma exacerbations.

What is the bottom line? Suppression of Toll-like receptor (TLR)7 by interleukin-5-induced lung eosinophilia impairs interferon responses to rhinovirus, leading to exaggerated inflammatory responses and exacerbation. Why read on? This provides a mechanism whereby therapies modulating TLR7 or targeting eosinophils may ameliorate virus-induced asthma exacerbations. Introduction Asthma is a complex and heterogenic inflammatory disease of the airways with increasing global prevalence. The most common trigger for asthma symptoms is immune cell activation against innocuous antigens (allergens) and respiratory viral infections.1 Upon antigen exposure, cytokines such as thymic stromal lymphopoietin (TSLP), granulocyte-macrophage colony-stimulating factor, interleukin (IL)-25, IL-33 and tumour necrosis factor-related apoptosis-inducing ligand are released by the airway epithelium, resulting in the activation of innate immune cells that promote T helper 2 (TH2) cell differentiation, leading to the release of TH2 cytokines such as IL-4, IL-5 and IL-13.1 2 IL-13 is a potent inducer of airway hyper reactivity (AHR) and mucus production in a signal transducer and activator of transcription-6 (STAT6)-dependent manner.3 IL-5 regulates maturation of eosinophils in the bone marrow and in concert with chemokines such as eotaxins the recruitment of these cells into the airways.4 Thus, TH2 cell activation underpins many clinical phenotypes including allergic and eosinophilic asthma.5 The proportion of asthmatics with high eosinophil numbers in their airways represent the majority of patients, and some of those unresponsive to corticosteroids have severe therapy-refractory asthma with a disproportionally large burden of disease. Importantly, they are also prone to asthma exacerbations, which can be partially alleviated by therapeutics that block IL-4, IL-5 or IL-13 and reduce eosinophilic inflammation in the lungs.5 However, the molecular mechanisms that link TH2-induced eosinophilia with susceptibility to exacerbation are yet to be defined.

. Importantly, they are also prone to asthma exacerbations, which can be partially alleviated by therapeutics that block IL-4, IL-5 or IL-13 and reduce eosinophilic inflammation in the lungs.5 However, the molecular mechanisms that link TH2-induced eosinophilia with susceptibility to exacerbation are yet to be defined. Viral respiratory infections are detected in up to 85% of asthma exacerbations and two-thirds of those are caused by rhinoviruses (RV).6 Notably a group of asthmatics have impaired release of innate interferons (IFNα, IFNβ and IFNλ) upon experimental RV infection and IFNλ inversely correlated with induced sputum eosinophils on day 3 of acute infection.7 8 IFNλ 2/3 inversely correlated with induced sputum eosinophils on day 3 of the acute infection (r=−0.53, p=0.05), when subjects were experimentally infected with RV16 in vivo.8 RV is sensed by a limited number of pattern recognition receptors, including Toll-like receptors (TLRs) and retinoid acid-induced gene 1 like receptors. TLR3 and TLR7 are endosomally localised and recognise viral nucleic acids, critically regulating antiviral IFN production. Recent studies have shown that impaired TLR3 function does not affect RV replication in vivo and TLR3 expression is not reduced in asthmatics.9 10 The in vivo role of functional TLR7 signalling in mounting antiviral responses to RV has yet to be determined. TLR7 is widely expressed in innate immune cells such as dendritic cells (DCs) and macrophages, as well as in structural lung cells, including airway epithelia.11 Treatment of mice with TLR7 agonists leads to a long-lasting protection from the development of allergic airways disease (AAD) and TLR7 activation also reduced airway smooth muscle contractility in mice.12 We and others have found reduced IFNα and IFNλ release upon TLR7 stimulation of peripheral blood mononuclear cells from asthmatics and a trend towards lower TLR7 expression was observed in bronchoalveolar lavage fluid (BALF) cells from asthmatics.9 However, the role of TLR7 in RV-induced exacerbation of AAD has yet to be fully elucidated.

ound reduced IFNα and IFNλ release upon TLR7 stimulation of peripheral blood mononuclear cells from asthmatics and a trend towards lower TLR7 expression was observed in bronchoalveolar lavage fluid (BALF) cells from asthmatics.9 However, the role of TLR7 in RV-induced exacerbation of AAD has yet to be fully elucidated. Methods Patient biopsies Endobronchial biopsies were obtained by bronchoscopy with samples taken from third-generation bronchi. Biopsies were stored in RNALater (Ambion) at −20°C until needed. RNA was extracted following the miRNeasy Mini Handbook (Qiagen)—purification of total RNA from tissue via homogenisation (Qiashredder). RNA was quantified by spectrophotometry (NanoDrop) and 200 ng of extracted RNA reverse transcribed using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems; ABI). The inflammatory phenotype was determined by induced sputum count, and patients with a RAST ImmunoCAP Specific IgE-positive blood test were designated as atopic. Animals Wild-type (WT), Tlr7−/−, Tlr4−/−, MyD88−/−, Stat6−/− and IL-5 transgenic mice, all on a BALB/c background (6–14 weeks of age), were obtained from Australian BioResources (Moss Vale, Australia) and housed in approved containment facilities within the HMRI Building, University of Newcastle (Newcastle, Australia). Mice had ad libitum access to food and water under a 12 h light and dark cycle. All experiments were approved by the Animal Care and Ethics Committee of the University of Newcastle.

s Vale, Australia) and housed in approved containment facilities within the HMRI Building, University of Newcastle (Newcastle, Australia). Mice had ad libitum access to food and water under a 12 h light and dark cycle. All experiments were approved by the Animal Care and Ethics Committee of the University of Newcastle. Induction of AAD and RV-induced exacerbation AAD and RV-induced exacerbation were performed as previously described.13 Briefly, lyophilised crude house dust mite (HDM) (Dermatophagoides pteronyssinus) extract (Greer Laboratories) was resuspended in sterile saline (SAL) and intranasally delivered to mice under light isoflurane anaesthesia. Sensitisation on days 0, 1 and 2 (HDM 50 µg/50 µL) was followed by daily challenge (HDM 5 µg/50 µL) on days 14, 15, 16 and 17 to induce AAD. On day 18, mice were euthanased via pentobarbital sodium overdose (Virbac) and samples collected. In other experiments, mice were intranasally infected with live minor group RV (RV1B), 50 µL containing 5×106 virions median tissue culture infective dose (TCID50) or UV-inactivated RV1B 24 h after last HDM exposure to exacerbate pre-existing AAD. Samples were collected 24 h after RV1B infection.

and samples collected. In other experiments, mice were intranasally infected with live minor group RV (RV1B), 50 µL containing 5×106 virions median tissue culture infective dose (TCID50) or UV-inactivated RV1B 24 h after last HDM exposure to exacerbate pre-existing AAD. Samples were collected 24 h after RV1B infection. Administration of recombinant cytokines and LPS Naive mice were intranasally administered recombinant mouse IL-13 (15 µg/35 µL; Biolegend), IL-5 (15 µg/35 µL; Biolegend) or a low or high dose (0.12 µg/50 µL and 10 µg/50 µL ,respectively) of lipopolysaccharide (LPS) (Escherichia coli, 0111:B4; Sigma-Aldrich), with samples collected 24 h later. HDM-sensitised and challenged Tlr7−/− mice, which also received live RV1B, were intranasally administered either recombinant mouse IFNα2 (10 000 IU/50 µL; eBioscience), IFNβ1 (10 000 IU/50 µL; Biolegend), IFNλ2 (1 µg/50 µL; R&D Systems)14 or a vehicle control (phosphate buffered saline) on day 18 (2 h following RV1B infection). Mice were sacrificed 24 h postinfection and samples collected. AHR measurement AHR was measured as previously described.15 Briefly, AHR was invasively assessed in separate groups of anaesthetised mice by measurement of total lung resistance and dynamic compliance (Buxco). Mice were mechanically ventilated, and AHR to nebulised methacholine (increased lung resistance) was expressed as a percentage change from control (baseline).

escribed.15 Briefly, AHR was invasively assessed in separate groups of anaesthetised mice by measurement of total lung resistance and dynamic compliance (Buxco). Mice were mechanically ventilated, and AHR to nebulised methacholine (increased lung resistance) was expressed as a percentage change from control (baseline). Analysis of lung inflammation BALF was collected and analysed as previously described.16 Enumeration of peribronchial/perivascular eosinophils and PAS-positive cells was performed as previously described.17 Flow cytometry Single lung cell suspensions were prepared and stained as previously described.18 Antibodies used were FITC-anti-TCRβ chain (BD, cat. no. 553171, clone H57-597), PE-anti-CD4 (BD, cat. no. 553652, clone H129.19), PerCP-anti-CD8a (BD, cat. no. 561092, clone 53-6.7), PerCP-Cy5.5-anti-CD11b (BD, cat. no. 561092, clone M1/70), FITC-anti-CD11c (BD, cat. no. 553801, clone HL3), PE-anti-MHCII (eBioscience, cat. no. 12-5321, clone M5/114.15.2), all at 1:15 dilution. Positive cells were identified using a FACSCanto (BD) by the following criteria: mDCs—CD11b+ CD11c+ MHCII+; T cells—TCRβ chain+ with CD4+ or CD8a+. Data analysed with BD FACsdiva.

), FITC-anti-CD11c (BD, cat. no. 553801, clone HL3), PE-anti-MHCII (eBioscience, cat. no. 12-5321, clone M5/114.15.2), all at 1:15 dilution. Positive cells were identified using a FACSCanto (BD) by the following criteria: mDCs—CD11b+ CD11c+ MHCII+; T cells—TCRβ chain+ with CD4+ or CD8a+. Data analysed with BD FACsdiva. Quantitative RT-PCR Trachea and lungs were extracted from euthanased mice and forceps used to separate the airways from the parenchyma by blunt dissection.19 Total mRNA was extracted using TRIzol (Ambion; Carlsbad, USA). cDNA was generated via reverse transcription using BioScript (Bioline; Alexandria, Australia). Quantitative PCRs (qPCR) were performed on cDNA generated from mouse airway tissue and human endobronchial biopsies with SYBR Green (Invitrogen; Mulgrave, Australia) using primers detailed in online supplementary table S1. CTs of the genes of interest were referenced HPRT or GAPDH for mouse and human tissue, respectively. All steps were performed according to manufacturer's instructions. Generation of BM-derived pDCs and adoptive transfer WT and TLR7-deficient bone marrow-derived plasmacytoid DCs (pDCs) were generated as previously described.20 pDCs were resuspended at 1×107 cells/mL and 0.5×106 pDC were administered intranasally to allergic Tlr7−/− mice, 2 h prior to RV1B infection.

Quantitative RT-PCR Trachea and lungs were extracted from euthanased mice and forceps used to separate the airways from the parenchyma by blunt dissection.19 Total mRNA was extracted using TRIzol (Ambion; Carlsbad, USA). cDNA was generated via reverse transcription using BioScript (Bioline; Alexandria, Australia). Quantitative PCRs (qPCR) were performed on cDNA generated from mouse airway tissue and human endobronchial biopsies with SYBR Green (Invitrogen; Mulgrave, Australia) using primers detailed in online supplementary table S1. CTs of the genes of interest were referenced HPRT or GAPDH for mouse and human tissue, respectively. All steps were performed according to manufacturer's instructions. Generation of BM-derived pDCs and adoptive transfer WT and TLR7-deficient bone marrow-derived plasmacytoid DCs (pDCs) were generated as previously described.20 pDCs were resuspended at 1×107 cells/mL and 0.5×106 pDC were administered intranasally to allergic Tlr7−/− mice, 2 h prior to RV1B infection. Quantification of lung cytokines and IFNs A single lung lobe from each mouse was excised and snap frozen before being homogenised in buffers recommended in the manufacturer's instructions. Levels of IL-5, IL-13, IFNγ (BD Biosciences; PharMingen), IFNα, IFNβ (Verikine; PBL Assay Science), CCL7/MCP3 and IFNλ2/3 (R&D Systems) were determined in clarified lung lysates by ELISA. CCL2/MCP1, CCL3/MIP1α, CCL4/MIP1β, CXCL9/MIG and CXCL10/IP10 were measured by employing a Multiplex Immunoassay (Millipore). All concentrations were normalised to lung weight.

; PharMingen), IFNα, IFNβ (Verikine; PBL Assay Science), CCL7/MCP3 and IFNλ2/3 (R&D Systems) were determined in clarified lung lysates by ELISA. CCL2/MCP1, CCL3/MIP1α, CCL4/MIP1β, CXCL9/MIG and CXCL10/IP10 were measured by employing a Multiplex Immunoassay (Millipore). All concentrations were normalised to lung weight. pDC culture and TCID50 Primary mouse pDCs were isolated from pooled WT or TLR7-deficient spleens via mechanical dissociation and isolation with PDCA-1 magnetic beads on an AutoMACS platform (Miltenyi Biotec) according to manufacturer's instructions. Following two isolation runs, positively selected cells were seeded into 96-well plates at 1×106 cells/well and cultured in the absence or presence of RV1B (multiplicity of infection =5).20 Supernatant was harvested 48 h postinfection and IFN production assessed with ELISA. Number of live RV1B virions was determined by standard TCID50 serial dilution on Hela-H1 cells, calculated via the Spearman–Karber method. Statistical analysis The significance of differences between groups was analysed using Student's t test, Mann–Whitney test, analysis of variance or Kruskal–Wallis with Dunn's test for multiple comparisons as appropriate using Graphpad Prism 6. A value of p<0.05 is reported as significant. Data are expressed as mean±SEM.

Statistical analysis The significance of differences between groups was analysed using Student's t test, Mann–Whitney test, analysis of variance or Kruskal–Wallis with Dunn's test for multiple comparisons as appropriate using Graphpad Prism 6. A value of p<0.05 is reported as significant. Data are expressed as mean±SEM. Results To investigate TLR7 as a regulator of AAD and antiviral responses in vivo, we repeatedly challenged WT and TLR7-deficient (Tlr7−/−) mice with HDM, then superimposed a RV infection to exacerbate their established AAD as described previously.15 Exposure of allergic Tlr7−/− mice to RV, compared with WT controls, resulted in impaired production of type I (α, β), II (γ) and III (λ) IFNs with higher RV replication in the lower airways (figure 1A,B). This was despite a lack of an increase in type II (γ) and III (λ) IFNs to RV infection in allergic WT mice. Deficient IFN responses coincided with exaggerated AHR (figure 1C), eosinophilic airways inflammation (figure 1D,E), accumulation of CD4+, CD8+ and myeloid (m) DCs but not pDCs in the lungs (figure 1F), and production IL-5 and CCL11 (eotaxin-1) (figure 1G). We also observed increased levels of the TH2-priming cytokines IL-25 and TSLP in allergic Tlr7−/− mice infected with RV (see online supplementary figure S1). CCL2, CCL3, CCL4 and CCL7 but not CXCL9 and CXCL10 were also increased in the absence of TLR7 signalling (see online supplementary figure S2). There was no change in IL-13 levels or numbers of mucus-producing cells (see online supplementary figure S3a), suggesting a novel and critical role of TLR7 signalling in RV-induced asthma exacerbation.

7 but not CXCL9 and CXCL10 were also increased in the absence of TLR7 signalling (see online supplementary figure S2). There was no change in IL-13 levels or numbers of mucus-producing cells (see online supplementary figure S3a), suggesting a novel and critical role of TLR7 signalling in RV-induced asthma exacerbation. Figure 1 Toll-like receptor (TLR)7-deficient mice experience exaggerated rhinovirus (RV)-induced exacerbation. Allergic wildtype (WT) and TLR7-deficient mice were inoculated with live or UV-inactivated RV1B, samples collected 24 h post infection. Levels of interferons (A) and interleukin (IL)-5 (G) in clarified lung homogenates as assessed by ELISA. (B) Lung viral titre in infected mice was quantified via quantitative RT-PCR (qRT-PCR) and 50% tissue culture infectious dose (TCID50). (C) Total lung resistance presented as percentage change in response to methacholine (n=4–6 mice per group). Peribronchial/perivascular eosinophils and cellular infiltrates in bronchoalveolar lavage (BAL) fluid enumerated by light microscopy (D and E). (F) Numbers of lung mDCs and CD4+/CD8+ T cells quantified by flow cytometry. (G) CCL11 expression in lower airway tissue enumerated by qRT-PCR and expressed as % increase over WT house dust mite (HDM)+UV group. Results are mean±SEM (n=3–6 mice per group) and are representative of two independent experiments. *p<0.05, **p<0.01, ***p<0.001 as compared to strain-matched HDM+UV group or otherwise indicated determined by students t test except for (C) where analysis of variance was used to compare RI curves.

t mite (HDM)+UV group. Results are mean±SEM (n=3–6 mice per group) and are representative of two independent experiments. *p<0.05, **p<0.01, ***p<0.001 as compared to strain-matched HDM+UV group or otherwise indicated determined by students t test except for (C) where analysis of variance was used to compare RI curves. Allergic Tlr7−/− mice received recombinant type I and III IFNs 2 h after infection with RV. Notably, one dose of IFNα2, β or λ2 suppressed RV-induced eosinophilic inflammation (figure 2A), as well as IL-5 and CCL11 production (figure 2B) but not IL-13 (see online supplementary figure S3b). All IFN treatments induced IFN-γ in the lung (figure 2C) and limited RV replication (figure 2D). There was no effect of IFN treatment on TSLP, IL-25 or IL-33 expression (data not shown). Thus IFN treatment promotes IFN-γ release and impairs IL-5 and CCL11 production and RV replication in the absence of TLR7.

gure S3b). All IFN treatments induced IFN-γ in the lung (figure 2C) and limited RV replication (figure 2D). There was no effect of IFN treatment on TSLP, IL-25 or IL-33 expression (data not shown). Thus IFN treatment promotes IFN-γ release and impairs IL-5 and CCL11 production and RV replication in the absence of TLR7. Figure 2 Exogenous interferon can protect Toll-like receptor (TLR)7-deficient mice from rhinovirus (RV)-induced exacerbation. Exacerbated TLR7-deficient mice received recombinant interferon (IFN)α2 (α), IFNβ (β), IFNλ2 (λ) or a vehicle control (−) 2 h post-RV1B infection on day 18. Samples were collected 24 h post infection. (A) Eosinophils present in bronchoalveolar lavage (BAL) fluid and per 100 µm2 of lung tissue. Levels of interleukin (IL)-5 (B) and interferons (C) in clarified lung homogenates as assessed by ELISA. CCL11 expression (B) and viral RV1B RNA (D) in lower airway tissue was quantified by quantitative RT-PCR. Results are mean±SEM (n=3–7 mice per group) and are representative of two independent experiments. *p<0.05, **p<0.01, ***p<0.001 as compared to vehicle control group determined by Student's t test.

sessed by ELISA. CCL11 expression (B) and viral RV1B RNA (D) in lower airway tissue was quantified by quantitative RT-PCR. Results are mean±SEM (n=3–7 mice per group) and are representative of two independent experiments. *p<0.05, **p<0.01, ***p<0.001 as compared to vehicle control group determined by Student's t test. pDCs release large quantities of IFN during viral infection, and here we show that splenic pDCs exposed to RV in vitro display impaired release of IFN-α and IFN-β but not λ2/3 in the absence of TLR7 (figure 3A). Adoptive transfer of TLR7-expressing pDCs into allergic Tlr7−/− mice re-established IFNα, IFNβ and IFNγ but not IFNλ release in the lungs upon RV infection and limited RV replication as compared to allergic Tlr7−/− mice that received TLR7-deficient pDCs (figure 3B, C). Transfer of WT pDCs also limited RV-induced AHR (figure 3D), eosinophilic airways inflammation (figure 3E) and production of IL-5 and CCL11 (figure 3F) but not IL-13 release (see online supplementary figure S3c). Thus adoptive transfer of TLR7-expressing pDCs—like exogenous IFN treatment—promotes IFNγ release and impairs eosinophilic airways inflammation and RV replication in the absence of TLR7.

ys inflammation (figure 3E) and production of IL-5 and CCL11 (figure 3F) but not IL-13 release (see online supplementary figure S3c). Thus adoptive transfer of TLR7-expressing pDCs—like exogenous IFN treatment—promotes IFNγ release and impairs eosinophilic airways inflammation and RV replication in the absence of TLR7. Figure 3 Adoptive transfer of Toll-like receptor (TLR)7-competent plasmacytoid dendritic cells (pDCs) to TLR7-deficient mice limits exacerbation of allergic airways disease (AAD). Purified Flt3-L-expanded pDCs from TLR7-deficient (−/−) and TLR7-competent (+/+) bone marrow were adoptively transferred to allergic Tlr7−/− recipients. Mice were inoculated with RV1B 2 h later and endpoints measured 24 h post infection. (A) Spleen-isolated pDCs were infected in vitro with RV1B and interferon (IFN) release in cell supernatants assessed by ELISA. Levels of IFNs (B), as well as interleukin (IL)-5 (F) in clarified lung homogenates as assessed by ELISA. Positive-strand RV1B RNA (C) and CCL11 expression (F) from lower airway tissue quantified by quantitative RT-PCR. (D) Total lung resistance presented as percentage change in response to methacholine (n=7–8 mice per group). (E) Eosinophils present in bronchoalveolar lavage (BAL) fluid. Results are mean±SEM (n=3–5 mice per group). *p<0.05, **p<0.01, ***p<0.001 determined by Student's t test except for (D) where analysis of variance compared the RI curves. All p values as compared to exacerbated mice that received TLR7-deficient pDCs.

oup). (E) Eosinophils present in bronchoalveolar lavage (BAL) fluid. Results are mean±SEM (n=3–5 mice per group). *p<0.05, **p<0.01, ***p<0.001 determined by Student's t test except for (D) where analysis of variance compared the RI curves. All p values as compared to exacerbated mice that received TLR7-deficient pDCs. TLR7 expression was assessed in a number of knockout mice strains as it was suppressed in allergic WT mice with TH2-mediated AAD (figure 4A). Notably this allergen-induced reduction in TLR7 expression was not observed in TLR4 (Tlr4−/−), MyD88 (Myd88−/−) or STAT6 (Stat6−/−)-deficient mice, suggesting that intact TH2-promoting signalling pathways and the presence of eosinophilic airways inflammation are required for suppression of TLR7. To investigate the specific role of TH2 cytokines in this in vivo observation, we delivered one dose of recombinant IL-13 or IL-5 intranasally to WT mice and compared those responses to ones challenged with LPS (a TLR4 agonist). Interestingly, in mice that constitutively overexpress IL-5 (IL-5Tg/Tg mice) TLR7 expression was significantly reduced (figure 4B) and was associated with accumulation of eosinophils in the airways in the absence of allergy (figure 4D). Intranasal administration of one dose of IL-5 or IL-13, however, had no effect on TLR7 expression or eosinophil recruitment although IL-13 did induce AHR and increased expression of Muc5AC (data not shown). In addition, one low or high dose of LPS increased TLR7 expression and resulted in the accumulation of neutrophils but not eosinophils in the lungs (figure 4C,D). Thus eosinophilic airways inflammation due to chronic IL-5 release is associated with a reduction in TLR7 expression while a single administration of IL-5 or IL-13 had no effects on TLR7.

ose of LPS increased TLR7 expression and resulted in the accumulation of neutrophils but not eosinophils in the lungs (figure 4C,D). Thus eosinophilic airways inflammation due to chronic IL-5 release is associated with a reduction in TLR7 expression while a single administration of IL-5 or IL-13 had no effects on TLR7. Figure 4 Toll-like receptor (TLR)7 expression is reduced during eosinophilic lung inflammation. (A) Wildtype (WT), TLR4−/−, MyD88−/− and Stat6−/− mice were sensitised and challenged with house dust mite (HDM) over 18 days and gene expression of TLR7 in lower airway tissue was quantified by quantitative RT-PCR. TLR7 airway expression in allergic airways as a percentage of sterile saline (SAL) expression for each strain, (B) as well as lung tissue eosinophils and (C) bronchoalveolar lavage (BAL) fluid neutrophils and eosinophils (D) from rIL-5, rIL-13, IL-5 Tg and lipopolysaccharide (LPS)-treated mice 24 h post treatment. Results are mean±SEM (n=4–6 mice per group), gene expression in mice expressed as a % compared to non-allergic SAL-treated wildtype mice. TLR7 (E) and interferon (IFN)λ2/3 (F) expression in bronchial biopsies collected from non-asthmatic (n=13) and asthmatic (n=20) subjects stratified into eosinophilic (>3% eosinophils) or non-eosinophilic (<3% eosinophils) phenotypes based on BAL cell counts. (G) TLR7 expression in bronchial biopsies collected from non-asthmatic (n=13) and asthmatic (n=20) subjects stratified according to atopic status. Lines indicate the median, boxes extend from 25th to the 75th percentile, and error bars extend to 10th and 90th percentiles. (H) Correlation between TLR7 expression from bronchial biopsies and percentage of sputum eosinophils. (I–K) Correlations between TLR7 and IFN expression from patient biopsies. *p<0.05, **p<0.01, ***p<0.001 as determined by Student's t test (A–D) or Kruskal–Wallis with Dunn's multiple comparisons test.

90th percentiles. (H) Correlation between TLR7 expression from bronchial biopsies and percentage of sputum eosinophils. (I–K) Correlations between TLR7 and IFN expression from patient biopsies. *p<0.05, **p<0.01, ***p<0.001 as determined by Student's t test (A–D) or Kruskal–Wallis with Dunn's multiple comparisons test. We next analysed TLR7 expression in bronchial biopsies collected from healthy subjects and patients with moderate-to-severe persistent asthma (clinical and demographic data in online supplementary table S2). Importantly, asthmatics with an eosinophilic airways inflammation, as determined by bronchial lavage,21 displayed significantly reduced TLR7 (figure 4E) and IFNλ2/3 expression (figure 4F), independent of atopic status (figure 4G). Furthermore, levels of TLR7 expression positively correlated with IFNα and λ2/3 expression (figure 4I–K) and inversely correlated with percentage of sputum eosinophils (figure 4H) but not macrophages or neutrophils (rs=0.04; p=0.81; n=33 and rs=0.31; p=0.08; n=33, respectively). These results suggest that suppression of TLR7 and IFN expression in moderate-to-severe asthmatics is specifically tied to eosinophilic airways inflammation in this cohort.

th percentage of sputum eosinophils (figure 4H) but not macrophages or neutrophils (rs=0.04; p=0.81; n=33 and rs=0.31; p=0.08; n=33, respectively). These results suggest that suppression of TLR7 and IFN expression in moderate-to-severe asthmatics is specifically tied to eosinophilic airways inflammation in this cohort. Discussion Impaired innate IFN responses to respiratory viruses have been proposed as one mechanism underlying the clinical observation of asthmatics being susceptible to RV-induced exacerbation.7 8 Innate IFN responses are instigated by a limited number of pattern recognition receptors, such as TLR7 whose cognate ligands include the RV viral genome.10 11 Experiments in human immortalised epithelial cell lines (BEAS-2B) have supported a role of TLR3 in mediating an antiviral and anti-inflammatory response.9 However, TLR3 expression in the airways22 and TLR3-induced responses did not vary in PBMCs derived from asthmatic or atopic patients by comparison to healthy subjects.22 We show here, for the first time in vivo, that a lack of TLR7 signalling under conditions that model a viral asthma exacerbation leads to impaired IFN production and exaggerated TH2-driven inflammation. Our findings, such as increased levels of the TH2-priming cytokines IL-25, IL-33 and TSLP in Tlr7−/− mice, mirror those published by Kaiko et al,23 infecting non-allergic mice with mouse pneumovirus, a mouse pathogen similar to human respiratory syncytial virus. These studies suggest that intact TLR7 signalling is required for sufficient IFN induction and restrainment of proinflammatory responses in the lung.

Tlr7−/− mice, mirror those published by Kaiko et al,23 infecting non-allergic mice with mouse pneumovirus, a mouse pathogen similar to human respiratory syncytial virus. These studies suggest that intact TLR7 signalling is required for sufficient IFN induction and restrainment of proinflammatory responses in the lung. As TLR7 activation is upstream of IFN production, we hypothesised that administration of exogenous IFN would protect allergic Tlr7−/− mice from RV-induced exacerbation. We observed that treatment with type I or III IFNs resulted in the Tlr7−/− mice developing a suppressed phenotype similar to that of the WT controls, whose TLR7 signalling is active. All IFN treatments induced the release of IFNγ in the lung, with type I and III IFNs promoting themselves exclusively, indicative of their known distinct signalling pathways.14

or III IFNs resulted in the Tlr7−/− mice developing a suppressed phenotype similar to that of the WT controls, whose TLR7 signalling is active. All IFN treatments induced the release of IFNγ in the lung, with type I and III IFNs promoting themselves exclusively, indicative of their known distinct signalling pathways.14 RV-induced pDC-derived IFN production was dependent upon intact TLR7 signalling in vitro. Additionally, adoptive transfer of TLR7-expressing pDCs into allergic Tlr7−/− hosts re-established type I and II IFN responses, limiting RV replication and eosinophilic airways inflammation. Notably TLR7 deficiency did not impair pDC recruitment into the lungs. This implicates activation of TLR7 signalling on pDCs and increasing host IFN release as important therapeutic strategies, either through the use of TLR7 agonists or targeting pathways upstream of IFN production. Other IFN-producing cells may also be relevant. We have recently shown that anti-CCL7 treatment is sufficient to inhibit IRF-7-dependent IFNβ expression in RV infection, which was associated with reduced macrophage inflammation but not pDC influx.24 This highlights the complexity of the inflammatory and antiviral immune response on a cellular level, particularly in an allergic environment.

hat anti-CCL7 treatment is sufficient to inhibit IRF-7-dependent IFNβ expression in RV infection, which was associated with reduced macrophage inflammation but not pDC influx.24 This highlights the complexity of the inflammatory and antiviral immune response on a cellular level, particularly in an allergic environment. WT allergic mice with TH2-driven AAD had significantly lower levels of TLR7 expression compared with non-allergic SAL-treated mice in the absence of RV infection, which is in line with the pattern seen clinically in non-infected eosinophilic asthmatics. TLR4 signalling is required for the development of a robust TH2-mediated allergic airways inflammation in response to HDM extract, which also contains low amounts of endotoxins.19 Allergic airways inflammation suppresses TLR7 expression, which is prevented by disruption of TH2-promoting signalling pathways governed by TLR4, MyD88 and STAT6. In contrast to allergic mice, TLR4 activation by LPS led to an upregulation of TLR7 expression in non-allergic mice. Thus the effects of TLR4 signalling on the regulation of TLR7 expression are determined by the presence or absence of TH2-dominant allergic airways inflammation. Interestingly, LPS also upregulated TLR3 expression in human monocytes, which was critical for antiviral responses.25 TLR3−/− mice infected with RV1B, however, displayed normal type I IFN responses, unchanged viral titres and reduced inflammatory responses.10 This is in marked contrast to our data generated in allergic TLR7−/− mice and previous data in non-allergic TLR7−/− mice.20 Further studies are now required to elucidate the clinical effect of endotoxins on TLR7-mediated responses in asthma.

I IFN responses, unchanged viral titres and reduced inflammatory responses.10 This is in marked contrast to our data generated in allergic TLR7−/− mice and previous data in non-allergic TLR7−/− mice.20 Further studies are now required to elucidate the clinical effect of endotoxins on TLR7-mediated responses in asthma. In the absence of TH2-dominant allergic airways inflammation, no suppression of TLR7 expression was observed by a single exposure to IL-13 or IL-5. However, chronic exposure to IL-5 alone, which is associated with lung eosinophilia and the development of AHR,26 markedly impaired TLR7 lung expression in the absence of allergy. This finding was of clinical relevance because we observed reduced TLR7 expression in endobronchial biopsies from asthmatic patients with eosinophilic inflammatory profiles. Eosinophilic asthmatics also expressed lower levels of innate IFNs in addition to TLR7, a difference that was not observed for expression of TLR3, retinoic acid-inducible gene I or melanoma differentiation-associated gene 5.27 These results are congruent with a recent study that mapped TLR7 expression in the airways of severe asthmatics.28

cs also expressed lower levels of innate IFNs in addition to TLR7, a difference that was not observed for expression of TLR3, retinoic acid-inducible gene I or melanoma differentiation-associated gene 5.27 These results are congruent with a recent study that mapped TLR7 expression in the airways of severe asthmatics.28 Our results suggest a reciprocal regulation between IL-5-induced eosinophilia and TLR7 expression that affects antiviral IFN responses to RV (summary illustrated in figure 5). In the absence of virus, it has been shown that TLR7 stimulation with synthetic agonists in vitro resulted in inhibition of IL-5 through IFNγ and Notch signalling pathways in antigen-presenting cells29 and upregulation of IFNγ production by memory CD4+ T cells30 and natural killer cells.31 Consistent with these findings, we show impaired IFNγ release in the absence of TLR7 promotes TH2 immune responses, which can be rescued by type I and III IFN therapy or adoptive transfer of TLR7-sufficient pDCs. This is directly relevant to subjects with moderate-to-severe asthma with a predominantly eosinophilic airways inflammation as these individuals have suppressed—but not fully deficient—TLR7 expression in their lungs, which correlated with reduced IFN expression. Notably, a human monoclonal antibody to IL-5 (mepolizumab) reduced exacerbation frequency in moderate-to-severe asthmatics with a predominantly eosinophilic airways inflammation,32 a strategy that may be of specific benefit to that clinical population as not all asthmatics appear to exhibit impaired IFN responses to RV.33 Our data suggest that TLR7 and its downstream signalling pathway limits TH2 responses in RV-induced asthma exacerbations and may be a promising therapeutic target for the prevention and treatment of viral exacerbation in eosinophilic asthmatics.

on as not all asthmatics appear to exhibit impaired IFN responses to RV.33 Our data suggest that TLR7 and its downstream signalling pathway limits TH2 responses in RV-induced asthma exacerbations and may be a promising therapeutic target for the prevention and treatment of viral exacerbation in eosinophilic asthmatics. Figure 5 Proposed role and regulation of Toll-like receptor (TLR)7 in rhinovirus infection and allergic airways disease. IL, interleukin; RV, rhinovirus. Supplementary Material Web supplement Web table 1 Web table 2 We would like to thank S Akira, Osaka University, for providing Tlr4−/−, MyD88−/− and TLR7−/− mice and thank L Dent for providing IL-5Tg/Tgmice. We appreciate the technical assistance from A Pereira de Siqueira, C Cesar de Souza Alves, J Grehan, H Macdonald, M Morten, L Sokulsky and the staff from the animal care facility of the University of Newcastle.

ka University, for providing Tlr4−/−, MyD88−/− and TLR7−/− mice and thank L Dent for providing IL-5Tg/Tgmice. We appreciate the technical assistance from A Pereira de Siqueira, C Cesar de Souza Alves, J Grehan, H Macdonald, M Morten, L Sokulsky and the staff from the animal care facility of the University of Newcastle. Contributors: LH and AC contributed equally, performed and designed mouse and cell culture experiments, analysed data, generated figures and edited the manuscript. JG and JL performed experiments and analysed data. JZ assisted in supervision. PABW and KP performed and supervised studies on healthy subjects and subjects with asthma, collected and processed biopsies, and performed cell culture experiments. SP assisted in the design and conceptualisation of some mouse experiments. DK supervised and interpreted cell culture experiments. NWB and SLJ assisted in design of mouse experiments, provided RV1B for further propagation and cDNA standards. PSF assisted in design, supervision and interpretation of mouse studies. JM conceptualised, coordinated, designed and supervised mouse and human studies, interpreted and analysed data, and drafted and edited the manuscript. All authors contributed to data discussion and revised the manuscript.

ation and cDNA standards. PSF assisted in design, supervision and interpretation of mouse studies. JM conceptualised, coordinated, designed and supervised mouse and human studies, interpreted and analysed data, and drafted and edited the manuscript. All authors contributed to data discussion and revised the manuscript. Funding: This study was supported by the National Health and Medical Research Council (NH & MRC 1011153) (JM, PSF), the Hunter Medical Research Institute (JM, PABW, PSF), the Hunter Children's Research Foundation (JM, PSF) and an NH&MRC Health Practitioner Research Fellowship to JM (455623). NWB was supported by a project grant from Asthma UK (06-050) and SLJ by a Chair from Asthma UK (CH1155). This work was supported in part by MRC Centre Grant G1000758 and ERC FP7 Advanced Grant 233015 (to SLJ). Competing interests: SLJ has received consulting fees from GlaxoSmithKline, Chiesi, Boehringer Ingelheim and Novartis. PABW and SLJ are authors on patents relating to use of IFNs in treatment of exacerbations of airway disease and holds share options in Synairgen, a company developing IFNs for treatment of exacerbations of airway disease. Patient consent Obtained. Ethics approval: Hunter New England Area Health Service Ethics Committee and the University of Newcastle Committee Safety Committee (205/2008). Provenance and peer review: Not commissioned; externally peer reviewed.

Key messages What is the key question? How do three different antibiotics (moxifloxacin, doxycycline and azithromycin) compare in their effects on total airway bacterial load in stable COPD? What is the bottom line? Airway bacterial load was similar to placebo after treatment with all three antibiotics, although increases in antimicrobial resistance were noted in all treatment arms. Why read on? This is the first trial to directly compare the effects of different antibiotic classes on airway bacterial load in stable COPD. Introduction COPD exacerbations are important events associated with poorer health status,1 lung function decline2 and mortality.3 Their prevention is an important goal of COPD treatment4 5 which is not currently met by existing inhaled therapies.6 7 Exacerbation frequency is linked to airway bacterial colonisation during the stable state,8 and long-term antibiotic therapy has been proposed to prevent exacerbations. The best evidence for this comes from the macrolide antibiotics, with both erythromycin9 and azithromycin10 shown to significantly reduce exacerbation frequency although at the expense of auditory decrements and a possible increase in cardiovascular risk.11 There is less evidence for other antibiotic classes in stable COPD, although the fluoroquinolone moxifloxacin has shown some efficacy when used in a pulsed dosing regimen,12 and with no comparative studies there are few data to inform the optimum antibiotic choice. There are also growing concerns regarding the development of antibiotic resistance in airway bacteria.13

s in stable COPD, although the fluoroquinolone moxifloxacin has shown some efficacy when used in a pulsed dosing regimen,12 and with no comparative studies there are few data to inform the optimum antibiotic choice. There are also growing concerns regarding the development of antibiotic resistance in airway bacteria.13 Although the presumptive mode of action of antibiotics is via reducing airway bacterial load, no trials to date have examined this in stable COPD. Furthermore, molecular techniques are now important tools in accurately quantifying the airway microbiome14 and may be able to detect more subtle changes with therapy than traditional culture-based methods. For the first time, we aimed to directly compare the effects of different antibiotics on airway bacterial load in patients with stable COPD, measured both by culture and 16S quantitative PCR (qPCR). We also examined important secondary clinical endpoints, bacterial resistance and changes in airway inflammation. Methods Study design This was an exploratory 13-week, single-centre, single-blind, placebo controlled, randomised controlled trial conducted at the Royal Free Hospital (London, UK). We compared the efficacy of pulsed moxifloxacin 400 mg daily for 5 days every 4 weeks, doxycycline 100 mg daily, azithromycin 250 mg three times per week, or placebo one tablet daily, at reducing bacterial numbers in spontaneously expectorated sputum after 13 weeks of treatment. Research ethics approval was granted by King's College Regional Ethics Committee (reference 11/LO/0932).

Methods Study design This was an exploratory 13-week, single-centre, single-blind, placebo controlled, randomised controlled trial conducted at the Royal Free Hospital (London, UK). We compared the efficacy of pulsed moxifloxacin 400 mg daily for 5 days every 4 weeks, doxycycline 100 mg daily, azithromycin 250 mg three times per week, or placebo one tablet daily, at reducing bacterial numbers in spontaneously expectorated sputum after 13 weeks of treatment. Research ethics approval was granted by King's College Regional Ethics Committee (reference 11/LO/0932). Participants Subjects were recruited from primary and secondary care; further details can be found in the online supplementary data. Screening and all study visits took place in the outpatient department at the Royal Free Hospital, London. Sputum processing and culture, sensitivity testing and inflammatory marker analysis were conducted in the laboratories of the Centre for Respiratory Medicine and the Centre for Clinical Microbiology, Royal Free Campus, University College London. DNA extraction and 16S qPCR were conducted in the laboratories of the National Heart and Lung Institute, Imperial College London.

inflammatory marker analysis were conducted in the laboratories of the Centre for Respiratory Medicine and the Centre for Clinical Microbiology, Royal Free Campus, University College London. DNA extraction and 16S qPCR were conducted in the laboratories of the National Heart and Lung Institute, Imperial College London. Inclusion and exclusion criteria Stable patients aged ≥45 years with chronic bronchitis (self-reported sputum expectoration on most days when clinically stable) and spirometrically confirmed COPD (defined by FEV1<80% predicted, FEV1 to FVC ratio <0.7 and a history of smoking) were included. Patients who reported either treatment for an exacerbation or an episode of symptom worsening in the 4 weeks prior to screening, or were unable to enrol for safety reasons, were excluded. Detailed inclusion and exclusion criteria can be found in the online supplementary data. Screening and run-in period After clinical assessment, spirometry was performed in accordance with American Thoracic Society/European Respiratory Society guidance15 using a Vitalograph Gold Standard spirometer (Vitalograph, Maids Morton, UK). Spontaneously expectorated sputum was collected for analysis prior to randomisation. Patients also completed the St George's Respiratory Questionnaire (SGRQ) to assess health status. Patients received diary cards and recorded any worsening of their respiratory symptoms each day for a 1-week run-in period; these have previously been shown to be a reliable index of exacerbation.16

prior to randomisation. Patients also completed the St George's Respiratory Questionnaire (SGRQ) to assess health status. Patients received diary cards and recorded any worsening of their respiratory symptoms each day for a 1-week run-in period; these have previously been shown to be a reliable index of exacerbation.16 Randomisation and masking Patients returned to the clinic 1 week after screening. The diary cards and screening results were checked and details sought regarding any adverse events. If the subject remained eligible and clinically stable after reassessment, and had been able to provide a sputum sample for analysis by the second visit, then internet randomisation into groups of 1:1:1:1 was performed using a computer-generated permuted block system of variable sizes (Sealed Envelope, UK). Patients remained blinded to treatment allocation. Follow-up of patients Patients were telephoned by the study doctor during weeks 5 and 9 of treatment and any adverse events or exacerbations recorded. Patients who suffered a COPD exacerbation during the trial were treated either by their treating physician or using self-administered treatment as appropriate; exacerbation treatment during the study was not standardised. To avoid drug interactions, the study medication was temporarily stopped if the patient took other antibiotics for any reason. Details of exacerbations were recorded on the daily diary cards and the study team notified.

very limited clinical validation, and no clinical validation has been performed in patients with CF. Moreover, limited pharmacokinetic (PK) data are now available for MAC lung disease to support breakpoint concentrations,122 there are no representative PK or pharmacodynamic data to guide treatment of patients with CF. Breakpoints for clarithromycin susceptibility of MAC have been validated in HIV-related disseminated MAC disease and in retrospective series of MAC lung disease.119 123 124 Since the presence of macrolide resistance predicts worse clinical outcomes125 126 and requires augmented treatment,126 susceptibility to macrolides should be tested on isolates prior to treatment initiation and during treatment in refractory cases defined as those individuals who (1) fail to culture convert after 6 months of NTM treatment; (2) reculture MAC after initial culture conversion while on NTM treatment or (3) reculture MAC after completion of NTM treatment. A very recent study has shown that amikacin MICs >64 mg/L are measured only in MAC isolates that have mutations associated with amikacin resistance, that is, in the 16S rRNA gene. These strains are cultured from patients with significant aminoglycoside exposure, such as individuals with CF, and for disease caused by these strains, amikacin is unlikely to have any beneficial effect.127

administered treatment as appropriate; exacerbation treatment during the study was not standardised. To avoid drug interactions, the study medication was temporarily stopped if the patient took other antibiotics for any reason. Details of exacerbations were recorded on the daily diary cards and the study team notified. Patients attended the clinic at the end of 13 weeks of treatment. A further spontaneously expectorated sputum sample was collected and the SGRQ administered, diary cards collected and any exacerbations or adverse events recorded. Unused medication was collected and pill counts performed for adherence analysis. Outcome measures The primary outcome was the change in sputum bacterial load, as assessed by quantitative culture. Prespecified secondary outcomes were changes in resistance to the three tested antibiotics, changes in FEV1, adherence to therapy, health status as measured by total SGRQ scores and adverse events. Further exploratory outcomes assessed changes in sputum bacterial load as assessed by 16S rRNA gene targeted qPCR and changes in sputum inflammation. Laboratory analysis of sputum A minimum of 0.25 g spontaneously expectorated sputum was analysed using a modification of the method described by Pye et al.17 Mean inhibitory concentrations (MICs) were established for each isolate against each of the three antibiotics by Etest (bioMérieux UK, Basingstoke, UK) and isolates were defined as sensitive, intermediate or resistant where breakpoints were available.

Laboratory analysis of sputum A minimum of 0.25 g spontaneously expectorated sputum was analysed using a modification of the method described by Pye et al.17 Mean inhibitory concentrations (MICs) were established for each isolate against each of the three antibiotics by Etest (bioMérieux UK, Basingstoke, UK) and isolates were defined as sensitive, intermediate or resistant where breakpoints were available. DNA was isolated from sputum and qPCR for 16S bacterial ribosomal RNA gene carried out using a ViiA 7 real-time PCR system (Life Technologies, Paisley, UK). Sputum supernatant was batch analysed for the cytokines interleukin (IL)-6, IL-8 and IL-1β using commercial high-sensitivity sandwich ELISA kits (RD Systems, Abingdon, UK). The lower limits of detection were 0.70, 3.5 and <1.0 pg/mL, respectively. Further detail on these methods is included in the online supplementary data.

Sputum supernatant was batch analysed for the cytokines interleukin (IL)-6, IL-8 and IL-1β using commercial high-sensitivity sandwich ELISA kits (RD Systems, Abingdon, UK). The lower limits of detection were 0.70, 3.5 and <1.0 pg/mL, respectively. Further detail on these methods is included in the online supplementary data. Statistical analysis Pre-existing comparative data showing the effect of antibiotics on sputum bacterial numbers using quantitative culture methods were unavailable. Using PCR analysis, data from within our group showed that, in patients demonstrating bacterial colonisation on ≥2 occasions, mean (SD) sputum bacterial load when stable was 6.89 (1.23) log10 cfu/mL, falling to 5.01 (3.71) after antibiotic treatment at exacerbation. Assuming a correlation of 0.5 between the baseline and 3-month follow-up measurements, we calculated that 44 patients in each of the four groups would be sufficient to have a 90% chance of detecting, as significant at the 5% level, a difference between untreated (placebo) and post-treatment bacterial load of 1.88 log10 cfu/mL.

a correlation of 0.5 between the baseline and 3-month follow-up measurements, we calculated that 44 patients in each of the four groups would be sufficient to have a 90% chance of detecting, as significant at the 5% level, a difference between untreated (placebo) and post-treatment bacterial load of 1.88 log10 cfu/mL. Analysis was by intention to treat. The primary endpoint of total bacterial numbers in sputum (the sum of the bacterial load for each isolate cultured from a sample) was analysed using multiple regression after log10 transformation to assess the treatment effect for each antibiotic compared with placebo. Where growth was below the limit of detection (6×106 cfu/mL), patients were treated as having a left-censored outcome and a parametric survival regression model was used. Spirometry, health status (SGRQ) and adherence to therapy were modelled using linear, Poisson or logistic regression as appropriate. Antibiotic resistance was modelled using a linear mixed effects model for log(MIC) to account for multiple detected isolates within each individual; a generalised mixed effects model was used to analyse resistance as a binary outcome. All analyses were adjusted for baseline values, with multivariable analyses further adjusted for age, sex, current smoking status, FEV1% predicted and previous exacerbation history. Exacerbations during the study period were defined using diary card criteria as previously reported18 as well as patient reporting to clinical health professionals or the study team, with the final decision made by consensus. A further methodological analysis compared the measurement techniques of quantitative culture and 16S qPCR using the Bland–Altman method19 for all samples where both results were available.

rted18 as well as patient reporting to clinical health professionals or the study team, with the final decision made by consensus. A further methodological analysis compared the measurement techniques of quantitative culture and 16S qPCR using the Bland–Altman method19 for all samples where both results were available. Results Patient characteristics at recruitment Two hundred and forty-eight patients were screened between February 2012 and May 2013; 99 were randomised, and sputum was analysed for the primary endpoint from 86. Figure 1 shows the screening, randomisation and follow-up of patients. Patient characteristics at baseline are summarised in table 1. This cohort of patients was representative of the population with moderate–severe COPD, with a male preponderance, overall mean age of 69·5 years and mean FEV1 50·5% predicted. Table 1 Patient characteristics at baseline (mean (SD) unless stated)

Results Patient characteristics at recruitment Two hundred and forty-eight patients were screened between February 2012 and May 2013; 99 were randomised, and sputum was analysed for the primary endpoint from 86. Figure 1 shows the screening, randomisation and follow-up of patients. Patient characteristics at baseline are summarised in table 1. This cohort of patients was representative of the population with moderate–severe COPD, with a male preponderance, overall mean age of 69·5 years and mean FEV1 50·5% predicted. Table 1 Patient characteristics at baseline (mean (SD) unless stated) Treatment group Moxifloxacin Doxycycline Azithromycin Placebo Total (n) 25 25 25 24 Gender (n, % male) 17 (68) 18 (72) 16 (64) 18 (75) Age (years) 70.9 (8.2) 70.4 (7.0) 67.9 (8.6) 68.7 (9.8) BMI (kg/m2) 26.3 (5.2) 28.4 (6.4) 26.6 (6.9) 26.9 (4.9) Current smoker, n (%) 16 (64) 10 (40) 7 (28) 8 (33) Pack-years 53 (27) 52 (50) 51 (25) 56 (50) Number of exacerbations in previous year* 2.5 (2.1) 2.1 (1.7) 2.8 (4.0) 1.5 (1.4) Inhaled corticosteroid use, n (%) 21 (84) 19 (76) 18 (72) 13 (57) Bacterial load, log10 cfu/mL† 8.3 (0.8) 8.1 (0.7) 8.1 (0.8) 7.8 (0.7) Bacterial load, log10 16S copies/g sputum 9.4 (0.8) 9.3 (0.73) 9.0 (0.6) 9.1 (0.8) FEV1 (L) 1.4 (0.5) 1.5 (0.5) 1.2 (0.5) 1.5 (0.6) FEV1:FVC 0.51 (0.10) 0.51 (0.13) 0.45 (0.12) 0.51 (0.12) FEV1, % predicted 52 (13) 53 (14) 44 (17) 53 (13) FVC (L) 2.8 (1.1) 3.0 (1.1) 2.7 (0.7) 3.0 (1.0) SGRQ: total score 51 (14) 47 (16) 48 (18) 46 (19) SGRQ: symptom score 64 (16) 62 (24) 59 (18) 55 (19) SGRQ: activity score 67 (21) 62 (19) 66 (25) 61 (24) SGRQ: impact score 37 (12) 35 (16) 35 (18) 34 (20) Il-1β, log10 pg/mL 2.3 (0.6) 1.9 (0.7) 2.2 (0.8) 2.1 (0.7) IL-6, log10 pg/mL 1.9 (0.7) 1.5 (0.8) 1.8 (0.8) 1.6 (0.6) IL-8, log10 pg/mL 4.3 (0.7) 3.9 (0.9) 4.1 (0.7) 3.8 (0.8) *Self-reported exacerbation frequency.

9) SGRQ: activity score 67 (21) 62 (19) 66 (25) 61 (24) SGRQ: impact score 37 (12) 35 (16) 35 (18) 34 (20) Il-1β, log10 pg/mL 2.3 (0.6) 1.9 (0.7) 2.2 (0.8) 2.1 (0.7) IL-6, log10 pg/mL 1.9 (0.7) 1.5 (0.8) 1.8 (0.8) 1.6 (0.6) IL-8, log10 pg/mL 4.3 (0.7) 3.9 (0.9) 4.1 (0.7) 3.8 (0.8) *Self-reported exacerbation frequency. †For those patients with culture results above the threshold of detection. BMI, body mass index; IL, interleukin; SGRQ, St George's Respiratory Questionnaire. Figure 1 CONSORT diagram for this study showing screening, patient recruitment and data flow for the primary endpoint. Primary endpoint: bacterial numbers In an analysis of bacterial load, adjusted only for baseline values, there was a mean change of −0·32 log10 cfu/mL (95% CI −0.81 to 0.17, p=0.20) from placebo in the moxifloxacin arm, −0·05 (95% CI −0.50 to 0.40, p=0.82) in the doxycycline arm and −0.17 (95% CI −0.62 to 0.29, p=0.47) in the azithromycin arm. The largest change in bacterial numbers was therefore seen with moxifloxacin, equivalent to a 62% reduction, although this was not statistically significant at the 5% level. The smallest change was seen with azithromycin. These results are summarised in table 2. Table 2 Primary and secondary outcome measures for this study

Primary endpoint: bacterial numbers In an analysis of bacterial load, adjusted only for baseline values, there was a mean change of −0·32 log10 cfu/mL (95% CI −0.81 to 0.17, p=0.20) from placebo in the moxifloxacin arm, −0·05 (95% CI −0.50 to 0.40, p=0.82) in the doxycycline arm and −0.17 (95% CI −0.62 to 0.29, p=0.47) in the azithromycin arm. The largest change in bacterial numbers was therefore seen with moxifloxacin, equivalent to a 62% reduction, although this was not statistically significant at the 5% level. The smallest change was seen with azithromycin. These results are summarised in table 2. Table 2 Primary and secondary outcome measures for this study Drug Estimated change from placebo Baseline adjusted estimate (95% CI), p value Fully adjusted* estimate (95% CI), p value Bacterial load by quantitative culture (log10 cfu/mL) Moxifloxacin −0.32 (−0.81 to 0.17) p=0.20 −0.42 (−0.91 to 0.08) p=0.10 Doxycycline −0.05 (−0.50 to 0.40), p=0.82 −0.11 (−0.55 to 0.33), p=0.62 Azithromycin −0.17 (−0.62 to 0.29), p=0.47 −0.08 (−0.54 to 0.38), p=0.73 Bacterial load by 16S qPCR (log10 copies/g of sputum) Moxifloxacin 0.14 (−0.42 to 0.69), p=0.63 0.30 (−0.30 to 0.89), p=0.33 Doxycycline 0.06 (−0.49 to 0.60), p=0.84 0.16 (−0.40 to 0.72), p=0.58 Azithromycin 0.28 (−0.27 to 0.83), p=0.33 0.32 (−0.24 to 0.88), p=0.27 FEV1, mL Moxifloxacin 39 (−84 to 161), p=0.54 58 (−74 to 190), p=0.39 Doxycycline 31 (−89 to 151), p=0.61 39 (−85 to 162), p=0.54 Azithromycin 1 (−123 to 124), p=0.99 −1 (−126 to 125), p=0.99 SGRQ total score Moxifloxacin −2.25 (−8.60 to 4.09), p=0.49 −1.88 (−8.59 to 4.84), p=0.59 Doxycycline 0.88 (−5.30 to 7.06), p=0.78 1.02 (−5.28 to 7.31), p=0.75 Azithromycin −2.35 (−8.44 to 3.73), p=0.45 −2.29 (−8.43 to 3.86), p=0.47 Adherence to treatment. OR (95% CI) Moxifloxacin 0.89 (0.05 to 15.09), p=0.94 0.74 (0.04 to 15.61), p=0.85 Doxycycline 0.64 (0.05 to 9.02), p=0.74 0.60 (0.04 to 9.68), p=0.72 Azithromycin 1.34 (0.07 to 26.06), p=0.84 2.42 (0.09 to 63.28), p=0.60 Exacerbation frequency.

.44 to 3.73), p=0.45 −2.29 (−8.43 to 3.86), p=0.47 Adherence to treatment. OR (95% CI) Moxifloxacin 0.89 (0.05 to 15.09), p=0.94 0.74 (0.04 to 15.61), p=0.85 Doxycycline 0.64 (0.05 to 9.02), p=0.74 0.60 (0.04 to 9.68), p=0.72 Azithromycin 1.34 (0.07 to 26.06), p=0.84 2.42 (0.09 to 63.28), p=0.60 Exacerbation frequency. Relative risk (95% CI) Moxifloxacin 1.36 (0.62 to 2.97), p=0.44 1.38 (0.62 to 3.10), p=0.43 Doxycycline 2.05 (0.98 to 4.29), p=0.06 2.07 (0.99 to 4.35), p=0.05 Azithromycin 0.83 (0.35 to 1.93), p=0.66 0.72 (0.30 to 1.71), p=0.45 Antibiotic resistance testing Factor change in MIC Moxifloxacin 4.82 (1.44 to 16.19), p=0.01 Doxycycline 3.74 (1.46 to 9.58), p=0.01 Azithromycin 6.23 (1.66 to 23.35), p=0.01 OR for resistant isolates Moxifloxacin 2.03 (0.36 to 11.54, p=0.42 Doxycycline 5.77 (1.40 to 23.74, p=0·02) Azithromycin 2.42 (0.61 to 9·62, p=0.21) Sputum inflammatory markers IL-1β (log10 pg/mL) Moxifloxacin −0.25 (−0.63 to 0.14), p=0.21 −0.15 (−0.56 to 0.26), p=0.47 Doxycycline −0.10 (−0048 to 0.28), p=0.61 −0.03 (−0.43 to 0.37), p=0.88 Azithromycin 0.04 (−0.35 to 0.43), p=0.86 0.09 (−0.31 to 0.49), p=0.66 IL-6 (log10 pg/mL) Moxifloxacin −0.16 (−0.50 to 0.18), p=0.37 −0.22 (−0.59 to 0.16), p=0.27 Doxycycline −0.21 (−0.56 to 0.13), p=0.22 −0.23 (−0.58 to 0.13), p=0.25 Azithromycin −0.19 (−0.43 to 0.27), p=0.65 −0.06 (−0.42 to 0.30), p=0.77 IL-8 (log10 pg/mL) Moxifloxacin −0.26 (−0.72 to 0.19), p=0.26 −0.29 (−0.78 to 0.19), p=0.24 Doxycycline −0.11 (−0.55 to 0.34), p=0.64 −0.08 (−0.54 to 0.38), p=0.74 Azithromycin 0.00 (−0.46 to 0.46), p=1.00 0.02 (−0.45 to 0.49), p=0.94 *The results in the right hand column were adjusted for age, sex, smoking status, FEV1% predicted and prior exacerbation history. p Values were calculated using the Wald test.

4 Doxycycline −0.11 (−0.55 to 0.34), p=0.64 −0.08 (−0.54 to 0.38), p=0.74 Azithromycin 0.00 (−0.46 to 0.46), p=1.00 0.02 (−0.45 to 0.49), p=0.94 *The results in the right hand column were adjusted for age, sex, smoking status, FEV1% predicted and prior exacerbation history. p Values were calculated using the Wald test. IL, interleukin; MIC, mean inhibitory concentration; qPCR, quantitative PCR; SGRQ, St George's Respiratory Questionnaire. A total of 395 isolates were cultured: 222 at study start and 173 after treatment. The most common isolates were Streptococcus spp other than Streptococcus pneumoniae. Figure 2 shows the species breakdown of these isolates. Figure 2 Species breakdown of all cultured isolates (n=395) before and after treatment. Secondary and exploratory endpoints Bacterial numbers by 16S qPCR One hundred and forty-two paired samples from 71 patients were available for this analysis. As with quantitative culture, there were no significant changes in overall bacterial load compared with placebo in any of the antibiotic treatment arms (table 2). Sputum inflammatory markers These were also measured in the subset of patients where extra paired sputum samples were available (n=71). No significant changes were seen in any of the three measured cytokines IL-6, IL-8 and IL-1β in any of the antibiotic treatment arms compared with placebo (table 2).

