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.

Mortality endpoints are accepted by EMA and G-BA. EMA accepted well established and clinically relevant morbidity endpoints (e.g. progression-free survival and response rate), which were mostly excluded by G-BA from their value decisions. The applicability of methods used for benefit assessments to HRQoL differs from the mortality and morbidity categories, and requires further clarification.

Objectives: Health technology assessments (HTA) rely on head-to-head comparisons. We searched for intraindividual comparisons (IIC) qualifying as head-to-head design to develop comparative evidence.Methods: Gemeinsamer Bundesausschuss (G-BA) appraisals between January 2011 and April 2020 were reviewed for inclusion of IIC. Identified IIC were grouped according to disease characteristics into nonprogressive, progressive, irregular, or symmetrical conditions. Evaluation of IIC by Institut für Qualität und Wirschaftlichkeit im Gesundheitswesen (IQWIG) and acceptance of IIC by G-BA were determined, and criteria for the usage and quality of IIC were developed.Results: A total of 483 appraisals finalized between January 2011 and April 2020 were reviewed. Eleven appraisals included IIC: nonacog beta (hemophilia B), turoctocog alpha (hemophilia A), emicizumab (2 appraisals: hemophilia A), pasireotide (unresectable pituitary tumor), lomitapid (homozygous familial hypercholesterolemia), glycerol phenylbutyrate (2 appraisals: urea cycle disorders), asfotase alfa (hypophosphatasia), lumacaftor (cystic fibrosis), and larotrectinib (NTRK 1 solid tumors). All those appraisals related to rare genetic conditions with hemophilia and its bleeding rate are considered mainly a nonprogressive condition. All the other diseases show progressive disease characteristics. None of the identified IIC has been accepted by G-BA. Inconsistencies of before/after study design, lack of clarity on treatments prior to the switch, and different time intervals were among the most commonly cited methodological concerns.Conclusions: IICs provide a rare opportunity to determine comparative effectiveness in distinct clinical settings that are not suitable or difficult to randomize into parallel groups. While manufacturers and researchers should aim for highest methodological standards when running an IIC, HTA bodies should accept IIC in distinct settings when determining relative effectiveness.

A575Objectives: There is a growing interest from health technology assessment agencies in determining the clinical outcomes assessments and endpoint strategies that can establish treatment benefits. We describe a systematic literature review of endpoints and outcomes used in schizophrenia trials to determine treatment benefit. MethOds: The therapies selected in the search strategy included pharmacological interventions, cognitive-behavioural therapies, family intervention, and music therapy. These were chosen to reflect the range of interventions in current use, and to allow comparison between outcomes reported for different therapies. The search terms were designed to include all outcomes for each therapy area, and were used to search four electronic databases for published English language studies. Randomised controlled trials (RCTs) were retrieved if they included patients with schizophrenia treated with the chosen therapies, and clinical outcomes from a predefined list (e.g. symptom improvement, functionality, quality of life, remission rates, response rates, and recovery). Results: Of 2,221 RCTs identified, 271 progressed to data extraction; 225 assessed pharmacological interventions and 46 nonpharmacological interventions. Approximately 76 outcomes were measured across the trials. The most common scale used in pharmacological trials was the Positive and Negative Syndrome Scale (PANSS) total score (76.9%), and the PANSS positive subscale in non-pharmacological trials (50%). However, even within the common outcomes, the specified level of reduction to define a relevant response varied; among trials reporting PANSS total, five different levels of reduction were defined (≥ 20%, ≥ 25%, ≥ 30%, ≥ 40%, ≥ 50%). Common outcomes were also measured differently in terms of improvement from baseline and proportion achieving response/ remission, with little consensus on clinical meaningfulness. cOnclusiOns: The RCTs included in this review reported a broad range of outcomes, making comparison of different therapies a complex task. The disparity in outcomes between pharmacological and non-pharmacological outcomes scales highlights the challenges in designing trials to demonstrate clinical benefit.