From the cognitive health one:

This study has several limitations. The predominantly male, overweight-to-obese sample may limit generalizability to broader populations, including women and individuals with lower baseline adiposity. Although initial fat loss occurred under randomized conditions within the original trials, the 5–16-year follow-up did not involve random assignment to post-trial behavior patterns, limiting causal inference. Dietary and lifestyle patterns at follow-up were systematically assessed, and the MED diet adherence score was included in all multivariable models to account for post-trial behaviors. This adjustment also addresses the possibility that participants who responded well to the intervention were inherently more likely to maintain healthy dietary patterns over time, which may have independently supported brain health. ... Additionally, while follow-up MRI scans were analyzed for participants across all four trials, baseline and end-of-intervention abdominal adipose tissue scans were available only from CENTRAL and DIRECT-PLUS trials, and brain MRI scans were available only from the DIRECT-PLUS trial. Consequently, long-term associations between adipose depots and brain outcomes were **limited to a 5- and 10-year follow-up for adipose depots and a 5-year follow-up for brain MRI. Importantly, cognitive function was not systematically measured at pre- or post-intervention time points, limiting the ability to assess individual-level cognitive changes over time. **A further limitation is the absence of quantitative MRI measures of ectopic fat in organs such as the liver and pancreas, which are recognized as relevant to cardiometabolic and brain health. ... [Emphasis mine]

The MED score mentioned is 14 questions including how many tablespoons of oil you consume in a week. Self reported data about diet covering a month would barely be reliable let alone the 5 or ten years, but it's brushed off as being good enough to lean towards causation rather than correlation.