On estimands and the analysis of adverse events in the presence of varying follow‐up times within the benefit assessment of therapies

Abstract

The analysis of adverse events (AEs) is a key component in the assessment of a drug's safety profile. Inappropriate analysis methods may result in misleading conclusions about a therapy's safety and consequently its benefit‐risk ratio. The statistical analysis of AEs is complicated by the fact that the follow‐up times can vary between the patients included in a clinical trial. This paper takes as its focus the analysis of AE data in the presence of varying follow‐up times within the benefit assessment of therapeutic interventions. Instead of approaching this issue directly and solely from an analysis point of view, we first discuss what should be estimated in the context of safety data, leading to the concept of estimands. Although the current discussion on estimands is mainly related to efficacy evaluation, the concept is applicable to safety endpoints as well. Within the framework of estimands, we present statistical methods for analysing AEs with the focus being on the time to the occurrence of the first AE of a specific type. We give recommendations which estimators should be used for the estimands described. Furthermore, we state practical implications of the analysis of AEs in clinical trials and give an overview of examples across different indications. We also provide a review of current practices of health technology assessment (HTA) agencies with respect to the evaluation of safety data. Finally, we describe problems with meta‐analyses of AE data and sketch possible solutions.

Improved confidence intervals for a difference of two cause‐specific cumulative incidence functions estimated in the presence of competing risks and random censoring

“…In the present paper, we assumed that the difference of cause-specific CIFs is the treatment effect parameter (estimand) of interest. While this is certainly true in many situations, there are various alternative definitions of treatment effect estimands for competing risk data, as discussed for example by Allignol, Beyersmann, & Schmoor (2016), Unkel et al (2019), and Lau, Cole, & Gange (2009). Commonly used ratio estimands include ratios of cause-specific hazards and ratios of the so-called subdistribution hazards, defined for example in Fine & Gray (1999) or Lau et al (2009).…”

“…A ratio of cause-specific hazards for the study event can be estimated based on the Cox model where the competing events are represented with a censoring indicator (Beyersmann & Scheike, 2014;Lau et al, 2009;Prentice et al, 1978). The ratios of sub-distribution hazards can be estimated based on the Fine-Gray regression model (Austin & Fine, 2017;Fine & Gray, 1999;Unkel et al, 2019). Unlike estimated differences of cause-specific CIFs that can be regarded as fully non-parametric effect estimates, hazard ratios estimated by the Cox and the Fine-Gary models assume that the respective hazard ratio parameters are time-invariant during the study period.…”

“…While Wald-type CIs for the difference of cause-specific CIFs result in valid inference when the number of study events is large, as was the case in the prostate cancer study of Bill-Axelson et al (2018), these intervals are known to suffer from substantial under-coverage when the number of study events is small (e.g., Fagerland, Lydersen, & Laake, 2015), as may be the case for example in the analysis of rare adverse events. See Unkel et al (2019) for discussion of cause-specific CIFs and related estimands in this context.…”

A new perspective on loss to follow‐up in failure time and life history studies

“…Such data should be collected, even if it requires some degree of tracing for certain subjects in order to ascertain the reason for LTF. We note the increasing interest in dependent and informative LTF in randomized trials and the recent expansion of clinical trial guidelines to reflect this . Indeed, to arrive at a comprehensive understanding of disease treatment, it is desirable to collect data on life histories after LTF.…”

“…We note the increasing interest in dependent and informative LTF in randomized trials and the recent expansion of clinical trial guidelines to reflect this. 37 Indeed, to arrive at a comprehensive understanding of disease treatment, it is desirable to collect data on life histories after LTF. This might be done for a selected subset of study subjects, perhaps as a new study.…”