Again, inferences of causation can be sanely derived from studies with large sample sizes and statistical heterogenity.

But fat loss studies are hard to run with thousands of people (esp if you have to control other factors)