Jump to comment:

• Study design and interpretation
  Imran Aziz
  Published on: 29 April 2016

• Published on: 29 April 2016

  Study design and interpretation

  • Imran Aziz, doctor

  Dear Editor

  The study by Duffy et al,(1) published in the September issue of Thorax, has a number of errors in the design and interpretation of the results.

  In the study design; 1- By definition COPD is a condition where FEV1 changes very little, so a study of COPD intervention based on change in FEV1 is not correct. 2- The study was in theory powered to show what amounts to about 30% improvement in FEV1...

  Show More

  Dear Editor

  The study by Duffy et al,(1) published in the September issue of Thorax, has a number of errors in the design and interpretation of the results.

  In the study design; 1- By definition COPD is a condition where FEV1 changes very little, so a study of COPD intervention based on change in FEV1 is not correct. 2- The study was in theory powered to show what amounts to about 30% improvement in FEV1 (baseline FEV1 approx 600mls and expected difference 200mls) between the 2 regimens, which is unlikely to occur in COPD patients. So in a way the results were already known prior to start of the study. 3- By leaving the decision of stopping treatment to individual consultants rather than using a fixed protocol makes the results prone to bias. 4- As this a study of acute exacerbation, it should have been analysed at the end of first 12 hours, as any changes seen later are not ‘acute’ but simply reflect differences in practice.

  In the results section; 1- The duration of therapy reflects the rate of clinical improvement between active drug and placebo, as the IV therapy would have been stopped once clinical improvement was seen. This difference is almost statistically significant, which has not been elaborated. 2- While all patients were assessed within a 4 hours of admission. But during an acute illness the symptom scores and other ‘baseline’ parameters are likely to have altered in 4 hours. Ideally these should have been recorded prior to receiving any treatment. 3- The calculations and graphs are based on 5 days of therapy where it is clear that on days 4/5 there was deterioration in the patients on active treatment with a fall in FEV1. So saying that FEV1 improved more in the placebo limb is incorrect. 4- The mean number of days of treatment in the placebo group is given as 2.3 with CI of 1 to 2 is statistically impossible as mean cannot be greater than 95% CI.

  In the discussion section; 1- Firstly saying that the study was adequately powered is incorrect. 2- As the recording of baseline parameters was done late in the management pathway, this would make the patients less breathless with a lower symptom score. So the potential of improvement between active and placebo groups will be small and would favour placebo group (type 1 error), but no mention of this has been made in the discussion. 3- The fact the previous studies like YONIV (2) have shown that pH is a very important predictor of mortality. But in the discussion section, the fact that patients on active treatment had a statistically significant improvement in pH, is described only as 'slightly higher'. This suggests that according to authors any significant result of any intervention could be overlooked if the result goes against their line of reasoning.

  The statement that ONE investigator saw each and every patient (320) within 4 hours of admission is physically impossible.

  Using professor Town to write the accompanying editorial showed poor judgement on behalf of the editors, as he is clearly biased against use of aminophylline. In the accompanying editorial (3), professor Town gives 2 references of guidelines, which recommend that aminophylline should not be used. I am sure that even professor Town knows that these recommendations are based on grade D evidence, which means that they are not based on evidence but on a group consensus and therefore, do not have true evidence base. Ideally grade D recommendations should be taken out of guideline writing process. The fact that Professor Town did not find any errors in the study (design / results / interpretation), suggests that he has let his personnel bias cloud his judgements.

  But all is not lost as there are 2 very important positive conclusions from this study namely:

  Firstly it has been shown that use of aminophylline is not associated with any major side effects. Secondly the fact that aminophylline therapy improved the pH in the active treatment group, suggests that aminophylline could be used as initial management of acidotic COPD patients, till NIV is being arranged.

  After performing such a poorly designed and biased study, expecting it to be published in the 2nd highest rated thoracic journal is bad but it was shocking to see it published.

  There is however, one very amusing fact, in a hospital where the COPD team obviously believes that aminophylline does not work, usage of aminophylline was the second commonest reason for 'non-eligibility for the trial' in patients with exacerbation of COPD.

  References

  1. Duffy N, Walker P, Diamantea F et al. Intravenous aminophylline in patients admitted to hospital with non-acidotic exacerbations of chronic obstructive pulmonary disease: a prospective randomised controlled trial. Thorax 2005; 60: 713-7

  2. Plank P, Owen J, Elliott M. Early use of non-invasive ventilation for acute exacerbations of chronic obstructive pulmonary disease on general respiratory wards: a multicentre randomised controlled trial. Lancet. 2000; 355:1931-5

  3. Town G. Aminophylline for COPD exacerbations? Not usually. Thorax 2005; 60: 709

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top