Secondary and exploratory endpoints Bacterial numbers by 16S qPCR One hundred and forty-two paired samples from 71 patients were available for this analysis. As with quantitative culture, there were no significant changes in overall bacterial load compared with placebo in any of the antibiotic treatment arms (table 2). Sputum inflammatory markers These were also measured in the subset of patients where extra paired sputum samples were available (n=71). No significant changes were seen in any of the three measured cytokines IL-6, IL-8 and IL-1β in any of the antibiotic treatment arms compared with placebo (table 2). Bacterial resistance Two hundred and forty-three isolates from 69 patients where ≥1 isolate was cultured both pre and post treatment were tested for bacterial resistance. There were measurable increases in the degree of antibiotic resistance of isolates in all three antibiotic arms. After adjusting for baseline MIC and whether the isolate was a species known to be associated with lower respiratory tract infection, moxifloxacin was associated with a factor increase in MIC of 4.82 (95% CI 1.44 to 16.19, p=0.01), doxycycline 3·74 (95% CI 1.46 to 9.58, p=0.01) and azithromycin 6.23 (95% CI 1.66 to 23.35, p=0.01) for isolates cultured in sputum taken from patients assigned to those antibiotics compared with placebo. Modelling the number of resistant isolates (where MIC break points were available for specific isolates to specific antibiotics), isolates from patients in the doxycycline group were more likely to be resistant to doxycycline than placebo (OR 5.77 (95% CI 1.40 to 23.74, p=0.02)). ORs for the other antibiotics were also >2 but these were not statistically significant. Boxplots demonstrating the mean MIC concentrations are shown in figure 3.

tics), isolates from patients in the doxycycline group were more likely to be resistant to doxycycline than placebo (OR 5.77 (95% CI 1.40 to 23.74, p=0.02)). ORs for the other antibiotics were also >2 but these were not statistically significant. Boxplots demonstrating the mean MIC concentrations are shown in figure 3. Figure 3 Boxplots for each treatment arm showing mean inhibitory concentrations (MICs) against that antibiotic compared with placebo before and after 3 months of treatment. Note that MICs for all detected isolates are shown, and the number of isolates before and after treatment is not necessarily comparable. Lung function There was a mean increase in FEV1 compared with placebo of 40 mL with moxifloxacin, 30 mL with doxycycline and 0 mL with azithromycin, although these changes were not significant at the 5% level, even after further adjustment (table 2). Health status The changes in SGRQ total score are listed in table 2. After adjustment, the largest improvement was seen with azithromycin compared with placebo, but these changes were again not significant at the 5% level. Adherence to therapy Adherence to therapy was high in all treatment arms (mean (SD) adherence 95%,13 93%,20 96%11 and 95%9 for moxifloxacin, doxycycline, azithromycin and placebo, respectively). No significant differences were detected between groups (table 2).

Health status The changes in SGRQ total score are listed in table 2. After adjustment, the largest improvement was seen with azithromycin compared with placebo, but these changes were again not significant at the 5% level. Adherence to therapy Adherence to therapy was high in all treatment arms (mean (SD) adherence 95%,13 93%,20 96%11 and 95%9 for moxifloxacin, doxycycline, azithromycin and placebo, respectively). No significant differences were detected between groups (table 2). Exacerbations During the study, 41 patients experienced 81 distinct exacerbations (21 on moxifloxacin, 32 on doxycycline, 13 on azithromycin and 15 on placebo). The relative risks of exacerbation are shown in table 2; there was an indication of increased exacerbations in the doxycycline group, although there was a small number of patients in the doxycycline and moxifloxacin groups who reported frequent exacerbations during the study period. Figure 4 shows the number of exacerbations by treatment group. Figure 4 Frequency of exacerbations experienced by patients during the study period, by treatment group. Adverse events The highest number of treatment-related adverse events, 10 (40%), was reported with moxifloxacin therapy. These were predominantly minor; therapy was withdrawn in four cases. There were two treatment-related adverse events reported with doxycycline and one with azithromycin. There were no treatment-related serious adverse events or suspected unexpected serious adverse reactions. Further detail is contained in the supplementary data.

predominantly minor; therapy was withdrawn in four cases. There were two treatment-related adverse events reported with doxycycline and one with azithromycin. There were no treatment-related serious adverse events or suspected unexpected serious adverse reactions. Further detail is contained in the supplementary data. Comparison of airway load measurements using quantitative culture and 16S qPCR Across all samples measured by both methods, 16S qPCR detected 10-fold more bacteria than quantitative culture (excluding samples where no organisms were detected on culture, n=13) (mean (SD) 9.15 (0.79) log10 copies/mL vs 8.12 (0.71) log10 cfu/mL, p<0.001). There was a significant correlation between the two techniques (Pearson’s r=0.465, p<0.001). Comparison of the measurement techniques using a Bland–Altman plot showed a 3.5-log variation in the measurement differences, although there was no proportional bias across the measured range of values (figure 5). Figure 5 Bland–Altman plot showing the differences between the measurement techniques of quantitative culture and 16S quantitative PCR (qPCR). The solid line is the mean measurement distance and dotted lines are mean±1.96 (SD), that is, the values between which 95% of the measurement differences lie.

Comparison of airway load measurements using quantitative culture and 16S qPCR Across all samples measured by both methods, 16S qPCR detected 10-fold more bacteria than quantitative culture (excluding samples where no organisms were detected on culture, n=13) (mean (SD) 9.15 (0.79) log10 copies/mL vs 8.12 (0.71) log10 cfu/mL, p<0.001). There was a significant correlation between the two techniques (Pearson’s r=0.465, p<0.001). Comparison of the measurement techniques using a Bland–Altman plot showed a 3.5-log variation in the measurement differences, although there was no proportional bias across the measured range of values (figure 5). Figure 5 Bland–Altman plot showing the differences between the measurement techniques of quantitative culture and 16S quantitative PCR (qPCR). The solid line is the mean measurement distance and dotted lines are mean±1.96 (SD), that is, the values between which 95% of the measurement differences lie. Discussion This study, for the first time, reports on the direct comparison of 13 weeks of therapy using different antibiotic classes and different measurement techniques against the endpoint of airway bacterial load in stable COPD. Most notably, there were no significant reductions in sputum bacterial load, by culture or qPCR, with any of these three antibiotics compared with placebo. Although pulsed moxifloxacin demonstrated the largest fall in bacterial numbers on culture, it also had the highest incidence of associated adverse events. There was evidence of an increase in antibiotic resistance of the cultured isolates across all treatment groups.

y of these three antibiotics compared with placebo. Although pulsed moxifloxacin demonstrated the largest fall in bacterial numbers on culture, it also had the highest incidence of associated adverse events. There was evidence of an increase in antibiotic resistance of the cultured isolates across all treatment groups. The primary endpoint of this study was total bacterial load, as the sum of the load of all isolates from each sample. Most of the bacteria isolated here were species classically considered to be non-pathogenic in airway disease, and this is in keeping with molecular studies of the airway microbiome which demonstrate a diverse flora dominated by firmicutes, including the Streptococcus and Rothia spp predominant here.20 While previous exacerbation studies have focused on loads of the three typical respiratory tract bacteria (S. pneumoniae, Haemophilus influenzae or Moraxella catarrhalis),21 22 which appear to increase at exacerbation,21 more recent investigation of the overall airway microbiome during naturally occurring exacerbations23 did not find an increase in total bacterial numbers. As in the gut, the lung microbiome is commensal and plays a role in immune function;24 these non-pathogenic microorganisms are therefore also important, and changes in their numbers are likely to influence exacerbation susceptibility. We did not find any support for our hypothesis that antibiotic therapy reduces total bacterial load in stable COPD, as well as potentially improving clinical outcome measures. Airway inflammation also did not change during treatment with any of these antibiotics and this is in keeping with previous reports9 and despite known anti-inflammatory effects of macrolide antibiotics. As discussed below, patients in the doxycycline and moxifloxacin arms suffered more exacerbations during the study period, and given that bacterial load,21 airway inflammation25 and quality of life1 all worsen significantly at exacerbation this may have attenuated any measured reductions. In addition, reductions in total bacterial load may only persist for a short time after quinolone therapy26 which might reduce the effectiveness of intermittent, ‘pulsed’ dosing regimens.12

way inflammation25 and quality of life1 all worsen significantly at exacerbation this may have attenuated any measured reductions. In addition, reductions in total bacterial load may only persist for a short time after quinolone therapy26 which might reduce the effectiveness of intermittent, ‘pulsed’ dosing regimens.12 Overall, however, antibiotic therapy did not appear to reduce total bacterial load after 3 months of therapy when measured either by culture or by 16S qPCR. The reduction in exacerbation frequency observed elsewhere is therefore likely to be mediated by other mechanisms. These may include either altering the background composition of the airway microbiome such that exacerbation susceptibility is reduced, by making the airway a less hospitable environment for the acquisition of new bacterial strains that may potentially cause exacerbation27 or alternatively by attenuating the rise in bacterial load when an exacerbation occurs rather than reducing it in the stable state. Further research is required to investigate these possibilities.

airway a less hospitable environment for the acquisition of new bacterial strains that may potentially cause exacerbation27 or alternatively by attenuating the rise in bacterial load when an exacerbation occurs rather than reducing it in the stable state. Further research is required to investigate these possibilities. For the first time, we have also directly analysed the measurement techniques of 16S qPCR and quantitative culture in the quantification of the airway microbiome. qPCR detected approximately 10 times as many bacteria as culture and, although significantly correlated, the wide variation between the measured values limits their direct comparability. Culture-based techniques are less sensitive than qPCR, and a normally cultivable species may not grow on culture media even when it is proved on sequencing to be the dominant organism in a sample.20 These findings therefore support the move away from culture-based to molecular techniques.

ts their direct comparability. Culture-based techniques are less sensitive than qPCR, and a normally cultivable species may not grow on culture media even when it is proved on sequencing to be the dominant organism in a sample.20 These findings therefore support the move away from culture-based to molecular techniques. Administration of each antibiotic was associated with an MIC increase of at least three times pretreatment values compared with placebo, with a corresponding increase in clinical resistance in those organisms for which this has been defined; this supports findings from other short-term studies of antibiotic therapy.28 However, in contrast to previous studies of antibiotics in COPD focusing on resistance in specific pathogens,10 12 we examined all isolates cultured from sputum, and we have examined resistance in greater detail. The information here therefore provides a much better indication of resistance in the wider airway microbiome. Macrolide resistance is rapidly inducible, highly transferrable, and increasing in prevalence commensurate with increasing macrolide prescriptions;13 resistance to quinolones and tetracyclines is also increasing globally.29 30 There is evidence that commensal airway streptococci act as a reservoir for S. pneumoniae resistance31 and the increased resistance demonstrated here may therefore drive future antibiotic-resistant disease. The approach of only examining resistance in those species classically thought to be pathogenic may be too narrow.

here is evidence that commensal airway streptococci act as a reservoir for S. pneumoniae resistance31 and the increased resistance demonstrated here may therefore drive future antibiotic-resistant disease. The approach of only examining resistance in those species classically thought to be pathogenic may be too narrow. There was an indication of a higher rate of treatment-related adverse events and treatment discontinuation with moxifloxacin therapy than previously reported,12 although the types of events reported were similar. This may be related to the higher frequency of administration here (4-weekly rather than 8-weekly) than in the previous study. By contrast, azithromycin and doxycycline were well tolerated. However, there is growing evidence of increased cardiovascular risk with macrolides,11 32 as well as high-frequency hearing loss,10 although the size and duration of this trial were not sufficient to examine these. Until these risks are further evaluated, advice to exercise caution in patients with cardiac risk factors33 still seems prudent.

ng evidence of increased cardiovascular risk with macrolides,11 32 as well as high-frequency hearing loss,10 although the size and duration of this trial were not sufficient to examine these. Until these risks are further evaluated, advice to exercise caution in patients with cardiac risk factors33 still seems prudent. There were no significant changes seen in the other measures of quality of life, lung function or adherence to therapy. There was an apparent increase in exacerbations with doxycycline therapy, although this was influenced by a small number of patients, and the study period was relatively short for exacerbation detection. Although the placebo group had a lower self-reported exacerbation frequency than the active treatment arms, this was adjusted for in the analysis. Furthermore our composite detection of exacerbations (using diary cards and/or treatment) may have contributed to the relatively higher recorded exacerbation frequency during the study. This study was not powered to detect changes in exacerbation frequency and this result should therefore be treated with caution.

ysis. Furthermore our composite detection of exacerbations (using diary cards and/or treatment) may have contributed to the relatively higher recorded exacerbation frequency during the study. This study was not powered to detect changes in exacerbation frequency and this result should therefore be treated with caution. There were a few limitations to this study. First, due to difficulty in double-blinding each of the three treatment regimens in a single trial, only one placebo was used and the trial was single blind. The final sample size was 50% lower than that initially specified (99 compared with 200), and some patients were unable to spontaneously expectorate sputum at their final study visit. This study was intentionally designed as a pilot exploratory study to assess airway bacterial changes with different antibiotics prior to a definitive study. Hence, the precision of the estimates shown here cannot fully discount a reduction in airway bacterial numbers of up to 0.9 log10 cfu/mL for moxifloxacin and 0.5 log10 cfu/mL for doxycycline and azithromycin. Nevertheless, we have shown that larger reductions in airway bacterial numbers (>1.0 log10, or 10-fold, cfu/mL) are unlikely. While the study was terminated early due to constraints with time lines, further recruitment would therefore have been unlikely to alter these conclusions. Third, as an exploratory study, this trial was not powered to detect changes in clinical endpoints and therefore further trials will be needed regarding the relative effectiveness of these antibiotics in practice.

raints with time lines, further recruitment would therefore have been unlikely to alter these conclusions. Third, as an exploratory study, this trial was not powered to detect changes in clinical endpoints and therefore further trials will be needed regarding the relative effectiveness of these antibiotics in practice. In summary, we found no significant changes in total bacterial load with antibiotic therapy in this, the first randomised, blinded, placebo-controlled study to directly compare the use of different antibiotics in stable COPD. We have also demonstrated an increase in the degree and rate of resistance of cultured airway bacteria following administration of these antibiotics. Supplementary Material Web supplement The authors would like to thank Claire Eckold (UCL Centre for Clinical Microbiology) for help with bacterial quantitative culture and resistance testing and interpretation, and Dan Jackson (MRC Biostatistics Unit, Cambridge) for advice regarding the statistical analysis of antimicrobial resistance. The authors would also like to thank Bayer Pharma for providing moxifloxacin for this study.

icrobiology) for help with bacterial quantitative culture and resistance testing and interpretation, and Dan Jackson (MRC Biostatistics Unit, Cambridge) for advice regarding the statistical analysis of antimicrobial resistance. The authors would also like to thank Bayer Pharma for providing moxifloxacin for this study. Contributors: SEB, EE-E, IN, JRH, GCD, PMAC, TDM, MJS and JAW contributed to the study design. SEB, EE-E, IN, GCD, PMAC, TDM, MJS and JAW contributed to the protocol and study materials. SEB, JPA, EE-E and IN contributed to patient recruitment. SEB and JPA collected the study data. TDM and VM provided the quantitative culture and bacterial resistance testing and assisted in the interpretation of these data. ML and MJS wrote the statistical analysis plan and performed the analyses. SEB and PJ performed the 16S qPCR analysis in conjunction with WOC and MFM. SEB performed the cytokine analysis, performed the analysis comparing 16S and qPCR as measurement techniques, and wrote the first draft. All authors contributed to interpretation of the data and revision of the manuscript.

rformed the analyses. SEB and PJ performed the 16S qPCR analysis in conjunction with WOC and MFM. SEB performed the cytokine analysis, performed the analysis comparing 16S and qPCR as measurement techniques, and wrote the first draft. All authors contributed to interpretation of the data and revision of the manuscript. Funding: This article presents independent research funded by the National Institute for Health Research (NIHR) under the Programme Grants for Applied Research programme (RP-PG-0109-10056) and the NIHR Royal Brompton Respiratory Biomedical Research Unit. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health. The moxifloxacin for the study was provided by Bayer Pharma AG, Berlin, Germany and the study Sponsor was University College London, UK. Neither Bayer, the funder, nor the Sponsor had any influence in the study design, collection, analysis and interpretation of the data, the writing of the report or the decision to submit for publication. MJS is supported by the Medical Research Council, grant number: G0800860.

Sponsor was University College London, UK. Neither Bayer, the funder, nor the Sponsor had any influence in the study design, collection, analysis and interpretation of the data, the writing of the report or the decision to submit for publication. MJS is supported by the Medical Research Council, grant number: G0800860. Competing interests: JRH has received honoraria, consulting fees and support for meetings from AstraZeneca, Bayer, Boehringer Ingelheim, Chiesi, GlaxoSmithKline, Pfizer and Takeda. JAW has received honoraria, consulting and board membership fees, support for meetings and institutional grant support from GlaxoSmithKline, Novartis, Pfizer, Takeda, Boehringer Ingelheim, Vectura, Chiesi, Johnson and Johnson and Bayer. PMAC has received honoraria, consulting and board membership fees and support for meetings from GlaxoSmithKline, Takeda, Boehringer Ingelheim, Novartis, Pfizer, Astrazeneca and Nycomed. Provenance and peer review: Not commissioned; externally peer reviewed.

Key messages What is the key question? Has the rapid rise in the use of electronic cigarettes by smokers led to the decline in the use of licensed nicotine products? What is the bottom line? The shapes of the trajectories of the use of electronic cigarettes and licensed nicotine products suggest that growth in the use of electronic cigarettes is probably not responsible for the decline in licensed nicotine product use by smokers but instead appears to have increased the market for non-tobacco nicotine-containing products. Why read on? This study provides the first evidence in any country assessing the potential trade-off between the use of electronic cigarettes and nicotine products licensed by medicines regulators. Introduction Electronic cigarettes are battery-powered devices that can provide inhaled doses of nicotine by way of a vaporised solution. Since they were introduced to the European market in 2006, there has been substantial growth in their use by smokers.1 Supermarket sales data showed a rise of over 40% in their use between 2013 and 2014.2 In contrast, licensed nicotine products appear to have become less popular, with sales in many European countries having decreased over the past few years (eg, France and the UK).3 4

s been substantial growth in their use by smokers.1 Supermarket sales data showed a rise of over 40% in their use between 2013 and 2014.2 In contrast, licensed nicotine products appear to have become less popular, with sales in many European countries having decreased over the past few years (eg, France and the UK).3 4 The use of licensed nicotine products for smoking reduction appears to promote quit attempts.5 Thus if electronic cigarette use is substituting for licensed nicotine product use among smokers, and does not promote cessation, there could be a negative net effect on public health. On the other hand, if electronic cigarette users are primarily smokers who would have tried to reduce without any nicotine product use or not tried to reduce at all, there may be a public health gain, as long as electronic cigarette use also promotes subsequent cessation. Although electronic cigarettes are almost certainly considerably safer than traditional cigarettes, and when used for cessation they probably improve the chances of success,1 6 7 evidence on the benefits or otherwise of electronic cigarette use while continuing to smoke is mixed.1 6–11

so promotes subsequent cessation. Although electronic cigarettes are almost certainly considerably safer than traditional cigarettes, and when used for cessation they probably improve the chances of success,1 6 7 evidence on the benefits or otherwise of electronic cigarette use while continuing to smoke is mixed.1 6–11 It is therefore important to determine how far electronic cigarette use has replaced or supplemented licensed nicotine product use. The introduction of new smoking cessation aids on to the market has previously resulted in more smokers using medications of some sort to help them stop.12 13 However, this may not be the case for electronic cigarettes and smoking reduction. To address the issue of how far electronic cigarettes are complementing licensed nicotine products or replacing them, one can assess whether the temporal trajectories in prevalence of use of the two types of product mirror each other. If they follow very different patterns of change, then it is unlikely that they are connected. Therefore, this study set out to answer the following questions: What has been the trajectory in growth of electronic cigarette use in current smokers, and how has this compared with the trajectory in decline in use of licensed nicotine products? What has been the resultant trajectory in use of any nicotine-containing product?

It is therefore important to determine how far electronic cigarette use has replaced or supplemented licensed nicotine product use. The introduction of new smoking cessation aids on to the market has previously resulted in more smokers using medications of some sort to help them stop.12 13 However, this may not be the case for electronic cigarettes and smoking reduction. To address the issue of how far electronic cigarettes are complementing licensed nicotine products or replacing them, one can assess whether the temporal trajectories in prevalence of use of the two types of product mirror each other. If they follow very different patterns of change, then it is unlikely that they are connected. Therefore, this study set out to answer the following questions: What has been the trajectory in growth of electronic cigarette use in current smokers, and how has this compared with the trajectory in decline in use of licensed nicotine products? What has been the resultant trajectory in use of any nicotine-containing product? Methods Design The study formed part of the Smoking Toolkit Study (STS), an ongoing population study designed to provide information on smoking and smoking cessation patterns among smokers and recent ex-smokers in England. Data for this paper were obtained between March 2011 and August 2014. The STS involves monthly household surveys using a random location sampling design, with initial random selection of grouped output areas (containing 300 households), stratified by ACORN (socio-demographic) characteristics http://acorn.caci.co.uk/ and region. Interviewers then choose which houses within these areas are most likely to fulfil their quotas and conduct face-to-face computer-assisted interviews with one member per household.14 Participants in the STS appear to be representative of the population in England, having a similar socio-demographic composition to other large national surveys, such as the Health Survey for England.14

likely to fulfil their quotas and conduct face-to-face computer-assisted interviews with one member per household.14 Participants in the STS appear to be representative of the population in England, having a similar socio-demographic composition to other large national surveys, such as the Health Survey for England.14 Measures Smokers were asked: ‘Are you using any of the following either to help you stop smoking, to help you cut down or for any other reason at all?’ Answer: nicotine patch; nicotine gum; nicotine lozenges/tablets; nicotine inhaler; nicotine nasal spray; mouth spray; electronic cigarettes; I don't know; none of these; other Smokers were also asked if they were attempting to cut down their cigarette consumption and: ‘Which, if any, of the following are you currently using to help you cut down the amount you smoke?’ Answer: nicotine patch; nicotine gum; nicotine lozenges/tablets; nicotine inhaler; nicotine nasal spray; mouth spray; electronic cigarettes; I don't know; none of these; other ‘Do you regularly use any of the following in situations when you are not allowed to smoke?’ Answer: nicotine patch; nicotine gum; nicotine lozenges/tablets; nicotine inhaler; nicotine nasal spray; mouth spray; electronic cigarettes; I don't know; none of these; other Respondents were classified accordingly: Using electronic cigarettes: reported using electronic cigarettes in response to question 1 and/or 2 and/or 3 Using licensed nicotine products: reported using any of the licensed products in response to question 1 and/or 2 and/or 3

‘Do you regularly use any of the following in situations when you are not allowed to smoke?’ Answer: nicotine patch; nicotine gum; nicotine lozenges/tablets; nicotine inhaler; nicotine nasal spray; mouth spray; electronic cigarettes; I don't know; none of these; other Respondents were classified accordingly: Using electronic cigarettes: reported using electronic cigarettes in response to question 1 and/or 2 and/or 3 Using licensed nicotine products: reported using any of the licensed products in response to question 1 and/or 2 and/or 3 Using nicotine-containing products: reported using licensed nicotine products and/or e-cigarettes in response to question 1 and/or 2 and/or 3. Contextual information was also gathered on socio-demographic and smoking-related characteristics (ie, gender, age, socio-economic status, cigarette consumption, cigarette dependence, daily versus non-daily nicotine-containing product use, attempts to quit smoking in the previous 12 months and attempts to cut down cigarette consumption). Socio-economic status was measured using the Social-Grade Classification Tool,15 which categorises individuals into one of five social grades: AB, C1, C2, D and E. Grades AB and C1 were classified as ‘non-manual’ and Grades C2 to E were classified as ‘manual’ occupational groups. Cigarette dependence was assessed using time to first cigarette of the day.16 A copy of the questionnaire is available on the STS website (http://www.smokinginengland.info)

l grades: AB, C1, C2, D and E. Grades AB and C1 were classified as ‘non-manual’ and Grades C2 to E were classified as ‘manual’ occupational groups. Cigarette dependence was assessed using time to first cigarette of the day.16 A copy of the questionnaire is available on the STS website (http://www.smokinginengland.info) Statistical analysis Analyses were undertaken using R V.3.1.1. Data were weighted to match the population in England (see Fidler et al14 for details). Differences in socio-demographic and smoking characteristics as a function of nicotine-containing product use were assessed with generalised linear models (for normally distributed outcomes) and χ2 tests (for dichotomous outcomes), using the ‘Survey’ R package.17 Post hoc analyses were conducted using multiple χ2 and t tests, and were adjusted using the Benjamini and Yekutieli false discovery rate.

otine-containing product use were assessed with generalised linear models (for normally distributed outcomes) and χ2 tests (for dichotomous outcomes), using the ‘Survey’ R package.17 Post hoc analyses were conducted using multiple χ2 and t tests, and were adjusted using the Benjamini and Yekutieli false discovery rate. Data were aggregated into quarters to reduce the sampling variation associated with each data point. Trends in prevalence of use of electronic cigarettes, licensed nicotine products and nicotine-containing products were first assessed using generalised linear models (specifying the binomial family and logit link function). However, as it was hypothesised that the trends may be inconsistent over time (eg, there may be an initial increase then decrease in prevalence or vice versa), segmented regression models were also applied. These are regression models where the relationship between the outcome and the predictor variables are piecewise linear, namely represented by at least two straight lines connected at ‘breakpoints’. For this paper, a breakpoint would occur when there was a change in the slope of the function relating prevalence to time. Segmented regressions were applied using the ‘segmented’ package,18 which allows both single and multiple breakpoints to be specified. This programme uses an iterative procedure whereby only starting values for the breakpoints are required. It also implements bootstrap restarting to make the algorithm less sensitive to starting values.

ions were applied using the ‘segmented’ package,18 which allows both single and multiple breakpoints to be specified. This programme uses an iterative procedure whereby only starting values for the breakpoints are required. It also implements bootstrap restarting to make the algorithm less sensitive to starting values. To determine whether a model with breakpoints provided a better fit than a standard generalised linear model, the Davies test assessed the null hypothesis that there was no difference in slopes before and after a breakpoint. In all cases, the difference in slopes was significantly greater than 0. As recommended for segmented regression analyses, the Bayesian Information Criterion was used to select the optimal number of breakpoints (ie, 1, 2 or 3).19 Because time is a predictor in the regression analyses, error terms of consecutive observations may be correlated. This may lead to underestimation of SEs and overestimation of statistical significance.20 The presence of correlations between consecutive time points was assessed using the Durbin-Watson statistic. Based on a critical value of 2.0, there was no evidence of significant autocorrelation (values ranged from 1.91 to 1.97).

This may lead to underestimation of SEs and overestimation of statistical significance.20 The presence of correlations between consecutive time points was assessed using the Durbin-Watson statistic. Based on a critical value of 2.0, there was no evidence of significant autocorrelation (values ranged from 1.91 to 1.97). There were missing data for four of the socio-demographic and smoking characteristic variables. For frequency of licensed nicotine product and/or electronic cigarette use, missing data ranged from 7% among electronic cigarette users to 20% among those using both products, while less than 3% of participants had missing data on age, previous attempts to quit smoking and attempts to cut down. Missing data were imputed using the ‘Amelia 11’ package.21 The number of imputed data sets was based on the recommendations of Graham et al22 and set to 10. Estimates and results from hypothesis testing were combined using Rubin's rules.23 STROBE guidelines for the reporting of observational studies were followed.24 Results Between March 2011 and November 2014 data were collected from 14 502 cigarette smokers. Overall, 76% (95% CI 75.3% to 76.7%) were not using any form of nicotine-containing product, 9.9% (95% CI 9.4% to 10.4%) were using electronic cigarettes, 11.6% (95% CI 11.1% to 12.2%) were using licensed nicotine products, and 2.4% (95% CI 2.2% to 2.7%) were using both of these types of product.

502 cigarette smokers. Overall, 76% (95% CI 75.3% to 76.7%) were not using any form of nicotine-containing product, 9.9% (95% CI 9.4% to 10.4%) were using electronic cigarettes, 11.6% (95% CI 11.1% to 12.2%) were using licensed nicotine products, and 2.4% (95% CI 2.2% to 2.7%) were using both of these types of product. Table 1 shows the characteristics of participants as a function of their nicotine-containing product use. Those not using nicotine-containing products were less likely to be female compared with those using licensed nicotine products (OR 0.81; χ2=17.23, df=1, p=0.003) and were more likely to be in a manual job than those using electronic cigarettes (OR=1.30; χ2=21.80, df=1, p<0.001) and those using licensed nicotine products (OR=1.18; χ2=10.27, df=1, p=0.035). Those using licensed nicotine products were older than those not using any nicotine-containing product (mean difference 0.94; t(12 835)=4.35, p<0.001) and those using electronic cigarettes (mean difference 2.21; t(2998)=2.72, p=0.003). Table 1 Characteristics of smokers as a function of their use of nicotine-containing products

Table 1 shows the characteristics of participants as a function of their nicotine-containing product use. Those not using nicotine-containing products were less likely to be female compared with those using licensed nicotine products (OR 0.81; χ2=17.23, df=1, p=0.003) and were more likely to be in a manual job than those using electronic cigarettes (OR=1.30; χ2=21.80, df=1, p<0.001) and those using licensed nicotine products (OR=1.18; χ2=10.27, df=1, p=0.035). Those using licensed nicotine products were older than those not using any nicotine-containing product (mean difference 0.94; t(12 835)=4.35, p<0.001) and those using electronic cigarettes (mean difference 2.21; t(2998)=2.72, p=0.003). Table 1 Characteristics of smokers as a function of their use of nicotine-containing products Not using licensed nicotine products or electronic cigarettes Using electronic cigarettes only Using licensed nicotine products only Using electronic cigarettes and licensed nicotine products N=10 431 N=1360 N=1603 N=330 Female % (n) 44.7 (4667)a 48.8 (664) 50.1 (803)b 51.2 (169) Age mean (SD) 41.5 (17.56)b 41.2 (16.13)b 43.3 (16.17)a 42.5 (15.76) Manual occupation % (n) 61.7 (6434)a 55.4 (752)b 57.6 (924)b 54.7 (180) Time to first cigarette mean (SD) 1.3 (1.16)a 1.4 (1.15)b,f 1.5 (1.17)b,f 1.7 (1.11)b,e Attempts to quit smoking in the previous 12 months % (n) 23.4 (2452)a 53.7 (729)b 63.2 (1013)c 72.1 (237)c Non-daily smoking % (n) 10.3 (1077) 11.8 (160) 10.7 (171) 11.6 (38) Attempting to cut down % (n) 41.7 (4349)a 76.1 (1034)b 77.9 (1248)b 78.5 (259)b Using nicotine-containing products for Smoking reduction % (n) NA 69.2 (942) 70.5 (1130) 74.8 (247) Temporary abstinence % (n) NA 71.4 (971)a 64.4 (1032)b,e 78.9 (260)b,f Daily nicotine-containing product use % (n) NA 56.8 (772)a 52.4 (841) 47.1 (155)b Data are weighted to match the English population (weighted n's are presented); n's were rounded up to the nearest whole number. As a consequence percentages may not sum to 100 and n's may not sum to total N. Social grade was measured using the social grades system: non-manual includes A: higher managerial, administrative or professional, B: intermediate managerial, administrative or professional, C1: supervisory or clerical and junior managerial administrative or professional; manual includes C2: skilled manual workers, D: semi and unskilled manual workers, E: casual or lowest grade workers, pensioners and others who depend on the welfare state for their income. AB and C1 were coded as non-manual and C2, D and E as manual.17 Time to first cigarette of the day was assessed as (1) >60 min, (2) 31–60 min, (3) 6–30 min or (4) ≤5 min. a, b, c and d all differ; e differs from f. NA, not applicable.

rade workers, pensioners and others who depend on the welfare state for their income. AB and C1 were coded as non-manual and C2, D and E as manual.17 Time to first cigarette of the day was assessed as (1) >60 min, (2) 31–60 min, (3) 6–30 min or (4) ≤5 min. a, b, c and d all differ; e differs from f. NA, not applicable. Those not using any nicotine-containing product had lower odds of reporting that they were attempting to cut down compared with those using electronic cigarettes (OR 0.27; χ2=607.12, df=1, p<0.001), those using licensed nicotine products (OR 0.18; χ2=762.93, df=1, p<0.001) and those using both products (OR 0.12; χ2=231.23, df=1, p<0.001). Those using licensed nicotine products had higher odds than those using electronic cigarettes (OR 1.48; χ2=28.22, df=1, p<0.001) of having made a quit attempt and electronic cigarette users had higher odds than those not using any product (OR 3.78; χ2=587.86, df=1, p<0.001). Those using both products were more cigarette dependent than those using only licensed nicotine products or electronic cigarettes (mean difference 0.11; t(1663)=3.29, p=0.020 and mean difference 0.18; t(1975)=2.73, p=0.005, respectively), while those not using any nicotine-containing product had lower cigarette dependence compared with those using electronic cigarettes, those using licensed nicotine products and those using both products (mean difference 0.09; t(12 523)=2.71, p=0.020, mean difference 0.07; t(12 834)=4.19, p<0.001, and mean difference 0.11; t(11 500)=5.08, p<0.001, respectively). There was no difference among groups in the percentage reporting non-daily smoking (χ2=3.22, df=3, p=0.463). Of those using some form of nicotine-containing product, those using both electronic cigarettes and licensed nicotine products were less likely to report daily use of electronic cigarettes and/or licensed nicotine products compared with those using electronic cigarettes only (OR 2.22; χ2=9.75, df=1, p=0.020). Those using both products also had higher odds of reporting that they were doing so during temporary abstinence relative to electronic cigarette users (OR 1.50; χ2=7.38, df=1, p=0.026), who had a higher odds of reporting they were using electronic cigarettes for temporary abstinence compared to those using licensed nicotine products only (OR 1.38; χ2=16.81, df=1, p<0.001). There was no difference among groups in reports of using nicotine-containing products and/or electronic cigarettes to reduce cigarette consumption (χ2=4.05, df=2, p=0.180).

re using electronic cigarettes for temporary abstinence compared to those using licensed nicotine products only (OR 1.38; χ2=16.81, df=1, p<0.001). There was no difference among groups in reports of using nicotine-containing products and/or electronic cigarettes to reduce cigarette consumption (χ2=4.05, df=2, p=0.180). Figure 1 shows the prevalence of electronic cigarette, licensed nicotine product and nicotine-containing product use, while figure 2 displays the results of the segmented regression analyses. The segmented regression analyses showed that there was a rapid increase in electronic cigarette use up to quarter 3 of 2013 from 2.2% (95% CI 1.4% to 3.2%) to 20.8% (95% CI 18.3% to 23.4%) (OR 1.31, 95% CI 1.28% to 1.35%, p<0.001) and little change thereafter (OR 0.96, 95% CI 0.91% to 1.01%, p=0.129) with a prevalence in quarter 4 of 2014 of 16.3% (95% CI 13.5% to 19.5%). In contrast, there was no change in licensed nicotine product use between quarter 2 of 2011 (17.4%, 95% CI 15.3% to 19.8%) and quarter 3 of 2013 (17.9%, 95% CI 15.62% to 20.5%; OR 1.01, 95% CI 0.99 to 1.03, p=0.277); and there was a continual slow decline in licensed nicotine product use thereafter (OR 0.90, 95% CI 0.84 to 0.96, p=0.003). The prevalence in quarter 4 of 2014 was 7.9% (95% CI 6.0% to 10.4%). Figure 1 The proportion of smokers using nicotine-containing products over time. Data weighted to match the English population. E-cigarettes, electronic cigarettes; LNP, licensed nicotine product; NCP, nicotine-containing product. Time is represented quarterly in years.

Figure 1 shows the prevalence of electronic cigarette, licensed nicotine product and nicotine-containing product use, while figure 2 displays the results of the segmented regression analyses. The segmented regression analyses showed that there was a rapid increase in electronic cigarette use up to quarter 3 of 2013 from 2.2% (95% CI 1.4% to 3.2%) to 20.8% (95% CI 18.3% to 23.4%) (OR 1.31, 95% CI 1.28% to 1.35%, p<0.001) and little change thereafter (OR 0.96, 95% CI 0.91% to 1.01%, p=0.129) with a prevalence in quarter 4 of 2014 of 16.3% (95% CI 13.5% to 19.5%). In contrast, there was no change in licensed nicotine product use between quarter 2 of 2011 (17.4%, 95% CI 15.3% to 19.8%) and quarter 3 of 2013 (17.9%, 95% CI 15.62% to 20.5%; OR 1.01, 95% CI 0.99 to 1.03, p=0.277); and there was a continual slow decline in licensed nicotine product use thereafter (OR 0.90, 95% CI 0.84 to 0.96, p=0.003). The prevalence in quarter 4 of 2014 was 7.9% (95% CI 6.0% to 10.4%). Figure 1 The proportion of smokers using nicotine-containing products over time. Data weighted to match the English population. E-cigarettes, electronic cigarettes; LNP, licensed nicotine product; NCP, nicotine-containing product. Time is represented quarterly in years. Figure 2 The proportion of smokers using nicotine-containing products over time: segmented regression results. Data weighted to match the English population; E-cigarettes, electronic cigarettes; LNP, licensed nicotine product; NCP, nicotine-containing product. Time represented quarterly in years.

Figure 1 The proportion of smokers using nicotine-containing products over time. Data weighted to match the English population. E-cigarettes, electronic cigarettes; LNP, licensed nicotine product; NCP, nicotine-containing product. Time is represented quarterly in years. Figure 2 The proportion of smokers using nicotine-containing products over time: segmented regression results. Data weighted to match the English population; E-cigarettes, electronic cigarettes; LNP, licensed nicotine product; NCP, nicotine-containing product. Time represented quarterly in years. The result was that there was no increase in nicotine-containing product use between quarter 2 of 2011 (19.1%, 95% CI 16.9% to 21.5%) and quarter 1 of 2012 (18.5%, 95% CI 16.3% to 21.0%; OR 0.97, 95% CI 0.91% to 1.04%, p=0.297), an increase to 33.3% (95% CI 30.4% to 36.3%) up to quarter 3 of 2013 (OR 1.16, 95% CI 1.12 to 1.21, p<0.001) and then a decrease to 22.7% (95% CI 19.3% to 26.3%) in quarter 4 of 2014 (OR 0.94, 95% CI 0.90 to 0.99, p=0.014).

) and quarter 1 of 2012 (18.5%, 95% CI 16.3% to 21.0%; OR 0.97, 95% CI 0.91% to 1.04%, p=0.297), an increase to 33.3% (95% CI 30.4% to 36.3%) up to quarter 3 of 2013 (OR 1.16, 95% CI 1.12 to 1.21, p<0.001) and then a decrease to 22.7% (95% CI 19.3% to 26.3%) in quarter 4 of 2014 (OR 0.94, 95% CI 0.90 to 0.99, p=0.014). Discussion There was a rapid increase in electronic cigarette use by smokers between quarter 2 of 2011 and quarter 3 of 2013 with little change thereafter. Over the same period, licensed nicotine product use remained stable and then dropped gradually between quarter 3 of 2013 and quarter 4 of 2014. The result was an initial growth in nicotine-containing product use up to quarter 3 of 2013 and a decrease thereafter. These trajectories suggest that electronic cigarette use is not associated with the reduction in licensed nicotine product use by smokers but may have instead increased the market for nicotine-containing products.

as an initial growth in nicotine-containing product use up to quarter 3 of 2013 and a decrease thereafter. These trajectories suggest that electronic cigarette use is not associated with the reduction in licensed nicotine product use by smokers but may have instead increased the market for nicotine-containing products. If the rise in electronic cigarette use has not been primarily responsible for the decline in the use of licensed nicotine products by smokers, this raises the question as to what has caused this decrease. There was no reduction in the percentage of smokers attempting to reduce their smoking, so that is unlikely to explain the trend.25 The stop smoking services recommend licensed nicotine products and offer them on prescription and use of these services has declined since 2011.26 However, this decline is unlikely to be an important factor as the large majority of smokers who previously used licensed nicotine products while smoking had not attended stop smoking services.27 Marketing of licensed nicotine products increased over the period of data collection, so reduced exposure to advertising does not appear to be a factor.28 It is possible that the trend reflects a longer term disillusionment with licensed nicotine products as aids to smoking reduction. This is something that warrants further investigation.

d nicotine products increased over the period of data collection, so reduced exposure to advertising does not appear to be a factor.28 It is possible that the trend reflects a longer term disillusionment with licensed nicotine products as aids to smoking reduction. This is something that warrants further investigation. In terms of what may underlie the trajectory in electronic cigarette use, the rapid rise could be explained by social contagion with more and more smokers being persuaded that these products could help them reduce the amount they smoke or ultimately stop altogether. The plateau could reflect a ceiling on the proportion of smokers who want to reduce their cigarette consumption enough to buy a nicotine product, or it could be that an increasing proportion of smokers have been led to believe that electronic cigarettes are as harmful as tobacco cigarettes,29 thus removing the incentive to use them. This is another important area for future study.

s who want to reduce their cigarette consumption enough to buy a nicotine product, or it could be that an increasing proportion of smokers have been led to believe that electronic cigarettes are as harmful as tobacco cigarettes,29 thus removing the incentive to use them. This is another important area for future study. This study has several strengths including the large number of participants, the representativeness of the sample and the use of frequent sampling. There were also a number of limitations. This study only considered electronic cigarette and licensed nicotine product use by current smokers. Use specifically as an aid to cessation and use among never and ex-smokers are important further topics for investigation. Beyond assessing daily versus non-daily use, this study did not examine the frequency of electronic cigarette and licensed nicotine product use. This study did not assess to what extent electronic cigarette or licensed nicotine product use was associated with a reduction in cigarette consumption or intake of tobacco toxins. Because electronic cigarette and licensed nicotine product users are likely to have been heavier smokers before they started using the products, this can only be assessed using prospective designs.1 The data series was limited to approximately 3 years. It will be important to continue to monitor the trends to track further changes as the electronic cigarette market continues to mature and the regulatory environment changes. As with all survey-based designs, there is a possibility of misreporting of nicotine-containing product use although there is no reason to believe that this varied over time.

nue to monitor the trends to track further changes as the electronic cigarette market continues to mature and the regulatory environment changes. As with all survey-based designs, there is a possibility of misreporting of nicotine-containing product use although there is no reason to believe that this varied over time. In conclusion, the rapid growth in electronic cigarette use between 2011 and 2013 appears to have increased the overall market for use of nicotine products while smoking and was not associated with a decline in use of licensed nicotine products over this period. Contributors: EB, JB and RW designed the study. EB wrote the first draft. All authors commented on this draft and contributed to the final version.

In conclusion, the rapid growth in electronic cigarette use between 2011 and 2013 appears to have increased the overall market for use of nicotine products while smoking and was not associated with a decline in use of licensed nicotine products over this period. Contributors: EB, JB and RW designed the study. EB wrote the first draft. All authors commented on this draft and contributed to the final version. Funding: The Smoking Toolkit Study is currently funded by Cancer Research UK (grant number: A14135). It has also previously been funded by Pfizer, GlaxoSmithKline and Johnson & Johnson, none of whom had any involvement in the design of the study, the analysis or interpretation of the data, the writing of the report, or the decision to submit the paper for publication. EB is funded by the School for Public Health Research (SPHR) of the National Institute for Health Research (NIHR) and JB by the Society for the Study of Addiction. The views are those of the authors(s) and not necessarily those of the NHS, the NIHR or the Department of Health. SPHR is a partnership between the Universities of Sheffield, Bristol, Cambridge and Exeter; UCL; The London School for Hygiene and Tropical Medicine; the LiLaC collaboration between the Universities of Liverpool and Lancaster; and Fuse: The Centre for Translational Research in Public Health, a collaboration between Newcastle, Durham, Northumbria, Sunderland and Teesside universities.

ol, Cambridge and Exeter; UCL; The London School for Hygiene and Tropical Medicine; the LiLaC collaboration between the Universities of Liverpool and Lancaster; and Fuse: The Centre for Translational Research in Public Health, a collaboration between Newcastle, Durham, Northumbria, Sunderland and Teesside universities. Competing interests: RW undertakes consultancy and research for and receives travel funds and hospitality from manufacturers of smoking cessation medications but does not, and will not, take funds from electronic cigarette manufacturers or the tobacco industry. RW is an honorary co-director of the National Centre for Smoking Cessation and Training and a trustee of the stop-smoking charity, QUIT. RW's salary is funded by Cancer Research UK. EB and JB have received unrestricted research funding from Pfizer. EB and JB are funded by Cancer Research UK. Ethics approval: UCL Ethics Committee approved this study. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: Data come from the Smoking Toolkit Study. For more details please contact the primary author or visit http://www.smokinginengland.info

Background Epidemiology of non-tuberculous mycobacteria in individuals with cystic fibrosis Non-tuberculous mycobacteria (NTM) are increasingly being isolated from the sputum of adults and children with cystic fibrosis (CF), both in North America and in Europe.1–17 Estimates of the prevalence of NTM in the CF population have ranged from 1.3% in the earliest study reported in 19841 to 32.7% in a review of individuals with CF over the age of 40 years in Colorado.9 To date, the largest studies published examined 986,6 121615 and 158217 individuals with CF and reported rates of NTM-positive cultures of 13.0%, 13.7% and 6.6%, respectively. Recently, analysis of US Cystic Fibrosis Foundation (CFF) registry data has shown prevalence rates for NTM-positive culture in the USA of 12%18 but with considerable variation between individual states (0–28%).19

217 individuals with CF and reported rates of NTM-positive cultures of 13.0%, 13.7% and 6.6%, respectively. Recently, analysis of US Cystic Fibrosis Foundation (CFF) registry data has shown prevalence rates for NTM-positive culture in the USA of 12%18 but with considerable variation between individual states (0–28%).19 The NTM species most commonly identified in individuals with CF from North America and Europe are the slow-growing Mycobacterium avium complex (MAC) (including M. avium, M. intracellulare and M. chimaera), which can be found in up to 72% of NTM-positive sputum cultures,6 and the rapid-growing M. abscessus complex (MABSC) (comprising the subspecies M. abscessus subsp abscessus (M. a. abscessus), M. a. bolletii20 and M. a. massiliense21 22 (the latter currently classified as part of M. a. bolletii)), which in many centres has now become the most common NTM isolated from individuals with CF.7 15 17 21 23–25 Other less commonly isolated species include M. simiae,11 M. kansasii and M. fortuitum.26 There are geographical differences in both the prevalence of NTM-positive cultures and also the relative frequency of different species seen between but also within countries.6 17 19 24 25 27

m individuals with CF.7 15 17 21 23–25 Other less commonly isolated species include M. simiae,11 M. kansasii and M. fortuitum.26 There are geographical differences in both the prevalence of NTM-positive cultures and also the relative frequency of different species seen between but also within countries.6 17 19 24 25 27 NTM acquisition is strongly associated with age in individuals with CF, with prevalence increasing from 10% in children aged 10 years, to over 30% in adults over the age of 40 years.9 In individuals with an adult diagnosis of CF, over 50% (mostly females) have NTM-positive airway cultures.9 There appear to be species-specific differences in age-related prevalence within CF cohorts, with MAC more commonly isolated from adults over 25 years of age,6 7 14 17 27 while MABSC is isolated from all age groups, but peaks between those 11 and 15 years of age in some studies.17 28 There may also be species-specific differences in virulence: individuals with MABSC-positive cultures are more likely to meet American Thoracic Society (ATS)/Infectious Diseases Society of America (IDSA) criteria for diagnosing NTM pulmonary disease (NTM-PD, see Diagnosis of NTM-PD in CF section), and have worse morbidity and mortality associated with a more rapid decline in lung function.15 27 29 30

-positive cultures are more likely to meet American Thoracic Society (ATS)/Infectious Diseases Society of America (IDSA) criteria for diagnosing NTM pulmonary disease (NTM-PD, see Diagnosis of NTM-PD in CF section), and have worse morbidity and mortality associated with a more rapid decline in lung function.15 27 29 30 There has been a rise over the last four decades in the reported prevalence of NTM-positive cultures in respiratory samples from individuals with CF,1 6 15 17 18 23 an increase in part mirroring temporal changes seen in the non-CF cohort.31–38 While increasing detection rates may reflect enhanced surveillance and/or improved microbiological detection,6 27 39–42 there are a number of lines of evidence suggesting a true rise in the frequency of NTM infection. A number of CF studies43 show year on year increases in NTM-positive cultures with no change in surveillance intensity or culture methodology. There has been an increase over time in rates of skin test reactivity to NTM antigens in US population-based testing studies,44 potentially indicating increasing exposure to NTM (see below). Furthermore, the relative frequency of M. abscessus detection in NTM-positive samples from individuals with CF has increased remarkably over time both in the USA and in Europe,2 6 15 17 23 27 suggesting real changes in NTM acquisition rates (rather than increased sampling).

ating increasing exposure to NTM (see below). Furthermore, the relative frequency of M. abscessus detection in NTM-positive samples from individuals with CF has increased remarkably over time both in the USA and in Europe,2 6 15 17 23 27 suggesting real changes in NTM acquisition rates (rather than increased sampling). Possible reasons for the potential increased frequency of NTM-positive cultures in individuals with CF include: increases in environmental exposure to NTM through more permissive temperature settings of home water heaters45 and more contact with shower aerosols,46 47 increased antibiotic usage creating more NTM favourable lung niches,27 greater chronic use of medications that might impair host immunity to NTM,43 and/or spread of NTM through person-to-person transmission.48 49

gh more permissive temperature settings of home water heaters45 and more contact with shower aerosols,46 47 increased antibiotic usage creating more NTM favourable lung niches,27 greater chronic use of medications that might impair host immunity to NTM,43 and/or spread of NTM through person-to-person transmission.48 49 NTM-PD in individuals with CF NTM can cause progressive inflammatory lung damage, a condition termed ‘NTM pulmonary disease’ (NTM-PD),50 51 which is defined by the presence of specific microbiological, clinical and radiological features described in Diagnosis of NTM-PD in CF section. However, it has become clear that NTM can also transiently, intermittently or permanently reside within the lungs of individuals with CF without causing NTM-PD, thus representing asymptomatic infection and creating considerable difficulties in deciding how best to screen for and diagnose NTM.30 Further challenges exist in knowing how best to identify NTM in respiratory samples, when and how to initiate treatment for NTM-PD (as highlighted by a recent Cochrane review52) and how NTM may impact individuals under consideration for lung transplantation. As a consequence, the CFF and European Cystic Fibrosis Society (ECFS) sought to generate a consensus recommendations document to support and standardise the management of NTM infection in individuals with CF, permitting prospective evaluation of current best practice and forming a foundation for future research programmes.

consequence, the CFF and European Cystic Fibrosis Society (ECFS) sought to generate a consensus recommendations document to support and standardise the management of NTM infection in individuals with CF, permitting prospective evaluation of current best practice and forming a foundation for future research programmes. These consensus statements have been developed to assist in the management both of adults and children with CF who are infected with NTM. Given the virtual absence of published evidence to guide paediatric care,53 recommendations for children with CF infected with NTM are based on extrapolated adult data, the practical experience of experts and appropriate adjustment of drug regimens, and are, except where stated, the same as for adults. Methods Expert committee structure The CFF and the ECFS invited experts to participate in the statement development process. The 19-member committee consisted of professionals (10 US and 9 European) with expertise in CF and NTM, and included adult and paediatric CF physicians, lung transplant physicians, microbiologists, infectious disease specialists and a parent of an individual with CF. The committee convened in May 2012 and was divided into five subgroups, each responsible for a specific topic: Epidemiology and Risk Factors, Screening, Microbiology, Treatment and Transplantation. Each subgroup developed topic-specific questions using the PICO format (Population, Intervention, Comparison, Outcome.54) Questions were reviewed and approved by the entire committee before systematic literature searches were conducted.

demiology and Risk Factors, Screening, Microbiology, Treatment and Transplantation. Each subgroup developed topic-specific questions using the PICO format (Population, Intervention, Comparison, Outcome.54) Questions were reviewed and approved by the entire committee before systematic literature searches were conducted. Review process and consensus vote The members of each subgroup used the PICO questions to guide literature searches in PubMed. Searches were limited to the English language and the period 1984 to 2013. Subgroup members also searched for topic-relevant guidelines through searches of the ATS website, the IDSA website, the Clinical Laboratory Standards Institute (CLSI) website and the UK CF Trust website. After reviewing the relevant literature and existing guidelines, subgroup members drafted recommendation statements. In October 2012, a second meeting was convened and subgroups finalised draft recommendation statements. The committee also voted to set 80% agreement of all 19 members as the threshold for acceptance of a recommendation statement and not to use the GRADE system of evaluating published evidence, given the paucity of clinical trial data.

a second meeting was convened and subgroups finalised draft recommendation statements. The committee also voted to set 80% agreement of all 19 members as the threshold for acceptance of a recommendation statement and not to use the GRADE system of evaluating published evidence, given the paucity of clinical trial data. Each subgroup submitted final draft questions for entry into an electronic survey tool (Survey Monkey) for the purposes of anonymous voting and comment by all members. A project coordinator administered the survey and committee members were asked to rate each statement on a scale of: 0, completely disagree, to 9, completely agree; with 80% or between 7 and 9 being considered ‘good’ agreement. Space for entering free text was also provided after each statement to allow members to cite literature in support of their opinions or suggested revisions. All committee members were required to vote on each statement regardless of their role or expertise. Multiple rounds of voting and revisions to the statements were conducted, and for each round committee members were requested to complete their voting within 3 weeks. The committee chairs reviewed the results from each round and updated the statements based on comments entered by respondents for subsequent rounds.

ise. Multiple rounds of voting and revisions to the statements were conducted, and for each round committee members were requested to complete their voting within 3 weeks. The committee chairs reviewed the results from each round and updated the statements based on comments entered by respondents for subsequent rounds. External review A draft of the recommendations was presented at the 2013 North American Cystic Fibrosis Conference and the European Cystic Fibrosis Society Meeting. Additionally, the committee solicited feedback from the CF communities in the USA and in Europe, which included physicians, nurses, physical and respiratory therapists, parents and individuals with CF. All comments collected from this process were reviewed and addressed by the committee in the development of the final recommendation statements. Results Final recommendations and results of the consensus vote Three rounds of voting were conducted to achieve 80% consensus for each statement. Fifty-three statements were included in the first round of voting and 50 statements in the second and third rounds. Final statements and the consensus are reported in table 1. Table 1 NTM recommendation statements

Results Final recommendations and results of the consensus vote Three rounds of voting were conducted to achieve 80% consensus for each statement. Fifty-three statements were included in the first round of voting and 50 statements in the second and third rounds. Final statements and the consensus are reported in table 1. Table 1 NTM recommendation statements Recommendation Consensus (%) Recommendation 1: The CF Foundation and the ECFS recommend that the potential for cross-infection of NTM (particularly Mycobacterium abscessus complex) between individuals with CF should be minimised by following national infection control guidelines 94 Recommendation 2: The CF Foundation and the ECFS recommend that cultures for NTM be performed annually in spontaneously expectorating individuals with a stable clinical course 94 Recommendation 3: The CF Foundation and the ECFS recommend that, in the absence of clinical features suggestive of NTM pulmonary disease, individuals who are not capable of spontaneously producing sputum do not require screening cultures for NTM 100 Recommendation 4: The CF Foundation and the ECFS recommend that culture and smears for AFB from sputum should be used for NTM screening 100 Recommendation 5: The CF Foundation and the ECFS recommend against the use of oropharyngeal swabs for NTM screening 100 Recommendation 6: The CF Foundation and the ECFS recommend that culture and smears for AFB from sputum, induced sputum, bronchial washings or bronchoalveolar lavage samples can be used to evaluate individuals with CF suspected to have NTM pulmonary disease. 100 Recommendation 7: The CF Foundation and the ECFS recommend against the routine use of transbronchial biopsies to detect NTM in individuals with CF suspected to have NTM pulmonary disease 100 Recommendation 8: The CF Foundation and the ECFS recommend against the use of oropharyngeal swabs to perform diagnostic smears and cultures in individuals with CF suspected to have NTM pulmonary disease 100 Recommendation 9: The CF Foundation and the ECFS recommend that respiratory tract samples should be cultured using both solid and liquid media 100 Recommendation 10: The CF Foundation and the ECFS recommend that the incubation duration for NTM cultures should be for a minimum of 6 weeks 100 Recommendation 11: The CF Foundation and the ECFS recommend that an NTM culture should be processed within 24 h of collection to optimise the detection of NTM in respiratory samples.

ndation 10: The CF Foundation and the ECFS recommend that the incubation duration for NTM cultures should be for a minimum of 6 weeks 100 Recommendation 11: The CF Foundation and the ECFS recommend that an NTM culture should be processed within 24 h of collection to optimise the detection of NTM in respiratory samples. If a delay in processing is anticipated, refrigeration of samples is advised 100 Recommendation 12: The CF Foundation and the ECFS recommend that respiratory tract samples should be decontaminated using the standard N-acetyl l-cysteine, NALC, (0.5%)–NaOH (2%) method 100 Recommendation 13: The CF Foundation and the ECFS recommend that, if a sample remains contaminated with Gram-negative bacteria after standard NALC-NaOH decontamination, it should be further treated with either 5% oxalic acid or 1% chlorhexidine 100 Recommendation 14: The CF Foundation and the ECFS recommend against the use of non-culture-based methods for detecting NTM in respiratory tract samples 100 Recommendation 15: The CF Foundation and the ECFS recommend that all NTM isolates from individuals with CF should undergo molecular identification 100 Recommendation 16: The CF Foundation and the ECFS recommend that all NTM isolates from individuals with CF should be identified to the species level, except for M. intracellulare, M. avium and M. chimaera, where identification can be limited to MAC, and M. abscessus complex, which should be subspeciated 83 Recommendation 17: The CF Foundation and the ECFS recommend that for MAC, clarithromycin susceptibility testing should be performed on an isolate recovered prior to initiation of treatment. Clarithromycin susceptibility testing should also be performed on subsequent isolates if the patient (a) fails to culture convert after 6 months of NTM treatment; (b) recultures MAC after initial culture conversion while on NTM treatment or (c) recultures MAC after completion of NTM treatment 94 Recommendation 18: The CF Foundation and the ECFS recommend that for M.

