Email (record): Prevalent cases in observational studies of cancer survival: do they bias hazard ratio estimates?