Problem of immortal time bias in cohort studies: example using statins for preventing progression of diabetes | Synapse