Meta-analysisindependent cross-checkfull text checked I-FRU-03 in-depth review
Fruits, vegetables and diabetes
Fruit and vegetable consumption and the risk of type 2 diabetes: a systematic review and dose-response meta-analysis of prospective studies.
Overview
Halvorsen RE et al.. BMJ Nutr Prev Health 2021 · PMID 35028521 · DOI 10.1136/bmjnph-2020-000218
Eating more fruit was linked to about 7% lower diabetes risk. Grapes/raisins and apples looked favourable; cantaloupe and sweetened fruit drinks were linked to higher risk, but in very few studies. 100% fruit juice showed no clear link.
Does the study support the specific claim? Partly · Methodological quality: moderate
Plain-language summary
Study explained simply
- What did the researchers want to find out?
F&V and diabetes
- How was the study done?
Meta-analysis
- How many people, animals or samples were studied?
23 cohorts
- What was actually found?
Weak inverse
- How large was the observed effect?
RR 0.93
- What do the results mean in everyday life?
Whole fruit good
- What are the study's limitations?
Few studies per fruit type, mostly the same US cohorts; no registered protocol; observational
- What does it say about Elawady's specific claim?
Mixed: weak, few-study signal for cantaloupe; favourable for grapes; sweetened fruit drinks (nectars) linked to higher risk
- Do the results apply to Germany and Europe?
Moderate
- What can explicitly not be concluded from this study?
Cause
Scientific in-depth review
1. Full reference
- Authors
- Halvorsen RE et al.
- Journal
- BMJ Nutr Prev Health
- Year
- 2021
- Identifier
- PMID 35028521 DOI 10.1136/bmjnph-2020-000218
- Verification
- verified – PubMed efetch 2026-10-25
- Retracted
- no
- Corrections
- none
- Extent checked
- full text
- Conflicts of interest and funding
No specific grant from any funding agency; authors declare no competing interests (full text).
2. Research question
Fruit/vegetable subtypes and T2D
3. Study design and population
- Study design
- Meta-analysis – Systematic review and dose-response meta-analysis
- Population
Adults free of T2D at baseline in 23 prospective cohort, case-cohort or nested case-control studies (42 publications): 10 from America, 5 Europe, 5 Asia, 3 Australia; follow-up 4-30 years
- Country
Multiple (10 American, 5 European, 5 Asian, 3 Australian studies)
- Period
PubMed and Embase searched to 20 October 2020
- Sample size
23 cohort studies (42 publications); fruit analysis 20 studies, 81,313 cases, 1,542,306 participants; fruit and vegetables combined 10 studies, 29,768 cases, 729,478 participants; subtype analyses mostly 2-6 studies
- Inclusion and exclusion criteria
Prospective cohort, case-cohort or nested case-control studies of fruit/vegetables, subtypes, fruit juices or fruit drinks and T2D; free of T2D at baseline; multivariable-adjusted RRs with CIs; for dose-response, intake levels, cases and person-years. Most-cases publication used for duplicate cohorts; no language restriction; excluded studies listed in supplementary table 2.
4. Methods and endpoints
- Exposure / intervention
Fruit and vegetables combined, total fruit, total vegetables and subtypes (incl. apples, grapes/raisins, bananas, cantaloupe, watermelon, citrus), fruit juice, 100% fruit juice, fruit drinks, potatoes
- Comparator
Lowest vs highest intake category; per 200 g/day (totals), per 100 g/day or 50 g/day (subtypes), per 250 g/day (juices/drinks)
- Endpoints
T2D
- Statistical methods
DerSimonian-Laird random effects; Greenland-Longnecker linear dose-response; restricted cubic splines for non-linearity; I2; subgroup/meta-regression for exposures with >=8 studies; Egger test (P<=0.1) and funnel plots; leave-one-out influence analyses; modified Newcastle-Ottawa Scale (0-8); WCRF evidence grading. No registered protocol.
5. Actual results
- Results
High vs low: fruit and vegetables RR 0.93 (0.89 to 0.98, I2 0%, n=10); fruit 0.93 (0.90 to 0.97, I2 9.3%, n=20); vegetables 0.95 (0.88 to 1.02). Per 200 g/day: F&V 0.98 (0.95 to 1.01); fruit 0.96 (0.92 to 1.00, I2 68.4%); vegetables 0.97 (0.94 to 1.01). Non-linear: 8-12% lower risk at 100-500 g/day fruit. Subtypes (dose-response): apples 0.91 (0.88 to 0.95)/100 g; grapes and raisins 0.74 (0.66 to 0.83)/50 g; blueberries 0.60 (0.49 to 0.73)/50 g; grapefruit 0.90 (0.82 to 0.99)/100 g; cantaloupe 1.18 (1.04 to 1.34)/100 g; watermelon 1.05 (1.00 to 1.11)/100 g (2 studies); bananas 0.93 (0.80 to 1.08); fruit drinks 1.13 (1.01 to 1.27)/250 g; fruit juice 1.08 (1.00 to 1.16)/250 g (I2 68.7%); 100% fruit juice 0.97 (0.91 to 1.03)/250 g (4 studies); potatoes total 1.08 (1.02 to 1.15)/100 g. WCRF grading: fruit 'probable' inverse; F&V limited-suggestive; vegetables limited-no conclusion; all subtypes limited-suggestive or limited-no conclusion.
