レコードをメール: Conditions for Valid Empirical Estimates of Cancer Overdiagnosis in Randomized Trials and Population Studies