An application of restricted mean survival time in a competing risks setting: comparing time to ART initiation by injection drug use

Abstract

BackgroundRestricted mean survival time (RMST) is an underutilized estimand in time-to-event analyses. Herein, we highlight its strengths by comparing time to (1) all-cause mortality and (2) initiation of antiretroviral therapy (ART) for HIV-infected persons who inject drugs (PWID) and persons who do not inject drugs.MethodsRMST to death was determined by integrating the Kaplan-Meier survival curve to 5 years of follow-up. To account for the competing risks of death and loss-to-clinic when estimating time to ART, we calculated RMST to ART initiation by estimating the area between the survival curve for ART initiation and the cumulative incidence curve for death or loss-to-clinic. We standardized all curves using inverse probability of exposure weights.ResultsWe followed 3044 HIV-positive, ART-naive persons from enrollment into the Johns Hopkins HIV Clinical Cohort from 1996 to 2014. PWID had a − 0.19 year (95% confidence interval (CI): − 0.29, − 0.10) difference in survival over 5 years of follow-up compared to persons who did not inject drugs. There was no difference between the two groups in time not on ART while alive and in clinic (RMST difference = 0.08, 95% CI: -0.10, 0.36).ConclusionsPWID have similar expected time to ART initiation after properly accounting for their greater risk of death and loss-to-clinic.Electronic supplementary materialThe online version of this article (10.1186/s12874-018-0484-z) contains supplementary material, which is available to authorized users.

A univariable and multivariable analysis of time-to-event data based on restricted mean survival time: A combination with traditional survival methods.

“…meanwhile, clinical interpretation of HR may not be straightforward for clinicians [12]. Thereby, RMST…”

A univariable and multivariable analysis of right censored time-to-event data based on restricted mean survival time: A combination with traditional survival methods.

“…It can be considered as the average survival time until an event of interest occurs during a de ned period and be estimated without the limitation from proportional hazard assumption. If non-proportional hazards appears, biased estimates will be given by Cox proportional hazard model because HR is inconstant over time; meanwhile, clinical interpretation of HR may not be straightforward for clinicians [14].…”

Estimating restricted mean survival time and expected life-years lost in the presence of competing risks within flexible parametric survival models

“…It follows that the RMFT measure can be further partitioned to give the expected life-years lost due to a particular cause of death. Difference in RMST between two groups has also been identified as a clinically useful measure alternative to the HR [23,9,20]. One way in which this can be estimated is by obtaining marginal estimates using standardisation as described in Section 2.2.3.…”

“…The choice of t * should be pre-determined and clinically motivated, and will vary by, for example, cancer types [5,6]. This is also often chosen at maximum follow-up time [9,19].…”

“…Royston and Parmar [5] proposes estimation of the restricted mean survival time (RMST), also known as the restricted mean lifetime (RMLT), as a useful alternative to the HR as a summary measure, particularly when the proportional hazards assumption does not hold [6,7]. Merits on the use of the RMST measure are also highlighted by others [8,9,6,5,10]. RMST is essentially the mean survival before a pre-specified time-point.…”