OBJECTIVES: Patients, physicians, and other decision-makers make implicit trade-offs among benefits and risks of different treatments. Many methods have been proposed to conduct quantitative benefit-risk analysis (BRA). We propose a framework for classifying BRA methods based on factors that matter most to patients. Using common mathematical notation, we compare the methods using a hypothetical example.  METHODS: We classified available BRA methods into three categories: (1) un-weighted metrics, that use only probabilities of benefits and risks (e.g., number needed to treat and number needed to harm [NNT|NNH]); (2) metrics that incorporate preference weights to account for the impact and duration of outcomes (e.g., Maximum Acceptable Risk [MAR], relative value-adjusted life-years [RVALYs], quality-adjusted life-years [QALYs]); and (3) metrics that incorporate ad hocweights based on decision makers’ opinions (e.g., Multi-criteria Decision Analysis, Benefit-Less-Risk Analysis).  We used two hypothetical antiplatelet drugs (A and B), probabilities of benefits (reduction in myocardial infarction and stroke) and harms (increases in major and minor bleeding) based on randomized trial data, and preference weights from the literature to compare the BRA methods within the proposed framework. RESULTS: Use of the framework and notation revealed BRA methods share substantial commonality. In the example, BRA using NNT|NNH indicated that -1.3% of patients would experience net benefit with drug A versus B, (an unfavorable benefit-risk balance for A). In contrast, 4.6% of patients would experience a net benefit with drug A if weighted using MAR. BRA using RVALYs and QALYs suggested gains of 3.8 RVALYs and 5.4 QALYs per 100 patient-years, respectively, with drug A versus B.   CONCLUSIONS: The proposed framework provides a unified, patient-centered approach to BRA methods classification. All methods impose trade-offs between probabilities of benefits and risks. The weights used in the metrics is a key differentiating feature and can lead to quantitatively and qualitatively different results.