Prospective Diagnostic Accuracy Study

In a prospective study, participants are enrolled before test results exist and followed forward in time, ensuring consecutive enrollment and enabling blinding. In a retrospective study, existing records are searched for patients who already received both tests. Retrospective studies are faster and cheaper but are more susceptible to selection bias, verification bias, and incorporation bias, and they typically receive lower quality ratings in systematic reviews.

Not in the traditional sense. Instead of a control group, you need a reference standard — the best available method for establishing true disease status — applied to all enrolled participants. Both test-positive and test-negative participants should receive the reference standard to avoid verification bias. There is no untreated or placebo group as in a randomized trial.

Sample size depends on the target sensitivity or specificity, acceptable confidence interval width, and expected disease prevalence. A common rule of thumb is at least 10 events (true positives) per estimated parameter, but formal power calculations are strongly recommended. For rare conditions, multicenter designs are often necessary to reach the required sample size within a feasible time frame.

STARD (Standards for Reporting Diagnostic Accuracy Studies) is an international reporting guideline — most recently updated in 2015 — that specifies 30 items essential for a transparent report, including a patient flow diagram. Adherence to STARD is required or recommended by major medical journals and is evaluated in systematic reviews using tools such as QUADAS-2. Following STARD improves the completeness and reproducibility of your study report.

No. A diagnostic accuracy study establishes how well a test predicts disease status; it is not designed to assess whether exposure to the test or any intervention causes an outcome. Causal inference requires a different study design, typically a randomized trial or a carefully controlled natural experiment.