gms | German Medical Science

Background: The SAVVY project aims to improve the analyses of adverse event (AE) data in clinical trials through the use of survival techniques appropriately dealing with varying follow-up times and competing events. Although statistical methodologies have advanced, in AE analyses often the incidence proportion, incidence densities or a non-parametric Kaplan-Meier are used which either ignore censoring or competing events. In an empirical study including randomized clinical trials from several sponsor companies, these potential sources of bias are investigated. The main purpose of the empirical study is to compare the estimators that are typically used in AE analysis to the non-parametric Aalen-Johansen estimator as the gold standard [1].

Methods: The comparisons of the standard estimators to the Aalen-Johansen estimator are done with descriptive quantities, plots and with a more formal assessment using a random effects meta-analysis on AE level. Factors that influence the bias are investigated in a meta-regression. The comparisons are not only conducted at the end of follow-up but also at three earlier time points. Similar methods are applied for group comparisons.

Results: In the empirical study 10 sponsor companies provided 17 clinical trials including 186 investigated types of AEs. The 1-Kaplan-Meier estimator is on average about 1.2-fold larger than the Aalen-Johansen estimator and the probability transform of the incidence density overestimates the AE probability even more. Although here the average bias using the incidence proportion is less than 5\%, the bias should not generally be neglected as its size strongly depends on the amount of censoring. Furthermore, the adequate consideration of non-constant hazards is less an issue. But the decision on how a competing event is defined is important. As not only death but all treatment-related terminations of follow-up may be considered as a competing event, this influences the amount of censoring and of competing events which are the leading forces influencing the bias.

Conclusion: The choice of the estimator in the analysis of AEs is crucial. There is an urgent need to improve the guidelines of reporting AEs by finally replacing the Kaplan-Meier estimator and the incidence proportion by the Aalen-Johansen estimator with appropriate definition of competing events.

VJ, KR and FL are employees of Novartis Pharma AG (Basel, Switzerland), F.Hoffmann-La Roche (Basel, Switzerland) and Pfizer Deutschland (Berlin, Germany), respectively. TF has received personal fees for consultancies (including data monitoring committees) from Bayer, Boehringer Ingelheim, Janssen, Novartis and Roche, all outside the submitted work. JB has received personal fees for consultancy from Pfizer, all outside the submitted work. CS has received personal fees for consultancies (including data monitoring committees) from Novartis and Roche, all outside the submitted work. The companies mentioned will contribute data to the meta-analysis. RS has declared no conflict of interest.

The authors declare that an ethics committee vote is not required.