o be performed on subsequent isolates if the patient (a) fails to culture convert after 6 months of NTM treatment; (b) recultures MAC after initial culture conversion while on NTM treatment or (c) recultures MAC after completion of NTM treatment 94 Recommendation 18: The CF Foundation and the ECFS recommend that for M. abscessus complex, susceptibility testing should include at least clarithromycin, cefoxitin and amikacin (and preferably also tigecycline, imipenem, minocycline, moxifloxacin and linezolid) 89 Recommendation 19: The CF Foundation and the ECFS recommend that drug susceptibility testing should be performed in accordance with CLSI guidelines 100 Recommendation 20: The CF Foundation and the ECFS recommend that ATS/IDSA criteria for the diagnosis of NTM pulmonary disease should be used in individuals with CF (ATS/IDSA 2007 Statement) 100 Recommendation 21: The CF Foundation and the ECFS recommend that other CF pathogens and comorbidities should be considered as potential contributors to a patient's symptoms and radiological features when determining the clinical significance of NTM-positive cultures 100 Recommendation 22: The CF Foundation and the ECFS recommend that NTM treatment should be considered for individuals with CF who have ATS/IDSA defined NTM pulmonary disease 100 Recommendation 23: The CF Foundation and the ECFS recommend that individuals receiving azithromycin as part of their CF medical regimen who have a positive NTM culture should not continue azithromycin treatment while evaluation for NTM disease is underway as azithromycin monotherapy may lead to resistance. A macrolide agent may be included in a multidrug treatment regimen if criteria are met for NTM disease 89 Recommendation 24: The CF Foundation and the ECFS recommend that treatment of M. abscessus complex pulmonary disease should involve an intensive phase followed by a continuation phase 100 Recommendation 25: The CF Foundation and the ECFS recommend that the intensive phase should include a daily oral macrolide (preferably azithromycin) in conjunction with 3–12 weeks of intravenous amikacin and one or more of the following: intravenous tigecycline, imipenem or cefoxitin, guided but not dictated by drug susceptibility testing.

e CF Foundation and the ECFS recommend that the intensive phase should include a daily oral macrolide (preferably azithromycin) in conjunction with 3–12 weeks of intravenous amikacin and one or more of the following: intravenous tigecycline, imipenem or cefoxitin, guided but not dictated by drug susceptibility testing. The duration of intensive phase therapy should be determined by the severity of infection, the response to treatment and the tolerability of the regimen 83 Recommendation 26: The CF Foundation and the ECFS recommend that the continuation phase should include a daily oral macrolide (preferably azithromycin) and inhaled amikacin, in conjunction with 2–3 of the following additional oral antibiotics: minocycline, clofazimine, moxifloxacin and linezolid, guided but not dictated by drug susceptibility testing 89 Recommendation 27: The CF Foundation and the ECFS recommend that individuals with M. abscessus complex pulmonary disease should be managed in collaboration with experts in the treatment of NTM and CF, as drug intolerance and drug-related toxicity occur frequently, and changes in antibiotic therapy are often required 89 Recommendation 28: The CF Foundation and the ECFS recommend that monotherapy with a macrolide or other antimicrobial should never be used in the treatment of M. abscessus complex pulmonary disease 100 Recommendation 29: The CF Foundation and the ECFS recommend the same antibiotic regimen for treatment of all species within the MAC 94 Recommendation 30: The CF Foundation and the ECFS recommend that clarithromycin-sensitive MAC pulmonary disease should be treated with a daily oral antibiotic regimen containing a macrolide (preferably azithromycin), rifampin and ethambutol 89 Recommendation 31: The CF Foundation and the ECFS recommend against the use of intermittent (three times per week) oral antibiotic therapy to treat MAC pulmonary disease 89 Recommendation 32: The CF Foundation and the ECFS recommend that monotherapy with a macrolide or other antimicrobial agent should never be used in the treatment of MAC pulmonary disease 100 Recommendation 33: The CF Foundation and the ECFS recommend that an initial course of intravenous amikacin should be considered for the treatment of MAC pulmonary disease in the presence of one or more of the following: AFB smear positive respiratory tract samples

r be used in the treatment of MAC pulmonary disease 100 Recommendation 33: The CF Foundation and the ECFS recommend that an initial course of intravenous amikacin should be considered for the treatment of MAC pulmonary disease in the presence of one or more of the following: AFB smear positive respiratory tract samples Radiological evidence of lung cavitation or severe infection Systemic signs of illness

r be used in the treatment of MAC pulmonary disease 100 Recommendation 33: The CF Foundation and the ECFS recommend that an initial course of intravenous amikacin should be considered for the treatment of MAC pulmonary disease in the presence of one or more of the following: AFB smear positive respiratory tract samples Radiological evidence of lung cavitation or severe infection Systemic signs of illness 94 Recommendation 34: The CF Foundation and the ECFS recommend that clarithromycin-resistant MAC pulmonary disease should be managed in collaboration with experts in the treatment of NTM and CF 89 Recommendation 35: The CF Foundation and the ECFS recommend that individuals with CF receiving NTM treatment should have expectorated or induced sputum samples sent for NTM culture every 4–8 weeks throughout the entire course of treatment to assess the microbiological response 94 Recommendation 36: The CF Foundation and the ECFS recommend that a schedule for detecting drug toxicity (including hearing loss, visual loss, renal impairment and liver function test abnormalities) should be set in place at the time of NTM treatment initiation and implemented throughout treatment based on the specific drugs prescribed 100 Recommendation 37: The CF Foundation and the ECFS recommend that an HRCT scan of the lungs should be performed shortly before starting NTM treatment and at the end of NTM treatment to assess the radiological response 94 Recommendation 38: The CF Foundation and the ECFS recommend that NTM antibiotic therapy should be prescribed for 12 months beyond culture conversion (defined as three consecutive negative cultures, with the time of conversion being the date of the first of the three negative cultures) as long as no positive cultures are obtained during those 12 months 94 Recommendation 39: The CF Foundation and the ECFS recommend that individuals who fail to culture convert despite optimal NTM therapy may benefit from long-term suppressive antibiotic treatment 94 Recommendation 40: The CF Foundation and the ECFS recommend that, when amikacin is given intravenously or when streptomycin is given intravenously or intramuscularly, serum levels should be monitored and dosing adjusted to minimise ototoxicity and nephrotoxicity 100 Recommendation 41: The CF Foundation and the ECFS recommend against routinely obtaining serum levels of other anti-mycobacterial drugs. However, absorption of oral medications is often reduced in CF.

or intramuscularly, serum levels should be monitored and dosing adjusted to minimise ototoxicity and nephrotoxicity 100 Recommendation 41: The CF Foundation and the ECFS recommend against routinely obtaining serum levels of other anti-mycobacterial drugs. However, absorption of oral medications is often reduced in CF. Therefore use of therapeutic drug monitoring should be considered for individuals failing to improve despite taking recommended drug regimens or for those on concomitant medications with significant interactions with NTM drugs 100 Recommendation 42: The CF Foundation and the ECFS recommend against the use of interferon γ as adjuvant therapy for NTM pulmonary disease in individuals with CF 89 Recommendation 43: The CF Foundation and the ECFS recommend that vitamin D should be supplemented according to national CF care guidelines 94 Recommendation 44: The CF Foundation and the ECFS recommend that lung resection should only be considered under extraordinary circumstances and in consultation with experts on the treatment of NTM and CF 83 Recommendation 45: The CF Foundation and the ECFS recommend that all individuals with CF being considered for lung transplantation should be evaluated for NTM pulmonary disease 100 Recommendation 46: The CF Foundation and the ECFS recommend that the presence of current or previous respiratory tract samples positive for NTM should not preclude individuals being considered for lung transplantation 94 Recommendation 47: The CF Foundation and the ECFS recommend that individuals with CF who have NTM pulmonary disease and are being evaluated for transplantation should start treatment prior to transplant listing 100 Recommendation 48: The CF Foundation and the ECFS recommend that individuals with CF receiving NTM treatment with sequential negative cultures may be eligible for transplant listing 100 Recommendation 49: The CF Foundation and the ECFS recommend that individuals with CF who have completed treatment for NTM pulmonary disease with apparent eradication of the organism may be eligible for transplant listing 100 Recommendation 50: The CF Foundation and the ECFS recommend that the presence of persistent M. abscessus complex or MAC infection despite optimal therapy is not an absolute contraindication to lung transplant referral 94 AFB, acid-fast bacilli; CF, cystic fibrosis; CLSI, Clinical Laboratory Standards Institute; ECFS, European Cystic Fibrosis Society; HRCT, High-resolution CT; MAC, M.

he presence of persistent M. abscessus complex or MAC infection despite optimal therapy is not an absolute contraindication to lung transplant referral 94 AFB, acid-fast bacilli; CF, cystic fibrosis; CLSI, Clinical Laboratory Standards Institute; ECFS, European Cystic Fibrosis Society; HRCT, High-resolution CT; MAC, M. avium complex; NTM, non-tuberculous mycobacteria. Risk factors Are there modifiable risk factors for the development of NTM-PD in individuals with CF? Recommendation 1: The CF Foundation and the ECFS recommend that the potential for cross-infection of NTM (particularly MABSC) between individuals with CF should be minimised by following national infection control guidelines. CF-related lung disease is a clear risk factor for the development of NTM-PD and is presumed to relate to the presence of structural lung damage, impaired mucociliary clearance and inflamed airways; all of which are thought to favour the development of chronic NTM infection.55 Cystic Fibrosis Transmembrane conductance Regulator (CFTR) dysfunction may, of itself, predispose to NTM infection (although the pathophysiology is unknown), since rates of heterozygosity for CFTR mutations within the non-CF population with pulmonary NTM disease are high (30–50%).56 57 However, other risk factors that predispose specific individuals with CF to acquire NTM or to develop NTM-PD are, for the most part, poorly understood, with many studies presenting conflicting results. Potential risk factors for NTM acquisition are listed below.

Risk factors Are there modifiable risk factors for the development of NTM-PD in individuals with CF? Recommendation 1: The CF Foundation and the ECFS recommend that the potential for cross-infection of NTM (particularly MABSC) between individuals with CF should be minimised by following national infection control guidelines. CF-related lung disease is a clear risk factor for the development of NTM-PD and is presumed to relate to the presence of structural lung damage, impaired mucociliary clearance and inflamed airways; all of which are thought to favour the development of chronic NTM infection.55 Cystic Fibrosis Transmembrane conductance Regulator (CFTR) dysfunction may, of itself, predispose to NTM infection (although the pathophysiology is unknown), since rates of heterozygosity for CFTR mutations within the non-CF population with pulmonary NTM disease are high (30–50%).56 57 However, other risk factors that predispose specific individuals with CF to acquire NTM or to develop NTM-PD are, for the most part, poorly understood, with many studies presenting conflicting results. Potential risk factors for NTM acquisition are listed below. Lung function There have been conflicting reports on whether an individual's spirometry results are related to the likelihood of finding NTM-positive samples, with some studies suggesting no association with lung function,13 a positive association of NTM acquisition with higher FEV1% predicted6 or, conversely, with worse lung function.11 15 30 Support for the possibility that NTM acquisition is more likely in CF individuals with severe lung disease comes from observations that the prevalence of NTM-positive sputum samples in patients referred for lung transplantation has been reported to be as high as 19.7%.29

r, conversely, with worse lung function.11 15 30 Support for the possibility that NTM acquisition is more likely in CF individuals with severe lung disease comes from observations that the prevalence of NTM-positive sputum samples in patients referred for lung transplantation has been reported to be as high as 19.7%.29 Lung infection with specific pathogens In some studies, individuals with CF with NTM-positive samples are more likely to have Staphylococcus aureus infection and less likely to have Pseudomonas aeruginosa chronic pulmonary infection.6 7 58 Other studies, however, have reported NTM positivity associated with higher rates of P. aeruginosa infection,11 and variably associated with S. maltophilia infection.6 58 In contrast, Aspergillus fumigatus has consistently been associated with the presence of NTM-positive cultures,11 15 59 with some reports indicating an association with allergic bronchopulmonary aspergillosis.7 27 60 Medications Corticosteroids The impact of systemic steroids on NTM acquisition is controversial. There have been suggestions that steroids may protect against58 or predispose towards NTM infection,60 or may not influence the risk of NTM acquisition.4 11 12 Recent data from non-CF populations, however, have suggested that oral as well as some types of inhaled corticosteroids are associated with increased risk of NTM acquisition.61–63

gestions that steroids may protect against58 or predispose towards NTM infection,60 or may not influence the risk of NTM acquisition.4 11 12 Recent data from non-CF populations, however, have suggested that oral as well as some types of inhaled corticosteroids are associated with increased risk of NTM acquisition.61–63 Proton pump inhibitors The impact of proton pump inhibitor (PPI) is unclear. PPI use has been reported to be associated with the development of MAC pulmonary disease in non-CF cohorts,64 and may promote gastrointestinal survival of NTM and subsequent lung infection through gastric aspiration. Azithromycin Particular attention has recently been paid to the role of long-term azithromycin use as a risk factor for the acquisition of NTM. In a single centre study of CF adults, Renna et al43 reported increases in annual rates of NTM infection associated with chronic azithromycin use, postulating, through in vitro studies and mouse infection models, that azithromycin blocked autophagic killing of NTM within macrophages. While supporting findings from a previous case–control study reporting increased azithromycin use in individuals with NTM,11 other large retrospective studies have shown no such association.12 13 59 65–67 This includes a recent nested case–control analysis within the CF registry, which suggested long-term azithromycin use may protect against infection with NTM.67

e–control study reporting increased azithromycin use in individuals with NTM,11 other large retrospective studies have shown no such association.12 13 59 65–67 This includes a recent nested case–control analysis within the CF registry, which suggested long-term azithromycin use may protect against infection with NTM.67 Acquisition of NTM through cross-infection Person-to-person transmission of NTM has traditionally been considered unlikely. Two separate studies have shown that patients, even siblings living in the same household for more than 10 years, have unique strains,7 68 suggesting a lack of person-to-person transmission. However, a case report from the University of Washington described a possible outbreak of M. a. massiliense in five patients48 with potential transmission occurring during synchronous clinic visits. Recently, whole genome sequencing and antimicrobial susceptibility testing performed on 168 consecutive isolates of M. abscessus from 31 patients attending an adult CF centre in the UK revealed frequent, probably indirect, transmission of M. a. massiliense between individuals with CF despite conventional cross-infection measures.69 The results of these studies indicate that cross-infection may be an important mechanism for the acquisition of M. abscessus (at least within the CF population). To date, there has been no published evidence suggesting person-to-person transmission of other NTM species.

despite conventional cross-infection measures.69 The results of these studies indicate that cross-infection may be an important mechanism for the acquisition of M. abscessus (at least within the CF population). To date, there has been no published evidence suggesting person-to-person transmission of other NTM species. Other factors extrapolated from data in non-CF populations or studies on M. tuberculosis that might contribute to NTM acquisition in individuals with CF include: low vitamin D,70 71 the presence of gastro-oesophageal reflux disease,64 72 low body mass index56 73 or malnutrition.74 Screening How often should individuals with CF be screened for NTM? Recommendation 2: The CF Foundation and the ECFS recommend that cultures for NTM be performed annually in spontaneously expectorating individuals with a stable clinical course. Recommendation 3: The CF Foundation and the ECFS recommend that, in the absence of clinical features suggestive of NTM-PD, individuals who are not capable of spontaneously producing sputum do not require screening cultures for NTM.

Screening How often should individuals with CF be screened for NTM? Recommendation 2: The CF Foundation and the ECFS recommend that cultures for NTM be performed annually in spontaneously expectorating individuals with a stable clinical course. Recommendation 3: The CF Foundation and the ECFS recommend that, in the absence of clinical features suggestive of NTM-PD, individuals who are not capable of spontaneously producing sputum do not require screening cultures for NTM. Over the past two decades, a number of expert opinions and reviews have urged routine screening for NTM in the general CF population. However, the optimal frequency and methodology for NTM surveillance in individuals with CF are not known. NTM are common in the environment, and are likely to be transiently introduced on a regular basis into the airways of individuals with CF. More frequent screening will, therefore, result in detection of more positive cultures,11 many of which will not be associated with the presence of NTM-PD,6 30 58 generating anxiety in patients and caregivers and initiating further (potentially invasive) investigations. However, signs and symptoms of NTM disease are often subtle and non-specific, and the diagnosis can be delayed for years or missed altogether in the absence of effective surveillance.4 Furthermore, systematic screening may help researchers more accurately identify factors influencing poorly understood host susceptibility, acquisition, transmission and virulence of NTM. It is important to emphasise that screening refers to obtaining samples from individuals with no clinical, microbiological or radiological suspicion of NTM infection, and should be distinguished from strategies to investigate and diagnose NTM disease (covered in Diagnosis of NTM-PD in CF section).

and virulence of NTM. It is important to emphasise that screening refers to obtaining samples from individuals with no clinical, microbiological or radiological suspicion of NTM infection, and should be distinguished from strategies to investigate and diagnose NTM disease (covered in Diagnosis of NTM-PD in CF section). While our understanding of those factors predisposing individuals with CF to NTM infection is incomplete, there is, nevertheless, agreement that certain patient populations are at greater risk and therefore probably require more frequent surveillance. These populations include: those with advanced lung disease and previous NTM-positive cultures, and those living in areas with high NTM prevalence. Conversely, in individuals with no recognised risk factors, the prevalence of NTM infection is likely to be low; thus less frequent, perhaps annual, surveillance is warranted. In addition, NTM screening is important before starting long-term azithromycin treatment to avoid inadvertent macrolide monotherapy in individuals with undiagnosed NTM infection (in keeping with published guidelines.75) How should screening for NTM be performed? Recommendation 4: The CF Foundation and the ECFS recommend that culture and smears for acid-fast bacilli (AFB) from sputum should be used for NTM screening. Recommendation 5: The CF Foundation and the ECFS recommend against the use of oropharyngeal swabs for NTM screening.

While our understanding of those factors predisposing individuals with CF to NTM infection is incomplete, there is, nevertheless, agreement that certain patient populations are at greater risk and therefore probably require more frequent surveillance. These populations include: those with advanced lung disease and previous NTM-positive cultures, and those living in areas with high NTM prevalence. Conversely, in individuals with no recognised risk factors, the prevalence of NTM infection is likely to be low; thus less frequent, perhaps annual, surveillance is warranted. In addition, NTM screening is important before starting long-term azithromycin treatment to avoid inadvertent macrolide monotherapy in individuals with undiagnosed NTM infection (in keeping with published guidelines.75) How should screening for NTM be performed? Recommendation 4: The CF Foundation and the ECFS recommend that culture and smears for acid-fast bacilli (AFB) from sputum should be used for NTM screening. Recommendation 5: The CF Foundation and the ECFS recommend against the use of oropharyngeal swabs for NTM screening. The majority of published reports describing the prevalence of NTM in the CF population utilised AFB smear and culture from sputum as the standard screening method.4 6 7 11 13 17 To date, there has been no direct comparison between the sensitivity of samples from spontaneously expectorated sputum samples, and sputum induced by use of hypertonic saline. Analysis of induced sputum provides equal or better detection of ‘standard’ CF pathogens76 and the procedure is in widespread use to collect samples for mycobacterial culture among CF Centres worldwide. However, the Consensus Committee felt that, due to its inconvenience, induced sputum collection should not be used as a screening tool in individuals with no features suggestive of NTM-PD who are incapable of spontaneously producing sputum. As discussed in Microbiology section, there are currently no other validated screening methods to detect NTM in individuals with CF. Although positive cultures have been detected through laryngeal suction, oropharyngeal swabs, or gastric aspirate, there are insufficient data to support their use. Skin testing for delayed-type hypersensitivity against NTM antigens does not appear sufficiently sensitive or specific to use for surveillance in the CF population. Serological assays, such as IgG against Mycobacterium antigen A60 for NTM surveillance, appear promising,42 but have not been validated in the CF population.

in testing for delayed-type hypersensitivity against NTM antigens does not appear sufficiently sensitive or specific to use for surveillance in the CF population. Serological assays, such as IgG against Mycobacterium antigen A60 for NTM surveillance, appear promising,42 but have not been validated in the CF population. Microbiology What respiratory tract samples should be used to evaluate individuals with CF for suspected NTM-PD? Recommendation 6: The CF Foundation and the ECFS recommend that culture and smears for AFB from sputum, induced sputum, bronchial washings or bronchoalveolar lavage samples can be used to evaluate individuals with CF suspected to have NTM-PD. Recommendation 7: The CF Foundation and the ECFS recommend against the routine use of transbronchial biopsies to detect NTM in individuals with CF suspected to have NTM-PD. Recommendation 8: The CF Foundation and the ECFS recommend against the use of oropharyngeal swabs to perform diagnostic smears and cultures in individuals with CF suspected to have NTM-PD.

Recommendation 7: The CF Foundation and the ECFS recommend against the routine use of transbronchial biopsies to detect NTM in individuals with CF suspected to have NTM-PD. Recommendation 8: The CF Foundation and the ECFS recommend against the use of oropharyngeal swabs to perform diagnostic smears and cultures in individuals with CF suspected to have NTM-PD. Currently, sputum, induced sputum, bronchial washings and bronchoalveolar lavage samples are routinely used to evaluate individuals for suspected NTM-PD.77 Samples for NTM should be processed for smear microscopy, preferably by fluorescence, and for culture. Microscopy allows for direct evaluation of the bacterial burden, and may indicate false-negative culture results through excessive sample decontamination or overgrowth of conventional bacteria. Oropharyngeal swabs should not be used for the detection of NTM, since they do not consistently provide sufficient material for culture.77

llows for direct evaluation of the bacterial burden, and may indicate false-negative culture results through excessive sample decontamination or overgrowth of conventional bacteria. Oropharyngeal swabs should not be used for the detection of NTM, since they do not consistently provide sufficient material for culture.77 A staged approach should be adopted for obtaining diagnostic samples; testing spontaneously expectorated or induced sputum (if available) before resorting to bronchoscopy. Although there are no published studies comparing the relative performance of these different methods for detection of NTM, the presence of negative sputum samples in individuals with radiological and clinical suspicion of NTM disease should prompt CT-guided bronchoscopic sampling, as, for example, in nodular bronchiectatic disease.78–80 While trans-bronchial biopsies can reveal NTM (on microscopy or culture) and may demonstrate granulomatous inflammation (supporting NTM disease rather than transient colonisation), they should not be obtained routinely in individuals with CF given the significant risks of bleeding and pneumothorax.81 How should respiratory tract samples from individuals with CF be cultured for NTM? Recommendation 9: The CF Foundation and the ECFS recommend that respiratory tract samples should be cultured using both solid and liquid media. Recommendation 10: The CF Foundation and the ECFS recommend that the incubation duration for NTM cultures should be for a minimum of 6 weeks.

How should respiratory tract samples from individuals with CF be cultured for NTM? Recommendation 9: The CF Foundation and the ECFS recommend that respiratory tract samples should be cultured using both solid and liquid media. Recommendation 10: The CF Foundation and the ECFS recommend that the incubation duration for NTM cultures should be for a minimum of 6 weeks. Recommendation 11: The CF Foundation and the ECFS recommend that an NTM culture should be processed within 24 h of collection to optimise the detection of NTM in respiratory samples. If a delay in processing is anticipated, refrigeration of samples is advised.

Recommendation 10: The CF Foundation and the ECFS recommend that the incubation duration for NTM cultures should be for a minimum of 6 weeks. Recommendation 11: The CF Foundation and the ECFS recommend that an NTM culture should be processed within 24 h of collection to optimise the detection of NTM in respiratory samples. If a delay in processing is anticipated, refrigeration of samples is advised. The most sensitive and rapid way to detect viable mycobacteria is to culture samples (following decontamination to remove conventional bacteria and fungi) in liquid media using an automated growth detection system (such as Mycobacteria Growth Indicator Tube (MGIT)77 82 83); a process widely used around the world. However, concomitant culture on solid media may increase the diagnostic yield since NTM can be detected despite incomplete sample decontamination.84 Since decontamination procedures substantially reduce the viability of mycobacteria in samples, attempts have been made to use highly selective agar for solid culture of unprocessed sputum. A recent study, using agar designed for Burkholderia cepacia complex culture,84 demonstrated an improvement in detection of rapidly growing mycobacteria from 0.7% with conventional liquid culture to 2.8%. The duration, both of liquid and solid culture methods, has not been rigorously tested but the vast majority of pathogenic NTM will grow by 6 weeks—the current recommended duration in US and European laboratories.77

ovement in detection of rapidly growing mycobacteria from 0.7% with conventional liquid culture to 2.8%. The duration, both of liquid and solid culture methods, has not been rigorously tested but the vast majority of pathogenic NTM will grow by 6 weeks—the current recommended duration in US and European laboratories.77 Laboratory processing of samples should ideally be performed within 24 h of collection to avoid overgrowth by conventional bacteria, which can reduce NTM viability85 and prevent successful decontamination.85 Studies have shown that refrigeration of samples may improve NTM detection from sputum samples86 and should be considered if delays longer than 24 h in processing are anticipated. How should respiratory tract samples from individuals with CF be decontaminated to optimise the detection of NTM? Recommendation 12: The CF Foundation and the ECFS recommend that respiratory tract samples should be decontaminated using the standard N-acetyl l-cysteine, NALC, (0.5%)-NaOH (2%) method. Recommendation 13: The CF Foundation and the ECFS recommend that, if a sample remains contaminated with Gram-negative bacteria after standard NALC-NaOH decontamination, it should be further treated with either 5% oxalic acid or 1% chlorhexidine.

How should respiratory tract samples from individuals with CF be decontaminated to optimise the detection of NTM? Recommendation 12: The CF Foundation and the ECFS recommend that respiratory tract samples should be decontaminated using the standard N-acetyl l-cysteine, NALC, (0.5%)-NaOH (2%) method. Recommendation 13: The CF Foundation and the ECFS recommend that, if a sample remains contaminated with Gram-negative bacteria after standard NALC-NaOH decontamination, it should be further treated with either 5% oxalic acid or 1% chlorhexidine. Adequate sample decontamination to remove conventional bacteria and fungi is essential to permit culture-based detection of mycobacteria,77 87 88 but often fails in CF samples given high densities of P. aeruginosa and other microbes.39–41 89 90 Since enhanced decontamination protocols adversely impact on NTM viability in samples,90 a two-step approach to sample processing should be adopted.41 Virtually all US and European clinical microbiology laboratories currently use an NALC-NaOH decontamination step prior to mycobacterial culture.41 87 88

obes.39–41 89 90 Since enhanced decontamination protocols adversely impact on NTM viability in samples,90 a two-step approach to sample processing should be adopted.41 Virtually all US and European clinical microbiology laboratories currently use an NALC-NaOH decontamination step prior to mycobacterial culture.41 87 88 The addition of a second decontamination step using oxalic acid has been shown to permit the recovery of NTM from persistently contaminated samples albeit with reduced sensitivity.40 Alternatively, use of 1% chlorhexidine as a first step may improve the recovery of mycobacteria, but at the expense of higher rates of residual sample contamination.89 Chlorhexidine negatively affects the performance of the MGIT automated liquid culture system, because it needs to be neutralised with lecithin; lecithin generates random fluorescence reactions from the MGIT system sensor, limiting its use.89 Should non-culture-based methods be used to detect NTM in respiratory tract samples from individuals with CF? Recommendation 14: The CF Foundation and the ECFS recommend against the use of non-culture-based methods for detecting NTM in respiratory tract samples.

The addition of a second decontamination step using oxalic acid has been shown to permit the recovery of NTM from persistently contaminated samples albeit with reduced sensitivity.40 Alternatively, use of 1% chlorhexidine as a first step may improve the recovery of mycobacteria, but at the expense of higher rates of residual sample contamination.89 Chlorhexidine negatively affects the performance of the MGIT automated liquid culture system, because it needs to be neutralised with lecithin; lecithin generates random fluorescence reactions from the MGIT system sensor, limiting its use.89 Should non-culture-based methods be used to detect NTM in respiratory tract samples from individuals with CF? Recommendation 14: The CF Foundation and the ECFS recommend against the use of non-culture-based methods for detecting NTM in respiratory tract samples. A number of studies have been published on the use of PCR-based detection methods for NTM from respiratory samples.91–95 To date, however, none have been robustly evaluated for CF sputum samples, nor have they demonstrated sufficiently high sensitivity and specificity on smear-negative samples91 to recommend their routine diagnostic use. Furthermore, the clinical significance of PCR-positive respiratory samples is currently unknown. How should NTM isolates from individuals with CF be identified? Recommendation 15: The CF Foundation and the ECFS recommend that all NTM isolates from individuals with CF should undergo molecular identification.

A number of studies have been published on the use of PCR-based detection methods for NTM from respiratory samples.91–95 To date, however, none have been robustly evaluated for CF sputum samples, nor have they demonstrated sufficiently high sensitivity and specificity on smear-negative samples91 to recommend their routine diagnostic use. Furthermore, the clinical significance of PCR-positive respiratory samples is currently unknown. How should NTM isolates from individuals with CF be identified? Recommendation 15: The CF Foundation and the ECFS recommend that all NTM isolates from individuals with CF should undergo molecular identification. Recommendation 16: The CF Foundation and the ECFS recommend that all NTM isolates from individuals with CF should be identified to the species level, except for M. intracellulare, M. avium and M. chimaera, where identification can be limited to MAC and MABSC, which should be subspeciated.

How should NTM isolates from individuals with CF be identified? Recommendation 15: The CF Foundation and the ECFS recommend that all NTM isolates from individuals with CF should undergo molecular identification. Recommendation 16: The CF Foundation and the ECFS recommend that all NTM isolates from individuals with CF should be identified to the species level, except for M. intracellulare, M. avium and M. chimaera, where identification can be limited to MAC and MABSC, which should be subspeciated. As individual NTM species differ in their potential to cause clinical disease in humans96 and in their response to specific antibiotics, correct species identification of NTM isolates is clinically important. Moreover, in the case of M. abscessus, the ability to identify isolates to the subspecies level (M. a. abscessus, M. a. bolletii, M. a. massiliense) may predict treatment response97 and potentially permit targeted therapy.98 M. a. massiliense harbours a partial erm41 gene deletion, preventing inducible macrolide resistance,97 99 and leads to more successful outcomes with macrolide-based antibiotic regimens than in infections with M. a. abscessus (which has a full length, functional erm41 gene).97

and potentially permit targeted therapy.98 M. a. massiliense harbours a partial erm41 gene deletion, preventing inducible macrolide resistance,97 99 and leads to more successful outcomes with macrolide-based antibiotic regimens than in infections with M. a. abscessus (which has a full length, functional erm41 gene).97 There is no gold standard for NTM species identification. Molecular methods have now surpassed biochemical tests for NTM identification in many laboratories.100–107 Although matrix-assisted laser desorption ionisation-time of flight mass spectrometry has shown promise in providing rapid speciation of NTM,108–112 the optimal method for protein extraction from mycobacteria and the exact discriminatory power of this method have yet to be established.

tion in many laboratories.100–107 Although matrix-assisted laser desorption ionisation-time of flight mass spectrometry has shown promise in providing rapid speciation of NTM,108–112 the optimal method for protein extraction from mycobacteria and the exact discriminatory power of this method have yet to be established. Among molecular methods, three techniques are in current clinical use. The first includes line probe assays,103–105 113 which are easy to perform but costly, and permits accurate identification of the most frequently encountered NTM species but not subspeciation of M. abscessus. The second technique is PCR product restriction analysis in which amplified gene fragments are restriction digested to yield different sized fragments, which are then resolved by gel electrophoresis and correlated with specific species.114 This technique is mostly used in low-resource settings and is at least comparable to the line probe assays.106 The third technique is (partial) gene sequencing, which permits a higher level of discrimination, often to subspecies level, but is only available in laboratories with access to sequencing facilities. The choice of the optimal sequencing strategy is not straightforward. Although partial 16S ribosomal RNA (rRNA) gene sequencing provides insufficient discrimination, particularly between M. abscessus and M. chelonae,115 a number of other gene sequences (such as partial hsp65 and rpoB gene sequences) have been successfully used.107 116 For subspeciation of M. abscessus, a multilocus sequence typing approach has recently been validated.116–118 An alternative strategy close to subspeciation is to measure erm gene associated inducible macrolide resistance by phenotypic drug susceptibility testing (DST). This does not distinguish accurately between M. abscessus subspecies but does offer the data for which the subspeciation is generally performed—whether or not there is inducible macrolide resistance.

iation is to measure erm gene associated inducible macrolide resistance by phenotypic drug susceptibility testing (DST). This does not distinguish accurately between M. abscessus subspecies but does offer the data for which the subspeciation is generally performed—whether or not there is inducible macrolide resistance. Should DST be performed on NTM isolates from individuals with CF? Recommendation 17: The CF Foundation and the ECFS recommend that for MAC, clarithromycin susceptibility testing should be performed on an isolate recovered prior to initiation of treatment. Clarithromycin susceptibility testing should also be performed on subsequent isolates if the patient (a) fails to culture convert after 6 months of NTM treatment; (b) recultures MAC after initial culture conversion while on NTM treatment or (c) recultures MAC after completion of NTM treatment. Recommendation 18: The CF Foundation and the ECFS recommend that for MABSC, susceptibility testing should include at least clarithromycin, cefoxitin and amikacin (and preferably also tigecycline, imipenem, minocycline, moxifloxacin and linezolid). Recommendation 19: The CF Foundation and the ECFS recommend that DST should be performed in accordance with CLSI guidelines.

Recommendation 18: The CF Foundation and the ECFS recommend that for MABSC, susceptibility testing should include at least clarithromycin, cefoxitin and amikacin (and preferably also tigecycline, imipenem, minocycline, moxifloxacin and linezolid). Recommendation 19: The CF Foundation and the ECFS recommend that DST should be performed in accordance with CLSI guidelines. Based on current published data, the exact role of DST and its potential to guide regimen selection and predict outcomes in NTM lung disease in patients with CF, remains unknown.119 The CLSI has published guidelines on DST of NTM.17 120 121 Its European counterpart, the European Committee on Antimicrobial Susceptibility Testing (EUCAST), presently has no guidelines for DST of NTM.77 It is important to appreciate that, although CLSI guidelines provide breakpoint concentrations to interpret minimum inhibitory concentrations (MICs) as ‘susceptible’ or ‘resistant’, these cut-offs have had very limited clinical validation, and no clinical validation has been performed in patients with CF. Moreover, limited pharmacokinetic (PK) data are now available for MAC lung disease to support breakpoint concentrations,122 there are no representative PK or pharmacodynamic data to guide treatment of patients with CF.

red only in MAC isolates that have mutations associated with amikacin resistance, that is, in the 16S rRNA gene. These strains are cultured from patients with significant aminoglycoside exposure, such as individuals with CF, and for disease caused by these strains, amikacin is unlikely to have any beneficial effect.127 For rapidly growing mycobacteria including M. abscessus, clinical validation has only been performed in series of extra-pulmonary disease,128 and only for cefoxitin, aminoglycosides and co-trimoxazole. In series of M. abscessus lung disease, the outcomes of macrolide-based treatment are generally poor and do not correlate well with in vitro susceptibilities119 129 potentially due to erm41-dependent inducible macrolide resistance and relative short duration of adequate regimens, which were often interrupted because of toxicity. Indeed, in the absence of a functional erm41 gene, response to macrolide-containing treatments has been good.94 The CLSI has recommended routine testing for inducible macrolide resistance by performing extended incubation of isolates in the presence of clarithromycin, as inducible resistance may predict treatment failure.120 For M. simiae, the role of DST is unknown, although the generally poor outcomes of treatment have been correlated with a lack of synergistic activity between rifampicin and ethambutol, an in vitro observation that still awaits clinical validation.130 Some molecular methods to assess drug susceptibility exist, but are not yet routinely available. For example, sequencing of the 16S rRNA and 23S rRNA genes can reveal mutations associated with high-level resistance to aminoglycosides and macrolides, respectively.119 127

rvation that still awaits clinical validation.130 Some molecular methods to assess drug susceptibility exist, but are not yet routinely available. For example, sequencing of the 16S rRNA and 23S rRNA genes can reveal mutations associated with high-level resistance to aminoglycosides and macrolides, respectively.119 127 Diagnosis of NTM-PD in CF Should the ATS/IDSA criteria for the diagnosis of NTM-PD be used in individuals with CF? Recommendation 20: The CF Foundation and the ECFS recommend that ATS/IDSA criteria for the diagnosis of NTM-PD should be used in individuals with CF (ATS/IDSA 2007 Statement). Recommendation 21: The CF Foundation and the ECFS recommend that other CF pathogens and comorbidities should be considered as potential contributors to a patient's symptoms and radiological features when determining the clinical significance of NTM-positive cultures. Recommendation 22: The CF Foundation and the ECFS recommend that NTM treatment should be considered for individuals with CF who have ATS/IDSA defined NTM-PD. Recommendation 23: The CF Foundation and the ECFS recommend that individuals receiving azithromycin as part of their CF medical regimen who have a positive NTM culture should not continue azithromycin treatment while evaluation for NTM disease is underway, as azithromycin monotherapy may lead to resistance. A macrolide agent may be included in a multidrug treatment regimen if criteria are met for NTM disease.

romycin as part of their CF medical regimen who have a positive NTM culture should not continue azithromycin treatment while evaluation for NTM disease is underway, as azithromycin monotherapy may lead to resistance. A macrolide agent may be included in a multidrug treatment regimen if criteria are met for NTM disease. In contrast to M. tuberculosis, a single positive culture of NTM does not necessarily indicate that an individual has NTM-PD. To address the difficulty of making a diagnosis of NTM-PD, the ATS/IDSA proposed a set of clinical, radiological and microbiological criteria required to define an individual as having NTM-PD (ref 22; box 1). Although these criteria have not been validated for individuals with CF, they have been widely adopted by NTM specialists around the world and provide an operational definition for NTM-PD, which supports clinical decision-making and facilitates research. The Statements Committee therefore concluded that, in the absence of an alternate, CF-validated definition, the ATS/IDSA criteria should be used for the definition of NTM-PD in individuals with CF. Box 1 ATS/IDSA clinical and microbiologic criteria for diagnosing non-tuberculous mycobacterial pulmonary disease (NTM-PD) (based on ref 22) Clinical (both required) Pulmonary symptoms with nodular or cavitary opacities on chest radiograph, or a high-resolution CT scan that shows multifocal bronchiectasis with multiple small nodules. Appropriate exclusion of other diagnoses.

In contrast to M. tuberculosis, a single positive culture of NTM does not necessarily indicate that an individual has NTM-PD. To address the difficulty of making a diagnosis of NTM-PD, the ATS/IDSA proposed a set of clinical, radiological and microbiological criteria required to define an individual as having NTM-PD (ref 22; box 1). Although these criteria have not been validated for individuals with CF, they have been widely adopted by NTM specialists around the world and provide an operational definition for NTM-PD, which supports clinical decision-making and facilitates research. The Statements Committee therefore concluded that, in the absence of an alternate, CF-validated definition, the ATS/IDSA criteria should be used for the definition of NTM-PD in individuals with CF. Box 1 ATS/IDSA clinical and microbiologic criteria for diagnosing non-tuberculous mycobacterial pulmonary disease (NTM-PD) (based on ref 22) Clinical (both required) Pulmonary symptoms with nodular or cavitary opacities on chest radiograph, or a high-resolution CT scan that shows multifocal bronchiectasis with multiple small nodules. Appropriate exclusion of other diagnoses. Microbiologic (one of the following required) Positive culture results from at least two expectorated sputum samples. If the results from samples are non-diagnostic, consider repeat sputum acid-fast bacilli (AFB) smears and cultures. Positive culture results from at least one bronchial wash or lavage.

Appropriate exclusion of other diagnoses. Microbiologic (one of the following required) Positive culture results from at least two expectorated sputum samples. If the results from samples are non-diagnostic, consider repeat sputum acid-fast bacilli (AFB) smears and cultures. Positive culture results from at least one bronchial wash or lavage. Transbronchial or other lung biopsy with mycobacterial histopathological features (granulomatous inflammation or AFB) and positive culture for NTM or biopsy showing mycobacterial histopathological features (granulomatous inflammation or AFB) and one or more sputum or bronchial washings that are culture positive for NTM. Expert consultation should be obtained when either infrequently encountered NTM or those usually representing environmental contamination are recovered. Patients who are suspected of having NTM-PD but who do not meet the diagnostic criteria should be followed until the diagnosis is firmly established or excluded. Making the diagnosis of NTM-PD does not, per se, necessitate the institution of therapy, which is a decision based on potential risks and benefits of therapy for individual patients.

Patients who are suspected of having NTM-PD but who do not meet the diagnostic criteria should be followed until the diagnosis is firmly established or excluded. Making the diagnosis of NTM-PD does not, per se, necessitate the institution of therapy, which is a decision based on potential risks and benefits of therapy for individual patients. Microbiological criteria for NTM-PD Individuals should have two or more positive sputum cultures of the same NTM species or one positive culture from bronchoscopic lavage or wash. The threshold for the number of positive sputum samples is derived from an observational study of individuals without CF with MAC in which 98% individuals with at least two positive sputum cultures developed progressive radiographic change compared to only 2% with one positive culture.131 The type of NTM species isolated is also important. Thus, isolation of M. abscessus is more likely to reflect NTM-PD than culturing usually non-pathogenic species such as M. gordonae and M. terrae complex.

sitive sputum cultures developed progressive radiographic change compared to only 2% with one positive culture.131 The type of NTM species isolated is also important. Thus, isolation of M. abscessus is more likely to reflect NTM-PD than culturing usually non-pathogenic species such as M. gordonae and M. terrae complex. Radiological criteria for NTM-PD In the context of CF-related lung disease, a chest radiograph is unlikely to be of use for the investigation of NTM-PD. High-resolution CT (HRCT) scan changes supporting a diagnosis of NTM-PD would include: inflammatory nodules, new tree-in-bud opacities (particularly in areas of mild underlying bronchiectasis) and cavitation.132 However, these changes are non-specific, particularly in individuals with severe CF-related lung disease, and may reflect infection with more common CF pathogens, inadequate airway clearance or the development of allergic bronchopulmonary aspergillosis (ABPA). Clinical criteria for NTM-PD NTM-PD should be suspected in individuals with worsening respiratory symptoms (breathlessness, increased cough and sputum production) and/or declining pulmonary function tests that do not respond to antibiotic therapy targeting conventional CF-associated bacteria and optimised airway clearance. Night sweats, fevers, chest pains and weight loss (although uncommon) may also suggest possible NTM-PD.

breathlessness, increased cough and sputum production) and/or declining pulmonary function tests that do not respond to antibiotic therapy targeting conventional CF-associated bacteria and optimised airway clearance. Night sweats, fevers, chest pains and weight loss (although uncommon) may also suggest possible NTM-PD. NTM treatment should be considered in individuals with CF who fulfil ATS/IDSA criteria for NTM-PD. However, the decision to start treatment is a clinical one based on an amalgamation of patient factors, the NTM species involved, the risks of treatment side effects, adherence concerns and the expected outcomes of treatment. Recommended clinical practice for diagnosis A suggested algorithm for the investigation of individuals with CF suspected of having NTM-PD is shown in figure 1. Figure 1 A suggested algorithm for the investigation of individuals with clinical suspicion of NTM-PD (AFB, acid-fast bacilli; CF, cystic fibrosis; FEV1, forced expiratory volume in 1 s; HRCT, high-resolution CT; NTM-PD, non-tuberculous mycobacteria pulmonary disease).

Recommended clinical practice for diagnosis A suggested algorithm for the investigation of individuals with CF suspected of having NTM-PD is shown in figure 1. Figure 1 A suggested algorithm for the investigation of individuals with clinical suspicion of NTM-PD (AFB, acid-fast bacilli; CF, cystic fibrosis; FEV1, forced expiratory volume in 1 s; HRCT, high-resolution CT; NTM-PD, non-tuberculous mycobacteria pulmonary disease). When being investigated for potential NTM-PD, individuals should discontinue drugs liable to compromise NTM culture (such as macrolides, fluoroquinolones, aminoglycosides, co-trimoxazole, linezolid and doxycycline) prior to sputum sample collection. In the case of azithromycin, intracellular accumulation within phagocytes may require a washout period of 2 weeks or more to allow for drug clearance.133 134 If sputum samples are persistently culture negative, but clinical or radiological suspicion of NTM-PD remains, bronchoscopy with targeted sampling of areas with suggestive HRCT changes may be indicated. Individuals receiving azithromycin as part of their CF medical regimen, who have a positive surveillance NTM culture, should not continue azithromycin treatment while evaluation for NTM disease is underway, as azithromycin monotherapy may lead to the development of macrolide resistance.

e HRCT changes may be indicated. Individuals receiving azithromycin as part of their CF medical regimen, who have a positive surveillance NTM culture, should not continue azithromycin treatment while evaluation for NTM disease is underway, as azithromycin monotherapy may lead to the development of macrolide resistance. Other CF pathogens and comorbidities should be considered as potential contributors to a patient's symptoms and radiological features when determining the clinical significance of NTM-positive cultures. All aspects of CF care should be reviewed and optimised in order to determine the clinical significance of NTM in the sputum. Specifically, consider a trial of NTM-sparing intravenous antibiotics (ie, avoid carbapenems, cefoxitin, tigecycline, fluoroquinolones, linezolid and amikacin) that target conventional bacteria; and assess for CF-related diabetes, uncontrolled gastrointestinal reflux disease, and clinical and immunological features of ABPA. Likewise, adequate treatment of sinus disease, nutritional support and effective airway clearance strategies should be implemented. Before starting NTM treatment, side effects, the importance of adherence to therapy and complications of treatment should be discussed with patients, and these discussions documented in the medical notes. Discussion of the risk of treatment failure should be clearly documented.

Other CF pathogens and comorbidities should be considered as potential contributors to a patient's symptoms and radiological features when determining the clinical significance of NTM-positive cultures. All aspects of CF care should be reviewed and optimised in order to determine the clinical significance of NTM in the sputum. Specifically, consider a trial of NTM-sparing intravenous antibiotics (ie, avoid carbapenems, cefoxitin, tigecycline, fluoroquinolones, linezolid and amikacin) that target conventional bacteria; and assess for CF-related diabetes, uncontrolled gastrointestinal reflux disease, and clinical and immunological features of ABPA. Likewise, adequate treatment of sinus disease, nutritional support and effective airway clearance strategies should be implemented. Before starting NTM treatment, side effects, the importance of adherence to therapy and complications of treatment should be discussed with patients, and these discussions documented in the medical notes. Discussion of the risk of treatment failure should be clearly documented. Treatment Which antibiotic regimen should be used in individuals with CF who have ATS/IDSA-defined MABSC pulmonary disease? Recommendation 24: The CF Foundation and the ECFS recommend that treatment of MABSC pulmonary disease should involve an intensive phase followed by a continuation phase.

Before starting NTM treatment, side effects, the importance of adherence to therapy and complications of treatment should be discussed with patients, and these discussions documented in the medical notes. Discussion of the risk of treatment failure should be clearly documented. Treatment Which antibiotic regimen should be used in individuals with CF who have ATS/IDSA-defined MABSC pulmonary disease? Recommendation 24: The CF Foundation and the ECFS recommend that treatment of MABSC pulmonary disease should involve an intensive phase followed by a continuation phase. Recommendation 25: The CF Foundation and the ECFS recommend that the intensive phase should include a daily oral macrolide (preferably azithromycin) in conjunction with 3–12 weeks of intravenous amikacin and one or more of the following: intravenous tigecycline, imipenem or cefoxitin, guided but not dictated by DST. The duration of intensive phase therapy should be determined by the severity of infection, the response to treatment and the tolerability of the regimen. Recommendation 26: The CF Foundation and the ECFS recommend that the continuation phase should include a daily oral macrolide (preferably azithromycin) and inhaled amikacin, in conjunction with 2–3 of the following additional oral antibiotics: minocycline, clofazimine, moxifloxacin and linezolid, guided but not dictated by DST.

ation 26: The CF Foundation and the ECFS recommend that the continuation phase should include a daily oral macrolide (preferably azithromycin) and inhaled amikacin, in conjunction with 2–3 of the following additional oral antibiotics: minocycline, clofazimine, moxifloxacin and linezolid, guided but not dictated by DST. Recommendation 27: The CF Foundation and the ECFS recommend that individuals with MABSC pulmonary disease should be managed in collaboration with experts in the treatment of NTM and CF, as drug intolerance and drug-related toxicity occur frequently, and changes in antibiotic therapy are often required. Recommendation 28: The CF Foundation and the ECFS recommend that monotherapy with a macrolide or other antimicrobial should never be used in the treatment of MABSC pulmonary disease.

Recommendation 27: The CF Foundation and the ECFS recommend that individuals with MABSC pulmonary disease should be managed in collaboration with experts in the treatment of NTM and CF, as drug intolerance and drug-related toxicity occur frequently, and changes in antibiotic therapy are often required. Recommendation 28: The CF Foundation and the ECFS recommend that monotherapy with a macrolide or other antimicrobial should never be used in the treatment of MABSC pulmonary disease. There are no published randomised controlled trials evaluating treatment outcomes in individuals with M. abscessus pulmonary infections. Current treatment recommendations from the ATS and IDSA recommend consideration of a multidrug treatment regimen, but note that long-term sputum conversion is difficult to achieve and thus, alternative goals such as symptomatic improvement, radiographic regression of opacities or microbiological improvement, may be more realistic.26 The ATS/IDSA recommendations were based primarily on a single large study of 154 patients with lung disease caused by rapidly growing mycobacteria, in which more than 80% of patients were infected by M. abscessus.135 Treatment outcomes were extremely poor; however, the patients did not receive the currently recommended combination of antibiotics.

were based primarily on a single large study of 154 patients with lung disease caused by rapidly growing mycobacteria, in which more than 80% of patients were infected by M. abscessus.135 Treatment outcomes were extremely poor; however, the patients did not receive the currently recommended combination of antibiotics. Since the publication of the last ATS/IDSA guidelines,26 there have been several studies that reported treatment outcomes in individuals without CF with pulmonary disease due to M. abscessus. Jeon et al136 described treatment outcomes in 65 non-CF adults, in South Korea, with M. abscessus lung disease, who received a standardised treatment regimen. The regimen included 4 weeks of amikacin (15 mg/kg/day in two divided doses) and cefoxitin (200 mg/kg/day in three divided doses) along with clarithromycin (1000 mg/day in two divided doses), ciprofloxacin (1000 mg/day in two divided doses) and doxycycline (200 mg/day in two divided doses). The total duration of therapy was 24 months and at least 12 months after sputum culture conversion. Fifty-four (83%) patients responded with improved symptoms and 48 (74%) with improved HRCT findings. Sputum conversion and maintenance of negative sputum cultures for more than 12 months was achieved in 38 (58%) patients. This rate was significantly lower (17%) in patients whose isolates were resistant to clarithromycin. In contrast, in the 14 (22%) patients who underwent resectional surgery, negative sputum cultures were achieved and maintained in 7 (88%) of 8 with preoperatively positive cultures. The authors concluded that a standardised regimen was moderately effective, but adverse reactions were frequent.

istant to clarithromycin. In contrast, in the 14 (22%) patients who underwent resectional surgery, negative sputum cultures were achieved and maintained in 7 (88%) of 8 with preoperatively positive cultures. The authors concluded that a standardised regimen was moderately effective, but adverse reactions were frequent. Among 107 patients with M. abscessus pulmonary infection at National Jewish Health in Denver, CO, 69 non-CF individuals were treated and followed for a mean duration of 34 months.129 Patients were treated with individualised treatment regimens following ATS/IDSA recommendations. Twenty (29%) patients remained culture positive, 16 (23%) converted but experienced relapse, 33 (48%) converted to negative and did not relapse, while 17 (16%) died during the study period. There were significantly more surgical patients than medical patients whose culture converted and remained negative for at least 1 year (57% vs 28%, p=0.022). As in the previous study from South Korea, surgery may have been beneficial. However, surgical management is less likely to be applicable in individuals with CF in whom focal pulmonary disease is uncommon.

nts than medical patients whose culture converted and remained negative for at least 1 year (57% vs 28%, p=0.022). As in the previous study from South Korea, surgery may have been beneficial. However, surgical management is less likely to be applicable in individuals with CF in whom focal pulmonary disease is uncommon. In a follow-up study, Koh et al97 reported significant differences in outcomes based on which subspecies of M. abscessus was causing the infection. Treatment response rates to a standardised multidrug regimen were much higher in patients with M. a. massiliense than in those with M. a. abscessus: sputum culture conversion occurred in 88% of patients with M. a. massiliense compared with 25% with M. a. abscessus (p<0.001). All of the M. a. abscessus isolates contained a full length, functional erm41 that was shown to result in inducible macrolide resistance when the isolates were incubated with clarithromycin. In contrast, the MIC of M. a. massiliense strains did not increase after incubation with the macrolide agent because the erm41 gene contained a deletion, making it non-functional. Recent data from this same group of investigators have indicated that clarithromycin is a much stronger inducer of erm41 than azithromycin, suggesting that the latter macrolide may be a better choice when treating M. a. abscessus infections.98

agent because the erm41 gene contained a deletion, making it non-functional. Recent data from this same group of investigators have indicated that clarithromycin is a much stronger inducer of erm41 than azithromycin, suggesting that the latter macrolide may be a better choice when treating M. a. abscessus infections.98 Despite the clinical significance of M. abscessus lung infection in patients with CF, data on treatment outcomes are extremely limited. There is one anecdotal report that describes eradication of M. abscessus in an individual with CF who received a prolonged course of therapy with alternating month inhaled amikacin plus oral clarithromycin.137 However, this appears to be an uncommon outcome in practice. A recent case series of 52 individuals, including 15 with CF, with M. abscessus and/or M. chelonae infection, suggests that tigecycline-based regimens may be of benefit, with 10/15 individuals with CF showing some improvement.138 Recommended clinical practice for antibiotic treatment for M. abscessus pulmonary disease in CF A typical treatment schedule for individuals with CF with M. abscessus infection is shown in figure 2. Antibiotic dosing regimens are listed in table 2 with important side effects/toxicities described in table 3. Table 2 Antibiotic-dosing regimens used to treat Mycobacterium avium complex and Mycobacterium abscessus complex pulmonary disease in cystic fibrosis

Recommended clinical practice for antibiotic treatment for M. abscessus pulmonary disease in CF A typical treatment schedule for individuals with CF with M. abscessus infection is shown in figure 2. Antibiotic dosing regimens are listed in table 2 with important side effects/toxicities described in table 3. Table 2 Antibiotic-dosing regimens used to treat Mycobacterium avium complex and Mycobacterium abscessus complex pulmonary disease in cystic fibrosis Antibiotic Route Dose suitable for children/adolescents Dose suitable for adults Amikacin* Intravenous Children: 15–30 mg/kg/dose once daily Adolescents: 10–15 mg/kg/dose once daily Maximum dose 1500 mg daily 10–30 mg/kg once daily or 15 mg/kg/day in two divided doses Daily to 3× weekly dosing Amikacin*†‡ Nebulised 250–500 mg/dose once or twice daily 250–500 mg once or twice daily Azithromycin Oral Children: 10–12 mg/kg/dose once daily Adolescents: adult dosing regimen Maximum dose 500 mg 250–500 mg once daily Cefoxitin Intravenous 50 mg/kg/dose thrice daily (maximum dose 12 g/day) 200 mg/kg/day in three divided doses (maximum dose 12 g/day) Clarithromycin Oral 7.5 mg/kg/dose twice daily (maximum dose 500 mg) 500 mg twice daily§ Clarithromycin Intravenous Not recommended 500 mg twice daily§ Clofazimine†¶ Oral 1–2 mg/kg/dose once daily (maximum dose 100 mg) 50–100 mg once a day Co-trimoxazole (sulfamethoxazole and trimethoprim) Oral 10–20 mg/kg/dose twice daily 960 mg twice daily Co-trimoxazole (sulfamethoxazole and Trimethoprim) Intravenous 10–20 mg/kg/dose twice daily 1.44 g twice daily Ethambutol Oral Infants and children: 15 mg/kg/dose once daily Adolescents: 15 mg/kg/dose once daily 15 mg/kg once daily Imipenem Intravenous 15–20 mg/kg/dose twice daily (maximum dose 1000 mg) 1 g twice daily Linezolid** Oral <12 years old: 10 mg/kg/dose thrice daily 12 years and older: 10 mg/kg/dose once or twice daily (maximum dose 600 mg) 600 mg once or twice daily Linezolid** Intravenous <12 years old: 10 mg/kg/dose thrice daily 12 years and older: 10 mg/kg/dose once or twice daily (maximum dose 600 mg) 600 mg once or twice daily Moxifloxacin Oral 7.5–10 mg/kg/dose once daily (maximum dose 400 mg daily) 400 mg once daily Minocycline Oral 2 mg/kg/dose once daily (maximum dose 200 mg) 100 mg twice daily Rifampin (Rifampicin) Oral 10–20 mg/kg/dose once daily (maximum dose 600 mg) <50 kg 450 mg once daily >50 kg 600 mg once daily Rifabutin Oral 5–10 mg/kg/dose once daily (maximum dose 300 mg) 150–300 mg once daily 150 mg if patient taking strong CYP3A4 inhibitor 450–600 mg if patient taking strong CYP3A4 inducer Streptomycin* Intramuscular/intravenous 20–40 mg/kg/dose once dail

imum dose 600 mg) <50 kg 450 mg once daily >50 kg 600 mg once daily Rifabutin Oral 5–10 mg/kg/dose once daily (maximum dose 300 mg) 150–300 mg once daily 150 mg if patient taking strong CYP3A4 inhibitor 450–600 mg if patient taking strong CYP3A4 inducer Streptomycin* Intramuscular/intravenous 20–40 mg/kg/dose once dail y (maximum dose 1000 mg) 15 mg/kg once daily (maximum dose 1000 mg) Tigecycline†,†† Intravenous 8–11 years: 1.2 mg/kg/dose twice daily (maximum dose 50 mg) 12 years and older: 100 mg loading dose and then 50 mg once or twice daily 100 mg loading dose and then 50 mg once or twice daily *Adjust dose according to levels. Usually, starting dose is 15 mg/kg aiming for a peak level of 20–30 µg/mL and trough levels of <5–10 micrograms/ml. †As tolerated. ‡Mixed with normal saline. §For individuals under 55 kg, many practitioners recommend 7.5 mg/kg twice daily. ¶Only available in the USA through an IND application to the FDA. **Usually given with high dose (100 mg daily) pyridoxine (vitamin B6) to reduce risk of cytopaenias. ††Many practitioners recommend pre-dosing with one or more anti-emetics before dosing and/or gradual dose escalation from 25 mg daily to minimise nausea and vomiting. IND, investigational new drug; FDA, Food and Drug Administration. Table 3 Important side effects/toxicities of antibiotics and advisable monitoring procedures for MAC and MABSC in CF

††Many practitioners recommend pre-dosing with one or more anti-emetics before dosing and/or gradual dose escalation from 25 mg daily to minimise nausea and vomiting. IND, investigational new drug; FDA, Food and Drug Administration. Table 3 Important side effects/toxicities of antibiotics and advisable monitoring procedures for MAC and MABSC in CF Drug Common side effects/toxicity Monitoring procedures Amikacin Nephrotoxicity Regular serum amikacin levels* Regular serum creatinine levels Auditory-vestibular toxicity (tinnitus, high-frequency hearing loss) Symptoms, baseline and interval audiograms Azithromycin Nausea, vomiting, diarrhoea Symptoms Auditory-vestibular toxicity Symptoms, audiogram Prolonged QT ECG Clarithromycin Hepatitis Liver function tests Taste disturbance Symptoms Inhibited hepatic metabolism of rifabutin Symptoms Cefoxitin Fever, rash Symptoms Eosinophilia, anaemia, leucopaenia, thrombocytopaenia Full blood count Interference with common assays to measure serum creatinine Use alternative assay Clofazimine Discoloration of skin† Symptoms Enteropathy (sometimes mimicking pancreatic insufficiency)† Symptoms Nausea and vomiting Symptoms Co-trimoxazole Nausea, vomiting, diarrhoea Symptoms Anaemia, leucopoenia, thrombocytopaenia Full blood count Fever, rash, Stevens-Johnson syndrome Symptoms Ethambutol Optic neuritis Symptoms (loss of colour vision/acuity) Baseline and interval testing for colour vision and acuity‡ Ophthalmology opinion if symptoms occur Peripheral neuropathy Symptoms; nerve conduction studies Imipenem Hepatitis Liver function tests Imipenem (cont) Nausea, vomiting, diarrhoea Symptoms Linezolid Anaemia, leucopaenia, thrombocytopaenia Full blood count Peripheral neuropathy Symptoms/clinical evaluation/electrophysiology Optic neuritis Symptoms (loss of colour vision/acuity) Baseline and interval testing for colour vision and acuity Ophthalmology opinion if symptoms occur Moxifloxacin Nausea, vomiting, diarrhoea Symptoms Insomnia, agitation, anxiety Symptoms Tendonitis Symptoms Photosensitivity Symptoms Prolonged QT ECG Minocycline Photosensitivity Symptoms Nausea, vomiting, diarrhoea Symptoms Vertigo Symptoms Skin discolouration Clinical evaluation Rifampin and rifabutin Orange discolouration of bodily fluids (can stain contact lenses) Symptoms Hepatitis Liver function tests Nausea, vomiting, diarrhoea Symptoms Fever, chills Symptoms Thrombocytopaenia Full blood count Renal failure (rifampin) Blood tests Increased hepatic metabolism of numerous drugs Dose adjustment of other m

ifabutin Orange discolouration of bodily fluids (can stain contact lenses) Symptoms Hepatitis Liver function tests Nausea, vomiting, diarrhoea Symptoms Fever, chills Symptoms Thrombocytopaenia Full blood count Renal failure (rifampin) Blood tests Increased hepatic metabolism of numerous drugs Dose adjustment of other m edications/serum levels where available Rifabutin Leucopaenia, Full blood count Anterior uveitis (when combined with clarithromycin) Symptoms Flu-like symptoms polyarthralgia, polymyalgia Symptoms Streptomycin Nephrotoxicity Regular serum streptomycin levels Regular serum creatinine levels Auditory-vestibular toxicity (tinnitus, high frequency hearing loss) Symptoms, baseline and interval audiograms Tigecycline Nausea, vomiting, diarrhoea Symptoms Pancreatitis Serum amylase§ Hypoproteinaemia Serum albumin Bilirubinaemia Serum bilirubin *Usually aiming for peak levels of 20–30 µg/mL and trough levels of <5–10 µg/mL. †It may take up to 3 months for toxicity to resolve following cessation of clofazimine due to its long half-life. ‡Monthly checks if receiving 25 mg/kg/day. §In individuals with pancreatic sufficiency. CF, cystic fibrosis; MABSC, Mycobacterium abscessus complex; MAC, Mycobacterium avium complex.

edications/serum levels where available Rifabutin Leucopaenia, Full blood count Anterior uveitis (when combined with clarithromycin) Symptoms Flu-like symptoms polyarthralgia, polymyalgia Symptoms Streptomycin Nephrotoxicity Regular serum streptomycin levels Regular serum creatinine levels Auditory-vestibular toxicity (tinnitus, high frequency hearing loss) Symptoms, baseline and interval audiograms Tigecycline Nausea, vomiting, diarrhoea Symptoms Pancreatitis Serum amylase§ Hypoproteinaemia Serum albumin Bilirubinaemia Serum bilirubin *Usually aiming for peak levels of 20–30 µg/mL and trough levels of <5–10 µg/mL. †It may take up to 3 months for toxicity to resolve following cessation of clofazimine due to its long half-life. ‡Monthly checks if receiving 25 mg/kg/day. §In individuals with pancreatic sufficiency. CF, cystic fibrosis; MABSC, Mycobacterium abscessus complex; MAC, Mycobacterium avium complex. Figure 2 Typical treatment schedules for individuals with CF with Mycobacterium abscessus or MAC pulmonary disease. (A) M. abscessus treatment is divided into an initial intensive phase with an oral macrolide (preferably azithromycin) and intravenous amikacin with one or more additional intravenous antibiotics (tigecycline, imipenem, cefoxitin) for 3–12 weeks (depending on severity of infection, response to treatment, and the tolerability of the regimen), followed by a continuation phase of oral macrolide (preferably azithromycin) and inhaled amikacin with 2–3 additional antibiotics (minocycline, clofazimine, moxifloxacin, linezolid). Antibiotic choices should be guided but not dictated by drug susceptibility testing. Baseline and interval testing for drug toxicity is essential (B). MAC treatment (for clarithromycin-sensitive disease) should be with a daily oral macrolide (preferably azithromycin), rifampin and ethambutol. An initial course of injectable amikacin or streptomycin should be considered in the presence of (i) AFB smear positive respiratory tract samples, (ii) radiological evidence of lung cavitation or severe infection and (iii) systemic signs of illness. Baseline and interval testing for drug toxicity is essential (AFB, acid-fast bacilli; CF, cystic fibrosis; HRCT, high-resolution CT; MAC, Mycobacterium avium complex).

