A Practical Overview and Reporting Strategies for Statistical Analysis of Survival Studies

Abstract

Survival (time-to-event) analysis is commonly used in clinical research. Key features of performing a survival analysis include checking proportional hazards assumptions, reporting CIs for hazards ratios and relative risks, graphically displaying the findings, and analyzing with consideration of competing risks. This article provides a brief overview of important statistical considerations for survival analysis. Censoring schemes, different methods of survival function estimation, and ways to compare survival curves are described. We also explain competing risk and how to model survival data in the presence of it. Different kinds of bias that influence survival estimation and avenues to model the data under these circumstances are also described. Several analysis techniques are accompanied by graphical representations illustrating proper reporting strategies. We provide a list of guiding statements for researchers and reviewers.

A systematic review of the quality of conduct and reporting of survival analyses of tuberculosis outcomes in Africa

“…When the time-toevent of interest is treatment success, it is plausible to assume other treatment outcomes such as 'death',`lostto-follow-up' and`transfer out' were informative censored and thus considered as competing events. Fine and Gray non-parametric test comparing the cumulative incidence functions without requirement of noninformative censoring could be used in such settings [10,[78][79][80][81]. However, application of this method was rare in this review and one of the studies using the method, incorrectly reported hazard ratios rather than subdistribution hazard ratios [52].…”

Symptom Control and Survival for People Severely ill With COVID: A Multicentre Cohort Study (CovPall-Symptom)

“…Using Cox proportional hazards modelling, we estimated multivariate-adjusted hazard ratios (HRs) of multiple risk factors on the survival function (short survival used to indicate rapid deterioration), censored when cases were still in care or discharged. Here the censoring was noninformative, where censoring times of the patients are not influenced by their times of their death 31 . We tested whether the proportional hazard assumption stands with inspection of Kaplan-Meier survival curves.…”

Initiation of Breastfeeding in Low- and Middle-Income Countries: A Time-to-Event Analysis

“…25 Therefore, we selected an accelerated failure time (AFT) model, which does not require an assumption of proportional hazards. 26,27 We tested 3 distributions of the AFT model (Weibull, log-normal, and loglogistic) for goodness of fit using Akaike's information criterion (AIC). We selected the log-logistic model because it produced the best fit model in every country but 2, wherein the AIC did not vary substantially between log-normal and log-logistic distributions.…”