Diets
mediterranean diet reduces-incidence-of type 2 diabetes
In plain terms: Does eating a Mediterranean diet lower your chance of developing type 2 diabetes?
Strong support Diets 💰 Industry COI noted
Part of: 🥗 mediterranean diet
RefutedContestedStrong support
consensus score 0.93
Yes — the landmark PREDIMED RCT showed a ~30% reduction and multiple cohort meta-analyses agree, making this one of the better-evidenced dietary-prevention claims.
📅 Last reviewed: 2026-07-15 ⓘ
Evidence ladder
How far up the ladder this claim has climbed. A high consensus on a low rung means "consistent so far," not "proven in people."
Top evidence so far: All trials, pooled (Meta-analysis)
MechanismIn-vitroAnimalObservationalRCTMeta-analysis
How the studies fall
8 support 0 contradict 0 tested null 1 mixed · 9 sources, 8 independent groups · 1 superseded — pooled inside a review, counted once
The evidence (13)
| Source | Grade | Stance | Quality | Finding |
|---|---|---|---|---|
| Toi 2020 · Nutrients | meta-analysis | supports | low | Umbrella review of 60 systematic reviews with meta-analyses; it pooled nothing of its own — 'As a limitation of the study, we did not consider the primary individual included studies—instead, the results were considered only from SRMAs.' On the Mediterranean diet it restates four cohort SRMAs, all protective: 'In contrast, Mediterranean ( n = 4) and DASH diet showed protective effects on T2DM with the pooled RRs ranging from 0.77 (95% CI: 0.66, 0.89) to 0.87 (95% CI: 0.82, 0.93)' (Supplementary Table S4: Koloverou 2014 RR 0.77 [0.66-0.89], Esposito 2014 0.80 [0.68-0.93], Schwingshackl 2015 0.81 [0.73-0.90], Jannasch 2017 0.87 [0.82-0.93]). Direction supports the claim, but this is a second-order restatement of meta-analyses the vault cites directly, and the paper's own quality audit undercuts its weight: 'Confidence in the findings of included SRMAs are critically low, low, moderate, and high at 81.7%, 15.0%, 1.7%, and 1.7%, respectively', dietary-pattern overlap was CCA = 9.7% so 'the included SRMAs could not much add in new evidence', and the Schwingshackl 2015 Mediterranean meta-analysis is flagged for excess significant findings (O = 9 vs E = 6, p = 0.000005587), the authors warning that 'meta-analyses among this group might overestimate the real effect size of the intervention.' |
| Hardy DS et al 2026 · Hum Genomics | observational | supports | moderate | Longitudinal cohort 1976-2015 (n=8283 EA, 1205 AA); MedDiet assoc. w/ ~9% lower incident T2DM risk in AA, ~5% in EA (p<0.025), interacting w/ PRS/metabolic burden. |
| Martinez-Gonzalez 2023 · Cardiovasc Diabetol | observational | supports | moderate | PREDIMED cohort (n=3541 free of T2D, 273 blindly adjudicated incident cases, median 4.1 y). NOT the randomized contrast: exposure is attained adherence (cumulative-average MEDAS, 8 yearly measures) in time-varying Cox models, with the randomized arm entered only as a covariate — hence graded observational. HR 0.80 (0.70–0.92) per +2 MEDAS points, p-trend 0.001; 0.46 (0.25–0.83) for 12–14 vs <8 points; incidence 27.8 → 13.8 per 1000 person-years across adherence categories. Robust to adjustment for randomized arm (HR 0.82, 0.71–0.96) and corroborated by a 67-metabolite plasma signature (HR 0.73, 0.54–0.99 per 2 points, n=889). Caveats: the association vanishes using baseline MEDAS alone (HR 0.95, 0.84–1.08, p=0.44), so the signal rests entirely on post-randomization attained adherence; authors concede residual confounding; high-CV-risk Spanish population; olive oil and nuts donated by industry bodies that also pay several authors. |
| Galbete 2018 · Eur J Epidemiol | meta-analysis | supports | low | Umbrella review of 4 cohort meta-analyses (11 primary studies) on MedDiet and T2D: all inverse on high-vs-low adherence — 'an inverse association between 13% (RR 0.87, 95% CI 0.82, 0.97) and 23% (RR 0.77, 95% CI 0.66, 0.89) for the risk of T2D was observed'. Credibility is the caveat: NutriGrade Low for 3 of the 4 T2D meta-analyses (Moderate for 1), and every T2D 95% prediction interval touches or crosses the null (0.77-1.00; 0.61-1.11; 0.62-1.15; 0.50-1.26) — 'the null value was excluded only in some of the associations', T2D not among them. The authors' own re-pooled subgroup analyses found the index choice immaterial ('Test for subgroup differences (tMedDiet vs aMedDiet): p = 0.83') but the aMedDiet arms were themselves non-significant (0.88 [0.75, 1.04]). No RCTs by design: 'we targeted meta-analyses on prospective observational studies and thus, no randomized controlled trials were included.' |
