This case study aims to investigate the difficulties that may be encountered when undertaking a benefit-risk assessment for an older medicinal product with well-established use. To assess the difficulties of undertaking a benefit-risk assessment for an older medicine, we appliedthe BRAT framework to a case study assessing the benefit-risk balance of warfarin for the treatment of atrial fibrillation. The framework ensured that the process was documented, and that the discussions were focused on outcomes relevant to the BR problem. One of the biggest challenges identified related to the large variety of data sources, a result of this was very broadly defined benefit-risk criteria, which can make it difficult to elicit preference values.

The assessment was supplemented with an analysis using the quantitative benefit-risk method SMAA. The SMAA method made it possible to include uncertainty related to date variation in the assessment, and it proved especially useful in exploring the effect of different preference values, which was particular relevant for this case study where criteria were broadly defined, and with problems of poor preference comparable criteria.

Using data from individual patients, a NCB/individual harm-benefit model was created.With this model, differences in the benefit risk balance of warfarin for stroke prevention in patients with atrial fibrillation were investigated. The benefit-risk balance was stratified based on baseline risk of stroke and bleeds. With the model it was possible to identify groups with higher and lower possibility of a favourable benefit-risk with warfarin.

Overview of warfarin case study
Drug of interest: Warfarin
Indication: Primary prevention of stroke in patients with atrial fibrillation
Severe side effect: Haemorrhage stroke
Regulatory history: 1954 Approved
Data source: EPARs, published literature, GPRD data
Comparators: Placebo, Aprixaban, Riveroxaban, Dabigatran
Questions addressed:

Based on clinical trial data are the BR balance of warfarin versus placebo deemed positive?

Does inclusion of observational data change the BR of warfarin?

Based on data from newer RCTs are the BR of warfarin deemed positive against Rivaroxaban, Dabigatran and Apixaban?

Does data from newer RCTs change our beliefs about the benefit-risk of warfarin versus placebo?

Perspectives: Regulator
Reasons for selection: This case study, undertaking a BR assessment for an older medicine was conducted to identify which specific challenges might exist for medicines in this category
Methods used: BRAT, SMAA, NCB/individual harm-benefit method

The benefit risk assessment of warfarin was carried out in three stages, with increasing complexity, work stream 1, 2 and 3.

Work stream 1 (WS1):The benefit-risk balance of warfarin versus placebo was assessed based on data from 5 older randomised clinical trials (RCT). Data from observational studies were used to evaluate if the benefit-risk balances based on RCTs could be considered valid in actual practice. For the benefit-risk question in WS1 the descriptive benefit-risk methodology BRAT and the quantitative methodology SMAA were used.

Work stream 2 (WS2): Changes in the benefit-risk profile over time was considered by using data for three recent clinical trial programmes for a new class of anticoagulant. First by assessing whether the benefit-risk balance of warfarin versus the newer products was positive, and then assessing whether the warfarin clinical trial data from these new trials were compatible with data in the first assessment. For this assessment BRAT was used.

Work stream 3 (WS3:) The final part of the assessment used individual patient level data to identify the benefit-risk profile for the product based on patient demographics. In WS3 the quantitative BR method net clinical benefit (NCB) was used.