We thank Drs Hanhart and Vinker for their interest in our study. The authors point out to several methodological issues in our paper. First, the difference between the incidence of stroke and myocardial infarction (MI) in the 2 populations (wet age-related macular degeneration [AMD] using intravitreal bevacizumab and those not taking bevacizumab) is expected. As discussed in our paper, control patients who were not taking bevacizumab in the cohort analysis had a higher baseline cardiovascular risk profile than bevacizumab users, which naturally led to different incident rates for MI and stroke. Hence a propensity score analysis was used to adjust for the differences in baseline covariates. Second, the authors make the assumption that retina specialists in British Columbia opt not to treat a patient with advanced wet AMD with intravitreal bevacizumab, which may have led some of these patients to be included in the control group. The authors do not demonstrate any robust evidence for this assumption. Intravitreal bevacizumab became available in British Columbia in June 2009 and was not available between 2000 and 2007 (the follow-up period for the control group). These patients would have received verteporfin therapy. In order to test the authors’ assumption, we further conducted a sensitivity analysis for an unmeasured confounder, which we refer to as “advanced AMD,” as patients with severe disease may receive a higher number of bevacizumab injections and may be more likely to develop a myocardial infarction or stroke. We further assume that 50% of the unexposed cohort have this unmeasured confounder (severe AMD), whereas only 10% of the intravitreal bevacizumab users have this condition. Using a sensitivity analysis for unmeasured confounders, the relative risk from our study shifts from 0.74 (original relative risk [RR] presented in the paper) to 1.2, still consistent with no increase.
In the nested case-control analysis, the 0.71 (95% CI = 0.41–1.22) is not suggestive of a protective effect for bevacizumab users (comparing those using ≥3 injections vs those using <3). The magnitude of the RR crosses 1, consistent with a statistically nonsignificant effect.