Sample size and power: why “more participants” is not a complete statistical strategy
Sample-size calculations are not just formulas. They are structured arguments about effect size, variability, event rates, dropout, errors and clinical meaning.
The idea in plain language
Power is the probability of rejecting the null under a specified alternative. Type I error is the chance of a false positive under the null. Sample size balances sensitivity, feasibility and participant exposure.
The statistical or technical layer
For simple mean comparisons, sample size increases with variance and decreases as the target effect grows. Event-driven studies depend heavily on event counts rather than participant counts alone.
For a statistician, the useful habit is to ask what generated the number. A rate can move because the underlying process changed, because the denominator changed, because the data cut moved, or because a configuration/model version changed. A probability can be poorly calibrated. A rank can move even when absolute risk barely changes. That is why clinical analytics needs both mathematics and process context.
Where RBQM enters
Planning assumptions such as enrollment, dropout and event accrual should be forecast during conduct so emerging threats are visible early.
Aomics' RBQM architecture deliberately separates detection, precision, prioritization, site context, workflow and governance. The separation matters: it reduces double counting, keeps methods testable and helps the user understand whether a value is a measurement, a statistical signal, a prioritization score or a human decision.
Practical example
A trial assumes 10% missing primary outcomes but several countries trend toward 18%, threatening both operations and information.
The right next step is usually not “act because the dashboard is red.” First verify the source and data cut. Then inspect the denominator, trend, uncertainty and peer context. Ask whether the pattern persists and whether another independent source supports it. Only then move to an investigation hypothesis or an operational proposal.
What I would check
- Is the effect clinically justified?
- Are variance/event assumptions evidence-based?
- How is dropout handled?
- Which assumptions will be monitored?
- Can I reproduce the result from the same data cut and configuration?
- Has the function that owns the underlying process reviewed the evidence?
The mistake I see most often
Choosing an optimistic effect size simply to make the trial affordable.
The broader lesson is to match the statistical tool to the decision question. A ranking helps answer “where should we look first?” It does not answer “what caused the problem?” A correlation can suggest a relationship; it does not establish responsibility. A forecast creates time to investigate; it does not make the future certain.
Aomics resources
Clinical data science and biometricsAdaptrials
Adaptive trial design and analyticsRBQM-ai
Clinical quality intelligence and monitoring
Aomics GmbH provides clinical data science, biostatistics, statistical programming and related biometrics services. Adaptrials is relevant to prospective study design and adaptive analytics. RBQM-ai focuses on Quality by Design, centralized monitoring, site oversight, QTL surveillance, predictive risk, signal-to-action and audit-ready evidence.
Frequently asked questions
What is the practical point of sample size power clinical trials?
Power is the probability of rejecting the null under a specified alternative. Type I error is the chance of a false positive under the null. Sample size balances sensitivity, feasibility and participant exposure.
Where does this sit in RBQM-ai?
This topic connects mainly to M01, M16, M11. The platform keeps the analytical responsibility separate from human decision authority.
Can the software make the final decision?
No. The system provides evidence and prioritization. Qualified study, medical, operational, safety, quality or regulatory professionals remain responsible for consequential decisions.
Scientific and regulatory references
- ICH E9 Statistical Principles for Clinical Trials
- ICH E8(R1) General Considerations for Clinical Studies
Editorial note: Educational content only. Always check the current study protocol, SAP, monitoring plan, SOPs and applicable official regulator/guideline sources.