What are the recommended sample size for biomarker analysis

Study Tier Sample Size (Smallest Group) Statistical Reliability
Exploratory Pilot 15–20 High risk of over-optimism. ROC curves may look great but won’t replicate in a new cohort.
Standard Discovery 30–40 AUC values begin to stabilize and the margin of error (Confidence Interval) narrows significantly.
Robust Validation 80–100+ Necessary for clinical-grade biomarkers. Allows for complex machine learning without overfitting.