(i) AFB smear positive respiratory tract samples, (ii) radiological evidence of lung cavitation or severe infection and (iii) systemic signs of illness. Baseline and interval testing for drug toxicity is essential (AFB, acid-fast bacilli; CF, cystic fibrosis; HRCT, high-resolution CT; MAC, Mycobacterium avium complex). Given the lack of clinical trial data to inform treatment decisions there is a lot of variation in how patients are treated. An initial intensive phase is typically used to rapidly decrease the bacterial load. A combination of two intravenous drugs with demonstrated in vitro activity is administered for several weeks to months in combination with one or more oral drugs. Intravenous drug regimens of amikacin with cefoxitin and/or imipenem and/or tigecycline are the most commonly used combinations. Oral drugs with demonstrated in vitro activity include the macrolides (clarithromycin and azithromycin), linezolid, clofazimine and, occasionally, ciprofloxacin and/or moxifloxacin. After the intensive phase of therapy, patients are usually treated with at least two oral drugs in addition to a macrolide with or without inhaled antibiotics.

strated in vitro activity include the macrolides (clarithromycin and azithromycin), linezolid, clofazimine and, occasionally, ciprofloxacin and/or moxifloxacin. After the intensive phase of therapy, patients are usually treated with at least two oral drugs in addition to a macrolide with or without inhaled antibiotics. However, there is growing concern that treatment of M. abscessus isolates that have either a functional erm41 gene (resulting phenotypically in inducible macrolide resistance) or a 23S rRNA mutation (leading to high level constitutive macrolide resistance) may be compromised by switching from intravenous to oral therapy (given the relatively poor efficacy of oral antibiotics) and, therefore, continuous/very extended intravenous therapy with two or more effective antibiotics may be indicated in these cases.

utation (leading to high level constitutive macrolide resistance) may be compromised by switching from intravenous to oral therapy (given the relatively poor efficacy of oral antibiotics) and, therefore, continuous/very extended intravenous therapy with two or more effective antibiotics may be indicated in these cases. The choice of intravenous agents is based on in vitro activity and the toxicity profile of the drug. In addition to amikacin, imipenem is perhaps the best choice as companion intravenous therapy; the drug shows in vitro activity and the side effect profile is better than that of cefoxitin and tigecycline. In the study reported by Jeon et al,136 60% of the patients started on cefoxitin had to have the drug discontinued due to drug-related toxicity, after a median of 22 days of treatment. Neutropaenia occurred in 51% and thrombocytopaenia in 6% of patients on cefoxitin. Tigecycline has a low MIC against M. abscessus and showed efficacy against M. abscessus in combination.138 However, it is associated with significant nausea and vomiting, which has made it difficult to administer for a prolonged period.138

utropaenia occurred in 51% and thrombocytopaenia in 6% of patients on cefoxitin. Tigecycline has a low MIC against M. abscessus and showed efficacy against M. abscessus in combination.138 However, it is associated with significant nausea and vomiting, which has made it difficult to administer for a prolonged period.138 There are few oral drugs with significant in vitro activity against M. abscessus; the macrolides are the only oral drugs with consistent activity although their use may be potentially limited by inducible resistance (as described above) or acquired point mutations in the 23S rRNA. There are no clinical trials comparing azithromycin to clarithromycin in M. abscessus infection, so the choice of which macrolide to use is typically based on the in vitro activity, side effects profile and consideration of drug interactions. Clarithromycin has slightly better in vitro activity than azithromycin but there are conflicting reports regarding the impact of erm41 gene expression with each of these drugs.98 139 140 Clarithromycin is a stronger inhibitor of the P450 enzyme system than is azithromycin, so drug interactions are more common.

actions. Clarithromycin has slightly better in vitro activity than azithromycin but there are conflicting reports regarding the impact of erm41 gene expression with each of these drugs.98 139 140 Clarithromycin is a stronger inhibitor of the P450 enzyme system than is azithromycin, so drug interactions are more common. Linezolid shows in vitro activity in approximately 50% of M. abscessus isolates (although there is considerable geographical variation); however, haematological (anaemia, thrombocytopaenia) and neurological (peripheral neuropathy, optic neuritis) toxicities are common, particularly when linezolid is dosed 600 mg two times a day for prolonged courses. For this reason, many practitioners give 600 mg once daily to reduce the risk of adverse effects. However, care should be exercised in individuals chronically co-infected with methicillin-resistant Staphylococcus aureus (MRSA) since long-term linezolid therapy may encourage MRSA resistance. The fluoroquinolones and minocycline/doxycycline rarely show in vitro activity although they were included in the standardised treatment regimen used in the report by Jeon et al.98 Finally, clofazimine has significant in vitro activity against M. abscessus.141 However, this drug, used to treat leprosy, is not readily available in the USA at this time, although it can be obtained with an IRB-approved protocol through submission of an individual patient use IND to the Food and Drug Administration. Instructions for this process can be found on the NTM Info and Research, Inc, website (http://www.ntminfo.org/clofazimine).

is not readily available in the USA at this time, although it can be obtained with an IRB-approved protocol through submission of an individual patient use IND to the Food and Drug Administration. Instructions for this process can be found on the NTM Info and Research, Inc, website (http://www.ntminfo.org/clofazimine). The lack of oral antibiotics with activity against M. abscessus has led clinicians to use inhaled amikacin, usually during the continuation phase of therapy. There are no studies correlating treatment outcomes in patients with M. abscessus infection with the dose of inhaled amikacin and, therefore, there is a great deal of variation in the dose used (250–500 mg), and frequency of administration (daily to twice daily). A recent study, targeting treatment refractory NTM patients, most of whom were without CF with M. abscessus, evaluated the effect of adding inhaled amikacin to their oral and/or intravenous drug regimens.142 Among the 20 patients with persistently positive cultures, 8 (40%) had at least one negative culture and 5 (25%) had persistently negative cultures after addition of inhaled amikacin. Inhaled amikacin was stopped in 7 (35%) due to toxicity. There is currently significant interest in the potential use of a liposomal formulation of amikacin (which may improve drug delivery within the lung and into infected macrophages) as part of a multidrug regimen for both M. abscessus and MAC. Large multicentre studies are ongoing.

was stopped in 7 (35%) due to toxicity. There is currently significant interest in the potential use of a liposomal formulation of amikacin (which may improve drug delivery within the lung and into infected macrophages) as part of a multidrug regimen for both M. abscessus and MAC. Large multicentre studies are ongoing. The optimum duration of therapy is not known. Based on studies in individuals without CF, even prolonged treatment regimens were associated with high rates of failure and recurrence. Many patients who do not convert their cultures to negative on therapy may still benefit from continuing or repeating courses of treatment. Treatment for MAC Which antibiotic regimen should be used in individuals with CF who have ATS/IDSA-defined MAC pulmonary disease? Recommendation 29: The CF Foundation and the ECFS recommend the same antibiotic regimen for treatment of all species within the MAC. Recommendation 30: The CF Foundation and the ECFS recommend that clarithromycin-sensitive MAC pulmonary disease should be treated with a daily oral antibiotic regimen containing a macrolide (preferably azithromycin), rifampin and ethambutol. Recommendation 31: The CF Foundation and the ECFS recommend against the use of intermittent (three times per week) oral antibiotic therapy to treat MAC pulmonary disease. Recommendation 32: The CF Foundation and the ECFS recommend that monotherapy with a macrolide or other antimicrobial agent should never be used in the treatment of MAC pulmonary disease.

Recommendation 31: The CF Foundation and the ECFS recommend against the use of intermittent (three times per week) oral antibiotic therapy to treat MAC pulmonary disease. Recommendation 32: The CF Foundation and the ECFS recommend that monotherapy with a macrolide or other antimicrobial agent should never be used in the treatment of MAC pulmonary disease. Recommendation 33: The CF Foundation and the ECFS recommend that an initial course of intravenous amikacin should be considered for the treatment of MAC pulmonary disease in the presence of one or more of the following: (i) AFB smear positive respiratory tract samples, (ii) radiological evidence of lung cavitation or severe infection and (iii) systemic signs of illness. Recommendation 34: The CF Foundation and the ECFS recommend that clarithromycin-resistant MAC pulmonary disease should be managed in collaboration with experts in the treatment of NTM and CF.

Recommendation 33: The CF Foundation and the ECFS recommend that an initial course of intravenous amikacin should be considered for the treatment of MAC pulmonary disease in the presence of one or more of the following: (i) AFB smear positive respiratory tract samples, (ii) radiological evidence of lung cavitation or severe infection and (iii) systemic signs of illness. Recommendation 34: The CF Foundation and the ECFS recommend that clarithromycin-resistant MAC pulmonary disease should be managed in collaboration with experts in the treatment of NTM and CF. There are very few published randomised controlled trials evaluating treatment for MAC pulmonary disease (MAC-PD) in non-HIV-positive patients and none in individuals with CF. In the pre-macrolide era, a UK trial of individuals without CF and with largely cavitary disease reported that those randomised to receive rifampin and ethambutol had a combined failure/relapse rate of 41% compared to 16% of patients randomised to receive rifampin, ethambutol and isoniazid (p=0.033).143 In a subsequent study on a similar cohort, patients randomised to receive rifampin, ethambutol and clarithromycin had an all-cause mortality of 48% compared to 30% of patients randomised to receive rifampin, ethambutol and ciprofloxacin.144 However, only 13% of patients in the clarithromycin group failed treatment or relapsed compared to 23% in the ciprofloxacin group.

randomised to receive rifampin, ethambutol and clarithromycin had an all-cause mortality of 48% compared to 30% of patients randomised to receive rifampin, ethambutol and ciprofloxacin.144 However, only 13% of patients in the clarithromycin group failed treatment or relapsed compared to 23% in the ciprofloxacin group. In addition, there have been several non-comparator studies evaluating outcomes in HIV-negative patients with MAC-PD. The majority utilised a three oral drug regimen including a macrolide (clarithromycin or azithromycin), a rifamycin (rifampin or rifabutin) and ethambutol, in combination with an initial course of an aminoglycoside (streptomycin, amikacin or kanamycin).124 125 145–148 The culture conversion rate varied considerably between studies (13–82%), but on the whole, in 55–65% of patients, the culture converted after 6–12 months treatment and, when reported, the mean time from starting treatment to culture conversion was 3–5 months.124 125 Treatment failure was associated with previous MAC-PD treatment, cavitary disease, smear positivity, clarithromycin resistance at initiation of treatment, intolerance of NTM therapy and acquired clarithromycin resistance.124 125 145 147–149

mean time from starting treatment to culture conversion was 3–5 months.124 125 Treatment failure was associated with previous MAC-PD treatment, cavitary disease, smear positivity, clarithromycin resistance at initiation of treatment, intolerance of NTM therapy and acquired clarithromycin resistance.124 125 145 147–149 An alternative regimen using clofazimine with a macrolide and ethambutol in a study of 30 patients resulted in a culture conversion rate of 87% and a treatment success rate of 67%.150 Although 5 (19%) patients relapsed an average of 17 months after completing treatment, all MAC isolates remained clarithromycin sensitive, raising the possibility of reinfection rather than treatment failure.151 In another case series utilising clofazimine in combination with clarithromycin and minocycline, the culture conversion rate was 64% in patients completing the study (47% overall), which may indicate the importance of ethambutol as part of the multidrug regimen in the treatment of MAC-PD.152

than treatment failure.151 In another case series utilising clofazimine in combination with clarithromycin and minocycline, the culture conversion rate was 64% in patients completing the study (47% overall), which may indicate the importance of ethambutol as part of the multidrug regimen in the treatment of MAC-PD.152 Clarithromycin resistance developed in up to 15% of patients receiving treatment for MAC-PD and this was generally associated with clarithromycin monotherapy or the prescription of inadequate companion medications.124–126 146–149 When taken in combination with ethambutol and a rifamycin, acquired clarithromycin resistance developed in only 12/303 (4%) of patients. In the context of clarithromycin resistance, the best treatment responses were seen in patients who underwent surgical resection and received >6 months of an injectable aminoglycoside (amikacin or streptomycin),124–126 as 11/14 (79%) so treated achieved culture conversion compared to 1/27 (4%) of those not surgically resected and not receiving injectables.123

ce, the best treatment responses were seen in patients who underwent surgical resection and received >6 months of an injectable aminoglycoside (amikacin or streptomycin),124–126 as 11/14 (79%) so treated achieved culture conversion compared to 1/27 (4%) of those not surgically resected and not receiving injectables.123 While intermittent and daily dosing regimens appear equally effective in several case series of individuals without CF, intermittent regimens may be associated with less toxicity, better tolerability and adherence, and lower cost.147 148 However, a large multicentre study utilising an intermittent dosing regimen in individuals with moderate or severe MAC-PD (including many with cavitary disease and with prior treatment failure), reported a culture conversion rate of only 13% after 12 months of treatment.146 151 There have, to date, been no studies in individuals with CF to determine an optimal dosing regimen for MAC therapy, but concerns about drug absorption and lung penetration in CF have meant that many centres have adopted daily dosing protocols.

d a culture conversion rate of only 13% after 12 months of treatment.146 151 There have, to date, been no studies in individuals with CF to determine an optimal dosing regimen for MAC therapy, but concerns about drug absorption and lung penetration in CF have meant that many centres have adopted daily dosing protocols. It is unclear if use of an aminoglycoside during the initial phase of MAC antibiotic therapy is beneficial. In a multicentre study involving 146 HIV-negative patients with MAC-PD, participants were randomised to receive intramuscular streptomycin (15 mg/kg) or placebo thrice weekly for the first 3 months of therapy, in addition to clarithromycin, rifampin and ethambutol. Streptomycin-treated patients had a significantly higher culture conversion rate after approximately 2 years of treatment than did placebo patients (71% vs 51%, p<0.05), but a third of patients in each group experienced sputum relapse and there were no significant differences in symptoms or radiological response between groups.153 Furthermore, there was no statistically significant difference in culture conversion rates between individuals who received an initial course of intramuscular kanamycin (78%) compared to those who did not (58%) in a non-randomised study involving patients with MAC-PD.125 Recently, the use of aerosolised amikacin in addition to standard multidrug macrolide-based regimens was reported in six HIV-negative individuals with MAC-PD who had failed standard therapy.154 While four patients were culture negative after 6 months of therapy, one later cultured M. chelonae (resistant to amikacin), two re-cultured MAC and one patient was unable to tolerate prolonged therapy with aerosolised amikacin. A recent case series of the impact of nebulised amikacin (250 mg once or twice daily)142 in 20 individuals without CF with treatment refractory NTM-PD (of whom 5 had MAC) reported adverse events in 35% of cases. Two patients discontinued therapy due to hearing loss. Studies examining the use of liposomal amikacin (which may have a better side effects profile) for the treatment of NTM in individuals with CF are ongoing.

uals without CF with treatment refractory NTM-PD (of whom 5 had MAC) reported adverse events in 35% of cases. Two patients discontinued therapy due to hearing loss. Studies examining the use of liposomal amikacin (which may have a better side effects profile) for the treatment of NTM in individuals with CF are ongoing. Recommended clinical practice antibiotic treatment for MAC-PD in CF A typical treatment schedule for individuals with CF with MAC infection is shown in figure 2. Antibiotic dosing regimens are listed in table 2 with important side effects/toxicities described in table 3. Individuals with clarithromycin sensitive MAC-PD should be treated with a daily oral antibiotic regimen that includes a macrolide, rifampin and ethambutol (15 mg/kg), consistent with the ATS/IDSA recommendations for individuals with severe nodular bronchiectatic disease.26 Intermittent oral antibiotic therapy is not recommended due to the nature of the underlying lung disease and concerns regarding antibiotic absorption in CF. While there are no head to head trials showing a difference in outcome between individuals with MAC-PD treated with clarithromycin or azithromycin, the latter may be the macrolide of choice in CF, as it can be taken once daily, its serum levels may be less affected by rifamycins122 and it has well established benefits in individuals with CF in addition to its effects on NTM.

ce in outcome between individuals with MAC-PD treated with clarithromycin or azithromycin, the latter may be the macrolide of choice in CF, as it can be taken once daily, its serum levels may be less affected by rifamycins122 and it has well established benefits in individuals with CF in addition to its effects on NTM. Individuals with a high bacterial load (suggested by smear positivity, radiological evidence of lung cavitation and/or significant inflammatory change or the presence of systemic symptoms) may benefit from an initial (1–3 month) course of injectable amikacin or streptomycin, in addition to the standard three-drug regimen for MAC-PD. While the available data do not show a difference in toxicity between amikacin regimens dosed at 15 mg/kg once daily or 25 mg/kg thrice weekly, ototoxicity was found in 37% of all participants (associated with older age and larger cumulative dose), vestibular toxicity in 8% (usually reversible) and nephrotoxicity in 15% (usually mild and reversible).155 Streptomycin, although less widely used for MAC-PD than amikacin, may have less ototoxicity than amikacin.155 The use of aerosolised amikacin in place of an intravenous aminoglycoside may be preferable in terms of reduced burden of care and toxicity, but outcome data are limited and it is unlikely to be helpful for patients with cavitary disease in whom drug levels at the site of infection may be subtherapeutic.

n amikacin.155 The use of aerosolised amikacin in place of an intravenous aminoglycoside may be preferable in terms of reduced burden of care and toxicity, but outcome data are limited and it is unlikely to be helpful for patients with cavitary disease in whom drug levels at the site of infection may be subtherapeutic. The major risk factors for the development of clarithromycin-resistant MAC-PD are macrolide monotherapy and prior macrolide treatment with inadequate companion medications. Thus, macrolides (often prescribed for their anti-inflammatory effects in CF) should be discontinued immediately following isolation of a mycobacterial species, and macrolides should never be prescribed in the treatment of MAC-PD without two appropriate companion antibiotics.

ment with inadequate companion medications. Thus, macrolides (often prescribed for their anti-inflammatory effects in CF) should be discontinued immediately following isolation of a mycobacterial species, and macrolides should never be prescribed in the treatment of MAC-PD without two appropriate companion antibiotics. Macrolide therapy is not generally recommended in the context of clarithromycin-resistant MAC-PD,26 but macrolides may still be beneficial in this context in CF due to their non-antibiotic properties. Individuals with clarithromycin-resistant MAC-PD may respond to a regimen including a parenteral aminoglycoside, a rifamycin (usually rifabutin) and ethambutol, in addition to one or more companion medications (accepting that there are limited data to guide practice)26 126 144 such as a quinolone or clofazimine. Rifabutin may be useful in the treatment of clarithromycin-resistant MAC-PD, but adverse events (particularly blood dyscrasias, gastrointestinal upset and polyarthralgia) are more common and often necessitate dose reduction or complete cessation of treatment.156–158 Surgical resection might also be helpful in selected individuals with localised severe bronchiectatic disease, but this management is less likely to be useful in CF as MAC-PD is more likely to be diffuse.

and polyarthralgia) are more common and often necessitate dose reduction or complete cessation of treatment.156–158 Surgical resection might also be helpful in selected individuals with localised severe bronchiectatic disease, but this management is less likely to be useful in CF as MAC-PD is more likely to be diffuse. Ethambutol ocular toxicity (optic or retrobulbar neuritis) may present with blurred vision, decreased acuity, central scotomas, impaired red-green colour discrimination and peripheral visual field defects. Ocular toxicity was identified in 6% of individuals without CF with MAC-PD receiving ethambutol at a dose of 25 mg/kg/day for the first 2 months followed by 15 mg/kg/day for the remainder of treatment.159 Ocular toxicity is more likely to occur in the context of MAC-PD than in patients receiving tuberculosis (TB) treatment due to the longer duration of therapy. While individuals prescribed ethambutol should have regular visual acuity and colour vision testing, visual symptoms often occur before measurable changes can be identified. Thus, patients should be educated about the potential side effects of ethambutol and encouraged to self-report changes in vision, following which ethambutol therapy should be discontinued until an ophthalmological assessment has taken place.

ting, visual symptoms often occur before measurable changes can be identified. Thus, patients should be educated about the potential side effects of ethambutol and encouraged to self-report changes in vision, following which ethambutol therapy should be discontinued until an ophthalmological assessment has taken place. It is not uncommon for more than one NTM species to be isolated from an individual with CF.6 30 In these circumstances, continued microbiological surveillance is advisable to determine which species is/are persistently positive and which is/are likely to be causing disease. NTM-PD is also commonly associated with ABPA and/or the identification of Aspergillus spp in sputum or lavage specimens. As rifamycins increase the hepatic metabolism of azole antifungal agents, the treatment of Aspergillus in the context of MAC-PD is more difficult. One approach is to use rifabutin in place of rifampin (as it is the rifamycin with the least cytochrome P450 enzyme induction) in conjunction with the usual companion medications for MAC and voriconazole, or posaconazole, which may be less affected by rifabutin co-medication than voriconazole is, with adjustment of drug doses according to levels.160 161 If therapeutic drug monitoring (TDM) is not available, other approaches include using nebulised amikacin or clofazimine in place of rifampin.150

or MAC and voriconazole, or posaconazole, which may be less affected by rifabutin co-medication than voriconazole is, with adjustment of drug doses according to levels.160 161 If therapeutic drug monitoring (TDM) is not available, other approaches include using nebulised amikacin or clofazimine in place of rifampin.150 Treatment: generic recommendations What outcome monitoring should be performed in individuals with CF receiving treatment for NTM-PD? Recommendation 35: The CF Foundation and the ECFS recommend that individuals with CF receiving NTM treatment should have expectorated or induced sputum samples sent for NTM culture every 4–8 weeks throughout the entire course of treatment to assess the microbiological response. Recommendation 36: The CF Foundation and the ECFS recommend that a schedule for detecting drug toxicity (including hearing loss, visual loss, renal impairment and liver function test abnormalities) should be set in place at the time of NTM treatment initiation and implemented throughout treatment based on the specific drugs prescribed. Recommendation 37: The CF Foundation and the ECFS recommend that an HRCT scan of the lungs should be performed shortly before starting NTM treatment and at the end of NTM treatment, to assess the radiological response.

Recommendation 36: The CF Foundation and the ECFS recommend that a schedule for detecting drug toxicity (including hearing loss, visual loss, renal impairment and liver function test abnormalities) should be set in place at the time of NTM treatment initiation and implemented throughout treatment based on the specific drugs prescribed. Recommendation 37: The CF Foundation and the ECFS recommend that an HRCT scan of the lungs should be performed shortly before starting NTM treatment and at the end of NTM treatment, to assess the radiological response. What duration of antibiotic therapy is recommended for individuals with CF receiving treatment for NTM-PD? Recommendation 38: The CF Foundation and the ECFS recommend that NTM antibiotic therapy should be prescribed for 12 months beyond culture conversion (defined as three consecutive negative cultures, with the time of conversion being the date of the first of the three negative cultures) as long as no positive cultures are obtained during those 12 months. Recommendation 39: The CF Foundation and the ECFS recommend that individuals who fail to culture convert despite optimal NTM therapy may benefit from long-term suppressive antibiotic treatment. Treatment: TDM Should TDM be performed in individuals with CF receiving treatment for NTM-PD? Recommendation 40: The CF Foundation and the ECFS recommend that, when amikacin is given intravenously or when streptomycin is given intravenously or intramuscularly, serum levels should be monitored and dosing adjusted to minimise ototoxicity and nephrotoxicity.

d in individuals with CF receiving treatment for NTM-PD? Recommendation 40: The CF Foundation and the ECFS recommend that, when amikacin is given intravenously or when streptomycin is given intravenously or intramuscularly, serum levels should be monitored and dosing adjusted to minimise ototoxicity and nephrotoxicity. Recommendation 41: The CF Foundation and the ECFS recommend against routinely obtaining serum levels of other anti-mycobacterial drugs. However, absorption of oral medications is often reduced in CF. Therefore use of TDM should be considered for individuals failing to improve despite taking recommended drug regimens or for those on concomitant medications with significant interactions with NTM drugs. TDM seeks to quantify the relationship between drug dose, serum (plasma) concentration and clinical response,162 and to thereby maximise therapeutic response while avoiding toxicity. The potential benefits of TDM during NTM treatment in individuals with CF include adjusting drug dosing to: Correct for drug–drug interactions that could adversely affect serum antibiotic levels: Drug–drug interactions frequently occur among agents used to treat NTM. Rifampin (more than rifabutin) may increase the metabolism of several drugs including clarithromycin, azithromycin and moxifloxacin, while rifabutin increases azithromycin levels and decreases moxifloxacin levels.122 163

ct serum antibiotic levels: Drug–drug interactions frequently occur among agents used to treat NTM. Rifampin (more than rifabutin) may increase the metabolism of several drugs including clarithromycin, azithromycin and moxifloxacin, while rifabutin increases azithromycin levels and decreases moxifloxacin levels.122 163 Maximise the PK and pharmacodynamic (PD) parameters of antibiotics to optimise efficacy: The PK/PD indices that correlate with clinical efficacy vary by antimicrobial agent.122 164 165 To exert maximal activity, drugs such as aminoglycosides and ethambutol require high peak concentrations relative to the pathogen's minimal inhibitory concentration (Cmax/MIC). Ciprofloxacin and rifampin require a high concentration time or area under the plasma concentration curve measured over 24 h to MIC ratio (AUC0–24/MIC) and β-lactam agents require as much time as possible whereby the concentration persists above the infecting organism's MIC (%T> MIC). Macrolide agents such as azithromycin have weak concentration-dependent effects and time effects, but these agents exert their activity through intracellular activity, tissue penetration and prolonged, persistent effects, due to their long half-life.166

ncentration persists above the infecting organism's MIC (%T> MIC). Macrolide agents such as azithromycin have weak concentration-dependent effects and time effects, but these agents exert their activity through intracellular activity, tissue penetration and prolonged, persistent effects, due to their long half-life.166 Overcome CF-related differences in absorption, distribution and clearance of drugs: Individuals with CF have different renal and non-renal clearance of several drugs when compared to individuals without CF, due to reduced bioavailability, increased volume of distribution and more rapid clearance. In addition, hepatic disease and diabetes may further influence drug metabolism and absorption. Several recent reviews have addressed evidence-based dosing for various agents used for treatment of pulmonary exacerbations in CF.167–170 While the relevance of the recommended dosing schedules is unknown for treatment of NTM, it is possible that individuals with CF would need higher dosages of mycobacterial drugs.

n. Several recent reviews have addressed evidence-based dosing for various agents used for treatment of pulmonary exacerbations in CF.167–170 While the relevance of the recommended dosing schedules is unknown for treatment of NTM, it is possible that individuals with CF would need higher dosages of mycobacterial drugs. With the exception of aminoglycosides, the clinical utility of TDM during treatment for NTM is unknown for individuals with and without CF due to a lack of rigorous studies, although some experts have recommended TDM for mycobacterial agents on a case-by-case basis.165 A recent retrospective study assessed the PK and pharmacodynamic parameters for 481 patients with disease caused by MAC.122 Peak serum concentrations within reference/normal ranges were only achieved for ethambutol, clarithromycin and azithromycin, in 52%, 44% and 65% of patients, respectively. In addition, pharmacodynamic targets for Cmax/MIC or AUC0–24/MIC were rarely achieved. However, these observations were not linked with clinical outcomes.

erum concentrations within reference/normal ranges were only achieved for ethambutol, clarithromycin and azithromycin, in 52%, 44% and 65% of patients, respectively. In addition, pharmacodynamic targets for Cmax/MIC or AUC0–24/MIC were rarely achieved. However, these observations were not linked with clinical outcomes. Another recent evaluation of the potential utility for TDM in 130 individuals without CF treated for MAC found no association between peak plasma/MIC ratios for clarithromycin, rifampin or ethambutol, and clinical outcomes.171 As previously observed, rifampin had a substantial impact on clarithromycin levels; those treated with both drugs had a median peak plasma clarithromycin concentration of 0.3 µg/mL, while those treated with rifabutin had a median peak plasma concentration of 1.8 µg/mL, and those with M. abscessus (n=60) treated without rifampin had a median peak plasma concentration of 3.8 µg/mL. In all, 97% of patients with MAC treated with daily therapy and 100% of patients on intermittent therapy reached the target of 2 µg/mL for clarithromycin. These experts concluded that TDM for treatment of MAC lung disease may not be beneficial (although the effects of dose optimisation on clinical outcomes were not evaluated).

97% of patients with MAC treated with daily therapy and 100% of patients on intermittent therapy reached the target of 2 µg/mL for clarithromycin. These experts concluded that TDM for treatment of MAC lung disease may not be beneficial (although the effects of dose optimisation on clinical outcomes were not evaluated). To the best of our knowledge, there is only one case series, published over a decade ago, examining the potential role of TDM in CF. Ten patients with CF with mycobacterial disease (6 with MAC, 3 with M. abscessus and 1 with M. tuberculosis) had serum drug concentration measurements performed 2 and 4 h after ingestion.164 Monitoring serum levels at two time points helped distinguish between poor absorption and delayed absorption. Half of the patients had inadequate serum levels for one or more drugs, and one patient clinically improved following dose adjustments that achieved target serum levels. However, target concentrations were not achieved for several patients. Notably, this study did not compare outcomes in patients with and without TDM.

Half of the patients had inadequate serum levels for one or more drugs, and one patient clinically improved following dose adjustments that achieved target serum levels. However, target concentrations were not achieved for several patients. Notably, this study did not compare outcomes in patients with and without TDM. Treatment: adjuvant therapy and surgery In the context of infectious disease, adjuvants have been defined as ‘therapies that act by rendering the organism more susceptible to attack by antibiotics or the host immune system, by rendering the organism less virulent or killing it by other means’.172 A number of approaches have been proposed as candidates for adjuvant therapy in NTM infection in CF, including interferon γ (IFNγ; or agents that promote IFNγ release) and vitamin D. Drug delivery vehicles, such as liposomes, may be considered adjuvants. Liposomes have been studied as a mode of delivering amikacin for infection with P. aeruginosa in CF,173 and this approach is also being evaluated (clinicaltrials.gov/show/NCT01315236) for NTM. Does IFNγ therapy improve treatment outcomes in individuals with CF who have NTM-PD? Recommendation 42: The CF Foundation and the ECFS recommend against the use of IFNγ as adjuvant therapy for NTM-PD in individuals with CF.

Treatment: adjuvant therapy and surgery In the context of infectious disease, adjuvants have been defined as ‘therapies that act by rendering the organism more susceptible to attack by antibiotics or the host immune system, by rendering the organism less virulent or killing it by other means’.172 A number of approaches have been proposed as candidates for adjuvant therapy in NTM infection in CF, including interferon γ (IFNγ; or agents that promote IFNγ release) and vitamin D. Drug delivery vehicles, such as liposomes, may be considered adjuvants. Liposomes have been studied as a mode of delivering amikacin for infection with P. aeruginosa in CF,173 and this approach is also being evaluated (clinicaltrials.gov/show/NCT01315236) for NTM. Does IFNγ therapy improve treatment outcomes in individuals with CF who have NTM-PD? Recommendation 42: The CF Foundation and the ECFS recommend against the use of IFNγ as adjuvant therapy for NTM-PD in individuals with CF. IFNγ plays a critical role in the host defence against NTM infection: (1) deficiencies in IFNγ signalling (caused by deleterious mutations174 or neutralising autoantibodies175) lead to (usually disseminated) NTM infection in individuals without CF; (2) inoculation of mice deficient in IFNγ or IFNγ receptors results in disseminated NTM infection;176 (3) addition of IFNγ to NTM-infected human macrophages in vitro enhances intracellular killing probably through autophagy stimulation.43

odies175) lead to (usually disseminated) NTM infection in individuals without CF; (2) inoculation of mice deficient in IFNγ or IFNγ receptors results in disseminated NTM infection;176 (3) addition of IFNγ to NTM-infected human macrophages in vitro enhances intracellular killing probably through autophagy stimulation.43 In non-CF individuals, adjuvant IFNγ therapy in NTM infection has been examined in several studies.177 178 An uncontrolled trial of IFNγ was conducted in seven patients with presumed primary immunodeficiency (three with familial susceptibility to MAC and four with idiopathic CD4 lymphopaenia) who had disseminated NTM disease refractory to conventional antibiotic therapy.177 All the patients improved with the introduction of subcutaneous IFNγ two or three times per week.

seven patients with presumed primary immunodeficiency (three with familial susceptibility to MAC and four with idiopathic CD4 lymphopaenia) who had disseminated NTM disease refractory to conventional antibiotic therapy.177 All the patients improved with the introduction of subcutaneous IFNγ two or three times per week. In a randomised, placebo controlled trial, 32 patients with pulmonary NTM disease (30 with MAC) were randomised to receive either intramuscular IFNγ (1×106 IU) or placebo once daily for 4 weeks and then three times weekly for 20 weeks178 in addition to daily oral azithromycin, ciprofloxacin, ethambutol and rifampin. The primary outcome (a composite end point of improvements in symptoms, radiology and microbiology) was achieved at 6 months by 72% (13/18) of patients in the IFNγ arm compared to 36% (5/14) receiving placebo (p=0.037). The greater response rate with IFNγ was sustained at 12 months after completion of treatment. However, the small study size, the use of composite end points and the lack of microbiological response after 6 months treatment mean that these data need to be interpreted with caution. Furthermore three large trials (ClinicalTrials.gov Identifiers NCT00001318, NCT00111397 and NCT00043355) examining IFNγ therapy for pulmonary NTM disease remain unpublished or have been terminated (potentially due to lack of efficacy), again questioning the role of IFNγ adjuvant therapy.

In a randomised, placebo controlled trial, 32 patients with pulmonary NTM disease (30 with MAC) were randomised to receive either intramuscular IFNγ (1×106 IU) or placebo once daily for 4 weeks and then three times weekly for 20 weeks178 in addition to daily oral azithromycin, ciprofloxacin, ethambutol and rifampin. The primary outcome (a composite end point of improvements in symptoms, radiology and microbiology) was achieved at 6 months by 72% (13/18) of patients in the IFNγ arm compared to 36% (5/14) receiving placebo (p=0.037). The greater response rate with IFNγ was sustained at 12 months after completion of treatment. However, the small study size, the use of composite end points and the lack of microbiological response after 6 months treatment mean that these data need to be interpreted with caution. Furthermore three large trials (ClinicalTrials.gov Identifiers NCT00001318, NCT00111397 and NCT00043355) examining IFNγ therapy for pulmonary NTM disease remain unpublished or have been terminated (potentially due to lack of efficacy), again questioning the role of IFNγ adjuvant therapy. Does vitamin D supplementation improve treatment outcomes in individuals with CF who have NTM-PD? Recommendation 43: The CF Foundation and the ECFS recommend that vitamin D should be supplemented according to national CF care guidelines.

Furthermore three large trials (ClinicalTrials.gov Identifiers NCT00001318, NCT00111397 and NCT00043355) examining IFNγ therapy for pulmonary NTM disease remain unpublished or have been terminated (potentially due to lack of efficacy), again questioning the role of IFNγ adjuvant therapy. Does vitamin D supplementation improve treatment outcomes in individuals with CF who have NTM-PD? Recommendation 43: The CF Foundation and the ECFS recommend that vitamin D should be supplemented according to national CF care guidelines. Vitamin D is thought to play a critical role in host defence against mycobacteria. In vitro and ex vivo treatment with vitamin D of human macrophages infected with M. tuberculosis enhances intracellular killing (through stimulating antimicrobial peptide production179 and autophagy.180) Furthermore, several epidemiological studies have shown an association of vitamin D deficiency with reactivation of TB181 182 and, recently, the presence of NTM-PD.71 However, interventional trials of vitamin D supplementation in patients with active pulmonary TB have had mixed results,183 and there are no trials of vitamin D as an adjuvant treatment for NTM disease. Should surgery be considered in individuals with CF who have NTM-PD? Recommendation 44: The CF Foundation and the ECFS recommend that lung resection should only be considered under extraordinary circumstances and in consultation with experts on the treatment of NTM and CF.

Vitamin D is thought to play a critical role in host defence against mycobacteria. In vitro and ex vivo treatment with vitamin D of human macrophages infected with M. tuberculosis enhances intracellular killing (through stimulating antimicrobial peptide production179 and autophagy.180) Furthermore, several epidemiological studies have shown an association of vitamin D deficiency with reactivation of TB181 182 and, recently, the presence of NTM-PD.71 However, interventional trials of vitamin D supplementation in patients with active pulmonary TB have had mixed results,183 and there are no trials of vitamin D as an adjuvant treatment for NTM disease. Should surgery be considered in individuals with CF who have NTM-PD? Recommendation 44: The CF Foundation and the ECFS recommend that lung resection should only be considered under extraordinary circumstances and in consultation with experts on the treatment of NTM and CF. Surgical resection has been used extensively in the management of pulmonary mycobacterial infection in order to excise localised infection, debulk severe disease, or excise cavities or damaged lung into which antibiotic penetration may be impaired. In no cases has surgery been used as a substitute for antibiotic therapy. There are no randomised trials of surgery for the treatment of pulmonary NTM disease in any patient group. While many publications report the use of lung resection (pneumonectomy, lobectomy or segmentectomy) with combination antibiotic therapy in NTM infection, most are case reports with no comparator group receiving only medical therapy thereby preventing objective assessment of the efficacy of surgery.

any patient group. While many publications report the use of lung resection (pneumonectomy, lobectomy or segmentectomy) with combination antibiotic therapy in NTM infection, most are case reports with no comparator group receiving only medical therapy thereby preventing objective assessment of the efficacy of surgery. Nonetheless, three series of individuals without CF do contain some comparison data although the potential for selection bias of patients considered suitable for surgery makes interpretation difficult. The first series136 comprised of 65 patients, from South Korea, with pulmonary M. abscessus infection. Surgery was performed in 14 patients who failed to achieve sputum culture conversion, became culture positive again after a period of culture negativity or experienced disease-related complications such as haemoptysis. Of the eight patients who were sputum culture positive before surgery, seven became culture negative postoperatively (compared to culture conversion rates of 38/65 for the group as a whole). A second study, from the USA,129 reported outcomes for 69 patients with pulmonary M. abscessus infection all treated with combination antibiotics, 23 of whom underwent additional surgical resection. Indications for surgery included the presence of localised bronchiectasis, cavitary disease and haemoptysis. In the surgical group, significantly more patients (13/23) became persistently sputum culture negative compared to those in the medical treatment only group (13/46). A third study, also from the USA, described outcomes in 51 patients with macrolide-resistant MAC-PD.126 Individuals receiving both surgical resection and injectable aminoglycoside therapy had greater sputum conversion rates (11/14 patients) than those receiving neither treatment modality (2/37 patients).

group (13/46). A third study, also from the USA, described outcomes in 51 patients with macrolide-resistant MAC-PD.126 Individuals receiving both surgical resection and injectable aminoglycoside therapy had greater sputum conversion rates (11/14 patients) than those receiving neither treatment modality (2/37 patients). A recent review of case series published over the last 40 years184 suggests that localised resection (lobectomy or segmentectomy) should be considered for severe, localised, unilateral NTM disease that has failed to respond to conventional antibiotic therapy. In the context of CF, however, localised NTM disease is extremely rare (or at least very difficult to identify), and the risks of thoracic surgery are high and therefore the potential benefits of surgical resection limited. Transplantation Should individuals with CF with current or previous NTM-positive cultures be referred for lung transplantation? Recommendation 45: The CF Foundation and the ECFS recommend that all individuals with CF being considered for lung transplantation should be evaluated for NTM-PD. Recommendation 46: The CF Foundation and the ECFS recommend that the presence of current or previous respiratory tract samples positive for NTM should not preclude individuals being considered for lung transplantation. Recommendation 47: The CF Foundation and the ECFS recommend that individuals with CF who have NTM-PD and are being evaluated for transplantation should start treatment prior to transplant listing.

Recommendation 46: The CF Foundation and the ECFS recommend that the presence of current or previous respiratory tract samples positive for NTM should not preclude individuals being considered for lung transplantation. Recommendation 47: The CF Foundation and the ECFS recommend that individuals with CF who have NTM-PD and are being evaluated for transplantation should start treatment prior to transplant listing. Recommendation 48: The CF Foundation and the ECFS recommend that individuals with CF receiving NTM treatment with sequential negative cultures may be eligible for transplant listing. Recommendation 49: The CF Foundation and the ECFS recommend that individuals with CF who have completed treatment for NTM-PD with apparent eradication of the organism may be eligible for transplant listing. Recommendation 50: The CF Foundation and the ECFS recommend that the presence of persistent MABSC or MAC infection despite optimal therapy is not an absolute contraindication to lung transplant referral. The International Society for Heart and Lung Transplantation (ISHLT) International Guidelines lists ‘colonisation with highly resistant or virulent mycobacteria’ as a relative contra-indication for selection as a lung transplant candidate.185 There is, however, limited published information on transplant outcomes for individuals with previous or concurrent NTM infection, with very few reports (usually from single centres) specifically examining CF cohorts.186 187

mycobacteria’ as a relative contra-indication for selection as a lung transplant candidate.185 There is, however, limited published information on transplant outcomes for individuals with previous or concurrent NTM infection, with very few reports (usually from single centres) specifically examining CF cohorts.186 187 The risk of NTM infection post-transplantation is not well defined. A study of 201 CF and non-CF transplant recipients188 suggested that postoperative NTM acquisition was associated with increased mortality (HR 2.61), independent of bronchiolitis obliterans syndrome. However, these data should be interpreted keeping the following in mind: very little data were available on the presence of pulmonary NTM pretransplant; the vast majority of patients did not have CF or even bronchiectasis; and non-NTM-related causes were major contributors to death in fatal cases. In contrast, a recent study of CF and non-CF transplant recipients189 reported that 53 of 237 individuals (22.4%) acquired NTM-positive cultures postoperatively (70% MAC, 10% MABSC), of whom two fulfilled ATS/IDSA criteria for NTM-PD. Although overall mortality was not affected by NTM acquisition, four patients developed persistent surgical site infection (three with M. abscessus), of whom one died of disseminated NTM infection. The potential for M. abscessus to cause significant postoperative complications is supported by a review of outcomes from 31 transplant centres190 indicating frequent soft tissue and surgical site infections, and two deaths, attributable to M. abscessus infection.

essus), of whom one died of disseminated NTM infection. The potential for M. abscessus to cause significant postoperative complications is supported by a review of outcomes from 31 transplant centres190 indicating frequent soft tissue and surgical site infections, and two deaths, attributable to M. abscessus infection. The largest CF-specific case series comes from the University of North Carolina Chapel Hill experience between 1990 and 2003.29 One hundred and forty-six patients with CF underwent lung transplantation and 31 listed for transplantation. Of those individuals referred, 19.7% were NTM culture positive pretransplant. Rates of NTM following lung transplantation were 3.4%. Pretransplant infection with M. abscessus was recognised as a significant risk factor for recurrence of NTM post-transplantation. Although there was no effect on mortality, post-transplant NTM infection caused significant morbidity as patients developed M. abscessus-associated skin and soft tissue infection or pulmonary disease caused by MAC and other NTM species. There are several published case series of successful outcomes for individuals with CF who have culture positive M. abscessus infection at the time of transplantation.186 However, NTM-related complications in this group may be more frequent, and include persistent soft tissue or wound infections,186 empyema and disseminated NTM infection.164 191 Although a small series, the University of North Carolina (UNC) report suggests no effect of the presence of pretransplant M. abscessus positive cultures on post-transplant mortality.187

group may be more frequent, and include persistent soft tissue or wound infections,186 empyema and disseminated NTM infection.164 191 Although a small series, the University of North Carolina (UNC) report suggests no effect of the presence of pretransplant M. abscessus positive cultures on post-transplant mortality.187 The Consensus statements Committee concluded that all individuals with CF should be evaluated for NTM disease prior to referral for lung transplantation, given the very high reported rates of NTM culture positivity for this group, and the fact that untreated NTM infection may represent an increased (and potentially modifiable) postoperative risk. Consequently, if NTM-PD is diagnosed, treatment should be started prior to transplant listing.

rral for lung transplantation, given the very high reported rates of NTM culture positivity for this group, and the fact that untreated NTM infection may represent an increased (and potentially modifiable) postoperative risk. Consequently, if NTM-PD is diagnosed, treatment should be started prior to transplant listing. Conclusion The management of individuals with CF infected with NTM is extremely challenging. The limited amounts of published research and clinical trial data provide inadequate evidence to base management decisions on how best to screen, diagnose, detect and treat NTM-PD. As a response to this urgent clinical need, the CF Foundation and the ECFS formed a committee of clinicians, scientists and infectious disease experts to develop recommendations to guide and assist clinicians in the management of NTM-PD in individuals with CF. The committee believes these recommendations should serve as a benchmark for current medical care while providing a framework to inform the development of clinical, translation and basic research studies to generate robust evidence on which to base future iterations of these management guidelines, leading to better outcomes for individuals with CF infected with NTM. Supplementary Material Web NTM-recommendations Web executive-summary The authors would like to thank David Young, Chair of the CFF Pharmacist Mentorship Committee, for review of treatment regimens and drug tables.

Conclusion The management of individuals with CF infected with NTM is extremely challenging. The limited amounts of published research and clinical trial data provide inadequate evidence to base management decisions on how best to screen, diagnose, detect and treat NTM-PD. As a response to this urgent clinical need, the CF Foundation and the ECFS formed a committee of clinicians, scientists and infectious disease experts to develop recommendations to guide and assist clinicians in the management of NTM-PD in individuals with CF. The committee believes these recommendations should serve as a benchmark for current medical care while providing a framework to inform the development of clinical, translation and basic research studies to generate robust evidence on which to base future iterations of these management guidelines, leading to better outcomes for individuals with CF infected with NTM. Supplementary Material Web NTM-recommendations Web executive-summary The authors would like to thank David Young, Chair of the CFF Pharmacist Mentorship Committee, for review of treatment regimens and drug tables. Contributors: All the authors contributed to the expert committee on guidelines development (Co-chairs: RAF, CSH; Steering committee: RAF, CSH, BCM, KNO, LS, KAS, SEH; Committee subgroups: Epidemiology and Risk Factors (KNO (lead), IS-G, KLW): Screening (JAN (lead), RLG, KK; Microbiology (J-LH (lead), RJW, JvI, RAF); Treatment (CLD (lead), DB, LS, ARS, CSH); and Transplantation (PGN (lead), PC).

s development (Co-chairs: RAF, CSH; Steering committee: RAF, CSH, BCM, KNO, LS, KAS, SEH; Committee subgroups: Epidemiology and Risk Factors (KNO (lead), IS-G, KLW): Screening (JAN (lead), RLG, KK; Microbiology (J-LH (lead), RJW, JvI, RAF); Treatment (CLD (lead), DB, LS, ARS, CSH); and Transplantation (PGN (lead), PC). Funding: Cystic Fibrosis Foundation; European Cystic Fibrosis Society, The Wellcome Trust and Cambridge NIHR BRC (RAF); Intramural programme of the National Heart, Lung, and Blood Institute, NIH (KNO); Vaincre La Mucoviscidose (VLMIC1014 and RF20120600689) and the Région Ile-de-France Domaine d'Intérêt Majeur Maladies Infectieuses et Emergentes (J-LH); CF Foundation Clinical Research Award (NICK13A0) (JAN); Imperial College London NIHR Respiratory BRU (DB). Competing interests: None declared. Provenance and peer review: Not commissioned; externally peer reviewed.

Introduction The majority of patients with lung cancer (80%) have advanced disease at the time of diagnosis, making them ineligible for curative treatment, which contributes to the low overall 5-year survival (around 7% in the UK). By contrast, the prospects for patients with early stage disease are much more favourable, with a 5-year survival of 60%–70% among those with stage Ia non-small cell lung cancer. In the 1970s screening for lung cancer with chest radiography identified more cancers than in the control groups, but there was no improvement in mortality from the randomised trials. In 2006, several studies of CT scanning were used to generate hypotheses after identifying peripheral adenocarcinomas in prevalence screens, with the numbers discovered varying by population and whether the study group were Asian, where peripheral adenocarcinomas are more common. Recent or ongoing lung cancer screening trials The National Lung Cancer Screening Trial (NLST), in the USA, was the first large randomised study to show that low-dose CT can identify lung cancers at an earlier stage in an unselected smoking population and, importantly, reduce mortality.1 The trial assigned 53 452 heavy smokers aged between 55 and 74 years to either three annual CT screens or chest radiographs. It showed a statistically significant 20% reduction in mortality from lung cancer with CT. However, CT screening is expensive and a cost-effectiveness analysis based on the NLST results suggested that the cost per quality adjusted life year could be as high as US$81 000.2

r three annual CT screens or chest radiographs. It showed a statistically significant 20% reduction in mortality from lung cancer with CT. However, CT screening is expensive and a cost-effectiveness analysis based on the NLST results suggested that the cost per quality adjusted life year could be as high as US$81 000.2 The results on mortality from the NELSON trial (conducted in Belgium and The Netherlands) are due in 2015–2016, and are expected to confirm the findings of the NLST. Eligible participants were aged 50–75 years, and moderate to heavy current or former smokers (who had quit less than 10 years before). 15 822 participants were enrolled of whom 7915 were assigned to low-dose CT screening at baseline, then at years 2 and 4; and 7907 individuals had no screening (only smoking cessation). Interim results indicate a high cancer detection rate (85%) and low false-positive rate (1.4%) after three screening rounds combined and 2-year follow-up; with a positive predictive value of 40%.3

o low-dose CT screening at baseline, then at years 2 and 4; and 7907 individuals had no screening (only smoking cessation). Interim results indicate a high cancer detection rate (85%) and low false-positive rate (1.4%) after three screening rounds combined and 2-year follow-up; with a positive predictive value of 40%.3 There are ongoing attempts to use CT more efficiently (including the LungSEARCH trial outlined below), or biomarkers. The UK Lung Screening randomised trial uses a validated Liverpool Lung Project risk model to identify individuals who have a risk of at least 5% of developing lung cancer in the next 5 years, and only these are invited to participate. They are then randomised to receive either a single (prevalence) CT scan, or no screening.4 Another ongoing trial (in Scotland) is using a set of immune antibody markers to identify those at high risk of lung cancer and then follow these subjects closely with CT. Lung cancer risk in smokers with COPD The first National Health and Nutrition Examination Survey showed that individuals with COPD were significantly more likely to develop lung cancer than those with normal lung function.5 The cumulative risk of developing incident lung cancers in patients with moderate COPD was 4% after 5 years, and several other studies have shown a significant association between impaired lung function and lung cancer risk. It is also known that patients with COPD (of all grades of severity) are at increased risk of developing lung cancer if they have abnormal sputum cytology. These findings were used to help justify the LungSEARCH trial.

her studies have shown a significant association between impaired lung function and lung cancer risk. It is also known that patients with COPD (of all grades of severity) are at increased risk of developing lung cancer if they have abnormal sputum cytology. These findings were used to help justify the LungSEARCH trial. Rationale for LungSEARCH Targeting high-risk subjects should increase the yield of cancers, making it more cost-effective. COPD represents an additional risk for developing lung cancer, as are previous head and neck cancers, asbestos exposure and more advanced age. When launched, LungSEARCH was the first trial to focus on high-risk individuals and to use an initial cheaper screen to identify those who may have predisposition to lung cancer or have severe dysplasia/carcinoma in situ and allow detailed investigation. In the UK, squamous cell tumours were prevalent, and since most of these arise from central airways, we wished to re-explore the value of sputum cytology plus cytometry as a tool to identify malignant cells and also dysplastic cells, which could be precursors of malignancy. There was evidence to suggest that the sensitivity of sputum screening may be significantly enhanced by the application of computer-assisted image analysis (automated image cytometry) to quantitatively analyse the nuclear structure and DNA content of individual cells.

tic cells, which could be precursors of malignancy. There was evidence to suggest that the sensitivity of sputum screening may be significantly enhanced by the application of computer-assisted image analysis (automated image cytometry) to quantitatively analyse the nuclear structure and DNA content of individual cells. LungSEARCH study design Eligible participants were predominantly well but had mild to moderate COPD (using the Global Initiative for Chronic Obstructive Lung Disease criteria), and were also either current smokers with ≥20 pack years smoking history and/or 20-year duration of smoking, or former smokers who had quit within 8 years with ≥20 pack years smoking history and/or 20-year duration of smoking.

to moderate COPD (using the Global Initiative for Chronic Obstructive Lung Disease criteria), and were also either current smokers with ≥20 pack years smoking history and/or 20-year duration of smoking, or former smokers who had quit within 8 years with ≥20 pack years smoking history and/or 20-year duration of smoking. Participants were randomised to the control group (no intervention) or annual screening for 5 years. The screened group were asked to provide early morning sputum as the initial screen, which is a cheap and accessible test. All those with abnormal sputum results (positive cytology or cytometry) would then be offered surveillance for up to 5 years using an annual CT scan and annual autofluorescence bronchoscopy (AFB); the pathway is summarised in figure 1. Those with normal sputum results would provide another sample yearly, unless they have an abnormal result. The majority of published studies on AFB have shown a significant increase in diagnostic sensitivity for dysplasia and carcinoma in situ. There is also evidence to suggest that the detection of invasive carcinoma is enhanced by fluorescence bronchoscopy and this technique has been used to assess patients with abnormal sputum cytology. The clinicians in each participating centre would decide from any abnormal bronchoscopic and/or CT appearances whether follow-up should be more than annual. Figure 1 Flow diagram showing what happens to individuals in the screened (surveillance) arm of the LungSEARCH trial. AFB, autofluorescence bronchoscopy.

Participants were randomised to the control group (no intervention) or annual screening for 5 years. The screened group were asked to provide early morning sputum as the initial screen, which is a cheap and accessible test. All those with abnormal sputum results (positive cytology or cytometry) would then be offered surveillance for up to 5 years using an annual CT scan and annual autofluorescence bronchoscopy (AFB); the pathway is summarised in figure 1. Those with normal sputum results would provide another sample yearly, unless they have an abnormal result. The majority of published studies on AFB have shown a significant increase in diagnostic sensitivity for dysplasia and carcinoma in situ. There is also evidence to suggest that the detection of invasive carcinoma is enhanced by fluorescence bronchoscopy and this technique has been used to assess patients with abnormal sputum cytology. The clinicians in each participating centre would decide from any abnormal bronchoscopic and/or CT appearances whether follow-up should be more than annual. Figure 1 Flow diagram showing what happens to individuals in the screened (surveillance) arm of the LungSEARCH trial. AFB, autofluorescence bronchoscopy. The control subjects were not contacted after randomisation and are only expected to have an exit chest X-ray at the end of 5 years. Any cancers or deaths would be identified through patient records or the national cancer registry.