| Zeraattalab-Motlagh 2022 · Eur J Nutr | meta-analysis | supports | moderate | Dose-response meta-analysis of 14 prospective cohorts (410,303 participants, 41,466 incident T2D cases; PROSPERO CRD42021246589): highest vs lowest MedDiet adherence RR 0.79 (95%CI 0.72-0.88), risk difference -21 per 1000 persons (95%CI -28, -12), GRADE moderate certainty; per 2-point adherence increment RR 0.86 (95%CI 0.82-0.91, n=13), inverse linear. Heterogeneity high (I2 = 82%); all input studies observational. |
| Galbete 2018 · BMC Med | observational | supports | high | EPIC-Potsdam prospective cohort, n=23,411, 1376 medically verified incident T2D cases over 246,219 person-years (mean 10.5 y). Both Mediterranean scores were inversely associated with T2D: 'Participants in the highest tertile of adherence to the MedPyr presented 20% lower risk of T2D in comparison with those in the lowest tertile (HR T3 vs. T1 0.80, 95% CI 0.70-0.92; HR per 1 SD 0.92, 95% CI 0.87-0.97).' For the tMDS, 'HR T3 vs. T1 0.84, 95% CI 0.73-0.97; HR per 1SD 0.93, 95% CI 0.88-0.98.' Effect is modest (~7-8% per SD) and partly alcohol-driven: omitting the alcohol component attenuated tMDS to 0.95 (0.90-1.01) and MedPyr to 0.94 (0.89-0.99). Single baseline FFQ, non-Mediterranean population; publicly funded, no competing interests. |
| Brlek A, Gregorič M 2023 · Br J Nutr 2023 Aug 28;130(4):709-718 (e-pub 2022 Nov 25) | observational # was meta-analysis — umbrella pools nothing (reread #399, precedent k safety net under the review role) | supports | low | Umbrella review of SEVEN SYSTEMATIC REVIEWS, not a meta-analysis: 'The inclusion criteria were met by seven systematic reviews, entirely based on prospective cohort studies and reviewing five different diet indices - alternate healthy eating index (AHEI), dietary approaches to stop hypertension (DASH), dietary inflammatory index (DII), healthy eating index (HEI) and Mediterranean diet (MedDiet).' It pools no primary data of its own — 'Our aim was to systematically synthesise data that was previously summarised separately for each diet index in one umbrella review' — and reports no pooled RR, CI or heterogeneity statistic anywhere in the retrievable text; PubMed/Europe PMC type it a Systematic Review, not a Meta-Analysis. DIRECTION SUPPORTS: 'All seven included systematic reviews showed that greater adherence to these diet indices reduces the risks of all-cause mortality, CVD incidence and mortality and type 2 diabetes mellitus incidence.' THE LIMIT IS ARM-SPECIFIC AND WAS PREVIOUSLY MIS-STATED: NutriGrade strength of evidence was 'Moderate meta-evidence was presented for AHEI and DASH for all outcomes, also for DII for all-cause mortality, CVD mortality and incidence, MedDiet for all-cause mortality and for HEI for CVD incidence and mortality' — MedDiet reached moderate meta-evidence for MORTALITY, not for T2D incidence, so the previous extract's 'type 2 diabetes with moderate meta-evidence' attached a rating earned on a different outcome. The paper then omits MedDiet from its own endorsement: 'Our umbrella review provides further evidence for AHEI, DASH, DII and HEI diet indices to be used as predictors of selected health outcomes.' Hence evidence_role: review (second-order restatement of reviews the vault can cite directly — and the third overview layer on this claim, with s30030684 and s32899917) and quality: low (MedDiet-T2D fell below moderate on the paper's own NutriGrade scale; underlying evidence entirely observational prospective cohorts). Abstract-grade: not open access, inEPMC N, no full text or supplement on this box; the included reviews are not named, so the specific overlap with this claim's other 12 sources cannot be enumerated. |
| Ruiz-Canela 2025 · Ann Intern Med | RCT | supports | high | PREDIMED-Plus (ISRCTN89898870), n=4746 Spanish adults 55-75 with metabolic syndrome and overweight/obesity: 'Participants were randomly assigned 1:1 to an intervention group receiving an erMedDiet (planned reduction of 600 kcal per day), increased physical activity, and behavioral strategies for reducing weight, or a control group receiving ad libitum MedDiet advice.' Over a median 6 years, 'diabetes incidence was 31% (CI, 18% to 41%) relatively lower in the intervention group' (9.5%, 280 cases vs 12.0%, 349 cases). Both randomized arms received a Mediterranean diet, so the tested exposure is the ADDED energy restriction, activity and weight loss - not the Mediterranean diet itself; off-scope for a MedDiet-versus-usual-diet claim. Prespecified secondary outcome, single-blinded, self-reported adherence; a correction was published (Ann Intern Med 2026;179:467-468). |
