A second-class science? A defence of observational epidemiology to make causal inferences.