The possibility of unmeasured confounding variables in observational studies: a forgotten fact?