longitudinal design is good but where are the spaghetti plots.
In my opinion, compared to numerous studies in the same category, the placebo seems very reasonable, and the dose-dependent effect size seems highly suggestive of a real effect.
I hope additional / larger studies confirm the effect, and if so that a metabolic ward study can help us determine where the effect is taking place! (Food diaries are better than nothing, but notoriously inaccurate.)