In Reply Dr Felson raises 3 points with regard to a potential role of confounding by indication to explain our findings.1
First, he comments on the lack of information on severity of comorbidities. While our propensity score–matched analysis was not able to account for severity level of all covariates, some were incorporated (eg, stages of chronic kidney disease). Nevertheless, a comprehensive propensity score–matched study design with many covariates included tends to balance even unmeasured and unadjusted variables (through their associations with adjusted variables). As such, we expect the unmeasured and unadjustable covariates, such as levels of disease severity, also to be balanced after a comprehensive propensity score matching. Prompted by this comment, we conducted propensity score matching without incorporating stages of chronic kidney disease and found the stages were well balanced among all comparison groups, although they were not included in the propensity score creation. This supports the expectation that other comorbidity severities should be balanced despite not being included in the propensity score generation.
Lei G, Choi HK, Zhang Y. Tramadol and Mortality in Patients With Osteoarthritis—Reply. JAMA. 2019;322(5):466. doi:10.1001/jama.2019.7224
Browse and subscribe to JAMA Network podcasts!
Customize your JAMA Network experience by selecting one or more topics from the list below.
Create a personal account or sign in to: