There are many controlled intervention studies into diet and exercise. It is done. They provide the only results worth any consideration. More often than not regression analyses and "risk-adjustment" models are invalidated.