Potential pitfalls of meta-analyses of observational studies in cardiovascular research