The impact of different censoring methods for analyzing survival using real-world data with linked mortality information: a simulation study

Abstract

Abstract Background Evaluating outcome reliability is critical in real-world evidence studies. Overall survival is a common clinical outcome in these studies; however, its capture in real-world data sources is incomplete and often supplemented with linked mortality information from external sources. There are conflicting recommendations for censoring overall survival in real-world evidence studies. We conducted a simulation study to understand the impact of these different methods on estimating median survival and log hazard ratios when external mortality information is not fully captured. Methods We used Monte Carlo simulation to emulate a non-randomized comparative effectiveness study of two treatments with real-world data from electronic health records and linked external mortality data. We simulated the time to death, the time to last database activity and the time to data cutoff. We attributed death events after the last database activity to linked external mortality data and randomly set them to missing to reflect the sensitivity and specificity of contemporary real-world data sources. Two censoring schemes were evaluated: (1) censor at the last activity date without an observed death, and (2) censor at the end of data availability (data cutoff). We used bias, coverage, and rejection rate to assess the performance of each method in estimating median survival and log hazard ratios under varying amounts of incomplete mortality information and varying treatment effects, length of follow-up, and sample size. Results When mortality information was captured, median survival estimates were unbiased when censoring at data cutoff and underestimated when censoring at the last activity. When linked mortality information was missing, censoring at the last activity date underestimated the median survival, while censoring at the data cutoff overestimated it. As missing linked mortality information increased, bias decreased when censoring at the last activity date and increased when censoring at data-off. Conclusions Researchers should understand the completeness of linked external mortality information when choosing how to censor the analysis of overall survival using real-world data. Substantial bias in median survival estimates can occur if an inappropriate censoring scheme is selected. We advocate for real-world data providers to perform validation studies of their mortality data and to publish their findings to inform methodological decisions better.