Figure 1 Flow diagram showing what happens to individuals in the screened (surveillance) arm of the LungSEARCH trial. AFB, autofluorescence bronchoscopy. The control subjects were not contacted after randomisation and are only expected to have an exit chest X-ray at the end of 5 years. Any cancers or deaths would be identified through patient records or the national cancer registry. The trial is powered to detect a stage shift between the screened and control groups; specifically to show that the proportion of cancers that are early stage (I and II) is 50% in the surveillance group and only 10% in the control group; and for this we require at least 74 lung cancers. The screened group were also asked to participate in Lung reSEARCH, where annual blood and sputum would be collected and stored in a central tissue bank at Leeds University. All those undergoing AFB would also be asked to consent to provide control samples of bronchial mucosa as well as two samples from any suspicious area, in addition to those samples taken for pathology. We therefore have the opportunity to follow the natural history of dysplastic lesions for up to 5 years, providing an important contribution on the natural history of dysplastic lesions, whose management remains controversial.

ll as two samples from any suspicious area, in addition to those samples taken for pathology. We therefore have the opportunity to follow the natural history of dysplastic lesions for up to 5 years, providing an important contribution on the natural history of dysplastic lesions, whose management remains controversial. Progress The study recruited 1568 individuals between August 2007 and March 2011 (785 in the screened group and 783 controls). Follow-up finishes in March 2016. Ten centres across the UK were involved, with COPD patients approached from either the hospital clinics or local general practitioners (GP) practices. The centres were in London (Royal Brompton, Chelsea and Westminster, and University College London Hospital), Leeds, Cambridge, Leicester, Manchester, Sunderland, Coventry and Belfast. Recruitment from general practices proved much more difficult than expected, due mainly to reluctance of the majority of practices to participate in research. Many demanded payment for access and a separate grant was obtained to provide funds for this. The level of GP co-operation was lowest in London. As of April 2015, 65 lung cancers have been found through screening, patient medical records or the national cancer registry. Initial results showed an incidence of abnormal sputum (using either cytology or cytometry) of 23% of those in the screened arm in the first year, and 15% in the second year (excluding those who had a previous abnormal sputum).

have been found through screening, patient medical records or the national cancer registry. Initial results showed an incidence of abnormal sputum (using either cytology or cytometry) of 23% of those in the screened arm in the first year, and 15% in the second year (excluding those who had a previous abnormal sputum). The study will inform on the value of a simple initial screen in a high-risk population, the value of annual CT surveillance in those with sputum dysplasia, the role of annual AFB, the effect of smoking cessation in this high-risk group and cost-effectiveness of this screening policy. If sputum testing in high-risk individuals identifies a group especially at risk of lung cancer, whose tumour is discovered during surveillance, and the study demonstrates a stage shift between the control and the active group, then a larger study looking at mortality could be performed to confirm the hypothesis generated in this study.

in high-risk individuals identifies a group especially at risk of lung cancer, whose tumour is discovered during surveillance, and the study demonstrates a stage shift between the control and the active group, then a larger study looking at mortality could be performed to confirm the hypothesis generated in this study. Collaborators: The following are members of the LungSEARCH collaborative Group and are contributing to this trial. Dr Pallav Shah, Royal Brompton Hospital, London; Professor Marco Novelli, University College Hospital, London; Dr Gabrijela Kocjan, University College Hospital, London; Dr Penny Shaw, University College Hospital, London; Dr Magali Taylor, University College Hospital, London; Dr Simon Padley, Royal Brompton Hospital, London; Professor Chris Griffiths, St Bartholomew's Hospital, London; Dr Mary Falzon, University College Hospital, London; Dr Paul Plant, St James's University Hospital, Leeds; Dr Matthew Callister, St James's University Hospital, Leeds; Dr Mick Peake, Glenfield Hospital, Leicester; Dr Robert Rintoul, Papworth Hospital, Cambridge; Dr Jeremy George, UCLH; Professor Sam Janes, UCLH; Dr Niall Keaney, Sunderland Royal Hospital; Dr Kishore Sridharan, Sunderland Royal Hospital; Dr Paul Dhillon, University Hospital Coventry; Dr Timothy Warke, Belfast City Hospital; Dr Nicholas Magee, Belfast City Hospital; Dr Richard Booton, Wythenshawe Hospital.

idge; Dr Jeremy George, UCLH; Professor Sam Janes, UCLH; Dr Niall Keaney, Sunderland Royal Hospital; Dr Kishore Sridharan, Sunderland Royal Hospital; Dr Paul Dhillon, University Hospital Coventry; Dr Timothy Warke, Belfast City Hospital; Dr Nicholas Magee, Belfast City Hospital; Dr Richard Booton, Wythenshawe Hospital. Contributors: Both authors have devised, written and supervised this trial. All contributors acknowledged have either been responsible for co-ordinating a trial centre, or the sputum, pathology or imaging analysis. Competing interests: None declared. Ethics approval: SW England MREC. Provenance and peer review: Not commissioned; internally peer reviewed. Data sharing statement: Once the trial is completed results and data will be summated for peer review publication.

Background Non-tuberculous mycobacteria (NTM) are increasingly being isolated from the sputum of adults and children with cystic fibrosis (CF) both in North America and Europe. Estimates of the prevalence of NTM in the CF population have ranged from 1.3% in the earliest study reported in 19841 to 32.7% in a review of patients with CF over age 40 in Colorado.2 The NTM species most commonly identified in individuals with CF from North America and Europe are the slow growing Mycobacterium avium complex (MAC, including M. avium, M. intracellulare and M. chimaera), which can be found in up to 72% of NTM-positive sputum cultures, and the rapid growing M. abscessus complex (comprising the subspecies M. abscessus subsp abscessus (M. a. abscessus), M. a. bolletii and M. a. massiliense (the latter currently classified as part of M. a. bolletii)), which in many centres has now become the most common NTM isolated from individuals with CF. There has been a rise in the prevalence of NTM-positive cultures in respiratory samples from individuals with CF over the last three decades, which probably reflects a true increase in the frequency of NTM infection. A number of CF studies (eg, Renna et al3) show year-on-year increases in NTM-positive cultures with no change in surveillance intensity or culture methodology.

cultures in respiratory samples from individuals with CF over the last three decades, which probably reflects a true increase in the frequency of NTM infection. A number of CF studies (eg, Renna et al3) show year-on-year increases in NTM-positive cultures with no change in surveillance intensity or culture methodology. Possible reasons for increased NTM-positive cultures in individuals with CF include: increases in environmental exposure to NTM through more permissive temperature settings of home water heaters and more contact with shower aerosols, increased antibiotic usage creating more NTM permissive lung niches, greater chronic use of medications that might impair host immunity to NTM3 and/or spread of NTM through person-to-person transmission.4 NTM can cause progressive inflammatory lung damage, a condition termed ‘NTM pulmonary disease’ (NTM-PD), which is defined by the presence of specific microbiological, clinical and radiological features.5 However, it has become clear that NTM can also transiently, intermittently or permanently reside within the lungs of CF individuals without causing NTM-PD, thus representing asymptomatic infection and creating considerable difficulties in deciding how best to screen for and diagnose NTM.

cal and radiological features.5 However, it has become clear that NTM can also transiently, intermittently or permanently reside within the lungs of CF individuals without causing NTM-PD, thus representing asymptomatic infection and creating considerable difficulties in deciding how best to screen for and diagnose NTM. Further challenges exist in knowing how best to identify NTM in respiratory samples, when and how to initiate treatment for NTM-PD and how NTM may impact individuals under consideration for lung transplantation. As a consequence, the Cystic Fibrosis Foundation (CFF) and European Cystic Fibrosis Society (ECFS) sought to generate a consensus recommendations document to support and standardise the management of NTM infection in children and adults with CF, permitting prospective evaluation of current best practice and forming a foundation for future research programmes.

ndation (CFF) and European Cystic Fibrosis Society (ECFS) sought to generate a consensus recommendations document to support and standardise the management of NTM infection in children and adults with CF, permitting prospective evaluation of current best practice and forming a foundation for future research programmes. Methods The CFF and the ECFS invited experts to participate in the statement development process. The 19 member committee consisted of professionals with expertise in CF and NTM and included adult and paediatric CF physicians, lung transplant physicians, microbiologists, infectious disease specialists and a parent of an individual with CF. The committee convened in May 2012 and divided into five subgroups, each responsible for a specific topic: epidemiology and risk factors, screening, microbiology, treatment and transplantation. Each subgroup developed topic-specific questions using the PICO format (population, intervention, comparison, outcome). Questions were reviewed and approved by the entire committee before systematic literature searches were conducted. The members of each subgroup used the PICO questions to guide literature searches in PubMed. Searches were limited to English language and the period 1984–2013. Subgroup members also searched for topic relevant guidelines through searches of the ATS website, the IDSA website, the Clinical Laboratory Standards Institute website and the UK CF Trust website.

PICO questions to guide literature searches in PubMed. Searches were limited to English language and the period 1984–2013. Subgroup members also searched for topic relevant guidelines through searches of the ATS website, the IDSA website, the Clinical Laboratory Standards Institute website and the UK CF Trust website. After reviewing relevant literature and existing guidelines, subgroup members drafted recommendation statements. In October 2012, a second meeting was convened, and subgroups finalised draft recommendation statements. The committee also voted to set 80% agreement of all 19 members as the threshold for acceptance of a recommendation statement.

literature and existing guidelines, subgroup members drafted recommendation statements. In October 2012, a second meeting was convened, and subgroups finalised draft recommendation statements. The committee also voted to set 80% agreement of all 19 members as the threshold for acceptance of a recommendation statement. Each subgroup submitted final draft questions for entry into an electronic survey tool (Survey Monkey) for the purposes of anonymous voting and comment by all members. A project coordinator administered the survey, and committee members were asked to rate each statement on a scale of 0 (completely disagree) to 9 (completely agree), with 80% or between 7 and 9 being considered ‘good’ agreement. Space for entering free text was also provided after each statement to allow members to cite literature in support of their opinions or suggested revisions. All committee members were required to vote on each statement regardless of their role or expertise. Multiple rounds of voting and revisions to the statements were conducted, and for each round committee members were requested to complete their voting within 3 weeks. The committee chairs reviewed the results from each round and updated the statements based on comments entered by respondents for subsequent rounds.

ise. Multiple rounds of voting and revisions to the statements were conducted, and for each round committee members were requested to complete their voting within 3 weeks. The committee chairs reviewed the results from each round and updated the statements based on comments entered by respondents for subsequent rounds. A draft of the recommendations was presented at the 2013 North American Cystic Fibrosis Conference and the ECFS Meeting. In addition, the committee solicited feedback from the CF communities in the USA and Europe, which included physicians, nurses, physical and respiratory therapists, parents and individuals with CF. Comments collected from this process were considered by the committee in the development of the final recommendation statements. Results Three rounds of voting were conducted to achieve 80% consensus for each statement. Fifty-three statements were included in the first round of voting and 50 statements in the second and third rounds. Final statements are shown in figure 1. Figure 1 Cystic Fibrosis Foundation and European Cystic Fibrosis Society recommendations on non-tuberculous mycobacteria (NTM) management in cystic fibrosis (CF).

Results Three rounds of voting were conducted to achieve 80% consensus for each statement. Fifty-three statements were included in the first round of voting and 50 statements in the second and third rounds. Final statements are shown in figure 1. Figure 1 Cystic Fibrosis Foundation and European Cystic Fibrosis Society recommendations on non-tuberculous mycobacteria (NTM) management in cystic fibrosis (CF). Discussion The management of individuals with CF infected with NTM is extremely challenging. The limited amounts of published research and clinical trial data provide inadequate evidence to base management decisions on how best to screen, diagnose, detect and treat NTM-PD. As a response to this urgent clinical need, the CF Foundation and ECFS formed a committee of clinicians, scientists and infectious disease experts to develop recommendations to guide and assist clinicians in the management of NTM-PD in individuals with CF. The committee believe these recommendations should serve as a benchmark for current medical care while providing a framework to inform the development of clinical, translation and basic research studies to generate robust evidence to base future iterations of these management guidelines leading to better outcomes for individuals with CF infected with NTM.

ations should serve as a benchmark for current medical care while providing a framework to inform the development of clinical, translation and basic research studies to generate robust evidence to base future iterations of these management guidelines leading to better outcomes for individuals with CF infected with NTM. Contributors: All authors contributed to the expert committee on guidelines development (cochairs: RAF, CSH; steering committee: RAF, CSH, BCM, KNO, LS, KAS, SEH); committee subgroups: epidemiology and risk factors (KNO (lead), IS-G, KLW); screening (JAN (lead), RLG, KK); microbiology (J-LH (lead), RJW, JvI, RAF); treatment (CLD (lead), DB, LS, ARS, CSH) and transplantation (PGN (lead), PC). Funding: Cystic Fibrosis Foundation; European Cystic Fibrosis Society, The Wellcome Trust & Cambridge NIHR BRC (RAF); Intramural programme of the National Heart, Lung, and Blood Institute, NIH (KNO); Vaincre La Mucoviscidose (VLMIC1014 and RF20120600689) and the Région Ile-de-France Domaine d'Intérêt Majeur Maladies Infectieuses et Emergentes (J-LH); CF Foundation Clinical Research Award (NICK13A0) (JAN) and Imperial College London NIHR Respiratory BRU (DB). Competing interests: None declared. Patient consent: Obtained. Provenance and peer review: Not commissioned; internally peer reviewed.

Anand Shah and Scott R Henderson (AS and SRH): a 72-year-old retired male physician was referred with poorly controlled asthma, necessitating oral glucocorticoid therapy. He had been diagnosed with occupational asthma 9 years earlier with initially good control on an inhaled corticosteroid (ICS). In the preceding year, his symptoms deteriorated with increasing wheeze, dyspnoea and persistent productive cough with 5 mL sputum daily associated with reduced exercise tolerance of 30–50 metres on the flat. He also reported a postnasal drip and allergic rhinitis. Medical history included two nasal polypectomies and paroxysmal tachycardias, with normal 24 h Holter, echocardiogram and exercise tolerance test. He had previously worked as an occupational physician for 15 years, monitoring coffee bean handlers for asthma and sensitivity to Aspergillus spp, commonly found in coffee bean cargoes. Interestingly, he himself was indirectly exposed to coffee bean dust (including green coffee beans prior to roasting and castor bean dust) through his interaction with workers and his place of work. He then developed asthma symptoms 6 years after taking up his Occupational Physician post, consistent with a diagnosis of occupational asthma, and a series of skin prick tests demonstrated sensitivity to Aspergillus fumigatus.

to roasting and castor bean dust) through his interaction with workers and his place of work. He then developed asthma symptoms 6 years after taking up his Occupational Physician post, consistent with a diagnosis of occupational asthma, and a series of skin prick tests demonstrated sensitivity to Aspergillus fumigatus. Examination at presentation revealed no significant abnormalities. Investigations showed FEV1 of 1.5 L (45% predicted), FVC of 3.0 (63% predicted) and peak expiratory flow (PEF) of 280 L/min (predicted). Sputum cytology revealed 12% eosinophils. Total IgE was elevated at 219 (0–120) kU/L with Aspergillus radioallergosorbent test (RAST) at 4.05 (0–0.34) IU/mL. Peripheral eosinophil count, on prednisolone, was normal. Skin prick tests were positive to A. fumigatus, grass mixture and house dust mite. Chest radiograph was unremarkable. Philip W Ind (PWI): lung function tests demonstrated an obstructive defect with an FEV1:FVC ratio of 50% and reduced PEF. In addition to probable IgE-related occupational asthma, his previous allergen exposure, documented immune reactivity to Aspergillus spp and failure to respond to conventional therapy raise the possibility of allergic bronchopulmonary aspergillosis (ABPA). The possible differential diagnosis of continued exposure to workplace allergens resulting in worsening of his condition did not apply as he was retired.

ocumented immune reactivity to Aspergillus spp and failure to respond to conventional therapy raise the possibility of allergic bronchopulmonary aspergillosis (ABPA). The possible differential diagnosis of continued exposure to workplace allergens resulting in worsening of his condition did not apply as he was retired. He was intolerant of long-acting β2 agonists, theophylline was relatively contraindicated by previous tachycardias and a tapering dose of prednisolone and high-dose ICS may be inadequate to control symptoms. A leukotriene antagonist (LTRA) was tried in view of his nasal symptoms. High resolution CT scan (HRCT) was also arranged to evaluate lung parenchyma. Susan J Copley (SJC): thoracic HRCT demonstrated right middle lobe and left lower lobe bronchiectasis (figure 1) which could be consistent with ABPA. Figure 1 Coronal (A) and sagittal (B and C) CT reconstructions demonstrating right middle lobe (C) and left lower lobe (A and B) bronchiectasis (arrows) consistent with allergic bronchopulmonary aspergillosis.

Susan J Copley (SJC): thoracic HRCT demonstrated right middle lobe and left lower lobe bronchiectasis (figure 1) which could be consistent with ABPA. Figure 1 Coronal (A) and sagittal (B and C) CT reconstructions demonstrating right middle lobe (C) and left lower lobe (A and B) bronchiectasis (arrows) consistent with allergic bronchopulmonary aspergillosis. AS and SRH: LTRA therapy was complicated by a severe rash over the legs that resolved on stopping the medication. Tapered prednisolone and high-dose ICS produced improvement. However, 3 months later, he re-presented with a 5-week history of severe muscle cramps, radiating from the buttocks to the lower legs. There was no associated urinary retention and no loss of perianal sensation, but he noted paraesthesia of both hands and feet and drenching night sweats. Neurological examination revealed reduced power (4+/5) in shoulder abduction, hip flexion and hip extension bilaterally, and diminished pinprick sensation over the fourth and fifth fingers in both hands and in a stocking distribution in his legs.

he noted paraesthesia of both hands and feet and drenching night sweats. Neurological examination revealed reduced power (4+/5) in shoulder abduction, hip flexion and hip extension bilaterally, and diminished pinprick sensation over the fourth and fifth fingers in both hands and in a stocking distribution in his legs. Investigations at this time showed elevated total white cell count of 17.5×109/L with peripheral eosinophilia of 7.1×109/L and elevated C reactive protein of 71 (0–10 mg/L). Multiple sputum samples were negative for bacteria and fungi. Chest X-ray was again normal. MRI of the spine did not reveal any neural compression, but mild degenerative changes. Antineutrophil cytoplasm antibody (ANCA) was positive with perinuclear immunofluorescence staining pattern (p-ANCA) and ELISA demonstrated elevated antimyeloperoxidase (MPO) antibody titre of 107 AU/mL (0–30).

X-ray was again normal. MRI of the spine did not reveal any neural compression, but mild degenerative changes. Antineutrophil cytoplasm antibody (ANCA) was positive with perinuclear immunofluorescence staining pattern (p-ANCA) and ELISA demonstrated elevated antimyeloperoxidase (MPO) antibody titre of 107 AU/mL (0–30). Alan D Salama (ASD): the history of severe myalgia with peripheral neuropathy and elevated inflammatory markers suggests the possibility of a systemic inflammatory process such as vasculitis. Previous history of worsening asthma, nasal polyps, elevated eosinophil count and positive ANCA with a specific anti-MPO antibody is consistent with eosinophilic granulomatosis with polyangiitis (EGPA). It is important to confirm the diagnosis histologically, if possible, and explore the extent of the disease. The patient's urine showed no evidence of blood or protein on dipstick testing. In addition, there were no cells or casts seen on microscopy. With a bland urine sediment, the most useful way to obtain a tissue diagnosis would be sural nerve biopsy. H Terence Cook (HTC): sural nerve biopsy (figure 2) shows two epineural arteries with fibrosis and luminal narrowing. One of these is surrounded by eosinophils. The appearances are consistent with arterial vasculitis, but there is no active fibrinoid necrosis in this specimen. The presence of eosinophils is consistent with EGPA.

Cook (HTC): sural nerve biopsy (figure 2) shows two epineural arteries with fibrosis and luminal narrowing. One of these is surrounded by eosinophils. The appearances are consistent with arterial vasculitis, but there is no active fibrinoid necrosis in this specimen. The presence of eosinophils is consistent with EGPA. Figure 2 Sural nerve biopsy showing two epineural arteries with fibrosis and luminal narrowing. One of these is surrounded by eosinophils. The appearances are consistent with arterial damage from vasculitis, but there is no active fibrinoid necrosis. The presence of the eosinophils is consistent with eosinophilic granulomatosis with polyangiitis. PWI: this patient has evidence of both ABPA and EGPA which may have occurred in relatively quick succession. Immunosuppression is important in managing both these conditions, and the patient should be started on a combination of a glucocorticoid and a steroid sparing agent such as azathioprine. AS and SRH: this patient's clinical symptoms markedly improved following immunosuppression with prednisolone 60 mg and azathioprine 100 mg. Anti-MPO antibody titres and inflammatory markers diminished rapidly and normalised after 2 months’ treatment. He continues to do well with maintenance low-dose prednisolone and azathioprine and high-dose ICS. It is, however, unclear whether his occupational asthma and diagnosis of ABPA and EGPA are pathologically linked.

i-MPO antibody titres and inflammatory markers diminished rapidly and normalised after 2 months’ treatment. He continues to do well with maintenance low-dose prednisolone and azathioprine and high-dose ICS. It is, however, unclear whether his occupational asthma and diagnosis of ABPA and EGPA are pathologically linked. PWI: ABPA represents a severe complication of A. fumigatus sensitisation which, in this patient, is likely to be related to prolonged occupational exposure to Aspergillus from coffee beans, leading to an IgE-directed immune response resulting in airway and parenchymal lung damage. Premorbid occupational asthma and atopic symptomatology support the patient's increased susceptibility to ABPA, the most common antigen-specific hypersensitivity disorder within the UK. In addition, a latency period can exist between exposure to coffee bean dust, A. fumigatus sensitivity and occupational asthma. This would be consistent with his occupational asthma initially occurring 6 years after exposure to the coffee bean plantation, and an additional disease process evolving 9 years later, such as the development of ABPA.

iod can exist between exposure to coffee bean dust, A. fumigatus sensitivity and occupational asthma. This would be consistent with his occupational asthma initially occurring 6 years after exposure to the coffee bean plantation, and an additional disease process evolving 9 years later, such as the development of ABPA. However, the association between ABPA and the emergence of Churg Strauss syndrome (CSS) is extremely unusual. To my knowledge, only two cases of ABPA ‘progressing’ to allergic granulomatosis and angiitis or EGPA have been reported.1 2 The precise mechanism(s) by which ABPA and possibly other allergic mycoses could relate to a systemic vasculitis such as EGPA are unclear. Both show evidence of disordered humoral immunity, with IgE elevation in allergic mycoses and presence of ANCA in some patients with EGPA; however, these are generally IgG isotypes. Eosinophilic reactivity is present in both but is most marked in cases of EGPA. ASD: autoantibodies, in particular antibacterial permeability increasing protein, are associated with infective lung disease such as cystic fibrosis and non-cystic fibrosis bronchiectasis. It is noteworthy that ANCA can also be a feature of infective lung disease. In the same way that reducing bacterial load is important, treatment of fungal colonisation/infection with anti-fungals (in addition to glucocorticoids) was considered in this patient with the aim of reducing inflammatory processes resulting in autoantibody production.

A can also be a feature of infective lung disease. In the same way that reducing bacterial load is important, treatment of fungal colonisation/infection with anti-fungals (in addition to glucocorticoids) was considered in this patient with the aim of reducing inflammatory processes resulting in autoantibody production. Numerous A. fumigatus antigens are recognised that stimulate both humoral and cellular immune responses mediated directly and indirectly by allergens binding to IgE/IgG and stimulating CD4+ Th2 cell subsets. Human leucocyte antigen (HLA) association has been found in ABPA, in small cohort studies, with increased frequencies of HLA-DR2 and DR5. Although recent data have also demonstrated genetic risk factors for EGPA, including HLA-DR genes (HLA-DRB4) and interleukin-10 production, the precise trigger initiating EGPA is unknown. Eotaxin-3, an eotactic chemokine (CCL26), has been shown to be a very specific marker in EGPA and may be a useful biomarker for disease monitoring, but whether elevated levels occur in patients with EGPA in response to certain environmental antigens, such as Aspergillus, is unknown.3 However, it is of interest to note that eotaxin-3 was shown to be the only eotaxin upregulated in asthmatics following respiratory allergen challenge,4 lending support to the notion that certain allergens may provoke production of this chemokine, which, in genetically susceptible individuals, may promote excessive eosinophilic recruitment and damage.

te that eotaxin-3 was shown to be the only eotaxin upregulated in asthmatics following respiratory allergen challenge,4 lending support to the notion that certain allergens may provoke production of this chemokine, which, in genetically susceptible individuals, may promote excessive eosinophilic recruitment and damage. The pathogenic significance of anti-MPO autoantibodies, found in 30%–40% of patients with EGPA at diagnosis, is less established. Presence of anti-MPO antibodies has a closer association with microscopic polyangiitis, presenting with necrotising glomerulonephritis, alveolar haemorrhage, purpura and mononeuritis multiplex. A previous suggestion of cross-reactivity between MPO and eosinophilic peroxidise has not been confirmed. Therefore, there may be common antigen/allergen presentation in genetically susceptible individuals mediating both ABPA and EGPA; however, this association remains uncommon. Of note, we did not measure this patient's eotaxin-3 level or have his HLA-DR typing.

ween MPO and eosinophilic peroxidise has not been confirmed. Therefore, there may be common antigen/allergen presentation in genetically susceptible individuals mediating both ABPA and EGPA; however, this association remains uncommon. Of note, we did not measure this patient's eotaxin-3 level or have his HLA-DR typing. Charles D Pusey (CDP): another possible link is the use of LTRA therapy in the management of ABPA. Numerous cases of EGPA have been reported following the introduction of LTRAs for the management of asthma. Although this association has often been attributed to therapeutic withdrawal or cessation of corticosteroids, unmasking EGPA, an increasing number of cases have since been reported showing a spectrum of EGPA development in the presence and absence of concomitant corticosteroid use. Recent systematic review of this possible iatrogenic association demonstrated a clear temporal relationship between LTRAs and onset of EGPA.5 The exact mechanism by which LTRAs may induce EGPA remains unknown, and there is a distinct lack of experimental work to guide understanding. Nonetheless, the occurrence of 63 cases of EGPA on LTRAs, reported to the Committee on Safety of Medicines, over the last 30 years highlights the need for cautious introduction of selective cys-LT1 receptor antagonists in patients with severe asthma.

there is a distinct lack of experimental work to guide understanding. Nonetheless, the occurrence of 63 cases of EGPA on LTRAs, reported to the Committee on Safety of Medicines, over the last 30 years highlights the need for cautious introduction of selective cys-LT1 receptor antagonists in patients with severe asthma. Author note: Occam's razor, named after William of Ockham a Franciscan friar of the early 14th century, refers to the principle of logic that states that the simplest explanation is the most likely correct; in science with competing hypotheses, the one with the fewest assumptions is preferred. Hickam's dictum is the counter-argument to Occam's razor in medicine; the statement “Patients can have as many diseases as they damn well please” attributed to John Hickam a great US physician and educator who originally published research on lung function in heart and lung disease. Contributors: SRH co-ordinated the writing of each section of the manuscript with all other authors having an equal role. Competing interests: None declared. Patient consent: Obtained. Provenance and peer review: Not commissioned; externally peer reviewed.

Key messages What is the key question? What is the burden of respiratory disease and utility of spirometry in aiding assessment of respiratory health and diagnosis of respiratory disease in community-living 85 year olds in the UK? What is the bottom line? The study reveals a substantial burden of respiratory disease and symptoms in 85 year olds but also considerable discordance between physician-diagnosed COPD and confirmatory spirometry evidence in the very old that have important implications for clinical practice. Why read on? This study represents the largest and most detailed assessment to date of respiratory health status and challenges of using spirometry criteria in respiratory diagnosis in the very old, aged 85 years and over, which are now the fastest growing sector of the population. Introduction The very old, aged 85 years and older, are now the most rapidly expanding age sector of most populations worldwide.1 Data from the 2011 England and Wales Census showed a doubling of the over 85 years age group between 1985 and 2010, from nearly 0.7 million to over 1.4 million,2 and numbers are projected to double again between 2010 and 2030.3 This age group frequently uses healthcare resource in primary and secondary care,4 and therefore understanding their health status and burden of disease is important for training of health professionals and for organisation of healthcare provision.

ion,2 and numbers are projected to double again between 2010 and 2030.3 This age group frequently uses healthcare resource in primary and secondary care,4 and therefore understanding their health status and burden of disease is important for training of health professionals and for organisation of healthcare provision. Symptoms relating to the respiratory system, in particular dyspnoea, are common in those 85 years and older with a prevalence of over 40%,5 and are frequently a reason for older people to seek healthcare. Although it is recognised that many chronic respiratory diseases increase in prevalence and severity with age, it is also clear that dyspnoea is non-specific and may be associated with non-pulmonary morbidities.6 In the very old, assessment of respiratory health is further complicated by the physiological changes that occur as part of ‘normal’ or ‘healthy’ ageing, such as loss of lung elasticity and reduced thoracic cage movement, which will have an effect on objective measures of lung function.7

ith non-pulmonary morbidities.6 In the very old, assessment of respiratory health is further complicated by the physiological changes that occur as part of ‘normal’ or ‘healthy’ ageing, such as loss of lung elasticity and reduced thoracic cage movement, which will have an effect on objective measures of lung function.7 Current national and international guidelines on the management of COPD have obstructive spirometry (FEV1/FVC ratio <0.7) as a key diagnostic test directing physicians towards the use of specific respiratory medications.8 9 However, the accuracy of lung function criteria for the diagnosis of airflow obstruction or restrictive lung disease in very old people has been questioned due to the intrapulmonary and extrapulmonary physiological changes that occur in this age group as part of normal ageing.10 This may cause misdiagnosis and inappropriate use of medications in this population. Moreover, a previous study in a population with a mean age of 73 years suggested that COPD may be either overdiagnosed or underdiagnosed depending on the approach taken to defining abnormal lung function.11

as part of normal ageing.10 This may cause misdiagnosis and inappropriate use of medications in this population. Moreover, a previous study in a population with a mean age of 73 years suggested that COPD may be either overdiagnosed or underdiagnosed depending on the approach taken to defining abnormal lung function.11 This study aimed to address the lack of knowledge about respiratory health, prevalence of lung disease and objective measures of lung function in the very old using baseline data from the Newcastle 85+ Study,4 12 a large population-based cohort of 85 year olds. Specifically the study aimed to: assess the extent of common respiratory symptoms and the prevalence of physician-diagnosed lung disease, particularly COPD; and to assess the accuracy of COPD diagnosis based on lung function measurements, respiratory symptoms and identification of risk factors, and the degree to which respiratory medication was appropriately prescribed. Finally, in a healthy reference group (HRG), the study aimed to evaluate the applicability of three standard methods of interpreting lung function measurements as normal or abnormal to disentangle the effects of lung disease and ‘normal’ or ‘healthy’ ageing on measured lung function.

ation was appropriately prescribed. Finally, in a healthy reference group (HRG), the study aimed to evaluate the applicability of three standard methods of interpreting lung function measurements as normal or abnormal to disentangle the effects of lung disease and ‘normal’ or ‘healthy’ ageing on measured lung function. Methods Full details of the Newcastle 85+ Study methodology have been reported.12 In brief, members of the 1921 birth cohort living in Newcastle upon Tyne or North Tyneside (North-East England) were recruited around their 85th birthday over a 17 month-period spanning 2006 and 2007. Participants included people living at home or in institutional care and regardless of their current health status. More detailed methods are available as online supplementary materials.

Tyne or North Tyneside (North-East England) were recruited around their 85th birthday over a 17 month-period spanning 2006 and 2007. Participants included people living at home or in institutional care and regardless of their current health status. More detailed methods are available as online supplementary materials. Existing diagnoses of respiratory disease, respiratory symptoms, respiratory medications and environmental risk factors Current and past respiratory diagnoses were identified from a general practice records review (GPRR) using a predetermined checklist of chronic respiratory diseases. Data on use, but not doses, of respiratory medications were also obtained from GPRR. Data on symptoms of breathlessness, cough, wheeze and sputum production were obtained by a structured questionnaire administered as part of a domiciliary multidimensional health assessment (MDHA) conducted by a research nurse in the participant's home or institution. Specifically, participants were asked whether shortness of breath limited their day-to-day activities and responses were then used to assign an Medical Research Council (MRC) dyspnoea score.8 Participants were asked about any relevant environmental exposure in their occupation or at home, specifically detailed smoking history and relevant occupational history (including exposure to heavy industry generally as well as the chemical industry, asbestos and coal mining). Two measures of disease burden were used: a disease count (maximum 18 diseases) previously determined in the cohort; and a non-respiratory disease count excluding COPD and other respiratory disease (maximum 16 diseases).4 Further details of the individual respiratory diagnoses, medications and chronic non-respiratory diseases included in the disease count are provided (see online supplementary methods).

usly determined in the cohort; and a non-respiratory disease count excluding COPD and other respiratory disease (maximum 16 diseases).4 Further details of the individual respiratory diagnoses, medications and chronic non-respiratory diseases included in the disease count are provided (see online supplementary methods). Lung function measurements Spirometry and peak flow measurements were performed at the participant's place of residence by a trained research nurse using MicroLab Spirometer and Spida V.5 software (Micro Medical, Rochester, UK). The aim was to obtain three technically satisfactory maximal effort ‘blows’ to generate reproducible FEV1, FVC and peak expiratory flow measurement (PEF); blows were repeated until this was achieved or maximum effort reached. Blows were assessed for technical adequacy using in-built Spida algorithms. All spirometry curves were assessed independently by a respiratory clinical physiologist and those able to produce at least two adequate blows were included in the analysis. If the necessary quality was lacking they were excluded from analysis. Demispan was measured as a surrogate for height13 (calculated using standard equations) and height used with age and gender to calculate predicted values for FEV1, FVC and peak flow using equations in the UK Department of Health guide.9 Spirometry was classified (see online supplementary table S2) as normal, obstructive or restrictive based on the FEV1/FVC ratio of 0.7 and the percentage of predicted values for FEV1 and FVC, with obstructive spirometry further classified as mild, moderate or severe based on Global Initiative in Obstructive Lung Disease (GOLD) criteria.10 In addition, we reanalysed the data using criteria presented by the Global Lung Function Initiative (GLI)14 which provides alternative prediction model equations validated for ages 3 years to 95 years (see online supplementary tables S3–S5).

vere based on Global Initiative in Obstructive Lung Disease (GOLD) criteria.10 In addition, we reanalysed the data using criteria presented by the Global Lung Function Initiative (GLI)14 which provides alternative prediction model equations validated for ages 3 years to 95 years (see online supplementary tables S3–S5). Healthy reference group To establish the distribution of normal lung function in people aged 85 years, we identified a HRG of participants with no respiratory symptoms, no respiratory diagnoses, no current use of respiratory medications and no non-respiratory diagnosis which might influence lung function (eg, Parkinson's disease, kyphoscoliosis, heart failure, ankylosing spondylitis) in their GPRR. Those with a BMI >30 were also excluded from HRG. Lung function in the HRG was compared against equation derived15 predicted values based on gender and height by three accepted methods: percentage predicted value; lower limit of normal (LLN) using American Thoracic Society/European Respiratory Society (ATS/ERS) criteria;16 and Z scores.

30 were also excluded from HRG. Lung function in the HRG was compared against equation derived15 predicted values based on gender and height by three accepted methods: percentage predicted value; lower limit of normal (LLN) using American Thoracic Society/European Respiratory Society (ATS/ERS) criteria;16 and Z scores. Statistical methods Gender differences in respiratory symptoms, diagnoses, environmental exposures and medications were examined using χ2 and Mann-Whitney U tests. Gender differences in lung function were investigated in the whole sample, COPD group and the HRG using Mann-Whitney U test for continuous measures, χ2 and Fisher's exact tests for categorised measures and Kruskal-Wallis test for ordered categorised measures. The relationship between FEV1 and PEF scores was assessed using Pearson's correlation coefficients. Sensitivity analyses were carried out to examine differences between those included and excluded from analysis due to lack of spirometry measures and those with and without an MRC dyspnoea score. All analyses were conducted using Stata V.12.0 (StataCorp; College Station, Texas, USA).

Pearson's correlation coefficients. Sensitivity analyses were carried out to examine differences between those included and excluded from analysis due to lack of spirometry measures and those with and without an MRC dyspnoea score. All analyses were conducted using Stata V.12.0 (StataCorp; College Station, Texas, USA). Results Sociodemographic, non-respiratory health characteristics and environmental exposures of the study population Details of the Newcastle 85+ Study population have been reported previously, and the study population was broadly sociodemographically representative of the local population, and of England and Wales, including the proportion in institutional care.4 Data from MDHA and GPRR was available for 845 participants, 58.2% (845/1453) of those eligible (figure 1); their mean (SD) age was 85.5 (0.4) years, 62.3% (526/845) were female and 99.6% (839/845) were of white ethnic group (table 1). Three-quarters were living in standard housing, 12.8% (108/845) in warden-supported accommodation and 10.2% (86/845) in institutional care. The median (IQR) chronic disease count was 5(3–6) with no significant gender difference (p=0.074). Although the 845 participants were a non-random sample of the eligible population, data from an additional 188 participants (18%) who opted for GPRR only showed no difference in respiratory diagnoses compared with those who participated fully. Table 1 Sociodemographic and health characteristics of the total Newcastle 85+ cohort (n=845) and by gender

Results Sociodemographic, non-respiratory health characteristics and environmental exposures of the study population Details of the Newcastle 85+ Study population have been reported previously, and the study population was broadly sociodemographically representative of the local population, and of England and Wales, including the proportion in institutional care.4 Data from MDHA and GPRR was available for 845 participants, 58.2% (845/1453) of those eligible (figure 1); their mean (SD) age was 85.5 (0.4) years, 62.3% (526/845) were female and 99.6% (839/845) were of white ethnic group (table 1). Three-quarters were living in standard housing, 12.8% (108/845) in warden-supported accommodation and 10.2% (86/845) in institutional care. The median (IQR) chronic disease count was 5(3–6) with no significant gender difference (p=0.074). Although the 845 participants were a non-random sample of the eligible population, data from an additional 188 participants (18%) who opted for GPRR only showed no difference in respiratory diagnoses compared with those who participated fully. Table 1 Sociodemographic and health characteristics of the total Newcastle 85+ cohort (n=845) and by gender Men (n=319) Women (n=526) Overall cohort (n=845) p Value* Ethnicity % (N) White 99.4 (316) 99.8 (523) 99.6 (839) 0.272† Living arrangements % (N) Standard housing 83.4 (266) 73.2 (385) 77.0 (651) 0.002† Sheltered housing 10.3 (33) 14.3 (75) 12.8 (108) Institutional care 6.3 (20) 12.6 (66) 10.2 (86) Smoking % (N) Never 25.6 (81) 42.0 (220) 35.8 (301) <0.001† Former 69.9 (221) 51.5 (270) 58.5 (491) Current 4.4 (14) 6.5 (34) 5.7 (48) Occupational exposures % (N) Heavy industry 41.2 (126) 16.6 (83) 25.9 (209) <0.001† Coal mining 11.4 (35) 0.0 (0) 4.3 (35) <0.001‡ Chemical industry 11.1 (34) 4.0 (20) 6.7 (54) <0.001† Asbestos exposure 28.9 (88) 1.6 (8) 12.0 (96) <0.001† Respiratory symptoms % (N) Cough 28.3 (88) 25.8 (129) 26.7 (217) 0.425† Wheeze 25.0 (78) 20.2 (101) 22.0 (179) 0.109† Sputum production 40.7 (127) 28.0 (140) 32.9 (267) <0.001 MRC dyspnoea score % (N) 1 50.2 (123) 40.5 (143) 44.5 (266) 0.048§ 2 11.4 (28) 19.0 (67) 15.9 (95) 3 20.4 (50) 17.6 (62) 18.7 (112) 4 15.1 (37) 17.0 (60) 16.2 (97) 5 2.9 (7) 6.0 (21) 4.7 (28) Respiratory diagnoses % (N) COPD 17.9 (57) 15.8 (83) 16.6 (140) 0.429† Asthma 6.9 (22) 12.7 (67) 10.5 (89) 0.007† Bronchiectasis 2.5 (8) 1.5 (8) 1.9 (16) 0.308† Pulmonary fibrosis 0.0 (0) 0.2 (1) 0.1 (1) 1.000‡ Asbestosis 1.6 (5) 0.0 (0) 0.6 (5) 0.008‡ Pneumoconiosis 1.3 (4) 0.0 (0) 0.5 (4) 0.020‡ TB 4.4 (14) 4.9 (26) 4.7 (40) 0.713† Respiratory medications Inhaled short-acting β-2 adrenoreceptor agonists 9.1 (29) 11.4 (60) 10.5 (89) 0.288† Inhaled muscarinic antagonists 3.8 (12) 3.8 (20) 3.8 (32) 0.976† Oral theophylline 0.3 (1) 0.5 (3) 0.5 (4) 0.598‡ Combination short-acting bronchodilators 0.6 (2) 0.0 (0) 0.2 (2) 0.142‡ Inhaled corticosteroids 5.3 (17) 7.8 (41) 6.9 (58) 0.169† Combination inhaled Corticosteroids and long-acting β-2 adrenoreceptor agonists 1.9 (6) 2.1 (11) 2.0 (17) 0.833† Oral leukotriene receptor antagonists 0.0 (0) 0.4 (2) 0.2 (2) 0.529‡ Oral mucolytics 0.6 (2) 0.2 (1) 0.4 (3) 0.560‡ At least one respiratory medication % (N) 12.2 (39) 14.5 (76) 13.6 (115) 0.361† Disease count median (IQR) 4 (3–6) 5 (4–6) 5 (3–6) 0.074§ Comorbid disease count median (IQR) 4 (3–6) 5 (4–6) 5 (3–6) 0.047§ *Comparison of men and women.

gonists 0.0 (0) 0.4 (2) 0.2 (2) 0.529‡ Oral mucolytics 0.6 (2) 0.2 (1) 0.4 (3) 0.560‡ At least one respiratory medication % (N) 12.2 (39) 14.5 (76) 13.6 (115) 0.361† Disease count median (IQR) 4 (3–6) 5 (4–6) 5 (3–6) 0.074§ Comorbid disease count median (IQR) 4 (3–6) 5 (4–6) 5 (3–6) 0.047§ *Comparison of men and women. §Mann–Whitney U test. †χ2 test. ‡Fisher's exact test,. Denominators vary due to missing values. Figure 1 Flow chart illustrating how the total cohort of Newcastle 85+ Study participants was subdivided in the respiratory study sample, demonstrating why different numbers of participants are included in the analyses. The derivation of the study groups are shown in the flow chart; note that for some variables the number of participants included is less than 845 due to missing data, the reasons for which are detailed. The basis for the healthy reference group (HRG) was the 845 participants who had multidimensional health assessment (MDHA) and general practice records review (GPRR) conducted. Of these, 786 (93.0%) had spirometry performed of whom 772 performed it adequately; a further 35 participants with missing demispan were removed (unable to calculate predicted blows), resulting in 737. Participants with at least one respiratory condition, those with respiratory symptoms and those on respiratory medication were excluded which reduced the group size to 170. Other conditions which have an effect on spirometry values were also taken into account leading to exclusion of a further 19 participants. The remaining 151 (17.9% of 845) participants formed the HRG.

those with respiratory symptoms and those on respiratory medication were excluded which reduced the group size to 170. Other conditions which have an effect on spirometry values were also taken into account leading to exclusion of a further 19 participants. The remaining 151 (17.9% of 845) participants formed the HRG. Almost three quarters (74.4%, 235/316) of men and over half of women (58.0%, 304/524) had smoked in their lifetime, although very few (men: 4.4%, 14/316; women: 6.5%, 34/524) were current smokers. A significant proportion of men and women had occupational exposures which may have influenced respiratory health, with much higher prevalence in men (heavy industry: 41.2%, 126/306; coal mining: 11.4%, 35/307; chemical industry: 11.1%, 34/306; asbestos: 28.9%, 88/305), reflecting common historical occupations in this region of the UK (table 1). Respiratory diagnoses, symptomatology and medication use The most common physician-diagnosed respiratory condition was COPD with a prevalence of 16.6% (140/845) and no significant gender difference (p=0.43) (table 1). A diagnosis of asthma had been made in 10.5% (89/845) with a predominance in women (men: 6.9%; women: 12.7%; p=0.007). Other respiratory diagnoses were rare.

e The most common physician-diagnosed respiratory condition was COPD with a prevalence of 16.6% (140/845) and no significant gender difference (p=0.43) (table 1). A diagnosis of asthma had been made in 10.5% (89/845) with a predominance in women (men: 6.9%; women: 12.7%; p=0.007). Other respiratory diagnoses were rare. Chronic cough was self-reported in 26.7% (217/812) and wheeze in 22.0% (179/812) of participants. Regular sputum production was more common in men (men: 40.7%, 127/312; women: 28.0%, 140/500; p<0.001). An MRC dyspnoea score was assigned in 598 (70.8%) participants since in the other participants their activity could be limited by other non-respiratory conditions. Half (123/245) of the men and 40.5% (143/353) of the women allocated an MRC dyspnoea score had no limitations to their daily activities due to breathlessness. The most frequently prescribed respiratory medications were inhaled short-acting β-2 adrenoreceptor agonists (10.5%, 89/845 of participants) followed by inhaled corticosteroids (6.9%, 58/845) (table 1). Only 2.0% (17/845) were taking a combination inhaler containing corticosteroid and a long-acting β-2 adrenoreceptor agonist. The use of other respiratory medications was unusual (table 1).

rt-acting β-2 adrenoreceptor agonists (10.5%, 89/845 of participants) followed by inhaled corticosteroids (6.9%, 58/845) (table 1). Only 2.0% (17/845) were taking a combination inhaler containing corticosteroid and a long-acting β-2 adrenoreceptor agonist. The use of other respiratory medications was unusual (table 1). Lung function measurements Spirometry was performed by 786 (93.0%) participants (figure 1), most of whom (98.2%, 772/786) provided at least two adequate blows conforming to ATS/ERS guidelines. Demispan was available for 737 participants with adequate expiratory effort and consistency allowing calculation of predicted spirometry values, with these 737 forming the spirometry group (table 2). Comparison of the spirometry group (n=737) with those excluded due to missing/inadequate spirometry and/or missing demispan (n=108) showed those excluded were more likely to be female, living in an institution and with previous exposure to the chemical industry, but not significantly different in smoking history; respiratory symptoms, diagnoses or medications; or dyspnoea scores (see online supplementary table S1). Table 2 Results of spirometry in the cohort completing spirometry with adequate reproducible blows and demispan available for calculation of predicted blows (n=737)

Lung function measurements Spirometry was performed by 786 (93.0%) participants (figure 1), most of whom (98.2%, 772/786) provided at least two adequate blows conforming to ATS/ERS guidelines. Demispan was available for 737 participants with adequate expiratory effort and consistency allowing calculation of predicted spirometry values, with these 737 forming the spirometry group (table 2). Comparison of the spirometry group (n=737) with those excluded due to missing/inadequate spirometry and/or missing demispan (n=108) showed those excluded were more likely to be female, living in an institution and with previous exposure to the chemical industry, but not significantly different in smoking history; respiratory symptoms, diagnoses or medications; or dyspnoea scores (see online supplementary table S1). Table 2 Results of spirometry in the cohort completing spirometry with adequate reproducible blows and demispan available for calculation of predicted blows (n=737) Men (n=293) Women (n=444) All (n=737) p Value* Actual spirometry median (IQR) FEV1 (l/s) 1.8 (1.4–2.2) 1.2 (1.0–1.5) 1.4 (1.1–1.8) <0.001† FVC (l/s) 2.7 (2.2–3.2) 1.8 (1.4–2.1) 2.0 (1.6–2.6) <0.001† FEV1/FVC 0.7 (0.6–0.8) 0.7 (0.6–0.8) 0.7 (0.6–0.8) 0.006† PEF (L/m) 441 (323–604) 283 (196–362) 328 (233–450) <0.001† % predicted median (IQR) FEV1 78.8 (62.4–94.3) 83.4 (68.1–98.8) 81.5 (65.6–97.1) 0.008† FVC 83.4 (70.3–99.6) 96.6 (79.1–113.7) 90.8 (74.1–108.4) <0.001† Spirometry % (N) Normal 28.0 (82) 33.3 (148) 31.2 (230) 0.108‡ Restrictive 13.7 (40) 16.2 (72) 15.2 (112) Obstructive 58.4 (171) 50.5 (224) 53.6 (395) Grading of obstructive spirometry§ % (N) Mild 35.7 (61) 43.3 (97) 40.0 (158) 0.059¶ Moderate 46.8 (80) 45.1 (101) 45.8 (181) Severe 14.6 (25) 9.8 (22) 11.9 (47) Very severe 2.9 (5) 1.8 (4) 2.3 (9) FEV1 % (N) Below LLN 25.9 (76) 13.3 (59) 18.3 (135) <0.001** Normal range 73.7 (216) 85.6 (380) 80.9 (596) Above ULN 0.3 (1) 1.1 (5) 0.8 (6) FEV1 Z-score median (IQR) 1.0 (0.2–1.7) 0.6 (0.0–1.2) 0.8 (0.1–1.4) <0.001† FVC % (N) Below LLN 21.2 (62) 9.2 (41) 14.0 (103) <0.001‡ Normal range 77.1 (226) 86.3 (383) 82.6 (609) Above ULN 1.7 (5) 4.5 (20) 3.4 (25) FVC Z-score median (IQR) 0.9 (0.0–1.5) 0.1 (−0.6–0.9) 0.4 (−0.4–1.2) <0.001† Oxygen saturation median (IQR) 97 (96–98) 97 (96–98) 97 (96–98) 0.513† *Comparison of men and women.

1† FVC % (N) Below LLN 21.2 (62) 9.2 (41) 14.0 (103) <0.001‡ Normal range 77.1 (226) 86.3 (383) 82.6 (609) Above ULN 1.7 (5) 4.5 (20) 3.4 (25) FVC Z-score median (IQR) 0.9 (0.0–1.5) 0.1 (−0.6–0.9) 0.4 (−0.4–1.2) <0.001† Oxygen saturation median (IQR) 97 (96–98) 97 (96–98) 97 (96–98) 0.513† *Comparison of men and women. †Mann–Whitney U test. ‡χ2 test. §This is based on the 395 participant subsample with obstructive spirometry. ¶Kruskal–Wallis test. **Fisher's exact test. LLN, lower limit of normal; PEF, peak expiratory flow; ULN, upper limit of normal. Of the whole spirometry group, 31.2% (230/737) had a normal FEV1/FVC ratio and 15.2% (112/737) had a restrictive pattern. Obstructive spirometry was the most common finding (men: 58.4%, 171/293; women: 50.5%, 224/444) but with no gender difference in the spread of severity (table 2). Measured values of FEV1, FVC and PEF in the spirometry group were normally distributed but with a much wider distribution range than that of the predicted values (figure 2). Scatter plots of the measured FEV1 and FVC against the predicted values showed more participants with measured values below the predicted values than above suggesting a downward shift in the population as a whole (figure 3). The spread of FEV1 measurements around the predicted values was much wider in men than women. Figure 2 Distribution curves of FEV1 and FVC in all participants in spirometry cohort (all, men and women) measured (blue) and predicted (green). Figure 3 Scatter plots of spirometry and peak expiratory flow in all participants in spirometry cohort.

Of the whole spirometry group, 31.2% (230/737) had a normal FEV1/FVC ratio and 15.2% (112/737) had a restrictive pattern. Obstructive spirometry was the most common finding (men: 58.4%, 171/293; women: 50.5%, 224/444) but with no gender difference in the spread of severity (table 2). Measured values of FEV1, FVC and PEF in the spirometry group were normally distributed but with a much wider distribution range than that of the predicted values (figure 2). Scatter plots of the measured FEV1 and FVC against the predicted values showed more participants with measured values below the predicted values than above suggesting a downward shift in the population as a whole (figure 3). The spread of FEV1 measurements around the predicted values was much wider in men than women. Figure 2 Distribution curves of FEV1 and FVC in all participants in spirometry cohort (all, men and women) measured (blue) and predicted (green). Figure 3 Scatter plots of spirometry and peak expiratory flow in all participants in spirometry cohort. Prevalence and accuracy of physician-diagnosed COPD Of the spirometry group, 16.7% (123/737) had physician-diagnosed COPD (COPD group) of whom 57.7% (71/123) were female and 23.8% (29/123) reported being ‘never smokers’ (table 3). More than half of the ‘never smokers’ with a COPD diagnosis had no occupational exposures either. Table 3 Descriptive characteristics of subset with physician-diagnosed COPD in general practitioner records

Prevalence and accuracy of physician-diagnosed COPD Of the spirometry group, 16.7% (123/737) had physician-diagnosed COPD (COPD group) of whom 57.7% (71/123) were female and 23.8% (29/123) reported being ‘never smokers’ (table 3). More than half of the ‘never smokers’ with a COPD diagnosis had no occupational exposures either. Table 3 Descriptive characteristics of subset with physician-diagnosed COPD in general practitioner records Men (n=52) Women (n=71) All (n=123) p Value* Smoking % (N) Never 21.2 (11) 25.7 (18) 23.8 (29) 0.637† Former 67.3 (35) 67.1 (47) 67.2 (82) Current 11.5 (6) 7.1 (5) 9.0 (11) Occupational exposure % (N) Heavy industry 49.0 (25) 19.7 (14) 32.0 (39) 0.001† Coal mining 17.7 (9) 0.0 (0) 7.4 (9) <0.001‡ Chemical 13.7 (7) 2.8 (2) 7.4 (9) 0.034‡ Asbestos 33.3 (17) 7.1 (5) 18.2 (22) <0.001† Non-smokers with no occupational exposures % (N) 3.9 (2) 18.3 (13) 12.2 (15) 0.023† Respiratory symptoms % (N) Cough 46.2 (24) 53.5 (38) 50.4 (62) 0.419† Wheeze 53.9 (28) 56.3 (40) 55.3 (68) 0.784† Sputum production 63.5 (33) 54.3 (38) 58.2 (71) 0.310† MRC dyspnoea score % (N) 1 26.8 (11) 12.5 (7) 18.6 (18) 0.035§ 2 9.8 (4) 16.1 (9) 13.4 (13) 3 34.2 (14) 19.6 (11) 25.8 (25) 4 22.0 (9) 33.9 (19) 28.9 (28) 5 7.3 (3) 17.9 (10) 13.4 (13) Comorbid respiratory diagnoses % (N) Asthma 25.0 (13) 49.3 (35) 39.0 (48) 0.006† Bronchiectasis 7.7 (4) 2.8 (2) 4.9 (6) 0.240‡ Asbestosis 7.7 (4) 0.0 (0) 3.3 (4) 0.030‡ Pulmonary fibrosis 0.0 (0) 0.0 (0) 0.0 (0) – Pneumoconiosis 3.9 (2) 0.0 (0) 1.6 (2) 0.177‡ TB 5.8 (3) 9.9 (7) 8.1 (10) 0.516‡ Medications % (N) Inhaled short-acting β-2 adrenoreceptor agonists 36.5 (19) 52.1 (37) 45.5 (56) 0.087† Inhaled muscarinic antagonists 17.3 (9) 22.5 (16) 20.3 (25) 0.477† Oral theophylline 1.9 (1) 4.2 (3) 3.3 (4) 0.637‡ Combination short-acting bronchodilators 1.9 (1) 0.0 (0) 0.8 (1) 0.423‡ Inhaled corticosteroids 17.3 (9) 38.0 (27) 29.3 (36) 0.013† Combination inhaled corticosteroids and long-acting β-2 adrenoreceptor agonists 11.5 (6) 12.7 (9) 12.2 (15) 0.849† Oral leukotriene receptor antagonists 0.0 (0) 1.4 (1) 0.8 (1) 1.000‡ Oral mucolytics 1.9 (1) 1.4 (1) 1.6 (2) 1.000‡ Oral glucocorticoid therapy 5.8 (3) 4.2 (3) 4.9 (6) 0.697‡ At least 1 respiratory medication % (N) 46.2 (24) 66.2 (47) 57.7 (71) 0.026† Disease count median (IQR) 5 (4–7) 6 (5–7) 6 (4–7) 0.156§ Non-respiratory disease count median (IQR) 5 (4–6) 6 (5–7) 6 (4–7) 0.064§ *Comparison of men and women.

