If is very, very, very difficult to control for all these factors. Apart from the data dredging problem, the difficulty of controlling for all confounding facto

If is very, very, very difficult to control for all these factors. Apart from the data dredging problem, the difficulty of controlling for all confounding factors is what make so few association studies hold up when prospectively tested. As a rough rule of thumb if the hazard ratio is not 3 or greater in an association study then it almost certainly won’t be replicated prospectively.