Effect size benchmark values for quantifying benefit/risk of medicinal products
This method provides standardized benchmark values for measuring the size of treatment effects in medicinal products, using the Mann-Whitney effect size (also called relative effect or stochastic superiority) as a robust, distribution-free measure. It allows comparison of benefits and risks across different types of data and statistical models, translating well-known benchmarks (like Cohen’s small, medium, large effects) into universally applicable values.
At a glance
Use when
Comparing treatment effects across diverse data types in drug evaluation, especially when distributional assumptions are uncertain or violated; in regulatory benefit-risk assessments requiring standardized, interpretable metrics.
Avoid when
When only simple descriptive statistics are needed; when stakeholders lack statistical expertise to interpret stochastic superiority; when effect heterogeneity is extreme and summary measures are misleading.
Inputs
Effect size estimates from various statistical models (e.g., mean differences, odds ratios, hazard ratios), confidence intervals, distribution type (normal, proportional odds, etc.), and context of medicinal product evaluation.
Outputs
Standardized effect size benchmarks (small, medium, large) expressed as Mann-Whitney U values or relative effects, with interpretation across distribution families and relevance assessment using confidence intervals.
How it works
The method introduces the relative effect (Mann-Whitney effect size) as a robust, nonparametric measure of stochastic superiority, applicable across distribution families including normal, proportional odds, and proportional hazards models. It derives global effect measures such as risk difference averages and extremums, and links binary data from 2×2 tables to the Mann-Whitney framework. Benchmark values based on Cohen’s effect sizes are translated into corresponding Mann-Whitney values, enabling standardized interpretation. Confidence intervals are incorporated to assess relevance and uncertainty in practical applications.
- HTA domains
- Clinical Effectiveness, Safety, Patient and Social Aspects
- Assumptions
- The method assumes that the concept of stochastic superiority (probability that a randomly selected individual from one group has a better outcome than one from another) is meaningful for the outcome of interest; it does not assume specific distributional forms, making it robust and widely applicable.
- Strengths
- Distribution-free and robust, requiring no parametric assumptions,Enables cross-distribution comparison of effect sizes,Translates familiar Cohen-like benchmarks into universal values,Integrates binary and continuous outcomes under one framework,Supports confirmatory analyses in regulatory and clinical contexts
- Limitations
- Interpretation may be less intuitive than mean differences for some audiences,Requires understanding of nonparametric statistics,Benchmark values are derived theoretically and may need contextual calibration
- Also known as
- Mann-Whitney effect size, Relative effect, Stochastic superiority, Effect size benchmarks for medicinal products
Questions this answers
- › How can treatment effects be measured in a way that is comparable across different types of data and distributions?
- › What are meaningful benchmark values for small, medium, and large effects in benefit-risk assessment of drugs?
- › How can Cohen's effect size benchmarks be translated into nonparametric, distribution-free measures?
- › How are binary outcome measures related to continuous effect size frameworks?
- › How can confidence intervals be used to interpret the relevance of an observed effect size?
- › What effect size measures are suitable for confirmatory, assumption-free statistical analyses?
References & sources
Similar by meaning
Beta record. Generated from the primary source via AI extraction and independent audit, pending final human review.