Article Text
Statistics from Altmetric.com
Guite and colleagues1 recently attempted to identify risk factors for certain causes of death among patients admitted to hospital with asthma. The data were originally collected for a different purpose,2 however, and problems of data quality and completeness limit the conclusions that can be drawn from this analysis.
The current study is based on follow up of 2242 (68%) of the original cohort of 3292 patients. An expert panel identified 22 patients who died from asthma, 14 from chronic obstructive pulmonary disease (COPD) and other respiratory diseases, and 14 from cardiovascular diseases.2 In the paper by Guite et al 1 the numbers of cases have changed significantly with 29 deaths from asthma, 21 from COPD, and 21 cardiovascular deaths. Because both papers relied on the same expert panel, these discrepancies raise fundamental questions about the interpretation of this paper.
The fact that drugs are prescribed to patients who differ according to baseline risk poses an extraordinary methodological challenge to any epidemiological study. Guite and colleagues1 concluded that “ipratropium bromide is associated with increased risk of death from asthma even after adjustment for a range of markers of COPD.” Even after including extra deaths not identified by the expert panel as asthma deaths and controlling for a marker of COPD co-morbidity but for no markers of asthma mortality, the 95% confidence interval for the odds ratio for ipratropium bromide and death from asthma is extremely imprecise (1.2 to 11). Reducing the number of cases to the original 22 asthma deaths identified by the expert panel and controlling for validated markers of asthma mortality would only further degrade the precision of this estimate.
Boehringer Ingelheim has marketed ipratropium bromide for 25 years and its impressive safety profile has been firmly established by hundreds of randomised, controlled studies and epidemiological studies evaluating tens of thousands of patients. Limitations of the current data—understandable from the fact that the study was not originally designed to meet this challenge—preclude them from providing a reliable empirical basis for suggesting an adverse effect of ipratropium.
authors' reply We understand the concern that representatives of Boehringer Ingelheim have expressed about the interpretation of our paper1-1 which finds an association between prescription of ipratroprium and death among asthmatic patients. In the paper we were also cautious about the interpretation and stated that “ the ratio of information to cases is high and therefore results should be regarded as preliminary and should be tested in a larger study”.
There are, however, a number of issues that deserve additional comment. They state that this study was not originally designed to address cause of death among patients admitted to hospital for asthma. The original study plan was always to conduct such a study. Our first report of the cohort1-2 described the identification of the cohort and established the accuracy of classification of cause of death. The focus of this paper on establishing the accuracy of recording of cause of death on the death certificates by an expert panel explains why only 22 deaths were discussed. Seven further patients who were certified as having died from asthma were not reviewed by the expert panel as there was insufficient further information on which to base a different opinion other than that given on the death certificate (table 1).1-2 However, the high specificity of the diagnosis on the death certificates that were studied leads us to believe that assigning these seven deaths to death from asthma, as on the certificate, is reasonable. As we are concerned with deaths from any cause, this is not a critical issue.
Lanes and Wilson assume that the cohort consisted of 3292 people, but this refers to all admissions including re-admissions. The cohort consisted of 2382 individuals, as stated in both papers,1-1 1-2 and follow up information was available for 2242 (94%).
They are concerned that the risk estimates might be confounded by the severity of the disease. Although this is possible, we believe that their argument is misdirected. Most studies of asthma deaths have looked at death due to asthma as defined on the death certificate and have adjusted for severity using proxy measures. These measures have most often been related to health service use and have included admissions to hospital for asthma, use of oral steroids, and use of more than two asthma medications. Our study was different in that it looked at death from all causes in a cohort of patients, all of whom had been admitted to hospital for asthma. To this extent all the patients came from a single stratum of severity. Further stratification in terms of asthma severity makes little difference. We defined “clinically severe asthma” as any history of drowsiness as a result of asthma, loss of consciousness, respiratory arrest, mechanical ventilation, or admission to the ITU for asthma treatment. Taking the results in table 21-1 which shows the association between death and five risk factors and further stratifying by the presence of “clinically severe asthma”, we lose one observation in whom severity could not be determined. The odds ratio associated with taking ipratroprium or Duovent changes from 2.9 (95% CI 1.2 to 7.0) to 2.7 (1.1 to 6.7) on losing this case, and to 2.7 (1.1 to 6.6) on adjustment for “clinically severe asthma”.
The analysis by cause of death is a sub-analysis and is probably less relevant to the general argument. It was, however, of some interest that the excess mortality among those taking ipratroprium was found for all major causes of death.
Our own view remains, as stated in the paper, that the most likely source of confounding is with concurrent COPD. We have good reason to believe that this is associated with a particularly bleak outlook for the patients and is likely to be associated with prescription of ipratroprium. The analysis of the data did not, however, support this view (table 4) and the elevated risk seemed to persist on adjusting for different markers of COPD, regardless of the certified “cause of death”.
Prescribing decisions need to be taken with an overall view of costs and benefits and we would not feel that the data produced in our paper necessarily warrant a change in prescribing in those cases where there is clear clinical benefit to be obtained from the use of ipratroprium. An excess death rate that is estimated to be two to three times that on other regimens does, however, in our opinion need to be taken seriously and further work must be done to assess whether the observed association is real and, if so, how it might be explained.