1.4 (1) 1.6 (2) 1.000‡ Oral glucocorticoid therapy 5.8 (3) 4.2 (3) 4.9 (6) 0.697‡ At least 1 respiratory medication % (N) 46.2 (24) 66.2 (47) 57.7 (71) 0.026† Disease count median (IQR) 5 (4–7) 6 (5–7) 6 (4–7) 0.156§ Non-respiratory disease count median (IQR) 5 (4–6) 6 (5–7) 6 (4–7) 0.064§ *Comparison of men and women. †χ2 test. ‡Fisher's exact test. §Mann–Whitney U test. Denominators vary due to missing values. In the COPD group, only 45.5% (56/123) were taking short-acting inhaled β-2 adrenoreceptor agonist bronchodilator therapy, 20.3% (25/123) were taking inhaled long-acting muscarinic antagonists, 41.5% (51/123) were on inhaled corticosteroids either as monotherapy (36/51) or in combination with a long-acting β-agonist (15/51). There was minimal use of theophylline preparations, oral mucolytics or oral leukotriene receptor antagonists and none of the COPD group used home oxygen (table 3). The proportion of the COPD group that were on at least one respiratory medication differed significantly between men and women (men: 46.2%, 24/52; women: 66.2%, 47/71; p=0.026), although a sizeable proportion (42.3%, 52/123) of those with a COPD diagnosis were not on any (table 3). There was a significant overlap in the diagnoses of asthma and COPD with 61% (48/78) of those with an asthma diagnosis also being diagnosed with COPD.

men and women (men: 46.2%, 24/52; women: 66.2%, 47/71; p=0.026), although a sizeable proportion (42.3%, 52/123) of those with a COPD diagnosis were not on any (table 3). There was a significant overlap in the diagnoses of asthma and COPD with 61% (48/78) of those with an asthma diagnosis also being diagnosed with COPD. Respiratory symptoms were common but not universal in the COPD group with 50.4% (62/123) reporting cough and 58.2% (71/123) sputum production. Nevertheless 26.8% (11/52) of men and 12.5% (7/71) of women with a COPD diagnosis had only minimal breathlessness (MRC dyspnoea score=1).

men and women (men: 46.2%, 24/52; women: 66.2%, 47/71; p=0.026), although a sizeable proportion (42.3%, 52/123) of those with a COPD diagnosis were not on any (table 3). There was a significant overlap in the diagnoses of asthma and COPD with 61% (48/78) of those with an asthma diagnosis also being diagnosed with COPD. Respiratory symptoms were common but not universal in the COPD group with 50.4% (62/123) reporting cough and 58.2% (71/123) sputum production. Nevertheless 26.8% (11/52) of men and 12.5% (7/71) of women with a COPD diagnosis had only minimal breathlessness (MRC dyspnoea score=1). Only 75.6% (93/123) of the COPD group had obstructive spirometry by GOLD criteria (table 4). There was no gender difference in severity of airflow obstruction (based on % predicted FEV1) and only 63.4% (78/123) of the COPD group fulfilled the UK National Institute of Health and Care Excellence (NICE) guidelines spirometry definition of moderate, severe or very severe disease (table 4). Furthermore, only 63.4% (78/123) of the COPD group fulfilled the UK NICE guidelines spirometry definition of moderate, severe or very severe disease. When FEV1 was classified by the LLN approach, 48.1% (25/52) of men and 33.8% (24/71) of women from the COPD group fell below the LLN with all other participants falling between the LLN and upper limit of normal, suggesting that a substantial proportion (60.2%, 74/123) of those with physician-diagnosed COPD had an FEV1 in the normal range and/or no airflow obstruction on spirometry measurement. When applying the GLI prediction models to the COPD group, 48.1% (25/52) men and 50.7% (36/71) women satisfied criteria for airflow obstruction (see online supplementary table S4). The degree of agreement between physician-diagnosed COPD and spirometric evidence of airflow obstruction using either GOLD or GLI criteria is poor when assessed by the McNemar test (see online supplementary table S6).

nd 50.7% (36/71) women satisfied criteria for airflow obstruction (see online supplementary table S4). The degree of agreement between physician-diagnosed COPD and spirometric evidence of airflow obstruction using either GOLD or GLI criteria is poor when assessed by the McNemar test (see online supplementary table S6). Table 4 Results of spirometry in the subgroup with physician-diagnosed COPD (n=123)

nd 50.7% (36/71) women satisfied criteria for airflow obstruction (see online supplementary table S4). The degree of agreement between physician-diagnosed COPD and spirometric evidence of airflow obstruction using either GOLD or GLI criteria is poor when assessed by the McNemar test (see online supplementary table S6). Table 4 Results of spirometry in the subgroup with physician-diagnosed COPD (n=123) Men (n=52) Women (n=71) All (n=123) p Value* Actual median (IQR) FEV1 1.4 (1.1–1.8) 1.0 (0.7–1.1) 1.1 (0.8–1.4) <0.001† FVC 2.4 (2.0–3.1) 1.6 (1.3–1.9) 1.9 (1.5–2.3) <0.001† FEV1/FVC 0.6 (0.5–0.7) 0.6 (0.5–0.7) 0.6 (0.5–0.7) 0.591† PEF 382.5 (243–519) 218 (144–290) 259 (191–380) <0.001† %predicted median (IQR) FEV1 63.5 (50.9–73.4) 64.2 (51.7–79.9) 64.2 (51.3–76.4) 0.609† FVC 77.4 (64.2–94.1) 87.6 (70.4–101.0) 82.8 (68.2–99.8) 0.040† Spirometry %(N) Normal 7.7 (4) 8.5 (6) 8.1 (10) 0.959‡ Restrictive 15.4 (8) 16.9 (12) 16.3 (20) Obstructive 76.9 (40) 74.7 (53) 75.6 (93) Obstructive spirometry§ %(N) Mild 10.0 (4) 20.8 (11) 16.1 (15) 0.190¶ Moderate 60.0 (24) 56.6 (30) 58.1 (54) Severe 27.5 (11) 20.8 (11) 23.7 (22) Very severe 2.5 (1) 1.9 (1) 2.2 (2) FEV1 %(N) Below LLN 48.1 (25) 33.8 (24) 39.8 (49) 0.137** Normal range 51.9 (27) 66.2 (47) 60.2 (74) Above ULN 0.0 (0) 0.0 (0) 0.0 (0) FEV1 Z-score median (IQR) 1.6 (1.2–2.2) 1.3 (0.7–2.0) 1.5 (0.9–2.0) 0.039† FVC %(N) Below LLN 30.8 (16) 14.1 (10) 21.1 (26) 0.043** Normal range 69.2 (36) 84.5 (60) 78.1 (96) Above ULN 0.0 (0) 1.4 (1) 0.8 (1) FVC Z-score median (IQR) 1.1 (0.3–1.8) 0.6 (0.0–1.2) 0.8 (0.0–1.6) 0.008† Oxygen saturation median (IQR) 97 (96–98) 97 (95–98) 97 (95–98) 0.521† *Comparison of men and women.

2.0) 0.039† FVC %(N) Below LLN 30.8 (16) 14.1 (10) 21.1 (26) 0.043** Normal range 69.2 (36) 84.5 (60) 78.1 (96) Above ULN 0.0 (0) 1.4 (1) 0.8 (1) FVC Z-score median (IQR) 1.1 (0.3–1.8) 0.6 (0.0–1.2) 0.8 (0.0–1.6) 0.008† Oxygen saturation median (IQR) 97 (96–98) 97 (95–98) 97 (95–98) 0.521† *Comparison of men and women. †Mann–Whitney U test. ‡χ2 test. §This is based on the 93 participant subsample with obstructive spirometry. ¶Kruskal–Wallis test. **Fisher's exact test. LLN, lower limit of normal; PEF, peak expiratory flow; ULN, upper limit of normal. Assessment of lung function in an HRG Figure 1 shows the derivation of the HRG which comprised 20.5% (151/737) of the spirometry cohort (table 5). The distribution of measured and predicted FEV1, FVC and PEF in this group, by gender, are shown in figure 4 and table 5, with scatter plots of measured versus predicted FEV1 and FVC by gender in figure 5. Table 5 Results of spirometry in healthy reference group of participants (n=151)

Assessment of lung function in an HRG Figure 1 shows the derivation of the HRG which comprised 20.5% (151/737) of the spirometry cohort (table 5). The distribution of measured and predicted FEV1, FVC and PEF in this group, by gender, are shown in figure 4 and table 5, with scatter plots of measured versus predicted FEV1 and FVC by gender in figure 5. Table 5 Results of spirometry in healthy reference group of participants (n=151) Men (n=57) Women (n=94) All (n=151) p Value* Actual median (IQR) FEV1 2.0 (1.7–2.4) 1.4 (1.2–1.6) 1.5 (1.2–2.0) <0.001† FVC 2.9 (2.4–3.5) 1.9 (1.6–2.2) 2.1 (1.8–2.8) <0.001† FEV1/FVC 0.7 (0.6–0.8) 0.7 (0.7–0.8) 0.7 (0.6–0.8) 0.244† PEF 515 (340–647) 329.5 (243–417) 367 (263–515) <0.001† %predicted median (IQR) FEV1 90.1 (67.6–103.8) 93.8 (78.6–106.0) 91.6 (76.0–106.0) 0.154† FVC 92.3 (72.0–107.7) 101.2 (85.2–121.7) 97.5 (80.6–115.2) 0.006† Spirometry %(N) Normal 38.6 (22) 44.7 (42) 42.4 (64) 0.764‡ Restrictive 14.0 (98) 12.8 (12) 13.3 (20) Obstructive 47.4 (27) 42.6 (40) 44.4 (67) Obstructive spirometry§ %(N) Mild 48.2 (13) 62.5 (25) 56.7 (38) 0.137¶ Moderate 33.3 (9) 32.5 (13) 32.8 (22) Severe 11.1 (3) 5.0 (2) 7.5 (5) Very severe 7.4 (2) 0.0 (0) 3.0 (2) FEV1 %(N) Below LLN 21.1 (12) 5.3 (5) 11.3 (17) 0.008** Normal range 77.2 (44) 93.6 (88) 87.4 (132) Above ULN 1.8 (1) 1.1 (1) 1.3 (2) FEV1 Z-score median (IQR) 0.5 (−0.2–1.6) 0.3 (−0.2–0.9) 0.3 (−0.2–1.0) 0.071† FVC %(N) Below LLN 19.3 (11) 1.1 (1) 8.0 (12) <0.001** Normal range 79.0 (45) 91.5 (86) 86.8 (131) Above ULN 1.8 (1) 7.5 (7) 5.3 (8) FVC Z-score median (IQR) 0.4 (−0.4–1.5) −0.1 (−0.9–0.6) 0.1 (−0.7–0.9) 0.004† Oxygen saturation median (IQR) 98 (96–98) 98 (97–98) 98 (96–98) 0.970† *Comparison of men and women.

) 0.071† FVC %(N) Below LLN 19.3 (11) 1.1 (1) 8.0 (12) <0.001** Normal range 79.0 (45) 91.5 (86) 86.8 (131) Above ULN 1.8 (1) 7.5 (7) 5.3 (8) FVC Z-score median (IQR) 0.4 (−0.4–1.5) −0.1 (−0.9–0.6) 0.1 (−0.7–0.9) 0.004† Oxygen saturation median (IQR) 98 (96–98) 98 (97–98) 98 (96–98) 0.970† *Comparison of men and women. †Mann–Whitney U test. ‡χ2 test. §This is based on the 67 participant subsample with obstructive spirometry. ¶Kruskal–Wallis test. **Fisher's exact test. LLN, lower limit of normal; PEF, peak expiratory flow; ULN, upper limit of normal. Figure 4 Distribution curves of FEV1 and FVC of participants in the healthy reference group (all, men and women) measured (blue) and predicted (green). Figure 5 Scatter plots of spirometry and peak expiratory flow in the healthy reference group.

LLN, lower limit of normal; PEF, peak expiratory flow; ULN, upper limit of normal. Figure 4 Distribution curves of FEV1 and FVC of participants in the healthy reference group (all, men and women) measured (blue) and predicted (green). Figure 5 Scatter plots of spirometry and peak expiratory flow in the healthy reference group. Approximately half of the HRG (men: 47.4%, 27/57; women: 42.6%, 40/94) had a spirometry definition of airflow obstruction by GOLD criteria (table 5) yet did not fulfil the requirements for a diagnosis of COPD through lack of symptoms. Interestingly 19.2% (29/151) fulfilled a spirometry definition of at least moderate COPD using NICE criteria (obstructive spirometry and an FEV1 <80% predicted). The measured best PEF median (IQR) for this group was 367 (263–515) L/min, significantly higher in men (515 (340–647) L/min) than in women (329.5 (243–417) L/min) (p<0.001), and highly correlated with FEV1 (figure 5). When applying the GLI criteria to HRG only 17.5% (10/57) men and 16% (15/94) women (see online supplementary table 5) fulfilled criteria for airflow obstruction suggesting that GLI offered superiority to GOLD in spirometry interpretation in this age group.

) L/min) (p<0.001), and highly correlated with FEV1 (figure 5). When applying the GLI criteria to HRG only 17.5% (10/57) men and 16% (15/94) women (see online supplementary table 5) fulfilled criteria for airflow obstruction suggesting that GLI offered superiority to GOLD in spirometry interpretation in this age group. The measured spirometry values in HRG were compared with equation-derived15 predicted values based on gender and height using three different accepted approaches: percentage predicted value, LLN and Z scores (table 5). The median (IQR) percentage predicted value FEV1 in HRG was 90.1% (67.6–103.8%) in men and 93.8% (78.6–106.0%) in women. The measured FEV1 fell below LLN in 11.3% (17/151) of participants with a large gender difference (men: 21.1%, 12/57; women: 5.3%, 5/94; p=0.008). A significant gender difference was also found for the proportion of measured FVC falling below LLN with observed gender difference in the median Z-scores (table 5).

women. The measured FEV1 fell below LLN in 11.3% (17/151) of participants with a large gender difference (men: 21.1%, 12/57; women: 5.3%, 5/94; p=0.008). A significant gender difference was also found for the proportion of measured FVC falling below LLN with observed gender difference in the median Z-scores (table 5). Discussion This study presents the first evaluation of respiratory symptomatology, respiratory disease prevalence and objectively measured lung function in a large UK population-based single-year birth cohort of 85 year olds. It provides insight into the burden of respiratory disease and degree of respiratory impairment in very old people in an urban setting, and illustrates a population with substantial environmental exposures and smoking history, even in women. Furthermore, despite the higher rate of cognitive impairment with age, 93% of our cohort performed spirometry and of these 98% did so successfully which challenges reluctance to use spirometry in the very old and dispels misconceptions that they cannot perform spirometry successfully.

nd smoking history, even in women. Furthermore, despite the higher rate of cognitive impairment with age, 93% of our cohort performed spirometry and of these 98% did so successfully which challenges reluctance to use spirometry in the very old and dispels misconceptions that they cannot perform spirometry successfully. The participants are long-lived, and survivors of some of the most remarkable historical periods of our time, starting in the year of their birth immediately post World War I and the 1918 Spanish influenza pandemic. There were high levels of deprivation, and unemployment across Britain reached 17% in 1921. This period was pre welfare state, Housing Act (1930), Clean Air Act and widespread use of penicillin (1940). Many of the participants would have been nearing retirement age when the 1986 WHO: Ottawa Charter for Health Promotion was introduced—smoking rates are particularly high for men.

ss Britain reached 17% in 1921. This period was pre welfare state, Housing Act (1930), Clean Air Act and widespread use of penicillin (1940). Many of the participants would have been nearing retirement age when the 1986 WHO: Ottawa Charter for Health Promotion was introduced—smoking rates are particularly high for men. It is therefore not unexpected that a high prevalence of physician-diagnosed COPD (16.7%) was identified compared with previous self-reports of COPD of 10% in 65–74 year olds in the 2010 Health Survey for England.17 Nevertheless there were signs of potential misdiagnosis of COPD with a significant proportion of those with physician-diagnosed COPD having no evidence of airflow obstruction on spirometry, no smoking or occupational history and minimal symptoms. At the same time, a high proportion of our HRG fulfilled spirometry criteria for COPD using current GOLD/NICE guidelines, though use of LLN and GLI criteria rather than GOLD or NICE guidelines might reduce levels of misdiagnosis.

irflow obstruction on spirometry, no smoking or occupational history and minimal symptoms. At the same time, a high proportion of our HRG fulfilled spirometry criteria for COPD using current GOLD/NICE guidelines, though use of LLN and GLI criteria rather than GOLD or NICE guidelines might reduce levels of misdiagnosis. The risk of respiratory impairment increases with age due to the cumulative lifetime effect of environmental insults from active and passive cigarette smoking, air pollution, occupational dusts and infections.18 19 When this risk is added to the changes which occur in the respiratory system as part of normal ageing, including reduced ventilatory control, reduced respiratory muscle strength, increased compliance and less favourable respiratory mechanics due to reduced movement of the chest wall,7 it is not surprising that symptoms of cough, wheeze and dyspnoea are common in older people. All of these factors are likely to reduce measured lung function, which has been shown to be an independent risk factor for frailty and death.20–22 Distinguishing physiological age-related loss of lung function from a pathological disease process in the lungs is further complicated by a reduced perception of respiratory symptoms that occurs with increasing age as demonstrated by significantly reduced awareness of measured bronchospasm after a methacholine challenge in older compared with younger patients.23 Despite the high prevalence of chronic lung disease and respiratory symptoms, we found a significant proportion, 50% of men and 40% of women, with no reported limitations due to breathlessness suggesting many are either able to function very well or have a poor perception of symptoms.

compared with younger patients.23 Despite the high prevalence of chronic lung disease and respiratory symptoms, we found a significant proportion, 50% of men and 40% of women, with no reported limitations due to breathlessness suggesting many are either able to function very well or have a poor perception of symptoms. The strengths of this work are the comprehensive assessment of respiratory health and lung disease in a large population-based cohort of 85 year olds, including those in institutional care and those with cognitive impairment, in a stable urban setting and with little ethnic diversity. The cohort of >800 participants was achieved through engagement with 83% of the general practices in the area and a consent rate of almost 60% in those approached to participate. Previous studies of respiratory health in older subjects have relied on self-reported diagnoses whereas in our study the use of general practice records significantly improves the validity of our findings.24 25 Furthermore by conducting spirometry in the participant's place of residence using trained research nurses we were able to achieve a very high uptake of this assessment, in contrast to the known selection bias if participants had been required to attend a clinic for assessment. Although participants opting in for the health assessment were not a random sample of those eligible, there was little evidence to suggest they had more or less respiratory disease than those refusing the health assessment. In addition they were sociodemographically representative of their England and Wales birth cohort.4 A potential limitation of the study is that those who agreed to participate may be healthier and less frail than those who declined to participate and those with cognitive impairment may have been under-represented. Although some information was collected about why those invited declined to participate, we obviously do not have objective data on their respiratory health or disease burden. However the prevalence of COPD of 16.7% in those who agreed to MDHA and GPRR (n=845) was very similar to the prevalence of 16.5% reported previously in all participants with GPRR data (n=1030),4 suggesting that in terms of COPD, those agreeing to MDHA had similar respiratory health profiles to the larger study population.

ever the prevalence of COPD of 16.7% in those who agreed to MDHA and GPRR (n=845) was very similar to the prevalence of 16.5% reported previously in all participants with GPRR data (n=1030),4 suggesting that in terms of COPD, those agreeing to MDHA had similar respiratory health profiles to the larger study population. While 85 year olds in this urban area in North-East England are sociodemographically and ethnically similar to the same birth cohort in England and Wales as a whole, they may differ from those in other parts of the world.

ever the prevalence of COPD of 16.7% in those who agreed to MDHA and GPRR (n=845) was very similar to the prevalence of 16.5% reported previously in all participants with GPRR data (n=1030),4 suggesting that in terms of COPD, those agreeing to MDHA had similar respiratory health profiles to the larger study population. While 85 year olds in this urban area in North-East England are sociodemographically and ethnically similar to the same birth cohort in England and Wales as a whole, they may differ from those in other parts of the world. This study has revealed a substantial burden of respiratory symptoms and respiratory disease, particularly COPD, in a cohort of the very old aged 85 years; a group with substantial environmental exposures recorded through smoking and occupational exposure, which are known risk factors for lung disease. Despite these observations, we show a good proportion of participants functioning well with no respiratory symptoms or diagnoses. Lung function tests revealed only 75.6% of the COPD group satisfied spirometry criteria whereas 44% of the healthy group satisfied spirometry criteria for COPD using GOLD criteria. Healthcare professionals need to recognise that spirometry can be reliably assessed in the vast majority of this age group but care is needed as to how this is interpreted. Current definitions of COPD based on spirometry may lead to overdiagnosis in a group with transient symptoms and ‘normal’ lung ageing, whereas at the same time failure to use spirometry to assess symptoms in this age group may lead to mislabelling those with breathlessness or cough as having COPD when there are other explanations.

nitions of COPD based on spirometry may lead to overdiagnosis in a group with transient symptoms and ‘normal’ lung ageing, whereas at the same time failure to use spirometry to assess symptoms in this age group may lead to mislabelling those with breathlessness or cough as having COPD when there are other explanations. Supplementary Material Web supplement The authors thank the 85 year olds of Newcastle and North Tyneside, and their families and carers, for the generous donation of their time and personal information. The authors also thank the research nurses, data manager, project secretary, and the North of England Commissioning Support Unit and local general practitioners and their staff.

85 year olds of Newcastle and North Tyneside, and their families and carers, for the generous donation of their time and personal information. The authors also thank the research nurses, data manager, project secretary, and the North of England Commissioning Support Unit and local general practitioners and their staff. Contributors: AJF: study design; data preparation; literature review; data analysis and interpretation; development and writing of the paper. MEY: literature review, statistical analysis and interpretation of data; development and critical review of paper drafts. JC: study design; supervision of data collection; data preparation; literature review; data analysis and interpretation; and the development and writing of the paper. TS: study design; supervision of data collection; literature review; development of paper and critical review of paper drafts. TBLK: overall leadership of the Newcastle 85+ Study; study design; and critical review of paper drafts. KD: study design; participant recruitment; supervision of data collection; data preparation; and critical review of paper drafts. CJ: study design; supervision of statistical analysis; data interpretation; and critical review of paper drafts. PAC: study design; data interpretation; and critical review of paper drafts. Funding: UK Medical Research Council and Biotechnology and Biological Sciences Research Council (G0500997), Dunhill Medical Trust (R124/0509); Newcastle Healthcare Charity; NIHR Newcastle Biomedical Research Centre.

Contributors: AJF: study design; data preparation; literature review; data analysis and interpretation; development and writing of the paper. MEY: literature review, statistical analysis and interpretation of data; development and critical review of paper drafts. JC: study design; supervision of data collection; data preparation; literature review; data analysis and interpretation; and the development and writing of the paper. TS: study design; supervision of data collection; literature review; development of paper and critical review of paper drafts. TBLK: overall leadership of the Newcastle 85+ Study; study design; and critical review of paper drafts. KD: study design; participant recruitment; supervision of data collection; data preparation; and critical review of paper drafts. CJ: study design; supervision of statistical analysis; data interpretation; and critical review of paper drafts. PAC: study design; data interpretation; and critical review of paper drafts. Funding: UK Medical Research Council and Biotechnology and Biological Sciences Research Council (G0500997), Dunhill Medical Trust (R124/0509); Newcastle Healthcare Charity; NIHR Newcastle Biomedical Research Centre. Competing interests: The authors declare: the authors’ institution had financial support from the UK Medical Research Council and the Biotechnology and Biological Sciences Research Council (G0500997), the Dunhill Medical Trust (a privately endowed foundation having no connection with tobacco industries) and the Newcastle Healthcare Charity in terms of funding the submitted work; CJ was supported by the AXA Research Fund from 2010 to 2015.

cil and the Biotechnology and Biological Sciences Research Council (G0500997), the Dunhill Medical Trust (a privately endowed foundation having no connection with tobacco industries) and the Newcastle Healthcare Charity in terms of funding the submitted work; CJ was supported by the AXA Research Fund from 2010 to 2015. Patient consent Obtained. Ethics approval: The research complied with the requirements of the Declaration of Helsinki. Ethical approval was obtained from the Newcastle and North Tyneside 1 Research Ethics Committee (reference number 06/Q0905/2). Provenance and peer review: Not commissioned; externally peer reviewed.

Key messages What is the key question? Can venous blood gas analysis replace arterial blood gas sampling in the initial assessment of patients with COPD exacerbations? What is the bottom line? Over two-thirds of arterial blood gas samples could be replaced by the simpler and safer use of venous blood gas analysis. Why read on? This paper describes agreement between arteriovenous measures for key blood gas parameters and presents a simple algorithm for the substitution of arterial blood gas sampling with venous in the initial management of patients admitted to hospital with a COPD exacerbation. Introduction Exacerbations of COPD are the second most common cause of emergency hospital admission in the UK, with an estimated 94 000 per year.1 COPD exacerbations have a very high risk of mortality; 50% of people with a severe exacerbation will die within 4 years of an admission.1

Why read on? This paper describes agreement between arteriovenous measures for key blood gas parameters and presents a simple algorithm for the substitution of arterial blood gas sampling with venous in the initial management of patients admitted to hospital with a COPD exacerbation. Introduction Exacerbations of COPD are the second most common cause of emergency hospital admission in the UK, with an estimated 94 000 per year.1 COPD exacerbations have a very high risk of mortality; 50% of people with a severe exacerbation will die within 4 years of an admission.1 The recognition that high flow oxygen therapy can induce hypercapnia in susceptible patients during exacerbations of COPD,2 and that respiratory acidosis is associated with a worse outcome3 4 led to a rise in arterial blood gas (ABG) sampling to measure pH, PaCO2, PaO2 and . The current National Institute for Health and Care Excellence (NICE) COPD guidelines recommend obtaining an ABG in all patients admitted to hospital with a COPD exacerbation.5 Arterial sampling is more technically difficult and reportedly more painful6 than venous blood gas (VBG) sampling. Administration of local anaesthetic prior to arterial sampling is recommended but seldom used, as shown in a recent survey of junior doctors where 91% never or rarely used local anaesthesia when performing arterial puncture.7 Using less invasive measures of pCO2 and SaO2 could greatly benefit patients by both decreasing pain and streamlining the care pathway.

to arterial sampling is recommended but seldom used, as shown in a recent survey of junior doctors where 91% never or rarely used local anaesthesia when performing arterial puncture.7 Using less invasive measures of pCO2 and SaO2 could greatly benefit patients by both decreasing pain and streamlining the care pathway. Recent meta-analysis data suggest good agreement between venous and arterial measurements of pH, and base excess.8–11 In diabetes care arterial sampling has been replaced with venous for the monitoring of diabetic ketoacidosis.12 The use of venous samples to guide treatment in COPD exacerbations has been limited, perhaps because the relationship between arterial and venous measures of CO2 is less strong, although a PvCO2 of >6 kPa has been shown to have 100% (95% CI 97% to 100%) sensitivity in identifying patients with clinically relevant hypercapnia.13 We set out to assess the relationship between arterial and venous measures of PCO2, pH and , and between arterial and pulse oximetry oxygen saturations during exacerbations of COPD, in order to establish whether VBG analysis combined with pulse oximetry could replace ABG analysis in the initial assessment of COPD exacerbations.

t to assess the relationship between arterial and venous measures of PCO2, pH and , and between arterial and pulse oximetry oxygen saturations during exacerbations of COPD, in order to establish whether VBG analysis combined with pulse oximetry could replace ABG analysis in the initial assessment of COPD exacerbations. Methods Patients admitted to Nottingham University Hospitals Trust with a doctor-diagnosed exacerbation of COPD were considered for inclusion. Patients were included in the study unless they refused. If they were unable to give informed consent an approved alternative decision maker was approached. Recruited patients had an ABG and pulse oximetry performed by a junior doctor as per routine care, and a parallel paired VBG sample. Care was guided by the ABG results as per current clinical guidelines. Arterial samples were collected via a heparinised needle and syringe system, and venous samples were aspirated into a separate heparinised blood gas syringe via a butterfly needle or needle. Samples were processed as soon as possible on the same ward-based blood gas analyser. Analysers were calibrated regularly in accordance with the department of medical physics standard operating procedure. Research nurses collected data on demographics, body mass index, smoking status, pain scores (visual analogue scores 0=no pain, 10=worst pain imaginable) for arterial and venous sampling and the number of attempts taken to acquire each sample.

ordance with the department of medical physics standard operating procedure. Research nurses collected data on demographics, body mass index, smoking status, pain scores (visual analogue scores 0=no pain, 10=worst pain imaginable) for arterial and venous sampling and the number of attempts taken to acquire each sample. The main outcomes were the agreements between ABG and VBG parameters, and between ABG and pulse oximetry measures of oxygen saturation. Secondary outcome measures included pain scores, and the number of attempts taken to obtain blood. Statistical analysis Agreements between venous and arterial samples for CO2, pH and , and between SaO2 and SpO2 were assessed using the Bland–Altman method. Previous studies found 40% of patients attending the Emergency Department with an exacerbation of COPD had arterial hypercarbia, and it was estimated that 200 patients would allow us to calculate the sensitivity of the venous CO2 screening threshold for detection of arterial hypercarbia with CIs of <5%.9 13 Using this sample size of 200, our 95% CI around the limits of agreement are estimated at 0.24× the SD of the mean difference for which pH is ±0.001.14 Missing data were not imputed. Receiver operating characteristic curves were used to estimate how venous CO2, pH, predicted arterial values and how SpO2 predicted SaO2. Area under the curve, sensitivity and specificity were calculated to predict an arterial pH ≥7.35 and bicarbonate ≥21, and for the pulse oximetry result to predict an SaO2 of ≥92%.

. Receiver operating characteristic curves were used to estimate how venous CO2, pH, predicted arterial values and how SpO2 predicted SaO2. Area under the curve, sensitivity and specificity were calculated to predict an arterial pH ≥7.35 and bicarbonate ≥21, and for the pulse oximetry result to predict an SaO2 of ≥92%. Results Over the time course of the study (from 3 February 2013 to 10 January 2014), there were 1376 admissions with a coded diagnosis of COPD exacerbation. Of these, 234 participants were recruited and had at least one paired sample of blood gases. Twelve patients declined study participation. The mean age of the population was 71 years (SD 10.8) and 50% of the population was male. Characteristics of the population are shown in table 1. Table 1 Characteristics of study population n N Age (years) mean (SD) 71.0 (10.8) 234 Sex N (%) Male 118 (50.4) Female 116 (49.6) BMI (kg/m2) mean (SD) 26.2 (7.9) 223 Smoking status N (%) 230 Never 14 (6.1) Ex 146 (63.5) Current 70 (30.4) Clinical measures mean (SD) Heart rate (bpm) 98.7 (20.8) 234 Respiratory rate 23.6 (5.9) 234 Systolic blood pressure (mm Hg) 130.7 (23.7) 234 Diastolic blood pressure (mm Hg) 71.3 (13.1) 234 BMI, body mass index.

MI (kg/m2) mean (SD) 26.2 (7.9) 223 Smoking status N (%) 230 Never 14 (6.1) Ex 146 (63.5) Current 70 (30.4) Clinical measures mean (SD) Heart rate (bpm) 98.7 (20.8) 234 Respiratory rate 23.6 (5.9) 234 Systolic blood pressure (mm Hg) 130.7 (23.7) 234 Diastolic blood pressure (mm Hg) 71.3 (13.1) 234 BMI, body mass index. There was good agreement between arterial and venous pH, and between arterial and venous (figures 1 and 2 and tables 2 and 3), however, the CO2 values varied significantly (figure 3). There was also a good agreement between SaO2 and SpO2 in those patients with an SpO2 of ≥80% (figure 4 and tables 2 and 3). These relationships were not significantly different in the 20 patients with an admission systolic blood pressure of <100 mm Hg (see online supplementary material). Table 2 Agreement between arterial and venous pCO2, pH and HCO3− ABG (mean) (SD) VBG (mean) (SD) Mean difference (ABG–VBG) (95% CI) 95% limits of agreement N pH 7.40 (0.09) 7.37 (0.08) 0.03 (0.02 to 0.04) −0.05 to 0.11 234 (mEq/L) 29.7 (6.3) 29.7 (6.4) −0.04 (−0.22 to 0.15) −2.90 to 2.82 232 pCO2 (kPa) 6.89 (2.40) 7.63 (2.41) −0.75 (CI −0.89 to −0.61) −2.91 to 1.41 225 ABG, arterial blood gas; VBG, venous blood gas. Table 3 Agreement between SaO2 and SpO2 SaO2 (mean) (SD) SpO2 (mean) (SD) Mean difference (SaO2–SpO2) (95% CI) 95% limits of agreement N Oxygen percentage saturation* 91.2 (6.0) 91.0 (4.0) −0.17 (CI −0.89 to 0.56) −11.12 to 10.78 224 *In patients with SpO2 ≥80%. Figure 1 Bland–Altman plot for arterial and venous blood pH levels.

Table 3 Agreement between SaO2 and SpO2 SaO2 (mean) (SD) SpO2 (mean) (SD) Mean difference (SaO2–SpO2) (95% CI) 95% limits of agreement N Oxygen percentage saturation* 91.2 (6.0) 91.0 (4.0) −0.17 (CI −0.89 to 0.56) −11.12 to 10.78 224 *In patients with SpO2 ≥80%. Figure 1 Bland–Altman plot for arterial and venous blood pH levels. Figure 2 Bland–Altman plot for arterial and venous blood bicarbonate levels. Figure 3 Bland–Altman plot for arterial and venous blood CO2 levels. Figure 4 Bland–Altman plot for SaO2 and SpO2 levels. (This graph only includes patients with a pulse oximetry value of >80%.) Venous blood cut points for managing acute exacerbations of COPD Given the relationships observed, we calculated the sensitivity and specificity of a VBG pH and to correctly identify an arterial pH of ≥7.35, and an arterial of ≥21, as well as an SpO2 to identify an SaO2 of ≥92% (table 4). A venous pH of 7.34, a venous of 21.45 and an SpO2 of 91.5 would have correctly classified 87% (95% CI 82% to 91%), 97% (95% CI 93% to 98%) and 71% (95% CI 65% to 77%) of patients, respectively. In terms of specificity, 96% of patients with an ABG pH of ≥7.35 also had a VBG pH of ≥7.35 (tables 5 and 6). Table 4 Pain score and number of venesection attempts n N Pain score (median and IQR) Arterial pain score 4 (2–5) 187 Venous pain score 1 (0–2) 205 Number of attempts (N and %) Arterial number of attempts 1 162 (69.2) 2 55 (23.5) 3 10 (4.3) ≥4 7 (3.0) 234 Venous number of attempts 1 211 (90.2) 2 18 (7.7) ≥3 5 (2.1) 234 Table 5 Predictive performance of venous blood gas parameters

score (median and IQR) Arterial pain score 4 (2–5) 187 Venous pain score 1 (0–2) 205 Number of attempts (N and %) Arterial number of attempts 1 162 (69.2) 2 55 (23.5) 3 10 (4.3) ≥4 7 (3.0) 234 Venous number of attempts 1 211 (90.2) 2 18 (7.7) ≥3 5 (2.1) 234 Table 5 Predictive performance of venous blood gas parameters Venous blood cut-off AUC Sensitivity % Specificity % Correctly classified %* N Arterial pH ≥7.35 7.34 0.92 88.9 95.6 87 234 (mEq/L) ≥21 21.45 0.98 96 100 97 232 *Correctly classified refers to percentage of patients correctly classified both above and below given parameter. AUC; area under curve. Table 6 Predictive performance of pulse oximetry SpO2 cut-off AUC Sensitivity % Specificity % Correctly classified* N SaO2 91.5% 0.75 78 72 71 233 *Correctly classified refers to percentage of patients correctly classified both above and below given parameter. We used current oxygen guidelines15 to calculate how many ABG samples may have been avoided if the venous cut points were used instead. Of our 234 patients, 72 (31%) had a venous pH <7.35. Of the 162 with a venous pH ≥7.35, only two had an arterial pH of <7.35. Consequently, we estimate approximately two-thirds of ABGs can safely be avoided in the initial assessment of COPD exacerbations. This figure does not factor in the repeat attempts needed to obtain arterial blood. Using these data, we suggest a new algorithm for the management of COPD exacerbations based on the current guidelines (figure 5).

te approximately two-thirds of ABGs can safely be avoided in the initial assessment of COPD exacerbations. This figure does not factor in the repeat attempts needed to obtain arterial blood. Using these data, we suggest a new algorithm for the management of COPD exacerbations based on the current guidelines (figure 5). Figure 5 Suggested algorithm for blood gas analysis during COPD exacerbation (adapted from Kelly10). ABG, arterial blood gas; ICU, intensive care unit; NIV, non-invasive ventilation. Timing of samples We assessed the mean difference in time between the paired arterial and venous samples to see if a delay between samples had an effect on the relationship between ABG and VBG parameters. The mean time difference was −4.18 min, SD 16.92, range −58.98 to 78.64 min (arterial–venous). Given the range, we repeated our analysis using the 168 paired samples that were performed within 15 min of each other. This did not affect the relationships (see online supplementary table). Pain score The median pain score was significantly higher for ABG sampling as compared with VBG (p<0.001). In addition, there was a significantly greater number of attempts taken to obtain an ABG sample (69.2% achieved at first attempt) compared with VBG, where 90.2% were obtained at the first attempt (p<0.001) (table 4).

Pain score The median pain score was significantly higher for ABG sampling as compared with VBG (p<0.001). In addition, there was a significantly greater number of attempts taken to obtain an ABG sample (69.2% achieved at first attempt) compared with VBG, where 90.2% were obtained at the first attempt (p<0.001) (table 4). Discussion Exacerbations of COPD are a major cause of morbidity and mortality worldwide,16 and our local figures reflect this. In 2010, there were 1343 admissions into Nottingham University Hospitals Trust. The management of COPD exacerbations depends upon quickly identifying acute hypercapnic respiratory failure. In this study, we set out to establish if ABG analysis obtained for the initial assessment of COPD exacerbations could be replaced by VBG analysis and pulse oximetry when assessing for acute hypercapnic respiratory failure. We examined the agreement between ABG and VBG parameters and between ABG and pulse oximetry measurements of oxygen saturation in COPD exacerbations and found acceptable agreement for pH, and for SaO2 at an SpO2 >80%. We found that 96% of patients with an ABG pH of <7.35 also had a VBG pH of <7.35 and that only two patients were misclassified as having a normal venous pH but a low arterial pH.

d pulse oximetry measurements of oxygen saturation in COPD exacerbations and found acceptable agreement for pH, and for SaO2 at an SpO2 >80%. We found that 96% of patients with an ABG pH of <7.35 also had a VBG pH of <7.35 and that only two patients were misclassified as having a normal venous pH but a low arterial pH. A meta-analysis of five studies examining the utility of peripheral VBG analyses in exacerbations of COPD in the emergency department found that there was agreement between arterial and venous pH and .9 The weighted average difference for pCO2 was 0.79 kPa (n=440), whereas those for pH and were 0.028 and 1.34 mmol (n=239), respectively. The relatively weaker relationship seen between SpO2 and SaO2 at lower levels is unsurprising as commercial pulse oximeters are more accurate at higher oxygen saturations and significantly less accurate below 80%.17 18 The proposed algorithm reflects this; any patient with an SpO2 of <80% needs an immediate ABG. Our study has limitations. There was a small time gap between sample acquisition and processing, although results did not change significantly when this was factored in. We also had a pragmatic approach to sample collection which depended upon our junior doctors and specialist nurses; consequently, it was difficult to fully exclude mixed arterial/venous stabs which may explain why the sensitivities and specificities to predict an arterial pH of <7.35 were not 100%. We stress in our algorithm that if there is a risk or actual clinical deterioration an arterial analysis should be performed.

st nurses; consequently, it was difficult to fully exclude mixed arterial/venous stabs which may explain why the sensitivities and specificities to predict an arterial pH of <7.35 were not 100%. We stress in our algorithm that if there is a risk or actual clinical deterioration an arterial analysis should be performed. Arterial sampling was more painful than venous and required more attempts. While the pain of arterial sampling can be reduced by using local anaesthetic, it is not widely used.7 Although capillary sampling is used in speciality wards, widespread adoption is difficult because of the extra training, resources and time needed. As patients with exacerbations of COPD almost always have venipuncture to obtain samples for full blood count and blood chemistry analysis, VBG analysis can be performed on the same sample. Our results suggest that the close relationship between venous and arterial acid base parameters, and between oxygen saturations obtained from pulse oximetry and arterial blood, could allow the initial assessment of acute COPD exacerbations to be based on a combined measurement of a VBG pH and SpO2. This would mean a change in practice as the current oxygen guidelines published by the British Thoracic Society15 state that any patient requiring supplemental oxygen to achieve a target SpO2 of 92%–94% should have an arterial blood or arteriolised capillary blood gas performed.

mbined measurement of a VBG pH and SpO2. This would mean a change in practice as the current oxygen guidelines published by the British Thoracic Society15 state that any patient requiring supplemental oxygen to achieve a target SpO2 of 92%–94% should have an arterial blood or arteriolised capillary blood gas performed. We suggest that arterial sampling is reserved for patients with a venous pH of <7.35. The approach of using venous blood first has obvious benefits. Only one blood draw would be required resulting in less pain and a lower risk of bruising and associated side effects. Less training would be required to initially assess acid/base status, and fewer attempts to draw blood would be needed, needing less equipment and simplifying the care pathway for COPD exacerbations. We conservatively estimate that >66% of ABG attempts would be avoided and replaced by VBG sampling. Conclusion There is a good agreement between pH and values derived from venous and arterial blood, and between pulse oximetry and ABG oxygen saturations. These agreements could allow the initial assessment of COPD exacerbations to be based on VBG analysis and pulse oximetry rather than ABG analysis, simplifying the care pathway. Supplementary Material Web supplement Correction notice: This article has been corrected since it was published Online First. Data in the abstract and in Table 2 has been corrected. Figure 3 has been updated due to data corrections.

Conclusion There is a good agreement between pH and values derived from venous and arterial blood, and between pulse oximetry and ABG oxygen saturations. These agreements could allow the initial assessment of COPD exacerbations to be based on VBG analysis and pulse oximetry rather than ABG analysis, simplifying the care pathway. Supplementary Material Web supplement Correction notice: This article has been corrected since it was published Online First. Data in the abstract and in Table 2 has been corrected. Figure 3 has been updated due to data corrections. The authors would like to acknowledge the expertise and help of Rachel King and Adeline Sheehan for recruiting the patients running the study. They would like to thank all the ward staff and junior doctors working in the Emergency Department, Emergency Admissions Unit, Medical High Dependency Unit and Respiratory Wards at Queens Medical Centre and Nottingham City Hospital for their help with the study, and the hospital at night nurse coordinator team. The authors also wish to acknowledge Dr Dale Gardiner for his expert blood gas input. Twitter: Follow Glenn Hearson at @NottinghamRRU

The authors would like to acknowledge the expertise and help of Rachel King and Adeline Sheehan for recruiting the patients running the study. They would like to thank all the ward staff and junior doctors working in the Emergency Department, Emergency Admissions Unit, Medical High Dependency Unit and Respiratory Wards at Queens Medical Centre and Nottingham City Hospital for their help with the study, and the hospital at night nurse coordinator team. The authors also wish to acknowledge Dr Dale Gardiner for his expert blood gas input. Twitter: Follow Glenn Hearson at @NottinghamRRU Contributors: TMM performed the bulk of the data analyses and commented on the manuscript. GH helped with data acquisition and cleaning and commented on the manuscript. GH helped with data acquisition and follow-up data analyses and commented on the manuscript. CR and WK helped with data acquisition and commented on the manuscript. TWH helped with study design, data interpretation and commented on the manuscript. A-MK helped with study design, data interpretation and commented on the manuscript. DES had the original idea for the study, designed the study and obtained funding. He wrote the first draft. Funding: This study was funded by the National Institute for Health Research's (NIHR) research for patient benefit programme, grant number PB-PG-0211-24049. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health. Competing interests: None declared. Ethics approval: Nottingham Research and Ethics Committee East Midlands Nottingham 1 12/EM/0323.

Funding: This study was funded by the National Institute for Health Research's (NIHR) research for patient benefit programme, grant number PB-PG-0211-24049. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health. Competing interests: None declared. Ethics approval: Nottingham Research and Ethics Committee East Midlands Nottingham 1 12/EM/0323. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: Data on mortality are available; these were not included at the request of reviewer 3 but will be presented in abstract form.

Key messages What is the key question? What is the impact of very long-term (>30 years) air pollution exposure on mortality? What is the bottom line? Historic air pollution exposure has long-term effects on mortality that persist over 30 years after exposure and these potentially also influence current estimates of associations between air pollution and mortality. Why read on? This is one of the longest running studies to look at health effects of air pollution, using air pollution estimates independently assessed at multiple time points using contemporaneous monitoring data in a large cohort followed for 38 years. Introduction While the impact of air pollution on mortality in the short term (days) and medium term (<10 years) is now well established, there are relatively few studies assessing the long-term (>10 years) impact of air pollution1–10 with even fewer assessing the very long term (25+ years).2 3 8–10 Only a small number of these4–6 had exposure data at more than one time point.

in the short term (days) and medium term (<10 years) is now well established, there are relatively few studies assessing the long-term (>10 years) impact of air pollution1–10 with even fewer assessing the very long term (25+ years).2 3 8–10 Only a small number of these4–6 had exposure data at more than one time point. Like many other developed countries the UK experienced high levels of air pollution in the past, including the infamous London smog episode of December 1952,11 since when air pollution levels have fallen to much lower levels. Changes in air pollution concentrations in the UK are well documented as, uniquely, the UK had a comprehensive national air quality monitoring network running from the 1950s to the 1990s measuring black smoke (BS) and sulfur dioxide (SO2) arising from domestic and industrial coal and fossil fuel combustion, then major sources of emissions. Thereafter, networks switched to monitor nitrogen dioxide (NO2) (from the early 1990s) and particulate matter with a diameter of 10µm or less (PM10) (from the mid-1990s), as transport emissions became the largest source of air pollution.12 13 The present study uses a very large nationally representative British cohort to consider impact of air pollution over 38 years of follow-up. Three a priori hypotheses were investigated: Historic air pollution (ie, of several decades previously) is associated with later mortality risk. The mortality risks associated with a given exposure decrease over subsequent decades. Air pollution exposures in previous decades interact with recent exposures to affect mortality risk.

Like many other developed countries the UK experienced high levels of air pollution in the past, including the infamous London smog episode of December 1952,11 since when air pollution levels have fallen to much lower levels. Changes in air pollution concentrations in the UK are well documented as, uniquely, the UK had a comprehensive national air quality monitoring network running from the 1950s to the 1990s measuring black smoke (BS) and sulfur dioxide (SO2) arising from domestic and industrial coal and fossil fuel combustion, then major sources of emissions. Thereafter, networks switched to monitor nitrogen dioxide (NO2) (from the early 1990s) and particulate matter with a diameter of 10µm or less (PM10) (from the mid-1990s), as transport emissions became the largest source of air pollution.12 13 The present study uses a very large nationally representative British cohort to consider impact of air pollution over 38 years of follow-up. Three a priori hypotheses were investigated: Historic air pollution (ie, of several decades previously) is associated with later mortality risk. The mortality risks associated with a given exposure decrease over subsequent decades. Air pollution exposures in previous decades interact with recent exposures to affect mortality risk. Methods This investigation used a long-running census-based study, the Office for National Statistics (ONS) Longitudinal Study, which contains linked census and life events data on a representative 1% sample of the population of England and Wales. The initial sample was drawn from the 1971 census.14 15

Air pollution exposures in previous decades interact with recent exposures to affect mortality risk. Methods This investigation used a long-running census-based study, the Office for National Statistics (ONS) Longitudinal Study, which contains linked census and life events data on a representative 1% sample of the population of England and Wales. The initial sample was drawn from the 1971 census.14 15 For this investigation, the study was restricted to members of the cohort of all ages present at the 1971 census, who were either present at each subsequent census (1981, 1991 and 2001) and either traced up to 2009 or had died, who were not identified through general practice (GP) registration as having left the country. Exclusions (figure 1) were made for data inaccuracies, those who died in 1971 and those not UK born (who may have had different previous air pollution exposures). By constructing a closed cohort, we were able to estimate air pollution exposures across the entire period of their life 1971–2009 for each individual. Figure 1 Identification of longitudinal survey (LS) participants. Source: Office for National Statistics Longitudinal Study (authors’ own work).

For this investigation, the study was restricted to members of the cohort of all ages present at the 1971 census, who were either present at each subsequent census (1981, 1991 and 2001) and either traced up to 2009 or had died, who were not identified through general practice (GP) registration as having left the country. Exclusions (figure 1) were made for data inaccuracies, those who died in 1971 and those not UK born (who may have had different previous air pollution exposures). By constructing a closed cohort, we were able to estimate air pollution exposures across the entire period of their life 1971–2009 for each individual. Figure 1 Identification of longitudinal survey (LS) participants. Source: Office for National Statistics Longitudinal Study (authors’ own work). Air pollution exposures in 1971, 1981, 1991 and 2001 Land use regression techniques were used to model BS and SO2 annual concentrations in 1971, 1981 and 1991 at 1 km grids. Models for BS and SO2 have been described in detail previously12 but were developed with a range of variables including information on land cover, major and minor roads, and X–Y coordinates of each monitoring site. Models were developed against concentration data from national monitoring station sites where operational days in the year exceeded 75%, which involved a total of 966 sites for BS and 825 sites for SO2. Model building used 80% of network sites; the remaining independent, randomly stratified 20% sample was retained for model validation. The validation statistics from the independent subset gave r2 values for BS of 0.41, 0.38 and 0.34 for 1971, 1981 and 1991, respectively, and for SO2 of 0.57, 0.26 and 0.31. Values of mean (fractional) bias were low in all years (ie, <−0.1), which suggests that predicted values were within 20% of observed monitored concentrations. Land use regression techniques were used to model PM10 at 100 m grids in 2001.16 Leave-one-out validation statistics for PM10 in 2001 gave r2 value of 0.37.

.31. Values of mean (fractional) bias were low in all years (ie, <−0.1), which suggests that predicted values were within 20% of observed monitored concentrations. Land use regression techniques were used to model PM10 at 100 m grids in 2001.16 Leave-one-out validation statistics for PM10 in 2001 gave r2 value of 0.37. Integration of air pollution and confounder data into the longitudinal survey Air pollution exposure estimates were produced for UK grids and wards (ONS small area geographical units) and Centre for Longitudinal Study Information & User Support (CeLSIUS) staff matched these to individuals; precise geolocation of longitudinal survey (LS) participants is not made available to researchers. For 1971, individuals were assigned the annual average BS and SO2 concentration of the 1 km grid in which their residence was located. For other years (1981, 1991 and 2001) the pollutant surfaces (ie, regular grids) were intersected with ward boundaries and area-weighting was used to calculate the average values of each pollutant within each ward. Information on smoking was not available at individual level in the cohort, so smoothed district-level lung cancer mortality relative risk 2002–2009 (International Classification of Diseases (ICD)-10 codes C33-C34) as a proxy measure for cumulative smoking over the past 20+ years17 were used. Lung cancer risks using Bayesian smoothing methods were calculated from ONS data held by the Small Area Health Statistics Unit.

l lung cancer mortality relative risk 2002–2009 (International Classification of Diseases (ICD)-10 codes C33-C34) as a proxy measure for cumulative smoking over the past 20+ years17 were used. Lung cancer risks using Bayesian smoothing methods were calculated from ONS data held by the Small Area Health Statistics Unit. Statistical analysis Statistical analyses of individual-level data were conducted in person at the ONS offices using Stata V.11 (Stata, College Station, Texas, USA). Descriptive analyses were conducted for all variables. Logistic regression analyses (died/survived) were used to investigate associations between BS and/or SO2 exposure in 1971, 1981 and 1991 and PM10 in 2001 and risk of death in 1972–1981, 1982–1991, 1992–2001 and 2002–2009, respectively. Mortality outcomes were all-cause excluding accidents; cardiovascular (CV) and respiratory mortality and by constituent subgroups of coronary heart disease (CHD), stroke, respiratory infections, COPD and lung cancer. For ICD codes used, see the online supplementary appendix A table A1. We conducted analyses by decade to align with census years, the periodicity of which reflects marked changes in air pollution sources in the UK, from predominantly fossil fuel burning in the 1970s to domination by traffic-based sources with increasing contribution from diesel engines by 2001.

ntary appendix A table A1. We conducted analyses by decade to align with census years, the periodicity of which reflects marked changes in air pollution sources in the UK, from predominantly fossil fuel burning in the 1970s to domination by traffic-based sources with increasing contribution from diesel engines by 2001. Adjustments were made: (i) for age and sex; (ii) additionally for social class of individual (Registrar General occupation) and area (quintiles of Carstairs deprivation index), population density (not used in any of the exposure models) and geographical region (in 1971). Sensitivity analyses were conducted adjusting for the smoking proxy (lung cancer risks), restricting analyses to non-movers in the 5 years prior to the 1971 census and adjusting for exposures in other years. All air pollution variables, population density and age variables were centred prior to regression analyses. To evaluate whether past BS or SO2 exposure modified the effect of PM10 on mortality in 2002–2009 we introduced interaction terms in statistical models between tertiles of BS/SO2 and PM10 and examined risks by exposure tertiles. As SO2 was highly correlated with the same-year BS, we did not conduct two pollutant analyses. Finally, through a piecewise linear model for the tertiles of BS/SO2 at 1971 and the three main outcomes (all-cause mortality, all respiratory mortality and all CV mortality) we visually assessed the presence of a concentration–response.

ighly correlated with the same-year BS, we did not conduct two pollutant analyses. Finally, through a piecewise linear model for the tertiles of BS/SO2 at 1971 and the three main outcomes (all-cause mortality, all respiratory mortality and all CV mortality) we visually assessed the presence of a concentration–response. Results Descriptive analyses The analyses included 367 658 individuals followed from 1971 to 2009 with non-missing data (figure 1), comprising 71.67% of the initial cohort. The main reasons for exclusion were emigration (n=77 265) and missing or incomplete data (n=25 992). Those excluded from the analysis were significantly (p<0.001) younger in 1971 than those included (mean age 25 vs 38 years), more likely to be male (54% vs 48%), more likely to have lived in most deprived areas in 1971 (23% vs 20%) and to have moved between 1966 and 1971 (59% vs 39%) (see online supplementary appendix A table A2). Median air pollution exposures to BS and SO2 were twofold to threefold higher in 1971 compared with 1991. Ranges (10th–90th centiles) for BS were 18.5–70.5 μg/m3 in 1971 and 3–19 μg/m3 in 1991 (table 1). Table 1 Descriptive analyses of the Longitudinal Study

Results Descriptive analyses The analyses included 367 658 individuals followed from 1971 to 2009 with non-missing data (figure 1), comprising 71.67% of the initial cohort. The main reasons for exclusion were emigration (n=77 265) and missing or incomplete data (n=25 992). Those excluded from the analysis were significantly (p<0.001) younger in 1971 than those included (mean age 25 vs 38 years), more likely to be male (54% vs 48%), more likely to have lived in most deprived areas in 1971 (23% vs 20%) and to have moved between 1966 and 1971 (59% vs 39%) (see online supplementary appendix A table A2). Median air pollution exposures to BS and SO2 were twofold to threefold higher in 1971 compared with 1991. Ranges (10th–90th centiles) for BS were 18.5–70.5 μg/m3 in 1971 and 3–19 μg/m3 in 1991 (table 1). Table 1 Descriptive analyses of the Longitudinal Study Mean (SD) Median 10th centile 90th centile Air pollution exposure (μg/m3) BS 1971 42.7 (20.4) 41 18.5 70.5 SO2 1971 85.2 (36.8) 77 44.5 137 BS 1981 16.2 (5.2) 16 8.5 25 SO2 1981 43.1 (12.1) 41.5 25.5 66 BS 1991 11.8 (4.7) 12 3 19 SO2 1991 29.6 (6.5) 29.5 19 40.5 PM10 2001* 20.7 (2.5) 20 18 24 Age in years Age in 1971 38 (22.9) 30 7 68 Age in 1981 43.3 (20.8) 42 13 77 Age in 1991 48.9 (18.6) 47 22 80 Age in 2001 54.3 (16) 53 32 82 Population density (per km) 1971 3486.2 (3075.1) 2915.6 206.4 7187.6 Number of deaths by year 1972–1981 1982–1991 1992–2001 2002–2009 All-cause excluding accidents 48, 834 47 775 45 736 31 744 Cardiovascular (CV) 26 140 23 923 20 054 11 876 Cardiovascular mortality: CHD 14 050 13 461 6,219 5613 Cardiovascular mortality: Stroke 6401 6189 4803 810 Respiratory 6959 5300 7302 4598 Respiratory mortality: Respiratory Infections 4219 2364 4272 2109 Respiratory mortality: COPD 2259 2413 2319 1644 Lung cancer 3185 3154 2731 1920 Gender Number Per cent Males 175 532 47.7 Females 192 126 52.3 Non-movers (1966–1971) 223 055 60.7 Region 1971 Northern 26 243 7.1 Yorkshire/Humber 39 853 10.8 North West 52 114 14.2 East Midlands 29 582 8.1 West Midlands 34 535 9.4 East Anglia 13 602 3.7 South East 67 538 18.4 South West 31 726 8.6 Wales 22 155 6.0 Inner London 17 863 4.9 Outer London 32 447 8.8 Individual-level social class 1971 (Registrar General occupational categories) I Professional 11 381 3.1 II Intermediate 49 427 13.5 III(N) Skilled non-manual 50 789 13.8 III(M) Skilled manual 90 082 24.5 IV Semi-skilled manual 53 415 14.5 V Unskilled manual 20 220 5.5 Armed forces and inadequately described 25 860 7.0 Other economically inactive/student/childcare/retired/permanently sick 66 260 18.0 Area-level deprivation 1971 (Carstairs quintile) 1 (least deprived) 69 608 19.0 2 74 140 20.2 3 75 428 20.6 4 75 511 20.6 5 (most deprived) 72 202 19.7 *Given the finer 100 m resolution for PM10, only integer values of PM10 were permitted by ONS for analysis because of concerns that more decimal places might allow identifiability in the linked dataset. Source: ONS Longitudinal Study (authors’ own work).

40 20.2 3 75 428 20.6 4 75 511 20.6 5 (most deprived) 72 202 19.7 *Given the finer 100 m resolution for PM10, only integer values of PM10 were permitted by ONS for analysis because of concerns that more decimal places might allow identifiability in the linked dataset. Source: ONS Longitudinal Study (authors’ own work). BS, black smoke; CHD, coronary heart disease; ONS, Office for National Statistics. The mean PM10 exposure in 2001 was 20.7 μg/m3 (10–90th centile 18–24). The highest exposures were seen in urban metropolitan areas: BS was highest in the northern regions of England and Wales, and highest SO2 and PM10 exposures were seen in London (see online supplementary appendix A table A3). All exposures decreased with increasing individual-level social class and increased with increased deprivation of area of residence (see online supplementary appendix A table A3). PM10 exposure in 2001 was weakly correlated with BS and SO2 in earlier years (all r <0.45) (table 2). Table 2 Correlations between participant air pollution estimates and potential confounders

The mean PM10 exposure in 2001 was 20.7 μg/m3 (10–90th centile 18–24). The highest exposures were seen in urban metropolitan areas: BS was highest in the northern regions of England and Wales, and highest SO2 and PM10 exposures were seen in London (see online supplementary appendix A table A3). All exposures decreased with increasing individual-level social class and increased with increased deprivation of area of residence (see online supplementary appendix A table A3). PM10 exposure in 2001 was weakly correlated with BS and SO2 in earlier years (all r <0.45) (table 2). Table 2 Correlations between participant air pollution estimates and potential confounders BS 1971 SO2 1971 BS 1981 SO2 1981 BS 1991 SO2 1991 PM10 2001 Age Population density RR lung cancer BS 1971 n=367 658 1.000 SO2 1971 n=367 658 0.730 1.000 BS 1981 n=305 471 0.696 0.531 1.000 SO2 1981 n=305 471 0.519 0.720 0.762 1.000 BS 1991 n=259 649 0.651 0.405 0.769 0.538 1.000 SO2 1991 n=259 649 0.645 0.452 0.750 0.600 0.866 1.000 PM10 2001 n=221 148 0.195 0.411 0.200 0.413 0.077 0.190 1.000 Age n=367 658 −0.002 0.012 −0.020 −0.005 −0.035 −0.034 −0.014 1.000 Population density n=367 658 0.441 0.689 0.252 0.438 0.044 0.107 0.380 0.031 1.000 RR lung cancer n=221 148 0.059 −0.162 0.010 −0.202 0.007 0.047 −0.191 0.004 −0.150 1.000 Population Density—population density 1 km grid. Source: Office for National Statistics Longitudinal Study (authors’ own work). BS, black smoke.

