← All chapters·Chapter 05 · free opening
What holds up
The complete list of dietary findings that survive review, with populations, effect sizes and the caveats their own authors wrote.
This is the shortest chapter in the publication. That is the finding, not an apology for it. Chapter 1 gives the context: 2% of the meta-analyses in this field reached high certainty. A long "what works" chapter would have to be manufactured.
Everything below is dietary. The three larger levers are in chapter 2, and they outrank every item here.
The one A-grade dietary finding
A A Mediterranean-style pattern lowers CRP and hs-CRP in adults with at least one chronic condition: −0.37 … −1.04 mg/L, certainty high-to-low across outcomes. Three independent syntheses with different methods, different databases and different populations agree — an umbrella review of 30 reviews (PMID 40657695), a second umbrella review using a different certainty framework (PMID 39364652), and a 2026 meta-analysis of 33 randomised trials, n = 3,476 (PMID 41211687).
B The same pattern lowers IL-6: −0.38 … −1.07 pg/mL (PMID 40657695, PMID 41211687).
B It raises adiponectin, +0.59 … +1.69 µg/mL, at low certainty (PMID 40657695).
It does not lower TNF-α (PMID 41211687, PMID 40657695).
B In the randomised-trial synthesis the effect appeared on hs-CRP but not on standard CRP (PMID 41211687). Anyone reporting "lowers CRP" without that qualifier has lost information that changes what a reader would do.
Boundaries, all three of them explicit. The population is adults with an existing condition. The effect is largest where baseline markers are highest. Trials ran for weeks to months, not days.