© 2003 American Thoracic Society
Inhaled corticosteroids and hip fractureDisease or drugs?To the Editor:Hubbard and colleagues report a study using the General Practice Research Database exploring the relationship between inhaled corticosteroids and the risk of hip fracture. In so doing, they replicate findings recently presented by our own group, but fail to interpret these with due care (1). Using the traditional epidemiological cohort study design, we identified all users of inhaled corticosteroids, and identified fracture incidence among them (2). We contrasted incidence rates with cohorts who used inhaled bronchodilator therapy, but did not use inhaled corticosteroids, and with a control cohort who were matched by age and sex to the inhaled corticosteroids users, but did not suffer from asthma. In contrast, Hubbard and colleagues commenced with the fractures and looked back at exposure history using a nested case–control design. The difference in methodology represents different types of sampling, and should give identical results if derived from the same population (3). The two studies did indeed produce almost identical results with respect to the relative rate of hip fracture in inhaled corticosteroid users, but varied in the evaluation of the role of the underlying disease. To ascribe the effect of inhaled corticosteroid use to the drug, as compared with the disease or its comorbid correlates, seems to us erroneous. The key finding of our study was that an increased risk of fracture was also observed when patients using noncorticosteroid bronchodilator drugs were compared with controls. The risks associated with noncorticosteroid bronchodilator use were comparable to those associated with inhaled corticosteroid use. Therefore, we concluded that the excess risk of fracture in inhaled corticosteroid users was more likely to be related to the underlying respiratory disease and its clinical correlates, than to the inhaled use of corticosteroids (2). Finally, Hubbard and colleagues inappropriately suggest that our study lacked statistical power. They fail to mention that the incidence was ascertained from a cohort of 170,000 corticosteroid users, whereas their own study only included 3,000 inhaled corticosteroid users. Indeed, we estimate that both studies had similar statistical power. It therefore seems to us that the failure to include a second control group using bronchodilators, but not inhaled corticosteroids, has led Hubbard and colleagues to an erroneous inference from their findings. Although we are delighted that Hubbard and colleagues were able to reproduce one of our main findings, it is disappointing to note that they have failed to consider adequately a critical issue in epidemiology: confounding by underlying disease severity.
a Utrecht University Utrecht, The Netherlands REFERENCES
From the Authors: We thank Dr. van Staa and colleagues for their interest in our paper, but a number of their comments are inaccurate. Our two studies differ with regard to design and choice of control group (1, 2). We choose a nested matched case–control study, with controls sampled from the general population (median follow-up time 2.7 years). van Staa and colleagues selected cohorts exposed to inhaled corticosteroids, inhaled bronchodilators but not inhaled corticosteroids, and neither inhaled corticosteroids nor inhaled bronchodilators, with median follow-ups of 0.9, 0.2, and 2.3 years, respectively. The main question we addressed was whether the use of inhaled corticosteroids was associated with hip fracture, and if so, was this explained by other risk factors—particularly oral corticosteroids. We examined an extensive list of potential confounders including all corticosteroid exposures, all comorbid diagnoses, and all other drug exposures—a more extensive list than that examined by van Staa and colleagues. With regard to hip fracture, van Staa and colleagues examined whether patients prescribed inhaled corticosteroids had increased incidence of fracture and whether this was explained by underlying lung disease, hence the need for two control groups. In fact, their results were similar regardless of which control group was used; the rate ratios for hip fracture were 1.22 (95% confidence interval [CI], 1.04 to 1.43) with unexposed controls and 1.20 (95% CI, 0.99 to 1.45) with bronchodilator patients as controls. Our adjusted odds ratio was almost identical at 1.19 (95% CI, 1.10 to 1.28), the narrower confidence intervals reflecting the greater statistical power in our study. The dose–response results were also similar between the two studies. Thus, the two different approaches have produced remarkably similar results for hip fracture. Lung function data are not included in the GPRD so the potential for residual confounding by disease severity is present in both studies, a point we discussed (1). Because most airflow obstruction in the general population is mild, however, we believe that disease severity is unlikely to explain all of the excess fracture risk associated with inhaled corticosteroids. It is more plausible that at least part of this association is due to an impact of inhaled corticosteroids on bone. This interpretation is supported by evidence from observational cohorts (3–5) and clinical trials (6), which suggest that these drugs are associated with a reduction in bone mineral density. Importantly, our present experience suggests that the increase in hip fracture risk associated with inhaled corticosteroids is small (1, 2, 6) and, as we discussed carefully, needs to be balanced against the marked beneficial effects of inhaled corticosteroids for patients with asthma.
a University of Nottingham Nottingham, United Kingdom REFERENCES
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