BS 1971 SO2 1971 BS 1981 SO2 1981 BS 1991 SO2 1991 PM10 2001 Age Population density RR lung cancer BS 1971 n=367 658 1.000 SO2 1971 n=367 658 0.730 1.000 BS 1981 n=305 471 0.696 0.531 1.000 SO2 1981 n=305 471 0.519 0.720 0.762 1.000 BS 1991 n=259 649 0.651 0.405 0.769 0.538 1.000 SO2 1991 n=259 649 0.645 0.452 0.750 0.600 0.866 1.000 PM10 2001 n=221 148 0.195 0.411 0.200 0.413 0.077 0.190 1.000 Age n=367 658 −0.002 0.012 −0.020 −0.005 −0.035 −0.034 −0.014 1.000 Population density n=367 658 0.441 0.689 0.252 0.438 0.044 0.107 0.380 0.031 1.000 RR lung cancer n=221 148 0.059 −0.162 0.010 −0.202 0.007 0.047 −0.191 0.004 −0.150 1.000 Population Density—population density 1 km grid. Source: Office for National Statistics Longitudinal Study (authors’ own work). BS, black smoke. Within-year BS and SO2 exposures were highly correlated (r>0.7). Correlations were also moderate to high (r∼0.6–0.7) for BS exposures between years, but there was a greater range for SO2 (r∼0.45–0.7). BS exposures in 1971–1991 There were statistically significant associations between BS exposure in 1971 and all-cause mortality in all subsequent decades through to 2002–2009 (figure 2 and table 3); CV and respiratory mortality showed similar patterns as all-cause mortality. Table 3 Logistic regression ORs (95% CI) per 10 μg/m3 for black smoke exposure in 1971 and mortality in subsequent decades

BS exposures in 1971–1991 There were statistically significant associations between BS exposure in 1971 and all-cause mortality in all subsequent decades through to 2002–2009 (figure 2 and table 3); CV and respiratory mortality showed similar patterns as all-cause mortality. Table 3 Logistic regression ORs (95% CI) per 10 μg/m3 for black smoke exposure in 1971 and mortality in subsequent decades Decade of outcome (i) Unadjusted (age and sex only) (ii) Adjusted (age, sex, social class, area-level deprivation, region, population density) All-cause mortality excluding accidents 1972–2009 1.07 (1.07 to 1.08) 1.03 (1.02 to 1.05) 1972–1981 1.10 (1.08 to 1.11) 1.05 (1.02 to 1.08) 1982–1991 1.09 (1.08 to 1.10) 1.03 (1.01 to 1.06) 1992–2001 1.07 (1.07 to 1.08) 1.04 (1.02 to 1.05) 2002–2009 1.05 (1.05 to 1.06) 1.02 (1.01 to 1.04) Cardiovascular mortality 1972–2009 1.08 (1.07 to 1.09) 1.03 (1.01 to 1.05) 1972–1981 1.11 (1.09 to 1.13) 1.03 (0.99 to 1.08) 1982–1991 1.10 (1.09 to 1.12) 1.04 (1.01 to 1.07) 1992–2001 1.08 (1.07 to 1.09) 1.04 (1.01 to 1.06) 2002–2009 1.05 (1.04 to 1.06) 1.01 (0.98 to 1.04) Respiratory mortality 1972–2009 1.11 (1.10 to 1.12) 1.07 (1.04 to 1.10) 1972–1981 1.12 (1.09 to 1.16) 1.10 (1.02 to 1.18) 1982–1991 1.14 (1.12 to 1.17) 1.05 (0.99 to 1.12) 1992–2001 1.10 (1.08 to 1.12) 1.08 (1.04 to 1.13) 2002–2009 1.09 (1.07 to 1.11) 1.05 (1.01 to 1.09) Source: Office for National Statistics Longitudinal Study (authors’ own work).

9 1.11 (1.10 to 1.12) 1.07 (1.04 to 1.10) 1972–1981 1.12 (1.09 to 1.16) 1.10 (1.02 to 1.18) 1982–1991 1.14 (1.12 to 1.17) 1.05 (0.99 to 1.12) 1992–2001 1.10 (1.08 to 1.12) 1.08 (1.04 to 1.13) 2002–2009 1.09 (1.07 to 1.11) 1.05 (1.01 to 1.09) Source: Office for National Statistics Longitudinal Study (authors’ own work). Figure 2 ORs (95% CI) per 10 μg/m3 for black smoke (BS) exposure in 1971, 1981 and 1991 and mortality in subsequent decades. Adjusted for age and sex, social class of individual and area, population density and geographical region. Source: Office for National Statistics Longitudinal Study (authors’ own work). BS exposures in 1981 and 1991 were also significantly associated with all-cause, CV and respiratory mortality in subsequent decades (see figure 2 and online supplementary appendix B table B1). Figure 3 shows stronger effects for more recent BS exposures. Largest associations were between BS exposure in 1991 and respiratory mortality in 1991–2001 with OR 1.38 (95% CI 1.22 to 1.57) (see online supplementary appendix B table B1). In the subgroup analyses (see online supplementary appendix B table B1), risks were marginally higher for CHD than stroke mortality and higher for COPD and lung cancer than for respiratory infections especially for more recent exposures. The highest risks observed were for COPD mortality. COPD mortality remained significantly associated with past exposures up to the most recent decade, while respiratory infections generally reduced in magnitude and became non-significant over time.

r than for respiratory infections especially for more recent exposures. The highest risks observed were for COPD mortality. COPD mortality remained significantly associated with past exposures up to the most recent decade, while respiratory infections generally reduced in magnitude and became non-significant over time. Figure 3 ORs (95% CI) per 10 μg/m3 for BS exposure in 1971, 1981 and 1991 and PM10 in 2001 and mortality in each subsequent decades. Adjusted for age and sex, social class of individual and area, population density and geographical region. Source: ONS Longitudinal Study (authors’ own work). CV, BS, black smoke; cardiovascular disease; ONS, Office for National Statistics. Adjustment for confounders slightly reduced ORs for exposures in 1971 (table 3) and 1991 (see online supplementary appendix B table B1) but showed a larger effect for BS in 1981 (see online supplementary appendix B table B1). Sensitivity analyses restricted to the non-movers made no difference in ORs. The association with respiratory mortality in 2002–2009 no longer reached statistical significance after additional adjustment for the smoking proxy in 2002–2009 but it made no difference to associations with all-cause and CV mortality (see online supplementary appendix B table B2).

on-movers made no difference in ORs. The association with respiratory mortality in 2002–2009 no longer reached statistical significance after additional adjustment for the smoking proxy in 2002–2009 but it made no difference to associations with all-cause and CV mortality (see online supplementary appendix B table B2). Concentration–response for each tertile of BS exposure in 1971 for the three main outcomes (figure 4) showed a steeper response in the highest tertiles that was most marked for respiratory mortality. By 1981, the 90th centile value of BS exposure (25 μg/m3) was within the lowest tertile range of BS exposure for 1971 (1971 tertile cut points 31 and 50.5 μg/m3). Figure 4 Concentration-response for tertiles of BS exposure in 1971 and subsequent mortality 1971-2009 for all-cause, CV and respiratory mortality. Adjusted for age and sex, social class of individual and area, population density and geographical region. Source: ONS Longitudinal Study (authors' own work). BS, black smoke; CV, cardiovascular disease. SO2 exposures in 1971–1991 Results for SO2 showed very similar patterns to those for BS (see online supplementary appendix C) including for concentration–response. As for BS, there were statistically significant associations between SO2 exposure in 1971 and mortality in all subsequent decades through to 2002–2009 (see online supplementary appendix C figure C1 and table C1). The largest association, as for BS, was between SO2 exposure in 1991 and respiratory mortality in 1991–2001 with OR 1.24 (95% CI 1.15 to 1.34) (see online supplementary appendix C table C1).

n 1971 and mortality in all subsequent decades through to 2002–2009 (see online supplementary appendix C figure C1 and table C1). The largest association, as for BS, was between SO2 exposure in 1991 and respiratory mortality in 1991–2001 with OR 1.24 (95% CI 1.15 to 1.34) (see online supplementary appendix C table C1). PM10 exposure in 2001 PM10 exposure in 2001 was associated with an increased risk of all-cause mortality in 2002–2009 after adjustment (OR 1.24 (95% CI 1.15 to 1.33)) (table 4), with higher ORs for respiratory and lower for CV mortality. Within the subgroup analyses, the highest risk observed was for lung cancer with OR 1.60 (1.29 to 1.99) (see online supplementary appendix B table B4). Table 4 Logistic regression ORs (95% CI) per 10 μg/m3 for PM10 exposure in 2001 and mortality in 2002–2009 adjusted for past black smoke (BS) exposure

PM10 exposure in 2001 PM10 exposure in 2001 was associated with an increased risk of all-cause mortality in 2002–2009 after adjustment (OR 1.24 (95% CI 1.15 to 1.33)) (table 4), with higher ORs for respiratory and lower for CV mortality. Within the subgroup analyses, the highest risk observed was for lung cancer with OR 1.60 (1.29 to 1.99) (see online supplementary appendix B table B4). Table 4 Logistic regression ORs (95% CI) per 10 μg/m3 for PM10 exposure in 2001 and mortality in 2002–2009 adjusted for past black smoke (BS) exposure 2002–2009 (i) Unadjusted (age and sex only) (ii) Adjusted (age, sex, social class, area-level deprivation, region, population density) Sensitivity analysis: adjusted (age, sex, social class, area-level deprivation, region, population density and lung cancer) All-cause mortality excluding accidents PM10 2001 1.37 (1.29 to 1.45) 1.24 (1.16 to 1.34) 1.24 (1.15 to 1.33) PM10 2001+BS 1971 1.27 (1.19 to 1.35) 1.26 (1.17 to 1.37) 1.23 (1.14 to 1.32) PM10 2001+BS 1981 1.25 (1.17 to 1.33) 1.19 (1.11 to 1.28) 1.19 (1.10 to 1.28) PM10 2001+BS 1991 1.29 (1.21 to 1.38) 1.20 (1.11 to 1.29) 1.19 (1.11 to 1.28) PM10 2001+BS all years 1.18 (1.11 to 1.26) 1.21 (1.11 to 1.31) 1.16 (1.07 to 1.25) Cardiovascular mortality PM10 2001 1.25 (1.14 to 1.37) 1.12 (1.01 to 1.25) 1.12 (1.01 to 1.25) PM10 2001+BS 1971 1.16 (1.06 to 1.27) 1.12 (1.01 to 1.25) 1.12 (1.00 to 1.24) PM10 2001+BS 1981 1.15 (1.03 to 1.27) 1.09 (0.98 to 1.22) 1.09 (0.97 to 1.22) PM10 2001+BS 1991 1.18 (1.07 to 1.29) 1.08 (0.97 to 1.20) 1.07 (0.96 to 1.20) PM10 2001+BS all years 1.09 (0.98 to 1.20) 1.06 (0.95 to 1.19) 1.06 (0.94 to 1.19) Respiratory mortality PM10 2001 1.55 (1.35 to 1.77) 1.23 (1.05 to 1.45) 1.22 (1.04 to 1.44) PM10 2001+BS 1971 1.36 (1.19 to 1.57) 1.22 (1.04 to 1.43) 1.21 (1.03 to 1.42) PM10 2001+BS 1981 1.38 (1.20 to 1.59) 1.18 (1.00 to 1.39) 1.17 (0.99 to 1.38) PM10 2001+BS 1991 1.50 (1.31 to 1.72) 1.21 (1.02 to 1.42) 1.20 (1.02 to 1.41) PM10 2001+BS all years 1.33 (1.14 to 1.54) 1.18 (0.99 to 1.40) 1.17 (0.98 to 1.39) Source: Office for National Statistics Longitudinal Study (authors’ own work).

2) PM10 2001+BS 1981 1.38 (1.20 to 1.59) 1.18 (1.00 to 1.39) 1.17 (0.99 to 1.38) PM10 2001+BS 1991 1.50 (1.31 to 1.72) 1.21 (1.02 to 1.42) 1.20 (1.02 to 1.41) PM10 2001+BS all years 1.33 (1.14 to 1.54) 1.18 (0.99 to 1.40) 1.17 (0.98 to 1.39) Source: Office for National Statistics Longitudinal Study (authors’ own work). Additional adjustment for past exposures to BS in 1971, 1981 and 1991 reduced ORs for each outcome and the OR for CV mortality (CV) lost statistical significance. The impact of adjusting for past air pollution exposures was greater than that produced by adjusting for individual social class and area-level deprivation. Similar effects were seen whether past BS, SO2 or both were adjusted for, as expected from the high correlations between BS and SO2 (see online supplementary appendix B table B3). There were no clear patterns in interaction terms between PM10 in 2001 in relation to mortality in 2002–2009 and past exposures in terms of tertiles of BS or SO2 exposure in 1971, 1981 or 1991, or always highest or always lowest tertiles of exposures. The exception to this was for CV with respect to BS in 1981 (p value for highest tertile of exposure=0.014), suggestive of lower OR for PM10 and CV confined to the highest tertile of BS exposure in 1981, which is an apparently protective effect.

1, 1981 or 1991, or always highest or always lowest tertiles of exposures. The exception to this was for CV with respect to BS in 1981 (p value for highest tertile of exposure=0.014), suggestive of lower OR for PM10 and CV confined to the highest tertile of BS exposure in 1981, which is an apparently protective effect. Discussion This study investigated air pollution exposures in 370 000 individuals in a national census-based cohort followed for 38 years. In line with our prior hypotheses, we found that historic exposures to BS and SO2 were associated with increased risks of all-cause, CV and respiratory mortality in England and Wales over 30 years later, and mortality risks associated with a given exposure generally decreased over time. Subgroup analyses showed highest risks for COPD and lung cancer mortality. Adjusting for past BS or SO2 exposures resulted in slightly lower observed mortality associations with recent PM10 exposure (suggestive of confounding), but there was no clear evidence that higher air pollution exposures in earlier life resulted in greater or lesser susceptibility to PM10 (effect modification).

tality. Adjusting for past BS or SO2 exposures resulted in slightly lower observed mortality associations with recent PM10 exposure (suggestive of confounding), but there was no clear evidence that higher air pollution exposures in earlier life resulted in greater or lesser susceptibility to PM10 (effect modification). We saw highest associations with respiratory mortality, consistent with other UK-based studies investigating long-term BS exposures in the 1950s and 1960s,18 1970s,7 1980s and 1990s5 and PM10 in the 2000s19 and with a population-registry-based study in the Netherlands examining PM10 exposure in 2001.20 In contrast, studies of large American cohorts have found highest associations of particulates with CV mortality,4 6 as did a recent study in Rome,21 while the large European Study of Cohorts and Air Pollution Effects (ESCAPE) analyses found associations with all-cause9 but not non-malignant respiratory8 or CV mortality.10 Reasons for these differences are unclear, but may include differences in death certification practices between countries.22 23

study in Rome,21 while the large European Study of Cohorts and Air Pollution Effects (ESCAPE) analyses found associations with all-cause9 but not non-malignant respiratory8 or CV mortality.10 Reasons for these differences are unclear, but may include differences in death certification practices between countries.22 23 Our effect sizes for BS were very similar to those of recently reported previous British studies5 7 especially for more recent exposures, despite differing exposure estimation methods and study design. The only other UK study to investigate recent exposure to PM10 by Carey et al19 reported a lower effect size of HR 1.07 (0.99 to 1.16) for all-cause mortality in 2003–2007, compared with OR 1.24 (1.16 to 1.34) in this study. Effect estimates are not incompatible but non-differential exposure misclassification in the Carey study (where PM10 concentration estimates were on 1 km grids, compared with 100 m grids in this study) may have biased estimates towards the null, and alternatively the lower estimates in the Carey study may relate to better control for individual-level confounders. Our effect size for PM10 exposure in 2001 and lung cancer mortality in 2002–2009, OR 1.60 (1.29 to 1.99), is larger than the combined estimate from 14 cohorts in the ESCAPE study (HR 1.22 (1.03 to 1.45)),24 although our CIs overlap. The ESCAPE study was able to adjust for smoking, which we were not able to do for this outcome as our smoking proxy was lung cancer. Some earlier UK studies of air pollution in the 1970s,1 25 including one using the LS,1 did not find associations between mortality and particulate air pollution. This may relate to previous less accurate air pollution assessment based on nearest monitoring station.

or this outcome as our smoking proxy was lung cancer. Some earlier UK studies of air pollution in the 1970s,1 25 including one using the LS,1 did not find associations between mortality and particulate air pollution. This may relate to previous less accurate air pollution assessment based on nearest monitoring station. Few studies have examined long-term effects of SO2 exposures. Studies conducted in the UK5 19 25 and the American Cancer Society in the USA26 have generally found statistically significant associations of mortality with SO2, while studies in other parts of Europe2 3 27 have not. In the present study, BS and SO2 levels were highly correlated (both originated from fossil fuel combustion), so it is difficult to clearly attribute mortality effects to one pollutant.

e generally found statistically significant associations of mortality with SO2, while studies in other parts of Europe2 3 27 have not. In the present study, BS and SO2 levels were highly correlated (both originated from fossil fuel combustion), so it is difficult to clearly attribute mortality effects to one pollutant. Was apparent persistence of air pollution mortality risk due to highly correlated exposures? Our results showed continued effects air pollution from 1971 on mortality in subsequent decades up to 2002–2009, suggesting long-term persistence of risk. Since 1966 (5 years prior to the start of the study in 1971) and 2001, 72% of individuals had moved at least once, so results are not merely a function of living in the same place. Air pollution exposures were assigned to an individual's ward of residence (or 1 km grid of residence) at each census year. BS and SO2 air pollution exposures were moderately correlated between decades (r∼0.6–0.7), so an alternative explanation for the apparent persistence of risks to 2002–2009 is that exposure in 1971 may have been acting as a proxy for more recent exposures. However, while adjustment for BS in subsequent years (1981 and 1991) did reduce effect estimates, those for all-cause and respiratory mortality remained statistically significant (see online supplementary appendix B table B2). It is also possible that 1971 levels of exposure may correlate highly with and be a proxy for earlier life exposures when levels were much higher.

(1981 and 1991) did reduce effect estimates, those for all-cause and respiratory mortality remained statistically significant (see online supplementary appendix B table B2). It is also possible that 1971 levels of exposure may correlate highly with and be a proxy for earlier life exposures when levels were much higher. Mortality risks associated with a given exposure decrease over time Mortality risks were generally highest in decades immediately following exposure. This is consistent with a previous UK study finding that in the last 4 years BS and SO2 exposure gave higher risks of respiratory mortality than exposures 12–16 years prior.5 It is also consistent with the Harvard Six Cities follow-up study,4 which found similar risks for annual compared with mean air pollution over the study period (1980–1998), suggesting that air pollution during the last year may be important. The follow-up of the American Cancer Society study6 did not find clear differences in risks for average PM2.5 and SO2 concentrations 1–5, 6–10 and 11–15 years before death, but high correlations between the time periods may have reduced the ability to detect differences.

pollution during the last year may be important. The follow-up of the American Cancer Society study6 did not find clear differences in risks for average PM2.5 and SO2 concentrations 1–5, 6–10 and 11–15 years before death, but high correlations between the time periods may have reduced the ability to detect differences. Increased risks per unit pollutant were observed for more recent exposures even though air pollution levels fell markedly (fourfold lower mean BS concentrations between 1971 and 1991). This might imply a steeper concentration–response curve at lower exposures,28 although this was not the case for BS exposure in 1971 where a steeper concentration–response curve was seen in higher exposures (figure 4); or be due to more accurate exposure estimates for more recent periods. However, the latter seems unlikely because exposure models performed better for earlier periods. Alternatively, changes in air pollution sources over time, with reductions in industry and household emissions and increases from road traffic,12 13 may have led to changes in toxicity. Due to qualitative changes in particulate composition over time, we did not adopt a conversion factor between PM10 and BS. However, the two measures are moderately highly (r=0.5–0.8) correlated29 and a number of previous studies have found associations between BS and CV30 and respiratory mortality31 with comparable effect sizes expressed per IQR of exposure.29 The larger size of the PM10 associations relative to the BS per unit mass in the present study may suggest greater toxicity of recent particles or changes in population susceptibility. A ‘harvesting’ effect of sensitive individuals in our closed cohort might at least partly account for waning effects of air pollution over time, but would be inconsistent with increased risks for more recent exposures, unless sensitivity also increased over time.

ecent particles or changes in population susceptibility. A ‘harvesting’ effect of sensitive individuals in our closed cohort might at least partly account for waning effects of air pollution over time, but would be inconsistent with increased risks for more recent exposures, unless sensitivity also increased over time. Interaction of past with recent air pollution exposures We did not find evidence for effect modification by past exposures to BS and SO2, or put another way, we did not find that higher exposures in earlier life has a multiplicative effect on mortality risk associated with more recent PM10 exposure. Our results did suggest that the relationship between mortality and PM10 exposure was confounded by past exposure (ie, that past air pollution exposures are independently associated with both PM10 exposure and with mortality outcomes, suggesting that not accounting for past exposures will affect observed risk estimates for more recent exposures), although the overall impact of this was small. This is one of the first studies directly to investigate this question. This confounding is unlikely to be solely due to correlation between the exposures as PM10 exposure in 2001 was not strongly correlated with BS and SO2 in earlier years (all r<0.45).

recent exposures), although the overall impact of this was small. This is one of the first studies directly to investigate this question. This confounding is unlikely to be solely due to correlation between the exposures as PM10 exposure in 2001 was not strongly correlated with BS and SO2 in earlier years (all r<0.45). Exposure assessment Previous studies investigating very long-term air pollution exposures have had limited historical exposure information. ESCAPE studies8–10 relied on back-extrapolation from modelled exposures in 2008–2011, studies in Stockholm used emissions data3 32 to estimate historic SO2 exposure without independent measured concentrations to evaluate models, while other studies have used data from the nearest monitoring station.2 5 26 We used land-use regression models, validated against contemporaneous monitored concentrations. Model performance was moderate (r2 0.3–0.5) with weakest performance for SO2 in 1981 (r2=0.26) and best performance for SO2 in 1971 (r2=0.57),12 but the lower r2 values in later years may be partly related to lower variability in concentrations rather than model performance.

st contemporaneous monitored concentrations. Model performance was moderate (r2 0.3–0.5) with weakest performance for SO2 in 1981 (r2=0.26) and best performance for SO2 in 1971 (r2=0.57),12 but the lower r2 values in later years may be partly related to lower variability in concentrations rather than model performance. We used air pollution estimates to the highest spatial resolution available, which were on 1 km (BS and SO2) and 100 m (PM10) exposure grids. This may have contributed to some exposure misclassification, though this is likely to be non-differential and result in bias towards the null. Because of limited access to location of individuals in the LS cohort, we were unable to adjust for spatial autocorrelation, but other studies suggest that impact of this is small.6 19

have contributed to some exposure misclassification, though this is likely to be non-differential and result in bias towards the null. Because of limited access to location of individuals in the LS cohort, we were unable to adjust for spatial autocorrelation, but other studies suggest that impact of this is small.6 19 Other study limitations While the cohort was originally a population-based sample, losses to follow-up occurred that were higher in those living in more deprived areas. We consider that this is more likely to have led to an underestimate than overestimate of associations between air pollution and mortality. We cannot rule out an effect of residual confounding on the observed associations but previous work suggests that it is unlikely to have a large impact on our conclusions. The LS does not have information on individual-level smoking, so, as in some other studies,20 we used area-level lung cancer risk as a proxy; adjustment for this did not affect observed effect estimates (see online supplementary appendix B table B2). While some comparable studies with individual-level information on smoking have found larger effects in non-smokers,26 other studies have found no or only small confounding effects of smoking,7 9 19 or that smoking was not related to the exposure,21 with larger impacts from adjustments for socioeconomic status.19 Association of smoking with air pollution exposures is likely to be through deprivation with which it is highly correlated33 and, in the present study, we adjusted for socioeconomic status at both individual and area level.

not related to the exposure,21 with larger impacts from adjustments for socioeconomic status.19 Association of smoking with air pollution exposures is likely to be through deprivation with which it is highly correlated33 and, in the present study, we adjusted for socioeconomic status at both individual and area level. Conclusions This study suggests that air pollution exposure may have persistent long-lasting impacts on mortality risk but that more recent air pollution exposures is associated with higher relative risks than past exposures. Concentration–response function estimates for recent long-term exposures may be slightly overestimated if previous exposures are not taken into account. Findings may be particularly relevant to countries such as China experiencing high but declining levels of particulate concentrations, with a transition from coal to cleaner fuels and increases in emissions from traffic.

long-term exposures may be slightly overestimated if previous exposures are not taken into account. Findings may be particularly relevant to countries such as China experiencing high but declining levels of particulate concentrations, with a transition from coal to cleaner fuels and increases in emissions from traffic. Supplementary Material Web supplement The permission of the Office for National Statistics to use the Longitudinal Study is gratefully acknowledged, as is the help provided by staff of the Centre for Longitudinal Study Information & User Support (CeLSIUS). CeLSIUS is supported by the ESRC Census of Population Programme (Award Ref: ES/K000365/1). The authors alone are responsible for the interpretation of the data. This work contains statistical data from ONS which is Crown Copyright. The use of the ONS statistical data in this work does not imply the endorsement of the ONS in relation to the interpretation or analysis of the statistical data. This work uses research datasets, which may not exactly reproduce National Statistics aggregates. Contributors: ALH conceived the study and ALH, JG, DB and MB designed the study. JG, CP, DV and KG were responsible for exposure data extraction and preparation. MB, REG and ALH were involved in data preparation and carried out the statistical analyses. The analyses were interpreted by ALH, REG, MB, JG and DB. ALH and REG drafted the initial report; all coauthors revised the report and approved the final version. ALH is the guarantor of this paper.

a extraction and preparation. MB, REG and ALH were involved in data preparation and carried out the statistical analyses. The analyses were interpreted by ALH, REG, MB, JG and DB. ALH and REG drafted the initial report; all coauthors revised the report and approved the final version. ALH is the guarantor of this paper. Funding: The work of the UK Small Area Health Statistics Unit is funded by Public Health England as part of the MRC-PHE Centre for Environment and Health, funded also by the UK Medical Research Council. The study also received support from a Wellcome Trust Intermediate Clinical Fellowship study on Chronic Health Effects on Smoke and Sulphur (CHESS), grant number 075883. Competing interests: None declared. Ethics approval: The study was approved by the ONS LS Research Board. Provenance and peer review: Not commissioned; externally peer reviewed.

The Global Burden of Disease Study estimated that in 2010 there were 6.3, 3.5 and 3.2 million deaths attributable to tobacco smoking, household air pollution from solid fuels and ambient particulate matter pollution, respectively.1 The inhalation of polluted air is the leading risk factor for death and disability globally. The adverse effects of these three categories of air pollution impact most severely on the vulnerable and poor. This is especially true for household air pollution from solid fuels, which almost exclusively affects the poor in the world's poorest countries.2 Around the world approximately three billion people use dirty burning solid fuels like animal dung, crop residues and wood for their day-to-day household energy needs. These fuels are typically burned in open fires for cooking, heating and lighting in or near the home environment. When burned in this way they emit substantial quantities of partial products of combustion like black carbon and carbon monoxide which are harmful to health. The 3.5 million deaths attributable to household air pollution each year are mainly from cardiovascular disease and COPD in adults and pneumonia in young children.3 4

burned in this way they emit substantial quantities of partial products of combustion like black carbon and carbon monoxide which are harmful to health. The 3.5 million deaths attributable to household air pollution each year are mainly from cardiovascular disease and COPD in adults and pneumonia in young children.3 4 In Thorax, Heinzerling et al5 describe the CRECER study, a prospective cohort study of the effects of woodsmoke exposure on children in Guatemala in relation to the introduction of a chimney stove intervention. CRECER builds on the Randomized Exposure Study of Pollution Indoors and Respiratory Effects (RESPIRE) trial that evaluated the effect of a chimney stove intervention on the incidence of pneumonia in young children.6 RESPIRE in turn built on decades of painstaking work by the senior authors of the CRECER paper and their collaborators. This is noteworthy from the perspective of appreciating the value of the CRECER paper. The key finding of CRECER is that delayed installation of a chimney stove intervention is associated with poorer lung growth (assessed using six monthly PEF and FEV1 measurements over an average of 1.3 years) than immediate stove installation. The authors recommend additional studies that include longer follow-up and cleaner stoves or fuels. An alternative view is that it is time to scale up and widely implement chimney stoves on health grounds, which I will argue against. The solution lies instead with the availability of clean air for all to breathe.

tion. The authors recommend additional studies that include longer follow-up and cleaner stoves or fuels. An alternative view is that it is time to scale up and widely implement chimney stoves on health grounds, which I will argue against. The solution lies instead with the availability of clean air for all to breathe. Why not chimney stoves? The main reason is that there is insufficient evidence that they are effective and cost-effective as health interventions. Although RESPIRE was a giant leap forward in many respects, the incidence of pneumonia in children (primary trial outcome) was not statistically significantly reduced in the chimney stove group.6 CRECER certainly adds evidence that chimney stoves may provide health benefits but it does not provide definitive data (note the non-significant finding in relation to FEV1) and it does not provide cost-effectiveness data that would allow policy and decision makers to weigh this intervention against other key health priorities in low-middle income countries. Furthermore, chimney stoves are by and large no more efficient than open fires; they simply vent emissions to the outdoor environment. It is notable that approximately 500 000 of the deaths in the Global Burden of Disease Study attributable to ambient particulate matter pollution are actually attributable to household air pollution-vented outdoors.1 Chimney stoves consume more or less the same amount of fuel as open fires and so contribute in the same way to local environmental destruction and adverse health effects associated with this. By emitting the same burden of climate changing black carbon, carbon dioxide and other partial products of combustion they contribute in the same way to global climate change and the adverse health effects associated with this.

same way to local environmental destruction and adverse health effects associated with this. By emitting the same burden of climate changing black carbon, carbon dioxide and other partial products of combustion they contribute in the same way to global climate change and the adverse health effects associated with this. Cleaner biomass-burning cookstoves have been developed that emit considerably less carbon monoxide and particulate matter per unit of cooking energy delivered compared with open fires and as such offer promise as health interventions.7 The United Nations Foundation Global Alliance for Clean Cookstoves (http://cleancookstoves.org) is a public–private partnership to “save lives improve livelihoods, empower women, and protect the environment by creating a thriving global market for clean and efficient household cooking solutions”. The Alliance has a target of facilitating the adoption of 100 million clean cookstoves to households by 2020. Major global efforts are now underway to scale up the availability of cleaner cookstoves although definitive data about how clean such stoves need to be to deliver health benefits and the cost-effectiveness of such approaches are lacking.

facilitating the adoption of 100 million clean cookstoves to households by 2020. Major global efforts are now underway to scale up the availability of cleaner cookstoves although definitive data about how clean such stoves need to be to deliver health benefits and the cost-effectiveness of such approaches are lacking. A small number of clinical trials of cookstove interventions are underway that will provide some evidence here. The Cooking And Pneumonia Study is a village level cluster randomised controlled trial of a fan-assisted biomass burning cookstove compared with continuation of traditional cooking methods on pneumonia in children under the age of 5 in rural Malawi (http://www.capstudy.org). The Ghana Randomised Air Pollution and Health Study is evaluating a biomass cookstove and liquefied petroleum gas against traditional cooking methods in 1225 maternal infant pairs on birth weight (http://www.kintampo-hrc.org/projects/graphs.asp).

in children under the age of 5 in rural Malawi (http://www.capstudy.org). The Ghana Randomised Air Pollution and Health Study is evaluating a biomass cookstove and liquefied petroleum gas against traditional cooking methods in 1225 maternal infant pairs on birth weight (http://www.kintampo-hrc.org/projects/graphs.asp). The bigger picture message I would take from CRECER in the context of the burden of disease caused by the inhalation of polluted air is that lungs everywhere need clean air to breathe. Although we do not yet have definitive evidence on which to base recommendations about specific cookstove interventions on health grounds we do have guidance about how clean air should be including the recently published WHO Indoor Air Quality Guidelines about household fuel combustion.2 Chimney stoves, cleaner burning biomass stoves, other types of stoves and cleaner fuels may all be part of the solution but in isolation may not lead to sufficiently clean air to deliver health benefits.

n air should be including the recently published WHO Indoor Air Quality Guidelines about household fuel combustion.2 Chimney stoves, cleaner burning biomass stoves, other types of stoves and cleaner fuels may all be part of the solution but in isolation may not lead to sufficiently clean air to deliver health benefits. The Sustainable Energy for All (http://www.se4all.org) partnership was launched in 2011 by UN Secretary General Ban Ki-moon to ensure universal access to modern energy services by 2030. This partnership will help drive global efforts to deliver on goal 7 of the post 2015 sustainable development agenda to ensure access to affordable, reliable, sustainable and modern energy for all. To quote Ban Ki-moon “saving our planet, lifting people out of poverty, advancing economic growth—these are one and the same fight.” Research on how to lift people out of poverty of access to clean energy while providing clean air for all to breathe is also part of this same fight. Competing interests: I am co-PI of the MRC/WT/DfID funded Cooking And Pneumonia Study in Malawi I refer to in the editorial. I collaborate with the senior authors of the Heinzerling paper. Provenance and peer review: Commissioned; externally peer reviewed.

Key messages What is the key question? What is the effect of population-based proactive tobacco treatment on use of evidence-based smoking cessation treatments and long-term quit rates compared with usual care among socioeconomically disadvantaged smokers? What is the bottom line? Population-based proactive tobacco treatment is effective in increasing engagement in evidence-based tobacco cessation treatments and for increasing long-term population quit rates among hard to reach, socioeconomically disadvantaged smokers. Why read on? Taken together with prior research, these findings suggest that the dissemination and large-scale adoption of proactive tobacco treatment approaches may reduce smoking prevalence and socioeconomic disparities in tobacco use.

What is the bottom line? Population-based proactive tobacco treatment is effective in increasing engagement in evidence-based tobacco cessation treatments and for increasing long-term population quit rates among hard to reach, socioeconomically disadvantaged smokers. Why read on? Taken together with prior research, these findings suggest that the dissemination and large-scale adoption of proactive tobacco treatment approaches may reduce smoking prevalence and socioeconomic disparities in tobacco use. Background Smoking rates are much higher in socioeconomically disadvantaged populations in the majority of developed countries.1 Among those experiencing multiple forms of disadvantage (eg, single-parent households, public housing, no access to a car, etc), smoking rates can be as high as 60%, while rates among the most affluent can be as low as 15%.2 3 In the USA, 28% of adults living below the federal poverty line smoke cigarettes compared with 17% of adults at or above the poverty level.4 Among adults younger than 65, 16% of those with private health insurance are current smokers, compared with 34% of Medicaid (a government public health insurance programme) recipients and 32% of the uninsured.5 Because socioeconomically disadvantaged populations smoke more than their more advantaged counterparts, they also suffer disproportionately from smoking-caused diseases.

th insurance are current smokers, compared with 34% of Medicaid (a government public health insurance programme) recipients and 32% of the uninsured.5 Because socioeconomically disadvantaged populations smoke more than their more advantaged counterparts, they also suffer disproportionately from smoking-caused diseases. With respect to smoking cessation, socioeconomically disadvantaged smokers are less likely to use evidence-based smoking cessation treatments (pharmacotherapy including nicotine replacement therapy (NRT), bupropion and varenicline and/or counselling either in-person or by telephone) than the general population of smokers.6–9 Contributing factors to the disparity in evidence-based treatment use include greater life stressors that reduce motivation to quit, lack of knowledge about the benefits of using pharmacotherapy and lack of awareness regarding Medicaid coverage for smoking cessation treatment.10 11 Other barriers include a lower likelihood of receiving preventive care services, difficulty taking time from work for cessation services, travel time and costs and an inability to pay out-of-pocket expenses for pharmacotherapy.12 Providers are also less likely to offer smoking cessation treatment to low-income smokers, perhaps related to provider bias or negative assumptions about interest in quitting.13 14

y taking time from work for cessation services, travel time and costs and an inability to pay out-of-pocket expenses for pharmacotherapy.12 Providers are also less likely to offer smoking cessation treatment to low-income smokers, perhaps related to provider bias or negative assumptions about interest in quitting.13 14 The objective of this randomised controlled trial was to test whether a proactive tobacco treatment intervention, designed to overcome these critical barriers to access and delivery of evidence-based smoking cessation treatment, would improve smoking cessation outcomes relative to usual care among a population-based sample of publicly insured smokers. We chose a population-based approach to fully examine effects among all smokers in the cohort, not just those who expressed interest in quitting.

ry of evidence-based smoking cessation treatment, would improve smoking cessation outcomes relative to usual care among a population-based sample of publicly insured smokers. We chose a population-based approach to fully examine effects among all smokers in the cohort, not just those who expressed interest in quitting. Methods Study design and participants The Offering Proactive Treatment Intervention (OPT-IN) study was approved by the Institutional Review Boards at the University of Minnesota and the Minnesota Department of Human Services (DHS). As previously described,15 the study was a two-group randomised controlled trial conducted among clients of the Minnesota Health Care Programs (MHCP) (Medicaid or MinnesotaCare). Medicaid is a joint federal-state programme that provides payment for medical care for people falling into certain categories, including poverty and certain disabilities. MinnesotaCare is for Minnesota residents without access to affordable healthcare coverage, but who have higher income than those covered by Medicaid/Medical Assistance. The sampling frame consisted of a random sample of non-institutionalised MHCP clients that was stratified by age group (18–24, 25–34, and 35–64), by gender and by MHCP. If two or more individuals had the exact same address, one individual was randomly selected to be in the sample. Study participants were recruited using a mailed, baseline tobacco use screening survey packet that included an informed consent statement and notice of privacy practices. Participants were informed that if they were currently smoking and returned a completed baseline survey they would be participating in a research study and would have a 50% chance of being offered a new smoking cessation programme.

rvey packet that included an informed consent statement and notice of privacy practices. Participants were informed that if they were currently smoking and returned a completed baseline survey they would be participating in a research study and would have a 50% chance of being offered a new smoking cessation programme. Randomisation and masking Individuals who returned a completed baseline survey and reported current cigarette smoking (defined as having smoked a cigarette in the past 30 days, even a puff) were randomised, with equal likelihood within each of the 12 age, gender and MHCP strata (Medicaid or MinnesotaCare), to receive either (1) proactive outreach intervention or (2) usual care. The target recruitment goal was 2500 current cigarette smokers. In contrast to aid-to-cessation trials testing the efficacy of an intervention in smokers interested in quitting, this population impact trial of smoking cessation outreach and treatment included all identified smokers, regardless of their interest in quitting.16–18 Participants were not blinded. However, study staff who administered the questionnaires to collect primary outcome data were blinded to participant's treatment allocation.

population impact trial of smoking cessation outreach and treatment included all identified smokers, regardless of their interest in quitting.16–18 Participants were not blinded. However, study staff who administered the questionnaires to collect primary outcome data were blinded to participant's treatment allocation. Procedures Usual care All MHCP enrolees are assigned a primary care provider and usual care participants could contact their provider to access smoking cessation treatment. However, tobacco treatment was variable and depended on the primary care provider's willingness and capacity to adhere to guidelines. Usual care participants also had access to smoking cessation medications (NRT, sustain-released bupropion or varenicline) at substantially reduced cost ($1–$5 co-pay) through MHCP insurance coverage by obtaining a prescription from their provider. Alternatively, participants could purchase over-the-counter NRT at retail costs. In addition, they could access free telephone counselling by calling the Minnesota state quitline (1-888-354-PLAN).

substantially reduced cost ($1–$5 co-pay) through MHCP insurance coverage by obtaining a prescription from their provider. Alternatively, participants could purchase over-the-counter NRT at retail costs. In addition, they could access free telephone counselling by calling the Minnesota state quitline (1-888-354-PLAN). Proactive outreach intervention Intervention participants were able to receive the same MHCP provider smoking cessation treatment components as usual care participants. Additionally, the proactive outreach tobacco treatment intervention included two primary elements: (1) personalised mailings and telephone calls and (2) facilitated access to a free, comprehensive, evidence-based treatment for tobacco dependence (NRT and intensive, telephone-based behavioural counselling). We designed the intervention to overcome both access barriers and psychosocial barriers experienced by socioeconomically disadvantaged smokers, which was delivered by study telephone counsellors trained in motivational interviewing and smoking cessation counselling.

intensive, telephone-based behavioural counselling). We designed the intervention to overcome both access barriers and psychosocial barriers experienced by socioeconomically disadvantaged smokers, which was delivered by study telephone counsellors trained in motivational interviewing and smoking cessation counselling. Personalised mailings included invitation materials: a letter and brochure describing the University of Minnesota Choose to Quit Smoking cessation programme and the services available to help MHCP enrolees quit smoking. Approximately 3 weeks later, study counsellors called participants with up to 12 contact attempts made at different times of the day over 4 weeks. The choice of 12 contact attempts was based on experience of our pilot of intervention, which was an increase from six call attempts in the original protocol. The purpose of the outreach call was to (1) deliver motivational advice to quit smoking, (2) promote self-efficacy, (3) encourage participants to engage in smoking cessation treatment and (4) provide information on the safety, efficacy and functional benefits of pharmacotherapy, particularly NRT. Employing motivational interviewing techniques, counsellors tailored the content of the call to the participant's readiness to quit and concerns about quitting.19 Motivational interviewing is patient-centred and an evidence-based counselling practice directed at identifying ambivalence and enhancing intrinsic motivation for behavioural change.20 21

ewing techniques, counsellors tailored the content of the call to the participant's readiness to quit and concerns about quitting.19 Motivational interviewing is patient-centred and an evidence-based counselling practice directed at identifying ambivalence and enhancing intrinsic motivation for behavioural change.20 21 After the outreach call, telephone care comprised free proactive telephone counselling, free NRT and a self-help quit smoking manual. Specifically, study counsellors used an adaptation of the evidence-based California Helpline protocol which consisted of seven calls initiated by the counsellor, scheduled in a relapse-sensitive fashion over a 2-month period for those ready to set a quit date (pre-quit, quit day, then 3 days, 1 week, 2 weeks, 1 months and 2 months after the quit date).22 Given variability in participants’ readiness to quit and prior experience with quitting, counselling calls were individually tailored to address the participant's needs. For example, participants who were thinking about quitting but not ready to set a quit date right away received motivational interviewing to enhance their readiness to quit and the call schedule was based on participant preferences. In addition, participants who relapsed to smoking were encouraged to set new quit dates and repeat the counselling programme. In total, a participant was eligible to receive up to 14 counselling calls.

tivational interviewing to enhance their readiness to quit and the call schedule was based on participant preferences. In addition, participants who relapsed to smoking were encouraged to set new quit dates and repeat the counselling programme. In total, a participant was eligible to receive up to 14 counselling calls. Participants were also provided a free 8-week course of NRT (patch, gum or lozenge). NRT was mailed directly to participants in anticipation of their quit date using a protocol based on the US Public Health Service Guideline recommendations.9 23 NRT was purchased from the GlaxoSmithKline Consumer Health Care Government Customer Direct Purchase Program. All participants who received telephone counselling were offered NRT unless they had one of the following contraindications: (1) recent (within 2 weeks) heart attack or severe arrhythmia, (2) unstable angina or (3) pregnancy. Participants were not required to participate in telephone counselling in order to receive NRT, although this practice was not promoted. Participants interested in bupropion or varenicline were referred to their primary care provider. Participants who relapsed and attempted to quit again were able to receive an additional 4 weeks of NRT.

ot required to participate in telephone counselling in order to receive NRT, although this practice was not promoted. Participants interested in bupropion or varenicline were referred to their primary care provider. Participants who relapsed and attempted to quit again were able to receive an additional 4 weeks of NRT. Data collection There were two episodes of data collection: baseline and 1 year following randomisation (participant surveys and DHS administrative data). The baseline and follow-up surveys used modified Dillman mail survey procedures and have been previously described.15 The 1-year follow-up survey followed similar procedures as the mailed baseline survey and included additional procedures to reduce attrition. These included a $10.00 incentive with the first mailing, telephone administration (mixed-mode protocol) for non-respondents to the mailed protocol, and tracking procedures for non-respondents including receipt of updated contact information from the Minnesota DHS.

nd included additional procedures to reduce attrition. These included a $10.00 incentive with the first mailing, telephone administration (mixed-mode protocol) for non-respondents to the mailed protocol, and tracking procedures for non-respondents including receipt of updated contact information from the Minnesota DHS. Outcomes The primary outcome was self-reported 6-month prolonged smoking abstinence at 1 year following randomisation. A person who smoked at least once on seven consecutive days or who smoked at least once on two consecutive weekends in the 6-month period was defined as a treatment failure. The choice and definition of the primary outcome follows recommendations of the Society for Research on Nicotine and Tobacco Measures Workgroup to report multiple measures of abstinence in which prolonged abstinence is the preferred measure.18 Since OPT-IN was a cessation-induction trial (ie, evaluation of an intervention to encourage cessation among a population-based sample of smokers, including those not currently trying to quit), follow-up was tied to the onset of the intervention (ie, time of randomisation). Secondary outcomes included self-reported 30-day point prevalence abstinence and 7-day point prevalence abstinence, use of behavioural counselling, use of smoking cessation medications and use of combination counselling and medication.

t), follow-up was tied to the onset of the intervention (ie, time of randomisation). Secondary outcomes included self-reported 30-day point prevalence abstinence and 7-day point prevalence abstinence, use of behavioural counselling, use of smoking cessation medications and use of combination counselling and medication. Statistical analysis As previously described,15 the goal sample size for this study was 2500 participants (1250 per group), which accounted for attrition in order to have observed smoking abstinence outcomes on 1500 respondents (750 per group). This sample size provides approximately 85% power or greater with a two-sided α of 0.05 to detect differences if the intervention raises quit rates by 4%. Baseline data were obtained for all participants (n=2406) using a baseline survey and DHS administrative records. The usual care (n=1206) and proactive outreach (n=1200) groups were compared across the stratification variables of age, sex and insurance type, as well as socio-demographic and smoking-related clinical variables presented in table 1 using Pearson's χ2 tests and two-sample t tests. Logistic regressions, adjusted for the stratification variables of age, sex and insurance type, modelled the odds of treatment use over the year of follow-up (table 2). Separate regression models were used for each treatment use outcome within the medication, counselling and combination categories. Table 1 Baseline demographic and smoking characteristics, according to treatment group

Baseline data were obtained for all participants (n=2406) using a baseline survey and DHS administrative records. The usual care (n=1206) and proactive outreach (n=1200) groups were compared across the stratification variables of age, sex and insurance type, as well as socio-demographic and smoking-related clinical variables presented in table 1 using Pearson's χ2 tests and two-sample t tests. Logistic regressions, adjusted for the stratification variables of age, sex and insurance type, modelled the odds of treatment use over the year of follow-up (table 2). Separate regression models were used for each treatment use outcome within the medication, counselling and combination categories. Table 1 Baseline demographic and smoking characteristics, according to treatment group Characteristic Usual care (n=1206) Proactive outreach (n=1200) Total (n=2406) Demographics Insurance programme Medicaid 878 (72.8%) 871 (72.6%) 1749 (72.7%) MinnesotaCare 328 (27.2%) 329 (27.4%) 657 (27.3%) Gender (female) 853 (70.7%) 846 (70.5%) 1699 (70.6%) Age category 18–24 249 (20.7%) 247 (20.6%) 496 (20.6%) 25–34 414 (34.3%) 410 (34.2%) 824 (34.3%) 35–64 543 (45.0%) 543 (45.3%) 1086 (45.1%) Race/ethnicity Non-Hispanic white 944 (78.3%) 941 (78.4%) 1885 (78.4%) Black 122 (10.1%) 134 (11.2%) 256 (10.6%) American Indian 87 (7.2%) 80 (6.7%) 167 (6.9%) Hispanic 19 (1.6%) 23 (1.9%) 42 (1.8%) Asian 34 (2.8%) 22 (1.8%) 56 (2.3%) Education Grade 11/lower 156 (13.2%) 166 (14.1%) 322 (13.7%) HS grad/GED 383 (32.5%) 398 (33.9%) 781 (33.2%) Some college 487 (41.3%) 490 (41.7%) 977 (41.5%) College grad/higher 154 (13.1%) 120 (10.2%) 274 (11.6%) Employment Employed/self-employed 608 (51.2%) 598 (51.0%) 1206 (51.1%) Student 75 (6.3%) 87 (7.4%) 162 (6.9%) Out of work 154 (13.0%) 153 (13.0%) 307 (13.0%) Unable to work/disabled 276 (23.2%) 277 (23.6%) 553 (23.4%) Homemaker 75 (6.3%) 58 (4.9%) 133 (5.6%) Yearly income Less than $10k 427 (36.9%) 430 (37.7%) 857 (37.3%) $10 001 to $20k 345 (29.8%) 375 (32.9%) 720 (31.4%) $20 001 to $40k 259 (22.4%) 233 (20.4%) 492 (21.4%) More than $40k 125 (10.8%) 103 (9.0%) 228 (9.9%) Child in home 665 (56.2%) 651 (55.6%) 1316 (55.9%) Smoking Cigarettes/day 13.8 (9.1) 13.4 (9.2) 13.6 (9.2) Time until first cigarette (min) ≤5 321 (26.6%) 296 (24.7%) 617 (25.6%) 6–30 536 (44.4%) 538 (44.8%) 1074 (44.6%) ≥30 349 (28.9%) 366 (30.5%) 715 (29.7%) Cigarette type Menthol 450 (37.5%) 442 (37.1%) 892 (37.3%) Non-menthol 750 (62.5%) 751 (63.0%) 1501 (62.7%) Motivation to quit 6.3 (2.8) 6.3 (2.9) 6.3 (2.9) Quit attempt (past year) 643 (54.0%) 644 (54.6%) 1287 (54.3%) Treatment used (past year) Counselling only 10 (0.8%) 9 (0.8%) 19 (0.8%) Medication only 325 (27.0%) 337 (28.1%) 662 (27.5%) Both 33 (2.7%) 38 (3.2%) 71 (3.0%) Neither 838 (69.5%) 816 (68.0%) 1654 (68.7%) Data are number (%) or mean (SD).

6.3 (2.9) 6.3 (2.9) Quit attempt (past year) 643 (54.0%) 644 (54.6%) 1287 (54.3%) Treatment used (past year) Counselling only 10 (0.8%) 9 (0.8%) 19 (0.8%) Medication only 325 (27.0%) 337 (28.1%) 662 (27.5%) Both 33 (2.7%) 38 (3.2%) 71 (3.0%) Neither 838 (69.5%) 816 (68.0%) 1654 (68.7%) Data are number (%) or mean (SD). Motivation to quit assessed using the contemplation ladder which asked participants to indicate their readiness to quit on a scale from 0 to 10, with higher values indicative of greater readiness to quit.24 A value of 0 corresponds with the statement, ‘No thought of quitting’, a value of 5 corresponds with the statement, ‘Think I should quit but not quite ready’ and a value of 10 corresponds with the statement, ‘Taking action to quit’ (eg, cutting down, enrolling in a programme). GED, General Educational Development; grad, graduation; HS, High school. Table 2 Treatment usage by treatment group over the 1-year follow-up period

Motivation to quit assessed using the contemplation ladder which asked participants to indicate their readiness to quit on a scale from 0 to 10, with higher values indicative of greater readiness to quit.24 A value of 0 corresponds with the statement, ‘No thought of quitting’, a value of 5 corresponds with the statement, ‘Think I should quit but not quite ready’ and a value of 10 corresponds with the statement, ‘Taking action to quit’ (eg, cutting down, enrolling in a programme). GED, General Educational Development; grad, graduation; HS, High school. Table 2 Treatment usage by treatment group over the 1-year follow-up period Treatment type Usual care (n=944) Proactive outreach (n=826) OR* p Value Medication Any medication† 278 (29.5%) 335 (40.6%) 1.63 (1.34–2.00) <0.001 NRT 192 (20.5%) 275 (33.8%) 1.99 (1.60–2.48) <0.001 Bupropion/varenicline 104 (11.1%) 105 (12.9%) 1.19 (0.89–1.59) 0.249 Counselling Any counselling 45 (4.8%) 174 (21.1%) 5.42 (3.83–7.66) <0.001 Phone 27 (2.9%) 155 (19.4%) 8.08 (5.29–12.33) <0.001 In-person 23 (2.6%) 72 (9.4%) 3.87 (2.38–6.29) <0.001 Combination None reported 655 (69.4%) 461 (55.8%) 0.55 (0.45–0.68) <0.001 Medication only 244 (25.9%) 191 (23.1%) 0.85 (0.68–1.06) 0.150 Counselling only 11 (1.2%) 30 (3.6%) 3.21 (1.60–6.47) 0.001 Medication and counselling 34 (3.6%) 144 (17.4%) 5.69 (3.85–8.40) <0.001 Any cessation treatment used 289 (30.6%) 365 (44.2%) 1.81 (1.48–2.21) <0.001 Data are n (%) or OR (95% CI). *Adjusted for stratification variables of age, sex and insurance type. †Participants could report using more than one medication.

Treatment type Usual care (n=944) Proactive outreach (n=826) OR* p Value Medication Any medication† 278 (29.5%) 335 (40.6%) 1.63 (1.34–2.00) <0.001 NRT 192 (20.5%) 275 (33.8%) 1.99 (1.60–2.48) <0.001 Bupropion/varenicline 104 (11.1%) 105 (12.9%) 1.19 (0.89–1.59) 0.249 Counselling Any counselling 45 (4.8%) 174 (21.1%) 5.42 (3.83–7.66) <0.001 Phone 27 (2.9%) 155 (19.4%) 8.08 (5.29–12.33) <0.001 In-person 23 (2.6%) 72 (9.4%) 3.87 (2.38–6.29) <0.001 Combination None reported 655 (69.4%) 461 (55.8%) 0.55 (0.45–0.68) <0.001 Medication only 244 (25.9%) 191 (23.1%) 0.85 (0.68–1.06) 0.150 Counselling only 11 (1.2%) 30 (3.6%) 3.21 (1.60–6.47) 0.001 Medication and counselling 34 (3.6%) 144 (17.4%) 5.69 (3.85–8.40) <0.001 Any cessation treatment used 289 (30.6%) 365 (44.2%) 1.81 (1.48–2.21) <0.001 Data are n (%) or OR (95% CI). *Adjusted for stratification variables of age, sex and insurance type. †Participants could report using more than one medication. NRT, nicotine replacement therapy.

Treatment type Usual care (n=944) Proactive outreach (n=826) OR* p Value Medication Any medication† 278 (29.5%) 335 (40.6%) 1.63 (1.34–2.00) <0.001 NRT 192 (20.5%) 275 (33.8%) 1.99 (1.60–2.48) <0.001 Bupropion/varenicline 104 (11.1%) 105 (12.9%) 1.19 (0.89–1.59) 0.249 Counselling Any counselling 45 (4.8%) 174 (21.1%) 5.42 (3.83–7.66) <0.001 Phone 27 (2.9%) 155 (19.4%) 8.08 (5.29–12.33) <0.001 In-person 23 (2.6%) 72 (9.4%) 3.87 (2.38–6.29) <0.001 Combination None reported 655 (69.4%) 461 (55.8%) 0.55 (0.45–0.68) <0.001 Medication only 244 (25.9%) 191 (23.1%) 0.85 (0.68–1.06) 0.150 Counselling only 11 (1.2%) 30 (3.6%) 3.21 (1.60–6.47) 0.001 Medication and counselling 34 (3.6%) 144 (17.4%) 5.69 (3.85–8.40) <0.001 Any cessation treatment used 289 (30.6%) 365 (44.2%) 1.81 (1.48–2.21) <0.001 Data are n (%) or OR (95% CI). *Adjusted for stratification variables of age, sex and insurance type. †Participants could report using more than one medication. NRT, nicotine replacement therapy. For the analysis of our primary outcome, we fit a stratified logistic regression equation modelling the odds that a participant reported 6-months prolonged abstinence at 1-year follow-up using intervention group, age, gender and MHCP strata as explanatory variables (table 3). Similar analyses assessed the effect of the proactive intervention on reported 30-day and 7-day abstinence. The initial analyses used data from those who responded to the follow-up survey. To address potential informative non-response bias in these initial analyses, we fit a series of selection model analyses. Two common, related, approaches for addressing informative missing data comprise selection models and pattern mixture models.25 Selection models jointly model the study outcomes and the missing of the outcomes, for all participants, by modelling (1) how outcomes are related to the available predictors and (2) modelling how whether the outcome measure is missing is related to the value of the outcome measure and the available predictors. We posited different assumptions for how follow-up survey response would be related to abstinence, the sampling strata and selected covariates. A given selection model analysis jointly fit the two models:

whether the outcome measure is missing is related to the value of the outcome measure and the available predictors. We posited different assumptions for how follow-up survey response would be related to abstinence, the sampling strata and selected covariates. A given selection model analysis jointly fit the two models: Table 3 Smoking abstinence by treatment group at 1 year Model-based estimate of association Abstinence outcome Usual care abstinence rate Proactive outreach abstinence rate OR (95% CI) p Value 6 month prolonged Analysis of observed data* 12.1% (113/937) 16.5% (135/820) 1.47 (1.12 to 1.93) 0.006 Selection model analysis† 7.8–9.0% 11.2–14.2% 1.50 to 1.68 <0.001–0.002 30-day point prevalence Analysis of observed data* 12.1 (114/940) 15.0 (124/826) 1.31 (0.99 to 1.73) 0.055 Selection model analysis† 7.7–7.8% 10.1–10.1% 1.33 to 1.34 0.030–0.033 7-day point prevalence Analysis of observed data* 16.3% (154/942) 17.4% (143/823) 1.11 (0.86 to 1.42) 0.439 Selection model analysis† 11.1–11.6% 11.2–11.5% 0.99 to 1.05 0.719–0.932 *Data are percentages (n/N) with model estimated OR (95% CI) for the intervention and corresponding p value from logistic regression of abstinence on intervention adjusted for age, sex and insurance type stratification measures. †Results presented are the range of least square mean type estimated percentages with range of estimate ORs for intervention and corresponding p values from the regression models for the abstinence outcome with lower AIC statistics.

Model-based estimate of association Abstinence outcome Usual care abstinence rate Proactive outreach abstinence rate OR (95% CI) p Value 6 month prolonged Analysis of observed data* 12.1% (113/937) 16.5% (135/820) 1.47 (1.12 to 1.93) 0.006 Selection model analysis† 7.8–9.0% 11.2–14.2% 1.50 to 1.68 <0.001–0.002 30-day point prevalence Analysis of observed data* 12.1 (114/940) 15.0 (124/826) 1.31 (0.99 to 1.73) 0.055 Selection model analysis† 7.7–7.8% 10.1–10.1% 1.33 to 1.34 0.030–0.033 7-day point prevalence Analysis of observed data* 16.3% (154/942) 17.4% (143/823) 1.11 (0.86 to 1.42) 0.439 Selection model analysis† 11.1–11.6% 11.2–11.5% 0.99 to 1.05 0.719–0.932 *Data are percentages (n/N) with model estimated OR (95% CI) for the intervention and corresponding p value from logistic regression of abstinence on intervention adjusted for age, sex and insurance type stratification measures. †Results presented are the range of least square mean type estimated percentages with range of estimate ORs for intervention and corresponding p values from the regression models for the abstinence outcome with lower AIC statistics. for a specified function g and selected set of covariates for the ith participant, to the observed data for all participants using the expectation–maximisation algorithm process proposed by Ibrahim and Lipsitz.25 For the specification of the function g, the series of selection models considered (1) a simple additive model, (2) a model adding an interaction between intervention and abstinence to this simple additive model and (3) seven different models adding different combinations of interactions between abstinence and the sampling strata to the model with the interaction between abstinence and intervention. In addition to age, gender and healthcare coverage programme strata, the models incorporated patient demographics, smoking history, quit attempt history and motivation to quit and measures of general health, alcohol use and mental health including those measures that differed between respondents and non-respondents to the follow-up survey. (Additional details can be found in the online supplementary appendix.)

patient demographics, smoking history, quit attempt history and motivation to quit and measures of general health, alcohol use and mental health including those measures that differed between respondents and non-respondents to the follow-up survey. (Additional details can be found in the online supplementary appendix.) 10.1136/thoraxjnl-2015-207904.supp1Supplementary data Role of the funding source The funder of the study had no role in study design, data collection, data analysis, data interpretation or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

10.1136/thoraxjnl-2015-207904.supp1Supplementary data Role of the funding source The funder of the study had no role in study design, data collection, data analysis, data interpretation or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. Results Study participants Study participants were recruited from July 2011 to August 2012. We mailed recruitment and tobacco use screening surveys to 21 181 MHCP clients. There were 9362 respondents of whom 2406 were current smokers (25.7%) (see figure 1). The blocked randomisation, implemented separately within the 12 age, gender and MHCP strata, assigned 1200 participants to proactive care and 1206 participants to usual care. The overall follow-up survey response rate was 74% (69% proactive outreach vs 78% usual care). As illustrated in figure 1, complete primary outcome data were available for 820 participants in proactive outreach (68%) and 937 participants in usual care (78%) and were used for the respondent analysis. All randomised participants (n=2406) were included in the selection models accounting for non-response. There were no significant differences in baseline characteristics between the intervention and usual care arms (table 1). Figure 1 Flow chart showing the enrolment, randomisation and follow-up of the study participants.

