I agree that a larger N would be preferable, but the effect size was enormous, and so the P value was <.001. Therefore, IMO, it's very likely that Vitamin D has a causal effect. Given that Vitamin D has at least some causal effect, it's much more likely that the observational studies that show a correlation are due to causation.