In Reply The study objective was to assess the association of testosterone use with patient outcomes using observational data from clinical practice. Because treatment with testosterone was not randomized, drawing conclusions based on raw event numbers in patients not receiving treatment vs those receiving treatment is not valid.
To account for differences in follow-up (659 940 days of follow-up for treated vs 6 653 821 days for untreated) and the timing of testosterone therapy initiation, we applied Kaplan-Meier methods with testosterone therapy initiation as a time-varying covariate, which considers patients as receiving therapy once they fill an initial prescription (median of 531 days after angiography [time 0]). This contrasts with the standard methods of categorizing use of testosterone therapy at baseline based on whether patients start therapy at any time during follow-up, which incorrectly classifies patients as receiving therapy before they receive it and attenuates treatment effects.1
Ho PM, Barón AE, Wierman ME. Deaths and Cardiovascular Events in Men Receiving Testosterone—Reply. JAMA. 2014;311(9):964-965. doi:10.1001/jama.2014.401