I read with interest the recent report by Barone-Rochette et al describing the 6-year follow-up of the EVASTENT (Cost-Effectiveness of Sirolimus-Eluting Stents in Patients With and Without Diabetes) matched-cohort registry. In my opinion, the investigators may have inappropriately used 1 minus the Kaplan-Meier curve in Figures 1 and 2 to estimate the probability of failure (very late stent thrombosis) in the setting of a competing risk (namely, death). The misuse of the 1 minus the Kaplan-Meier curve in the setting of competing risks is well described and is known to cause biased results. To create the 1 minus the Kaplan-Meier curve in the report labeled as (b), I assume the investigators treated death (a competing risk) as censored at the time it occurred. This incorrectly assumes that patients who die and are subsequently censored are equally as likely to have very late stent thrombosis as the patients still at risk. Stata (the investigators’ chosen statistical package) has computational methods for cumulative incidence estimation to account for competing risks. I suggest that the results of this study may be more valid and interpretable if the data are analyzed to consider death as a competing risk when considering very late in-stent thrombosis.