The problem is pretty clear - good data. We need to run randomized clinical trials to tell if drugs work. We’d need the same to make any definitive conclusion

The problem is pretty clear - good data. We need to run randomized clinical trials to tell if drugs work. We’d need the same to make any definitive conclusion about the impact of diet. But the problem is unless say, cancer treatments, where an effect is obvious after months, the impact of diet can take decades to show up. Can you run randomized clinical trials for decades? Not really. So you do non-randomized, observational, retrospective studies that have so many holes in them you question the effort spent doing them. That’s why we’re in the state we are.