| Sarsangi 2022 · Adv Nutr | meta-analysis | supports | moderate | Systematic review + dose-response meta-analysis of 16 prospective cohorts (17 effect sizes), searched to January 2022; PROSPERO CRD42021265332. Greatest vs lowest MedDiet adherence: pooled RR 0.83 (95% CI 0.77-0.90), I2 = 79%. Linear dose-response: each 1-point MedDiet-score increase HR 0.97 (95% CI 0.96-0.98, P < 0.001), i.e. ~3% lower risk per point; nonlinear relation also present (P-nonlinearity = 0.001), authors conclude 'even modest adherence to the MD was linked to a decreased incidence of type 2 diabetes'. All pooled inputs are observational cohorts; heterogeneity substantial (I2 = 79%) and, with the full text paywalled, the risk-of-bias, publication-bias and certainty-of-evidence assessments are unverified - hence quality moderate, not high. |
| Sobiecki JG, Imamura F, Davis CR, Sharp SJ, Koulman A, Wareham NJ, Forouhi NG, et al. 2023 · PLoS Med | observational | supports | high | EPIC-InterAct case-cohort (22,202 participants, 9,453 incident T2D cases, 8 European countries, mean 9.7 y follow-up) using a 23-biomarker carotenoid/fatty-acid score of Mediterranean-diet adherence derived in the MedLey partial-feeding RCT: 'The score was inversely associated with incident T2D in EPIC-InterAct: the hazard ratio (HR) per standard deviation of the score was 0.71 (95% CI: 0.65 to 0.77) following adjustment for sociodemographic, lifestyle and medical factors, and adiposity.' Top vs bottom fifth HR 0.38 (0.30 to 0.50), inverse in all countries, I2 = 67%, 95% prediction interval 0.55 to 0.91. Grade stays observational despite 'randomised controlled trial' in the title: the MedLey RCT only derived the classifier (C-statistic 0.88 for diet arm) and never measured T2D. Two honest caveats, which is why this is a supports and not a proof: the authors state the biomarkers 'may have represented a metabolic response to a healthy diet, rather than a specific biomarker profile indicative of adherence to the Mediterranean diet', and the directly on-subject self-reported Mediterranean score in the same participants gives a much weaker HR 0.90 (0.86 to 0.95) per SD. |
| Wallerer 2025 · Adv Nutr | meta-analysis | supports | high | Updated SR + dose-response meta-analysis, 24 prospective cohorts + 1 RCT, 991,878 participants and 68,325 incident T2D cases, mean follow-up 12.2 y. Cohorts: 'Each 2-point increment in the MedDiet adherence score was likely associated with a decrease in T2D risk by 8%' (HR 0.92; 95% CI 0.90, 0.94; MODERATE GRADE certainty; 95% prediction interval 0.85 to 1.01). Dose-response: HR 0.79 (0.73, 0.85) at score 8 vs 1. The single RCT (PREDIMED) is congruent but imprecise and LOW certainty: HR 0.75 (95% CI 0.56, 1.01); the authors did not pool it with the cohorts. Robust to excluding the 9 high-RoB cohorts and to EPIC overlap; Egger P = 0.55; no funding source. |
| Eichelmann 2024 · Nat Med | observational | mixed | moderate | PREDIMED olive-oil MedDiet reduced diabetes incidence chiefly in those with unfavorable baseline lipid profiles, indicating a conditional rather than uniform effect. |
| Rossi 2013 · Diabetologia | observational | supports | moderate | Greek EPIC cohort, 22,295 participants and 2,330 incident cases over a median 11.34 years: 'A higher MDS was inversely associated with diabetes risk (HR 0.88 [95% CI 0.78, 0.99] for MDS ≥ 6 vs MDS ≤ 3).' Small effect whose CI reaches 0.99; exposure is a single baseline FFQ. ↩ SUPERSEDED — pooled in the review above, counted once |
Disagree, or know a study we missed?
We grade by evidence, not opinions. The way to weigh in is to point us to a study we haven't cited (check the evidence table above first), or to flag a problem with one we have. Every submission is reviewed; if it holds up, the grade updates and shows in Science Changes Its Mind.
Opens a short form. You'll sign in with Google so submissions are tied to a real account — we don't display your identity, and we only accept a link we can verify (PubMed, DOI, ClinicalTrials.gov).
Educational only, not medical advice. Grades and scores reflect published evidence weighted by study design and quality; see the methodology.