The secret of immortal time bias in epidemiologic studies.