Sample sizes are too small, and later studies usually negate the originally observed effect. It also means most headline news about health is useless.