A quantitative bias analysis of the confounding effects due to smoking on the association between fluoroquinolones and risk of aortic aneurysm

摘要

Purpose: Epidemioiogic studies consistently report an increased risk of aortic aneurysm (AA) among users of fluoroquinolones (FQ), but confounding by smoking could explain all or some of the observed risk. Therefore, to better elucidate the potential causal impact of FQ on AA, we quantitatively evaluated the potential confounding effect of smoking on this observed association. Methods: We conducted a series of quantitative bias analyses using three previously published approaches: the E-value approach, the rule-out approach, and the array approach. We additionally conducted a numerical comparison between the rule-out approach and the E-value approach. Results: For an apparent relative risk of 2, the E-value is 3.41, suggesting that smoking needs to be associated with both FQ and AA with a minimal magnitude of 3.41 to explain away the observed twofold FQ-AA association. The array approach found that the prevalence of smoking among FQ users would need to be at least 2.9 times higher (43%) than the nonusers (15%), assuming smoking increases the risk of AA by 7.6-fold. A numerical comparison demonstrated that the results from the rule-out approach are similar to that of the E-value approach when there is a lack of prior data on bias parameters. Conclusions: Using three different approaches, we demonstrate that the strengths of association between smoking and both FQ and AA need to be unusually strong to fully account for the twofold increased risk between FQ and AA. Therefore, it is unlikely that smoking alone would explain away the association reported in the epidemioiogic studies.