Sample size calculations for comparative studies of medical tests for detecting presence of disease.

摘要: Technologic advances give rise to new tests for detecting disease in many fields, including cancer and sexually transmitted disease. Before a screening test is approved public use, its accuracy should be shown better than or at least not inferior an existing test. Standards do yet exist designing analysing studies address this issue. Established principles the design of therapeutic can adapted tests. In particular, drawing upon methods superiority non-inferiority agents, we propose that confidence intervals relative dichotomous drive comparative We derive sample size formulae variety designs, where patients undergo several receive only one under evaluation. Both cohort case-control study designs are considered. Modifications discussed accommodate where, because invasive nature definitive testing, true status obtained subjects who positive on more The proposed applied comparing modified pap conventional cervical screening. impact error gold standard reference evaluation also discussed.