The analysis of safety data poses many statistical challenges, especially when done on clinical trials that had been designed with a focus on efficacy. Sample sizes are often inadequate for conclusive statements about rare, but serious events, imposing the need for cross-study comparisons. Furthermore, serious adverse of events of special interest will often require closer scrutiny than routinely collected, less serious adverse events (AEs). Hence, both data collection and analyses methods might differ for these two types of AEs.

In this talk, we will give an overview of methods which can be applied to the analysis of AEs. Special attention will be given to the fact that the assessment of safety risks often has to rely on non-randomized comparisons, which complicates causal interpretations. In platform trials, for example, safety data on different treatments will be collected non-concurrently, requiring the handling of potential time trends in the emergence of safety signals.

Duration of exposure to a risk (e.g. a treatment) is also an important component of the risk assessment. Exposure to risk might differ between treatment arms and between studies. In addition, its observation can be masked by intercurrent events such as withdrawal of a patient from the study or by the occurrence of AEs, including other AEs which prevent observation of the AE of interest. A decision then must be made if the intercurrent event induces censoring or if it should be viewed as a competing risk.

Depending on the importance of time and the presence of competing risks, simple estimates of event probabilities in a given time frame, estimates of incidence rates, or cumulative incidence functions might be the most appropriate summary of risk. Regarding inference, available tools range from meta-analyses of crude event probabilities via Cox-regression models for time to AE and models for restricted mean survival times, for the cause-specific or the subdistribution hazard to multistate models.

To aid the decision about the most appropriate analysis tool, it is important to obtain clarity about the safety estimand which is targeted. As an example, the risk of (non-lethal) strokes may appear increased relative to untreated patients by a treatment which reduces the probability of dying from a stroke, but does not influence the frequency of strokes. Hypothetical and composite estimands can be considered, but have different interpretations and require different analysis methods.

The talk will illustrate possible approaches to dealing with these topics with examples from various therapeutic areas.

Aggregate safety assessment planning (ASAP) has recently be proposed (Hendrickson et al 2021). ASAP provides a guide for systematic product-level safety planning, standardized data collection and analyses, knowledge gap assessment and safety related communications. The concept can be extended to Benefit Risk Assessment Planning by taking efficacy and tradeoff between efficacy and safety into consideration. Both activities require an interdisciplinary collaboration along the life-cycle of a medical product.

One the of key features of both ASAP and BRAP is the understanding that this planning is not sufficient at the clinical trial level. Both a comprehensive safety assessment and the benefit-risk assessment require an ordered development programme that systematically addresses knowledge gaps. Master protocols by definition deal with generation and use of information across trials while not necessarily focusing on a single compound and/or single indication. Similar to the traditional approach to drug development, the discussion of the value of master protocols seems to focus on efficacy.

This presentations explores challenges and opportunities when extending the current concepts of ASAP and BRAP to master protocol.

References

Hendrickson, B.A., Wang, W., Ball, G. et al. Aggregate Safety Assessment Planning for the Drug Development Life-Cycle. Ther Innov Regul Sci 55, 717–732 (2021). https://doi.org/10.1007/s43441-021-00271-2

Bronson A, Chase MK, Fisher K, Millar D, Perlmutter J, Richardson N. Mobilizing the clinical trial ecosystem to drive adoption of master protocols. Clin Trials. 2022 Dec;19(6):690-696. doi: 10.1177/17407745221110199. Epub 2022 Sep 9. PMID: 36086812; PMCID: PMC9679560.

European Medicine Agency. Complex clinical trials – Questions and answers. 2022. https://health.ec.europa.eu/system/files/2022-06/medicinal_qa_complex_clinical-trials_en.pdf (assessed om 20 Feb 2023)

US Food and Drug Administration. 2022. Master Protocols: Efficient Clinical Trial Design Strategies to Expedite Development of Oncology Drugs and Biologics.

https://www.fda.gov/media/120721/download (assessed om 20 Feb 2023)

Master protocols are becoming increasingly utilized as the clinical trial designs of choice for the new precision medicine paradigm. This can come in the form of master protocols with the same experimental therapies in cohorts of different disease populations (i.e., basket trials) or multiple experimental arms with a shared control arm (i.e., umbrella trials). Many adaptive features are available to tailor the design for efficient evidence generation. As customary, the bulk of these methods focus solely on efficacy providing methodological tools able to identify the most promising treatments and/or populations. At the same time, safety is usually investigated by means of descriptive incidences rather than using more advanced methods.

An overarching goal of any drug development program is to evaluate benefit-risk profile of pharmaceutical products. Our research will focus on (1) how to leverage information across sub-components – e.g., multi-cohorts or multi-arms – of the master protocol for efficient aggregate safety evaluation and (2) how to systematically combine safety and efficacy evaluations in master protocols by means of structured benefit-risk assessment (BRA) tools. Key concepts will be introduced to highlight different considerations to be made through the various phases of any drug development lifecycle. Available qualitative and quantitative approaches to BRA will also be outlined.