Bisphosphonates and mortality: confounding in observational studies?