Results Study participants Study participants were recruited from July 2011 to August 2012. We mailed recruitment and tobacco use screening surveys to 21 181 MHCP clients. There were 9362 respondents of whom 2406 were current smokers (25.7%) (see figure 1). The blocked randomisation, implemented separately within the 12 age, gender and MHCP strata, assigned 1200 participants to proactive care and 1206 participants to usual care. The overall follow-up survey response rate was 74% (69% proactive outreach vs 78% usual care). As illustrated in figure 1, complete primary outcome data were available for 820 participants in proactive outreach (68%) and 937 participants in usual care (78%) and were used for the respondent analysis. All randomised participants (n=2406) were included in the selection models accounting for non-response. There were no significant differences in baseline characteristics between the intervention and usual care arms (table 1). Figure 1 Flow chart showing the enrolment, randomisation and follow-up of the study participants. Proactive outreach engagement Telephone outreach was successful in contacting 836 participants (70%) in the proactive treatment group. Among those who participated in the outreach call, 49 (6%) had already quit smoking, 397 (47%) expressed interest in participating in telephone coaching and 291 (34%) subsequently completed a counselling call. Among this latter group, 212 (73%) were ready to set a quit date, 79 (27%) wanted to talk more about their smoking before setting a quit date and the average number of total completed telephone counselling calls per participant was 4.7.

ting in telephone coaching and 291 (34%) subsequently completed a counselling call. Among this latter group, 212 (73%) were ready to set a quit date, 79 (27%) wanted to talk more about their smoking before setting a quit date and the average number of total completed telephone counselling calls per participant was 4.7. Tobacco treatment use Table 2 details tobacco treatment use by the intervention and usual care arms. Proactive outreach participants were much more likely to use smoking cessation medications compared with usual care (40.6% vs 29.4%), particularly NRT. Telephone counselling for smoking cessation was higher in the proactive outreach intervention group compared with usual care (19.4% vs 2.9%). Furthermore, the rate of combined behavioural counselling and medication treatment was significantly higher in the proactive outreach intervention group (17.4% vs 3.6%, OR 5.69 (95% CI 3.85 to 8.40)).

for smoking cessation was higher in the proactive outreach intervention group compared with usual care (19.4% vs 2.9%). Furthermore, the rate of combined behavioural counselling and medication treatment was significantly higher in the proactive outreach intervention group (17.4% vs 3.6%, OR 5.69 (95% CI 3.85 to 8.40)). Smoking abstinence Participants in the proactive outreach intervention group were significantly more likely to quit smoking than usual care participants. The primary outcome, 6-month prolonged smoking abstinence rate at 1 year, was significantly higher for proactive outreach compared with usual care (16.5% vs 12.1%, OR 1.47 (95% CI 1.12 to 1.93)) (table 3). The number needed to treat (NNT) to gain one participant with 6-month prolonged abstinence was 23. In table 3, we present the selection model analyses which consist of a range of estimated ORs for intervention and estimated abstinence rates across the series of fitted models. The selection models with lower Akaike information criterion (AIC) statistics for modelling smoking abstinence yielded the more plausible estimated abstinence rates. Among these models, estimated prolonged abstinence rates were lower than in the observed data but the differences in rates between outreach and usual care were generally consistent with the observed difference with a range of 3.3%–5.2% while the estimated ORs ranged from 1.50 (1.15 to 1.96) to 1.68 (1.319 to 2.16). Among secondary outcomes, similar effects were found for 30-day abstinence favouring the proactive outreach intervention in analysis of observed data and selection models. However, results for 7-day point prevalence abstinence were insignificant.

he estimated ORs ranged from 1.50 (1.15 to 1.96) to 1.68 (1.319 to 2.16). Among secondary outcomes, similar effects were found for 30-day abstinence favouring the proactive outreach intervention in analysis of observed data and selection models. However, results for 7-day point prevalence abstinence were insignificant. Discussion Proactive outreach was effective at markedly increasing the use of tobacco cessation treatments among socioeconomically disadvantaged smokers, particularly telephone counselling and the combination of counselling and medication. Furthermore, the proactive tobacco treatment intervention was effective at increasing long-term quit rates compared with usual care. This randomised controlled trial adds evidence supporting the effectiveness of proactive tobacco treatment for increasing the population impact of tobacco cessation treatment. An absolute increase in long-term smoking cessation rates of 4.4% is highly significant from a public health perspective. In the USA, the Medicaid programme provides health coverage for 11 million non-elderly low-income adults27 and assuming a 34% prevalence of current smoking an estimated population of 3.74 million smokers.5 28 If such results were generalised to the entire current smoking Medicaid population, there would be nearly 165 000 fewer cigarette smokers in Medicaid.

ides health coverage for 11 million non-elderly low-income adults27 and assuming a 34% prevalence of current smoking an estimated population of 3.74 million smokers.5 28 If such results were generalised to the entire current smoking Medicaid population, there would be nearly 165 000 fewer cigarette smokers in Medicaid. The current study contributes new evidence to support the use of population-based proactive approach to deliver tobacco cessation treatment and confirms that this approach is feasible, effective and applicable to diverse settings and populations. Rigotti et al found increased NRT use (OR 3.47) and short-term abstinence (7-day point prevalence abstinence at 3 months, 5.3% vs 1.1%, OR 5.35) using free telephone consultation with a tobacco coordinator providing 8 weeks of NRT and proactive referral to a state quitline.29 Fu et al found higher 6-month prolonged abstinence at 1 year (13.5% vs 10.9%, OR 1.27) compared with usual care among veteran smokers using proactive outreach programme offering telephone coaching and Veterans Health Administration smoking cessation services.30 Using interactive voice response technology to proactively offer telephone counselling, free NRT and community-based referrals among a disadvantaged population, Haas et al found increased 7-day abstinence at 9 months (17.8% vs 8.7%, OR 2.5).31

ne coaching and Veterans Health Administration smoking cessation services.30 Using interactive voice response technology to proactively offer telephone counselling, free NRT and community-based referrals among a disadvantaged population, Haas et al found increased 7-day abstinence at 9 months (17.8% vs 8.7%, OR 2.5).31 Those engaging in telephone counselling completed an average 4.7 counselling calls, which is substantially higher than the average number of calls reported by publicly funded quitlines.32 This is striking given that for most state-based quitlines, initial contact to the quitline requires action on the part of the tobacco user indicating a relatively high degree of interest in using the quitline's services. This study recruited tobacco users at all stages of readiness to quit. It may be that there was previously unmet demand for such services that the proactive outreach approach made more accessible or salient. It may also be that tobacco users who were not yet ready to make a quit attempt at the start of the study responded well to opportunities to discuss the quitting process and were willing to engage in a higher number of counselling interactions by phone. The increase in use of quitting medications is understandable given the perceived lack of access to free or low-cost medication alternatives for the study population. Despite state-wide media campaigns promoting the Minnesota QUITPLAN Helpline during the study period, it may be that the proactive approach used for the study was particularly effective at increasing awareness of, and interest in, using evidence-based cessation tools.

t medication alternatives for the study population. Despite state-wide media campaigns promoting the Minnesota QUITPLAN Helpline during the study period, it may be that the proactive approach used for the study was particularly effective at increasing awareness of, and interest in, using evidence-based cessation tools. We observed significant intervention effects on 6-month prolonged abstinence, the primary outcome, and 30-day abstinence, a secondary outcome. However, we did not observe significant effects on 7-day point prevalence abstinence, which was another secondary outcome. Since 7-day point prevalence abstinence also includes participants who made a quit attempt after the intervention period, this measure may have underestimated the effectiveness of the proactive outreach intervention. Alternatively, this may suggest that usual care participants were able to achieve similar short-term abstinence but not sustained abstinence, perhaps related to the lower use of treatment by the usual care group.

tion period, this measure may have underestimated the effectiveness of the proactive outreach intervention. Alternatively, this may suggest that usual care participants were able to achieve similar short-term abstinence but not sustained abstinence, perhaps related to the lower use of treatment by the usual care group. This study has several limitations. First, smoking abstinence outcome relied on self-report and was not biochemically verified; however, this approach is similar to other population-based interventions.33 In addition, biochemical verification is not possible for 6-month prolonged abstinence, the study's primary outcome. Second, the follow-up survey response rate was 74%. While this is an excellent response rate considering the low socioeconomic characteristics of the population, there was differential response by intervention and usual care arms and potential for non-response bias. We conducted a series of selection model analyses to account for non-response and observed similar effects, suggesting that our findings are robust. Third, the intervention consisted of a discrete episode of care and it is possible a longitudinal or chronic disease model of care for tobacco use would be more effective. Further research is needed to assess the effects of proactive treatment as part of chronic disease management. Fourth, it is possible that there could have been a Hawthorne effect for usual care participants due to their awareness of and participation in the study as a result of completing the baseline survey. In the unlikely event that the baseline survey did exert a therapeutic effect, it would have been conservative and attenuated the observed effects. Finally, our study did not include high socioeconomic groups and it is unknown if the intervention would have differential effects between income population groups. According to the ‘fundamental cause’ perspective on health disparities, public health interventions that reduce morbidity and mortality can create health disparities because advantaged groups are often better poised to take advantage of opportunities afforded by the intervention. In light of evidence supporting the fundamental cause perspective, it might be advisable to target interventions such as the current proactive outreach intervention to socioeconomically disadvantaged groups to reduce tobacco-related health disparities.

ed to take advantage of opportunities afforded by the intervention. In light of evidence supporting the fundamental cause perspective, it might be advisable to target interventions such as the current proactive outreach intervention to socioeconomically disadvantaged groups to reduce tobacco-related health disparities. In conclusion, this population-based clinical trial demonstrates the effectiveness of proactive tobacco treatment for increasing engagement in evidence-based tobacco cessation treatments and for increasing long-term population quit rates among hard to reach, socioeconomically disadvantaged smokers. Results of this trial may be of particular interest for public insurance agencies, European quitlines and the North American Quitline Consortium, a membership organisation for all publicly funded quitlines in the USA and Canada, whose goals are to maximise the effectiveness, broaden the reach and increase service capacity of quitlines. Administrators of publicly funded quitlines, public insurance agencies (eg, Medicaid) and other health-related organisations should consider adopting a proactive outreach strategy for tobacco users in addition to mass media promotions as a way of effectively recruiting and enrolling socioeconomically disadvantaged tobacco users to smoking cessation services.

nes, public insurance agencies (eg, Medicaid) and other health-related organisations should consider adopting a proactive outreach strategy for tobacco users in addition to mass media promotions as a way of effectively recruiting and enrolling socioeconomically disadvantaged tobacco users to smoking cessation services. The authors would like to gratefully acknowledge the support and assistance of Alan Rodgers and the Minnesota Department of Human Services (DHS). The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the US government. Contributors: SSF, MvR, DN, DJB, JLT, JAN and AMJ designed the study. SSF, MvR and DN supervised the collection of the data. DN, BC and PH analysed the data. SSF, MvR, DN, DJB, JLT, JAN and AMJ interpreted the data. JS, PH and SSF did the literature search. JS, PH and SSF prepared the figures, tables and wrote the first draft of the manuscript. All authors critically reviewed and edited the manuscript. Funding: This study was funded by the National Cancer Institute (5R01CA141527), NIH. This material is the result of work supported with resources and the use of facilities at the Minneapolis VA HSR&D Center for Chronic Disease Outcomes Research. Disclaimer: The authors declare that this manuscript is an honest, accurate and transparent account of the study being reported, that no important aspects of the study have been omitted and that any discrepancies from the study as planned have been explained. Competing interests: None declared.

Funding: This study was funded by the National Cancer Institute (5R01CA141527), NIH. This material is the result of work supported with resources and the use of facilities at the Minneapolis VA HSR&D Center for Chronic Disease Outcomes Research. Disclaimer: The authors declare that this manuscript is an honest, accurate and transparent account of the study being reported, that no important aspects of the study have been omitted and that any discrepancies from the study as planned have been explained. Competing interests: None declared. Ethics approval: University of Minnesota institutional review board (IRB) and Minnesota DHS IRB. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: A dataset stripped of identifiers will be made available on reasonable request for sharing under a data sharing agreement that provides for: (1) a commitment to use the data only for research purposes; (2) a commitment to secure the data using appropriate computer and server technologies, and (3) a commitment to destroy or return the data after analyses are completed. Depending on the data requested, the data sharing agreement and the specific details of the agreement would have to be approved by the Minnesota DHS.

ontinuous variables were expressed as mean values and compared using one-way analysis of variance (ANOVA) with Bonferroni's post hoc test for comparison between individual conditions. All analyses were performed using Prism (Graphpad Software, La Jolla, California, USA). p≤0.05 was considered statistically significant. Ethical approval The samples from patient cohorts 1 and 2 were collected in studies approved by the Lothian Research Ethics Committee (REC) (LREC/2002/8/19) and NRES North East REC (11/NE/0242) and Scotland A REC (11/SS/0089), respectively, with the informed consent/assent from the next of kin. The healthy volunteer BAL fluid was obtained from volunteers in a study approved by Lothian REC (06/S1101/50) with all volunteers giving written informed consent. The monocyte and MDM experiments were conducted using cells obtained under the authorisation of the Lothian REC (08/S1103/38) following written, informed consent.

ealthy volunteer BAL fluid was obtained from volunteers in a study approved by Lothian REC (06/S1101/50) with all volunteers giving written informed consent. The monocyte and MDM experiments were conducted using cells obtained under the authorisation of the Lothian REC (08/S1103/38) following written, informed consent. Results Settings and patients In total there were 159 patients (67 from cohort 1, 92 from cohort 2) for analysis, of whom 43 (13 from cohort 1 and 30 from cohort 2) had confirmed VAP. The demographics of patient cohorts 1 and 2 have been described previously;13 14 briefly the patients were 69% male, with a median age of 60 years (IQR 48–71), 45% were admitted to ICU with surgical diagnoses with the remainder admitted with medical diagnoses. Median APACHE II score was 20 (IQR 16–24), ICU mortality was 28% and hospital mortality 35%. Further demographic details are shown in the online supplementary table S1. Thirty-six healthy donors underwent BAL using the same protocol as cohort 1.13 10.1136/thoraxjnl-2015-208050.supp1Supplementary data

Key messages What is the key question? What are the effects of metformin in severe COPD exacerbations, particularly in relation to lowering blood glucose concentration? What is the bottom line? In this randomised, placebo-controlled trial, metformin did not ameliorate in-hospital hyperglycaemia among non-diabetic patients admitted for COPD exacerbations nor did it affect the secondary end points of fructosamine, C reactive protein and patient-reported outcomes. Why read on? Randomised controlled trials testing novel treatments for severe COPD exacerbations are urgently needed but exceptionally difficult and, as a result, are few and far between; in this trial we explored the anti-hyperglycaemic, anti-inflammatory and clinical effects of metformin in an acutely ill, inpatient COPD population and demonstrated that it is unlikely to offer benefit. Introduction Exacerbations are major events for patients with COPD. They are associated with deconditioning, an increased risk of becoming housebound and reduced physical activity for weeks after onset of symptoms.1 Exacerbations often necessitate hospital admission, when they may be designated ‘severe’.2 Severe exacerbations are associated with a high risk of early mortality and a median survival of only 3.6 years.3 Specific medical treatment for most patients with severe COPD exacerbations comprises systemic corticosteroids, antibiotics and bronchodilators.4 These are at best modestly effective.5 6 New strategies are urgently needed to improve outcomes of patients with severe COPD exacerbations.

edian survival of only 3.6 years.3 Specific medical treatment for most patients with severe COPD exacerbations comprises systemic corticosteroids, antibiotics and bronchodilators.4 These are at best modestly effective.5 6 New strategies are urgently needed to improve outcomes of patients with severe COPD exacerbations. Hyperglycaemia is an unexplored therapeutic target in COPD exacerbations. Elevated blood glucose concentrations occur in the majority of patients admitted to hospital for COPD exacerbations.7–9 The pathogenesis of this is probably multifactorial, including effects from systemic corticosteroid and inhaled β-agonist therapy,10 11 hypoxia,12 acidosis,13 and stress-related increases in glucose-elevating hormones.14 15 Elevated blood glucose concentration is associated with prolonged hospital stay and death, the risk of which increases by 7%–15% for each 1 mmol/L increment in blood glucose concentration,7 16 17 and with failure of non-invasive ventilation.18 Whether these associations are causal is unknown. However, given the well-established adverse effects of hyperglycaemia on oxidative stress, inflammation, immune function, endothelium and thrombosis,19–21 such a relationship is plausible and—subject to the availability of a suitable anti-hyperglycaemic agent—warrants investigation.

associations are causal is unknown. However, given the well-established adverse effects of hyperglycaemia on oxidative stress, inflammation, immune function, endothelium and thrombosis,19–21 such a relationship is plausible and—subject to the availability of a suitable anti-hyperglycaemic agent—warrants investigation. The ideal anti-hyperglycaemic agent would promote euglycaemia without causing hypoglycaemia, be widely available, inexpensive, free from serious adverse effects and amenable to administration both in hospital and at home. Metformin could fulfil these criteria, but its effects in acute medical illness, and COPD exacerbations in particular, are unknown. The present trial was conducted chiefly to test the anti-hyperglycaemic effects of metformin in severe COPD exacerbations, and secondarily to explore its safety, tolerability and effects on inflammation and clinical outcome. Methods Overview This was a randomised, double-blind, placebo-controlled trial, conducted at nine acute NHS hospitals in the UK. The trial was designed, led and implemented entirely by academic investigators and NHS clinicians. It was approved by the South East Research Ethics Committee (reference 10/H1102/62) and implemented in accordance with Good Clinical Practice guidelines. It was publicly registered with ClinicalTrials.gov (NCT01247870) and the International Standard Randomised Controlled Trial Number (ISRCTN) Registry (ISRCTN66148745).

It was approved by the South East Research Ethics Committee (reference 10/H1102/62) and implemented in accordance with Good Clinical Practice guidelines. It was publicly registered with ClinicalTrials.gov (NCT01247870) and the International Standard Randomised Controlled Trial Number (ISRCTN) Registry (ISRCTN66148745). Participants Patients were eligible to participate if they were aged ≥35 years, had COPD diagnosed by physician assessment and/or spirometry and were admitted to hospital for an acute exacerbation, with features as described by Anthonisen et al.22 To facilitate supervised initiation of study treatment, patients were also required to have an expected inpatient stay of ≥48 h. Exclusion criteria included pre-existing, pharmacologically treated diabetes mellitus, respiratory or metabolic acidaemia, risk factors for hypoglycaemia or lactate accumulation and other factors, such as confusion, which precluded trial participation (figure 1). Operational definitions for exclusion criteria are detailed in the trial protocol (see online supplementary data). Figure 1 Flow diagram showing patient disposition. NYHA, New York Heart Association. 10.1136/thoraxjnl-2015-208035.supp1Supplementary data

Participants Patients were eligible to participate if they were aged ≥35 years, had COPD diagnosed by physician assessment and/or spirometry and were admitted to hospital for an acute exacerbation, with features as described by Anthonisen et al.22 To facilitate supervised initiation of study treatment, patients were also required to have an expected inpatient stay of ≥48 h. Exclusion criteria included pre-existing, pharmacologically treated diabetes mellitus, respiratory or metabolic acidaemia, risk factors for hypoglycaemia or lactate accumulation and other factors, such as confusion, which precluded trial participation (figure 1). Operational definitions for exclusion criteria are detailed in the trial protocol (see online supplementary data). Figure 1 Flow diagram showing patient disposition. NYHA, New York Heart Association. 10.1136/thoraxjnl-2015-208035.supp1Supplementary data Interventions Study drugs were supplied by Sharp Clinical Services (formerly Bilcare GCS (Europe), Powys, UK), which over-encapsulated metformin 500 mg tablets and produced visually identical placebo capsules. The random allocation sequence, with a block size of six, was generated by the manufacturer and implemented through sequentially numbered containers. Neither participants nor investigators were aware of treatment assignment until after data lock. Emergency code breaks were administered independently of the trial investigators by staff of the St George's Healthcare NHS Trust Clinical Trials Pharmacy.

ufacturer and implemented through sequentially numbered containers. Neither participants nor investigators were aware of treatment assignment until after data lock. Emergency code breaks were administered independently of the trial investigators by staff of the St George's Healthcare NHS Trust Clinical Trials Pharmacy. Participants were randomly assigned to metformin 500 mg capsules or matched placebo, starting at one capsule on day 1, incrementing to one capsule twice daily on day 2, two capsules in the morning and one in the evening on day 3 and two capsules twice daily from day 4 onwards. This was continued until a follow-up appointment at 1 month. Initially, the trial was designed with an equal allocation ratio. This was modified early in the recruitment phase to an active:placebo ratio of 2:1 on recommendation of a panel of external experts, on the basis that increased exposure to metformin would permit a more informed assessment of safety and tolerability. This modification was performed without revealing the treatment assignment for enrolled participants, and with a commensurate increase in sample size to preserve statistical power. The protocol recommended that participants should receive oral prednisolone 30 mg/day for at least 7 days. Open-label oral anti-hyperglycaemic therapy was prohibited. All other care was at the discretion of the treating physicians.

Participants were randomly assigned to metformin 500 mg capsules or matched placebo, starting at one capsule on day 1, incrementing to one capsule twice daily on day 2, two capsules in the morning and one in the evening on day 3 and two capsules twice daily from day 4 onwards. This was continued until a follow-up appointment at 1 month. Initially, the trial was designed with an equal allocation ratio. This was modified early in the recruitment phase to an active:placebo ratio of 2:1 on recommendation of a panel of external experts, on the basis that increased exposure to metformin would permit a more informed assessment of safety and tolerability. This modification was performed without revealing the treatment assignment for enrolled participants, and with a commensurate increase in sample size to preserve statistical power. The protocol recommended that participants should receive oral prednisolone 30 mg/day for at least 7 days. Open-label oral anti-hyperglycaemic therapy was prohibited. All other care was at the discretion of the treating physicians. Assessments and end points Scheduled study visits took place at baseline, hospital discharge and 1 month and telephone and postal contact with participants was made at 3 months. In addition, participants were reviewed regularly throughout the in-hospital phase to ascertain and encourage compliance with study procedures. They had access to 24 h telephone support post discharge. Capillary blood glucose concentration was measured seven times per day (pre-mealtime and 2 h post-mealtime and late evening) during the in-hospital phase, using point-of-care glucometers. Staff performing these measurements were blinded to treatment allocation but not necessarily to adverse events. Fructosamine and high-sensitivity C reactive protein (hsCRP) concentrations were measured at baseline, discharge and 1 month. The COPD Assessment Test (CAT) was also administered at these time points, and again at 3 months. The Exacerbations of Chronic Pulmonary Disease Tool (EXACT) was completed on paper every evening from randomisation for 1 month. Recurrent healthcare-utilisation events (systemic antibiotic and corticosteroid prescriptions and unplanned hospital admissions) were ascertained at 1 and 3 months. Plasma lactate concentration was monitored as a safety parameter, and details of adverse reactions and serious adverse events were sought at all contact opportunities.

althcare-utilisation events (systemic antibiotic and corticosteroid prescriptions and unplanned hospital admissions) were ascertained at 1 and 3 months. Plasma lactate concentration was monitored as a safety parameter, and details of adverse reactions and serious adverse events were sought at all contact opportunities. The primary end point was mean in-hospital capillary glucose concentration, which has been found to be the most practical metric of hyperglycaemia-associated risk.23 Prespecified secondary end points were selected to supplement assessment of anti-hyperglycaemic effects (fructosamine), inflammation (hsCRP) and clinical recovery (CAT, EXACT and recurrent healthcare-utilisation events). Operational definitions for all end points, including time points for comparisons, were specified prospectively in the trial protocol.

ere selected to supplement assessment of anti-hyperglycaemic effects (fructosamine), inflammation (hsCRP) and clinical recovery (CAT, EXACT and recurrent healthcare-utilisation events). Operational definitions for all end points, including time points for comparisons, were specified prospectively in the trial protocol. Sample size The sample size was calculated to provide at least 90% power at the 5% significance level to detect a 1.5 mmol/L difference in the mean in-hospital capillary blood glucose concentration between groups, using an independent samples t test. This margin was selected to ensure the difference in blood glucose concentration observed between metformin and placebo in the settings of diabetes mellitus (1.9–3.6 mmol/L24–26) and burns injuries (2.3–4.5 mmol/L27 28) would readily be captured. Based on local audit data, in which the SD of mean blood glucose concentration in inpatients with COPD was 1.5 mmol/L, we calculated that with equal allocation, 22 participants per group would be required. An initial target sample size of 23 per group was set to allow for up to two non-analysable participants. With the amendment to 2:1 allocation, described above, a minimum of 32 metformin-treated and 18 placebo-treated participants was required to preserve the a priori power specification. An actual target sample size of 46 and 23 metformin-treated and placebo-treated participants, respectively, was set arbitrarily (the original active treatment group was doubled) so as to offer additional data on the safety and tolerability of metformin in this context.

ired to preserve the a priori power specification. An actual target sample size of 46 and 23 metformin-treated and placebo-treated participants, respectively, was set arbitrarily (the original active treatment group was doubled) so as to offer additional data on the safety and tolerability of metformin in this context. Statistical analysis The primary analysis was performed in accordance with the intention-to-treat principle. No interim analyses were planned or conducted. The prospectively defined statistical analysis plan specified the independent samples t test for the primary analysis of blood glucose concentration. Fructosamine was analysed by calculating the changes between baseline and follow-up, and comparing these between groups with independent samples t tests. CRP data were transformed to a natural log scale for analysis, with comparisons made between groups using the Wilcoxon signed-rank test, then anti-logged to permit the expression of changes over time as a percentage of the baseline value. Changes in CAT and EXACT scores were compared between groups with independent samples t tests. The prospectively defined time points for comparison were follow-up and 12 weeks for the CAT, and days 5, 10 and 28 for the EXACT (day 1 was defined by the first dose of the study drug). The significance of within-group changes in fructosamine, CRP, CAT and EXACT over the various study time points were evaluated using paired-samples t tests. Post hoc sensitivity analyses were conducted to test the robustness of findings to alternative analytical approaches. Statistical analyses were performed using SPSS (V.21.0; IBM Corp.).

thin-group changes in fructosamine, CRP, CAT and EXACT over the various study time points were evaluated using paired-samples t tests. Post hoc sensitivity analyses were conducted to test the robustness of findings to alternative analytical approaches. Statistical analyses were performed using SPSS (V.21.0; IBM Corp.). Results Participants Between January 2011 and March 2014, we recruited 52 participants, all of whom were included in the primary analysis (figure 1). This exceeded the minimum number required for the prospectively defined power specification, but did not reach the larger target sample size set to enhance assessment of safety and tolerability. A decision was taken by the trial steering committee at this point, without reference to study data, to stop recruitment due to time and funding constraints, and imminent expiry of study drugs.

ned power specification, but did not reach the larger target sample size set to enhance assessment of safety and tolerability. A decision was taken by the trial steering committee at this point, without reference to study data, to stop recruitment due to time and funding constraints, and imminent expiry of study drugs. The groups were similar in baseline demographic and clinical characteristics (table 1) and concomitant treatment during the in-hospital phase (see online supplementary table S1). Spirometry, performed where possible at discharge and 1 month, was also similar between groups (see online supplementary table S2). Patients with diabetes mellitus already established on treatment were not eligible for inclusion, and no participant had pre-existing untreated diabetes mellitus. However, baseline haemoglobin A1c (HbA1c) concentrations indicated a possible diagnosis of diabetes mellitus (≥48 mmol/mol) in 7 (18%) of the 39 patients in whom this measurement was obtained (median (IQR) 42 (40–44) and 42 (39–44) mmol/mol in metformin and placebo groups, respectively). Moreover, 41 participants (79%) had a blood glucose ≥10 mmol/L at least once during the in-hospital phase; in 29 (56%) this was ≥11.1 mmol/L. Table 1 Baseline characteristics

The groups were similar in baseline demographic and clinical characteristics (table 1) and concomitant treatment during the in-hospital phase (see online supplementary table S1). Spirometry, performed where possible at discharge and 1 month, was also similar between groups (see online supplementary table S2). Patients with diabetes mellitus already established on treatment were not eligible for inclusion, and no participant had pre-existing untreated diabetes mellitus. However, baseline haemoglobin A1c (HbA1c) concentrations indicated a possible diagnosis of diabetes mellitus (≥48 mmol/mol) in 7 (18%) of the 39 patients in whom this measurement was obtained (median (IQR) 42 (40–44) and 42 (39–44) mmol/mol in metformin and placebo groups, respectively). Moreover, 41 participants (79%) had a blood glucose ≥10 mmol/L at least once during the in-hospital phase; in 29 (56%) this was ≥11.1 mmol/L. Table 1 Baseline characteristics Characteristic* Metformin group (n=34) Placebo group (n=18) p Value Age, year 69±10 65±8 0.258 Sex, M/F 21/13 11/7 0.963 Body mass index, kg/m2 26.0±5.9 25.6±6.1 0.834 Past medical history, n (%) Myocardial infarction 6 (18%) 1 (6%) 0.399 Heart failure 3 (9%) 1 (6%) >0.99 Atrial fibrillation 4 (12%) 2 (11%) >0.99 Cerebrovascular disease 3 (9%) 0 (0%) 0.543 Charlson comorbidity index 0.5 (0–1) 0.0 (0–1) 0.424 Smoking history, pack-years 54±52 54±21 0.961 Pulse, bpm 93±13 92±12 0.769 Systolic blood pressure, mm Hg 133±21 137±23 0.544 Diastolic blood pressure, mm Hg 75±10 80±13 0.133 Respiratory rate, breaths/min 19±2 19±3 0.705 Arterial blood gases pH 7.42±0.05 7.41±0.05 0.854 pCO2, kPa 5.5±1.2 5.6±1.1 0.649 pO2, kPa 9.2±2.5 9.4±2.2 0.824 Biochemistry Urea, mmol/L 5.7±1.7 5.4±2.3 0.693 Creatinine, μmol/L 73±17 72±23 0.861 Haemoglobin A1c, mmol/mol 42 (40–44) 42 (39–44) 0.821 C reactive protein, mg/L 14 (5–77) 17 (5–59) 0.840 Radiographic consolidation, n (%) 7 (21%) 3 (17%) 0.827 *Continuous data are presented as mean±SD or median (IQR), as applicable.

Biochemistry Urea, mmol/L 5.7±1.7 5.4±2.3 0.693 Creatinine, μmol/L 73±17 72±23 0.861 Haemoglobin A1c, mmol/mol 42 (40–44) 42 (39–44) 0.821 C reactive protein, mg/L 14 (5–77) 17 (5–59) 0.840 Radiographic consolidation, n (%) 7 (21%) 3 (17%) 0.827 *Continuous data are presented as mean±SD or median (IQR), as applicable. 10.1136/thoraxjnl-2015-208035.supp2Supplementary tables Primary end point Mean capillary blood glucose measurements were derived from a mean (±SD) of 24±16 individual measurements per participant. The number of individual measurements did not differ between metformin and placebo groups (24±17 and 23±14, respectively; p=0.830). Mean in-hospital blood glucose concentration was 7.1±0.9 and 8.0±3.3 mmol/L, respectively, in the metformin and placebo groups (difference −0.9 mmol/L, 95% CI −2.1 to +0.3; p=0.273; figure 2). Figure 2 Plot showing mean in-hospital capillary blood glucose concentrations. Individual participant means are indicated by shaded circles, and group level means are indicated by solid circles with error bars denoting SD. The difference between the metformin and placebo groups was −0.9 mmol/L (95% CI −2.1 to +0.3; p=0.273). In a sensitivity analysis specified after data collection but before unblinding, in which an extreme outlier was excluded (mean blood glucose concentration 20.2 mmol/L), the difference between groups was −0.2 mmol/L (95% CI −0.8 to +0.45; p=0.566).

rmin and placebo groups was −0.9 mmol/L (95% CI −2.1 to +0.3; p=0.273). In a sensitivity analysis specified after data collection but before unblinding, in which an extreme outlier was excluded (mean blood glucose concentration 20.2 mmol/L), the difference between groups was −0.2 mmol/L (95% CI −0.8 to +0.45; p=0.566). Sensitivity analyses During data entry, it was noted that one participant had a mean glucose concentration of >10 SDs from the cohort mean. On the advice of the trial statistician, and without knowledge of this or any other participants’ treatment assignment, a sensitivity analysis excluding this participant was appended to the analysis plan. In this, the mean glucose concentration in the placebo group became 7.2±1.4 mmol/L, and the difference between metformin and placebo groups was −0.2 mmol/L (95% CI −0.8 to +0.4; p=0.566).

r any other participants’ treatment assignment, a sensitivity analysis excluding this participant was appended to the analysis plan. In this, the mean glucose concentration in the placebo group became 7.2±1.4 mmol/L, and the difference between metformin and placebo groups was −0.2 mmol/L (95% CI −0.8 to +0.4; p=0.566). A post hoc mixed-effects analysis, incorporating trial site as a random factor, did not materially change the findings. Per protocol analysis, accounting for metformin suspension or early discontinuation, revealed no significant differences between groups. Substituting mean blood glucose with alternative summary metrics of glucose concentration similarly did not reveal any significant differences (see online supplementary table S3). In a subgroup analysis (see online supplementary table S4), a trend towards a glucose-lowering effect was more evident among participants whose mean day-1 blood glucose concentration was above the cohort median. However, the difference was not statistically significant (−2.3 mmol/L, 95% CI −6.1 to +1.5; p=0.201) and it was heavily influenced by the outlying observation (when excluded, difference −0.8, 95% CI −2.4 to +0.8; p=0.166). An analysis to investigate for the possibility of a delayed treatment effect identified no interaction between treatment allocation and study day (see online supplementary figure S1; p=0.832, analysis of variance). However, mean daily capillary blood glucose concentration fell over the first 7 inpatient study days (p<0.001).

lysis to investigate for the possibility of a delayed treatment effect identified no interaction between treatment allocation and study day (see online supplementary figure S1; p=0.832, analysis of variance). However, mean daily capillary blood glucose concentration fell over the first 7 inpatient study days (p<0.001). 10.1136/thoraxjnl-2015-208035.supp3Supplementary figure Mean (standard deviation) capillary blood glucose concentrations for the first 7 inpatient study days. Secondary end points Fructosamine Overall, serum fructosamine concentration fell over the three time points examined in the study, but there were no significant differences between groups in this respect. In the metformin group, mean fructosamine concentration (expressed in µmol/L) was 225±39 at baseline, 210±23 at hospital discharge (p=0.031 for change from baseline) and 205±21 at 1 month (p=0.015). In the placebo group, the corresponding values at baseline, discharge and 1 month were 241±47, 245±50 (p=0.482) and 226±40 (p=0.175), respectively. The difference between metformin and placebo groups in respect of the change from baseline was −6 µmol/L (95% CI −22 to +10) at discharge and −1 µmol/L (−19 to +16) at 1 month.

he placebo group, the corresponding values at baseline, discharge and 1 month were 241±47, 245±50 (p=0.482) and 226±40 (p=0.175), respectively. The difference between metformin and placebo groups in respect of the change from baseline was −6 µmol/L (95% CI −22 to +10) at discharge and −1 µmol/L (−19 to +16) at 1 month. C reactive protein The distribution of serum hsCRP measurements at baseline, discharge and 1 month is presented in figure 3. There were no differences between groups at any time point. Overall, expressed as percentages of the baseline value, hsCRP concentration fell to a median (IQR) of 30% (19%–52%) at discharge and 40% (17%–131%) at 1 month (p<0.001 and p=0.007 for changes from baseline to discharge and 1 month, respectively). However, there were no significant differences between groups in this respect (p=0.190 for the change between baseline and discharge, and 0.780 between baseline and 1 month). Figure 3 Box and whiskers plot showing high-sensitivity C reactive protein concentration. Medians are denoted by horizontal bars, IQRs by boxes and ranges by whiskers, save for any outliers which are denoted by circles. The lower limit of detection for the assay is indicated by a dashed line (0.3 mg/L).There were no significant differences in the absolute concentrations between the metformin and placebo groups at any time points.

horizontal bars, IQRs by boxes and ranges by whiskers, save for any outliers which are denoted by circles. The lower limit of detection for the assay is indicated by a dashed line (0.3 mg/L).There were no significant differences in the absolute concentrations between the metformin and placebo groups at any time points. Clinical end points Overall, mean CAT scores were 28±6 points at baseline, falling to 24±7 at discharge (p<0.001 for change from baseline), 25±7 at 1 month (p=0.001) and 25±7 at 3 months (p=0.063). This did not differ between groups (figure 4). Measured from baseline, the proportion of participants with at least a 5-point improvement was 50% at discharge, 43% at 1 month and 41% at 3 months; this was similar between groups. Figure 4 COPD Assessment Test (CAT) scores. CAT scores are on a 40-point scale, with higher scores indicating worse COPD-related health status. Medians are denoted by horizontal bars, IQRs by boxes and ranges by whiskers, save for any outliers which are indicated as circles. There were no significant differences between the groups at any time point. EXACT scores were measured daily, but days 5, 10 and 28 were the predefined time points for comparisons (day 1 was defined by the first study drug dose). Overall, the EXACT score was 52±9 at baseline, falling to 48±11 on day 5 (p=0.041 for change from baseline), 48±11 on day 10 (p=0.054) and 47±10 (p=0.042) on day 28. The groups were similar in this respect (figure 5).

d 28 were the predefined time points for comparisons (day 1 was defined by the first study drug dose). Overall, the EXACT score was 52±9 at baseline, falling to 48±11 on day 5 (p=0.041 for change from baseline), 48±11 on day 10 (p=0.054) and 47±10 (p=0.042) on day 28. The groups were similar in this respect (figure 5). Figure 5 Exacerbations of Chronic Pulmonary Disease Tool (EXACT) scores. EXACT scores are on a 100-point logit scale, with higher scores indicating worse symptoms. Error bars indicate SDs. There were no significant differences between the groups at the three prospectively defined time points for comparison (indicated by solid markers). Day 1 was defined as the first day on which the study medication was administered. Taking the study cohort as a whole, in the 3 months from index admission, 23 participants (44%) experienced 37 further hospital admission episodes, 23 (44%) required 30 further systemic corticosteroid prescriptions and 27 (52%) required 39 further antibiotic prescriptions. The median times to any event (metformin group 46 days vs placebo group 40 days, p=0.682); readmission (76 vs 50 days, p=0.814) and systemic corticosteroid and/or antibiotic prescription (49 vs 63 days, p=0.769) were similar between groups. Median length of stay was also similar, at 6.5 days (4.3–11.5) from admission in the metformin group and 7.0 days (4.8–9.3) in the placebo group (p=0.968). Measured from study entry, corresponding figures were 5.0 days (3.0–8.8) and 5.0 days (4.0–7.3) (p=0.737).

ion (49 vs 63 days, p=0.769) were similar between groups. Median length of stay was also similar, at 6.5 days (4.3–11.5) from admission in the metformin group and 7.0 days (4.8–9.3) in the placebo group (p=0.968). Measured from study entry, corresponding figures were 5.0 days (3.0–8.8) and 5.0 days (4.0–7.3) (p=0.737). Safety and tolerability Twelve participants (23%) experienced at least one hypoglycaemic event, defined as blood glucose concentration of ≤3.9 mmol/L, during the in-hospital phase. There was no significant difference between groups in this respect, with six patients (18%) in the metformin group and six (33%) in the placebo group experiencing an event (p=0.202). Mean lactate concentration was 2.0 mmol/L in both groups (mean difference 0.0, 95% CI −0.3 to +0.3 mmol/L, p=0.879). There were no cases of lactic acidosis, defined prospectively as a lactate concentration of ≥5 mmol/L with acidaemia, in either group.

and six (33%) in the placebo group experiencing an event (p=0.202). Mean lactate concentration was 2.0 mmol/L in both groups (mean difference 0.0, 95% CI −0.3 to +0.3 mmol/L, p=0.879). There were no cases of lactic acidosis, defined prospectively as a lactate concentration of ≥5 mmol/L with acidaemia, in either group. There were 69 adverse events in the metformin group and 25 in the placebo group (event rates 2.0 and 1.4 per participant, respectively; p=0.149; table 2). Related, non-serious adverse events were more common in the metformin group. This was largely due to an excess of gastrointestinal symptoms, which affected 24 patients (71%) in the metformin group and five patients (28%) in the placebo group (relative risk 2.54, 95% CI 1.17 to 5.52, p=0.003). Changes to study treatment (dose reduction, suspension or early discontinuation) due to an adverse event were required for 13 patients (38%) in the metformin group and three patients (17%) in the placebo group (relative risk 2.29, 95% CI 0.75 to 7.02, p=0.109). Table 2 Adverse events

There were 69 adverse events in the metformin group and 25 in the placebo group (event rates 2.0 and 1.4 per participant, respectively; p=0.149; table 2). Related, non-serious adverse events were more common in the metformin group. This was largely due to an excess of gastrointestinal symptoms, which affected 24 patients (71%) in the metformin group and five patients (28%) in the placebo group (relative risk 2.54, 95% CI 1.17 to 5.52, p=0.003). Changes to study treatment (dose reduction, suspension or early discontinuation) due to an adverse event were required for 13 patients (38%) in the metformin group and three patients (17%) in the placebo group (relative risk 2.29, 95% CI 0.75 to 7.02, p=0.109). Table 2 Adverse events Frequency (events per participant) Event classification Metformin (n=34) Placebo (n=18) p Value Relationship with study treatment considered at least possible (adverse reactions) Non-serious 31 (0.91) 6 (0.33) 0.011 Serious* 3 (0.09) – 0.199 Required or prolonged hospital admission 3 (0.09) – Relationship with study treatment considered unlikely Non-serious 23 (0.68) 16 (0.89) 0.417 Serious* 12 (0.35) 3 (0.17) 0.366 Resulted in death 3 (0.09) – Resulted in significant disability or incapacity – 1 (0.06) Required or prolonged hospital admission 9 (0.26) 2 (0.11) This table presents the frequency of adverse events in each group and, due to the unequal group sizes and the potential for participants to experience more than one adverse event, the event rates per participant. Classification by relatedness and seriousness is based on the designations assigned by local investigators according to the Good Clinical Practice adverse event definitions.

and, due to the unequal group sizes and the potential for participants to experience more than one adverse event, the event rates per participant. Classification by relatedness and seriousness is based on the designations assigned by local investigators according to the Good Clinical Practice adverse event definitions. *Individual events could have more than one reason for a ‘serious’ designation; in these instances, the event is classified according to the most severe reason specified. Three participants (6%) died during the study period; all deaths were due to respiratory deterioration and none were judged to be related to the study. Discussion We conducted a multicentre, randomised, double-blind, placebo-controlled trial to explore the effects of metformin in patients admitted to hospital for exacerbations of COPD. We found no difference in the primary end point of mean in-hospital capillary glucose concentration, despite participants having a clear tendency towards hyperglycaemia. There were no differences in any of the prespecified secondary end points, including fructosamine, hsCRP, patient-reported outcome measures (CAT and EXACT) and recurrent healthcare-utilisation events. Adverse reactions, particularly gastrointestinal effects, were more common in metformin-treated participants. Lactate concentration did not differ between groups and there were no cases of lactic acidosis.

ructosamine, hsCRP, patient-reported outcome measures (CAT and EXACT) and recurrent healthcare-utilisation events. Adverse reactions, particularly gastrointestinal effects, were more common in metformin-treated participants. Lactate concentration did not differ between groups and there were no cases of lactic acidosis. Our finding that metformin did not ameliorate blood glucose elevations in COPD exacerbations contrasts with its well-established anti-hyperglycaemic effects in diabetes.29 This may be due to the differing pathophysiology of hyperglycaemia in COPD exacerbations. In this context it is notable that, despite metformin having been available for well over half a century, few studies have explored its glucose-lowering effect in acute illnesses. The only studies to address this question directly were two small trials in a burns unit,27 28 in which a glucose-lowering effect was observed, and an open-label study in critically ill patients,30 in which metformin appeared to be insulin-sparing. The present study differs from these in several important respects. First and foremost, it was conducted in the distinct context of COPD exacerbations, where the pathogenesis of hyperglycaemia may be different to that in burn injuries and critical illness. Second, the sample size was larger, and with its multicentre, double-blind design, the methodology was arguably more robust. Third, the participants in the earlier trials had a greater propensity to hyperglycaemia than those in the present trial. Using the placebo groups as comparators, the mean glucose concentrations in the earlier trials were in the range of 10.2–11.3 mmol/L,27 28 30 as compared with 8.0 mmol/L in the present study. Given that metformin acts to reduce blood glucose towards, but not below, the normal range,29 it is plausible that there was less scope in the present study for metformin to exert a detectable, clinically meaningful, glucose-lowering effect. Supporting this, a post hoc analysis suggested a trend towards a more pronounced glucose-lowering effect in participants with above-median baseline glucose concentrations.

it is plausible that there was less scope in the present study for metformin to exert a detectable, clinically meaningful, glucose-lowering effect. Supporting this, a post hoc analysis suggested a trend towards a more pronounced glucose-lowering effect in participants with above-median baseline glucose concentrations. Inflammation is central to the pathophysiology of COPD. It is evident both in the lungs and the serum, and markers of inflammation, notably CRP, rise during exacerbations and fall with recovery.31 Studies in type 2 diabetes32 33 and polycystic ovarian syndrome34 35 suggest that metformin may have anti-inflammatory effects, evident particularly as a fall in the serum CRP concentration. This finding was not reproduced in the present study. CRP was a secondary end point which did not contribute to the sample size determination and, as such, the negative results do not confirm the absence of an effect. It is also possible that a longer treatment period may be necessary, given that the positive studies lasted months to years, whereas this and other studies,36 37 of shorter duration, reported negative results.

e to the sample size determination and, as such, the negative results do not confirm the absence of an effect. It is also possible that a longer treatment period may be necessary, given that the positive studies lasted months to years, whereas this and other studies,36 37 of shorter duration, reported negative results. The study was not powered to detect changes in clinical end points. However, these were explored to assist in planning future effectiveness studies, if indicated. Overall, scores on the patient-reported outcome instruments improved over the follow-up period, with the trends in favour of metformin, but no significant differences were identified between groups. Likewise, there were trends in favour of metformin, but no significant differences, in the risk of recurrent healthcare-utilisation events. Integrating these non-significant trends with the negative mechanistic findings on anti-hyperglycaemic and anti-inflammatory effects, our interpretation is that a meaningful short-term clinical benefit with metformin in this population is unlikely.

nt differences, in the risk of recurrent healthcare-utilisation events. Integrating these non-significant trends with the negative mechanistic findings on anti-hyperglycaemic and anti-inflammatory effects, our interpretation is that a meaningful short-term clinical benefit with metformin in this population is unlikely. This study has some important limitations. Despite a recruitment period lasting over 3 years and involving nine sites—a considerable escalation over what had originally been envisaged—we were unable to enrol as many participants as intended. In large part this was due to the high prevalence of serious acute and chronic comorbidities among the patients screened. Our experience illustrates the challenges faced when conducting research in this setting, as is reflected in the scarcity of clinical trials testing novel treatments for COPD exacerbations. However, the sample size was sufficient to meet the prospectively defined power specification; its main impact was in constraining our assessment of tolerability and safety. The study cohort was also highly selected, which inevitably brings into question the generalisability of the findings. However, the tendency to hyperglycaemia seen in the present study was similar to that observed in two other inpatient COPD cohorts,7 18 and the mean baseline HbA1c (43 mmol/mol or 6.1%) was the same as that seen in another similar cohort.8

ted, which inevitably brings into question the generalisability of the findings. However, the tendency to hyperglycaemia seen in the present study was similar to that observed in two other inpatient COPD cohorts,7 18 and the mean baseline HbA1c (43 mmol/mol or 6.1%) was the same as that seen in another similar cohort.8 Capillary blood glucose concentration was selected as the primary outcome because it is the glycaemic metric most extensively studied in COPD exacerbations, and which has been clearly associated with adverse outcomes.7 8 16 17 However, point-of-care glucometers are subject to greater analytical error than laboratory analysis of venous blood, which may reduce measurement precision. This is mitigated by the fact that, being less burdensome and costly, capillary glucose can be measured more frequently and thus better capture biological fluctuations. Moreover, the negative primary end point data were supported by similar findings on the supplementary glycaemic metric of fructosamine, which reflects average blood glucose concentration over the preceding 2–3 weeks.38 In conclusion, in this randomised controlled trial conducted in non-diabetic patients admitted to hospital for COPD exacerbations, who were prone to elevated blood glucose concentrations, metformin had no detectable anti-hyperglycaemic effect and did not significantly alter CRP or clinical end points, although these secondary end points were not powered. Collectively, the findings suggest that as an acute treatment for COPD exacerbations in non-diabetic individuals, metformin is unlikely to be beneficial.

tformin had no detectable anti-hyperglycaemic effect and did not significantly alter CRP or clinical end points, although these secondary end points were not powered. Collectively, the findings suggest that as an acute treatment for COPD exacerbations in non-diabetic individuals, metformin is unlikely to be beneficial. The authors are grateful to the National Institute for Health Research (NIHR) Clinical Research Network Portfolio and the Comprehensive Local Research Networks (CLRNs) for their support in conducting this study.

tformin had no detectable anti-hyperglycaemic effect and did not significantly alter CRP or clinical end points, although these secondary end points were not powered. Collectively, the findings suggest that as an acute treatment for COPD exacerbations in non-diabetic individuals, metformin is unlikely to be beneficial. The authors are grateful to the National Institute for Health Research (NIHR) Clinical Research Network Portfolio and the Comprehensive Local Research Networks (CLRNs) for their support in conducting this study. Collaborators: Metformin in COPD Trial Team Members: Dr Srikanth Akunuri (Co-investigator), Mr Michael Tumilty (Research Coordinator) and Dr Jan Poloniecki (Statistician); Chelsea and Westminster Hospital, London—Dr Anjali Balasanthiran (Co-investigator), Ms Cielito Caneja (Co-investigator), Ms Kylie Norrie, Ms Theresa Weldring and Mr Jaime Carungcong (Research Nurses); University Hospital of North Tees—Dr Richard Harrison (Principal Investigator), Ms Nicola Bateman and Ms June Battram (Research Nurses); Kings Mill Hospital—Dr Sam Kemp (Principal Investigator), Ms Karen Whysall (Research Nurse); Royal Lancaster Infirmary—Dr Mark Wilkinson (Principal Investigator), Ms Jayne Craig, Ms Clare Tibke and Ms Rebecca Jeffery (Research Nurses); Blackpool Victoria Hospital—Dr Mohammed Paracha (Principal Investigator), Dr Tarek Saba (Co-investigator), Ms Gemma Swarbrick and Ms Judith Saba (Research Nurses); Royal Preston Hospital—Dr Aashish Vyas (Principal Investigator), Ms Janet Mills and Ms Sarah Goddard (Research Nurses); Freeman Hospital, Newcastle—Dr Anthony De Soyza (Principal Investigator); Ms Therese Small, Ms Ashley Dodds (Research Nurses); Conquest Hospital, Hastings—Dr Umesh Dashora (Principal Investigator), Ms Rachael Golton (Research Nurse).

Vyas (Principal Investigator), Ms Janet Mills and Ms Sarah Goddard (Research Nurses); Freeman Hospital, Newcastle—Dr Anthony De Soyza (Principal Investigator); Ms Therese Small, Ms Ashley Dodds (Research Nurses); Conquest Hospital, Hastings—Dr Umesh Dashora (Principal Investigator), Ms Rachael Golton (Research Nurse). Contributors: AWH wrote the study protocol and associated documents, coordinated the implementation of the study, collated and managed the study data, conducted data analysis and drafted the manuscript. EHB conceived the study and supported AWH in its design, implementation and analysis. PWJ provided advice on the design, implementation and analysis of the study, and the use of patient-reported outcome instruments. DL contributed to the implementation of the study. All authors contributed to the interpretation of study findings, serial revisions of the manuscript and final approval of the submitted version. Funding: The authors wish to acknowledge the support of the British Lung Foundation (grant number COPD10/7) and the Medical Research Council (MR/J010235/1), who jointly funded this trial. The funders contributed expert advice on initial study design as part of their peer review processes. Competing interests: PWJ is employed as a Global Medical Expert by GlaxoSmithKline. EHB reports receiving grants from the Medical Research Council/AstraZeneca, GlaxoSmithKline and Boehringer Ingelheim to support research outside the submitted work. Patient consent: Obtained.

Funding: The authors wish to acknowledge the support of the British Lung Foundation (grant number COPD10/7) and the Medical Research Council (MR/J010235/1), who jointly funded this trial. The funders contributed expert advice on initial study design as part of their peer review processes. Competing interests: PWJ is employed as a Global Medical Expert by GlaxoSmithKline. EHB reports receiving grants from the Medical Research Council/AstraZeneca, GlaxoSmithKline and Boehringer Ingelheim to support research outside the submitted work. Patient consent: Obtained. Ethics approval: This study was approved by the South East Research Ethics Committee (reference 10/H1102/62) and implemented in accordance with Good Clinical Practice guidelines. All participants gave written informed consent. Provenance and peer review: Not commissioned; externally peer reviewed.

Key messages What is the key question? What is the prevalence of Mycoplasmataceae among patients with ventilator-acquired pneumonia (VAP), and does this have any pathophysiological relevance? What is the bottom line? Patients with VAP have a high prevalence of Mycoplasmataceae compared with similar patients without VAP, and the most common species found can impair immune cell function. Why read on? These findings provide a novel insight into the biology of VAP and suggest new potential prevention strategies. Introduction Ventilator-acquired pneumonia (VAP) remains a significant cause of morbidity and mortality in patients admitted to the intensive care unit (ICU).1 There is a growing recognition of failure of immune cell function (immunoparesis) in the lung2 and peripherally3 4 in the pathogenesis of VAP. The mediators of this immunoparesis remain incompletely understood, although it is likely that it arises from multifactorial insults involving both host and microbial factors. It is now widely recognised that the predominant route for infection is via ‘microaspiration’ of organisms from the hypopharynx, allowing colonisation and subsequent infection of the lower airways.1

y understood, although it is likely that it arises from multifactorial insults involving both host and microbial factors. It is now widely recognised that the predominant route for infection is via ‘microaspiration’ of organisms from the hypopharynx, allowing colonisation and subsequent infection of the lower airways.1 The organisms which cause VAP have traditionally been thought to be conventional bacterial species, such as Staphylococcus aureus and Pseudomonas aeruginosa.5 However our understanding of the microbiology of VAP is heavily influenced by the relative ease of microbial culture. In recent years there has been a greater understanding of the human microbiome and the influence of bacteria that are not cultured.6 Recent reports have suggested that non-classical organisms, such as Mycoplasma spp., may be present in patients with VAP.7 8 The methods of detection used in these studies have been variable and there was little evidence of the impact these species may have on host immunity.

nfluence of bacteria that are not cultured.6 Recent reports have suggested that non-classical organisms, such as Mycoplasma spp., may be present in patients with VAP.7 8 The methods of detection used in these studies have been variable and there was little evidence of the impact these species may have on host immunity. Mycoplasmataceae are members of the mollicutes class of bacteria, which lack a cell wall and have the smallest genomes found in self-replicating organisms.9 Their limited biosynthetic capability makes culture and detection problematic, needing specialised growth media and prolonged incubation or reliance on indirect methods such as serology.9 A range of Mycoplasmataceae such as Mycoplasma salivarium and M. orale are generally considered to be human commensals,10 while others such as M. pneumoniae and Ureaplasma urealyticum are recognised pathogens of the respiratory and genitourinary tract, respectively. However, even the Mycoplasma spp. commonly thought to be commensals have been shown to cause infections in immunocompromised hosts.11 Additionally, co-infection with Mycoplasma spp. has been implicated in accelerated progression of HIV/AIDS.12

cognised pathogens of the respiratory and genitourinary tract, respectively. However, even the Mycoplasma spp. commonly thought to be commensals have been shown to cause infections in immunocompromised hosts.11 Additionally, co-infection with Mycoplasma spp. has been implicated in accelerated progression of HIV/AIDS.12 We hypothesised that patients with confirmed VAP would have a higher prevalence of Mycoplasmataceae than patients without VAP, and that the dominant organism found (M. salivarium) would impair the ability of monocytes and macrophages to respond to further bacterial stimuli. Accordingly, this study had two aims; first, to quantify the prevalence of Mycoplasmataceae in two cohorts of patients with suspected VAP and determine which species were present. The second aim was to examine the effect of a classically non-pathogenic Mycoplasma, M. salivarium, on the ability of monocytes and macrophages to respond to pathogenic stimuli. Methods Patients and volunteers We constructed the subject cohorts for this study from clinical samples and data from two recent patient studies and a healthy volunteer cohort.13 14 The first patient cohort was recruited from two general, teaching hospital ICUs from 2005 to 2009. Patients were included if they fulfilled clinical criteria for suspected VAP, that is, mechanical ventilation for >48 h, new infiltrates on chest radiograph, and at least two of the following: temperature >38°C, leucocyte count >11×109 per litre of peripheral blood or purulent tracheal secretions.

Us from 2005 to 2009. Patients were included if they fulfilled clinical criteria for suspected VAP, that is, mechanical ventilation for >48 h, new infiltrates on chest radiograph, and at least two of the following: temperature >38°C, leucocyte count >11×109 per litre of peripheral blood or purulent tracheal secretions. The second patient cohort was recruited from 12 ICUs across the UK from 2012 to 2013, and met similar enrolment criteria to those in patient cohort 1. Both cohorts 1 and 2 underwent a standardised bronchoscopy and bronchoalveolar lavage (BAL) as described.13 14 In both cohorts VAP was confirmed by growth of organisms at >104 colony forming units per ml of BAL fluid (CFU/mL) on conventional culture. The healthy volunteers were recruited from the University of Edinburgh and from a primary care practice. Volunteers underwent either bronchoscopy and BAL using the same protocol as used in patient cohort 1, or phlebotomy for extraction of peripheral blood leucocytes for the leucocyte function experiments detailed below.