- Absolute risks
not reported in full text
- Authors' conclusion
Weak inverse association of fruit and vegetable intake with T2D; supports recommendations to increase fruit intake; some subtypes (apples, blueberries, grapefruit, grapes and raisins) may be particularly beneficial while potatoes and fruit juice may increase risk; subtype findings rest on few studies and possible selective reporting and need confirmation.
6. Strengths
- Two databases, no language restriction, duplicate screening and quality assessment, excluded studies listed
- Large number of cohorts and participants for total fruit
- High vs low, linear and non-linear dose-response, subgroup, influence and publication-bias analyses
- Transparent WCRF evidence grading per exposure
7. Weaknesses
- No registered protocol (stated by authors)
- Subtype analyses based on 2-6 studies, many from the same US cohorts (NHS, NHS II, HPFS); possible selective reporting
- Observational: residual confounding, few studies adjusted for other dietary factors; FFQ measurement error; mostly baseline-only diet
- Fruit result attenuated/absent in studies adjusting for waist circumference
- Several results sensitive to single studies (fruit dose-response, fruit drinks, potatoes)
- Funding of included studies not reported; overlap with other fruit meta-analyses (e.g. I-FRU-01) means not independent confirmation
- No Middle Eastern cohorts; dates, figs and mango not analysed
8. Risk of bias
- Orienting tool
- AMSTAR 2
- Type of assessment
- structured assessment based on the full text by a single reviewer (no formal double review)
- Full formal assessment
- no
- Overall risk of bias
- some concerns
- Rationale
Strict AMSTAR 2: low confidence (one critical flaw: no protocol). Overall bias unchanged. Subtype results (cantaloupe, watermelon, grapes) rest on very few, overlapping US cohorts and are graded limited by the authors. Simplified single-reviewer judgement from the full text; formal AMSTAR 2/ROBIS requires two independent reviewers.
| Domain | Judgement | Rationale |
|---|---|---|
| Item 1 PICO | low | Exposures, outcome and designs defined. |
| Item 2 (critical) protocol | high | Authors state there was no registered protocol. |
| Item 3 design selection | low | Prospective designs only, rationale given (recall/selection bias). |
| Item 4 (critical) comprehensive search | some concerns | PubMed and Embase without filters, reference lists, no language restriction; no grey literature (partial yes). |
| Items 5-6 duplicate selection and extraction | low | Screening and selection in duplicate; extraction by one author double-checked by a second. |
| Item 7 (critical) excluded studies | low | Excluded studies listed in supplementary table 2. |
| Item 8 description of included studies | low | Supplementary table 4 describes study characteristics. |
| Item 9 (critical) RoB of individual studies | some concerns | Modified Newcastle-Ottawa Scale in duplicate; not a domain-based RoB tool for confounding (partial yes). |
| Item 10 funding of included studies | high | Not reported. |
| Item 11 (critical) meta-analytic methods | low | Random effects, appropriate dose-response and spline methods; assumptions for open categories stated. |
| Item 12 impact of RoB | low | Subgroup by study quality and by confounder adjustment. |
| Item 13 (critical) RoB in interpretation | low | Authors caution about few studies, selective reporting and confounding; WCRF grading. |
| Item 14 heterogeneity | low | Explored by subgroup and meta-regression. |
| Item 15 (critical) publication bias | low | Egger tests and funnel plots for >=8 studies; sensitivity analyses for outliers; subtype analyses too small to test. |
| Item 16 review COI | low | No funding, no competing interests declared. |
Full-text review (25 October 2026): Added dose-response and subtype estimates incl. watermelon 1.05 (1.00-1.11), grapes/raisins 0.74, 100% fruit juice null 0.97 (0.91-1.03) versus fruit drinks 1.13, WCRF grading (subtypes only limited evidence), no registered protocol, no funding. Bias overall unchanged (some_concerns; AMSTAR 2 low confidence) with itemised domains.
9. Assessment of the statistical interpretation
Small relative risk reductions for total fruit (7% high vs low); linear dose-response for fruit borderline (0.96, 0.92-1.00) with high heterogeneity. Subtype estimates are imprecise and based on few studies. Mixed 'fruit juice' category is borderline (1.08, 1.00-1.16), whereas 100% fruit juice alone is null (0.97, 0.91-1.03) and sweetened fruit drinks positive (1.13, 1.01-1.27).
10. Relevance to the specific claim
Context for C-FRU-01/03/05.
11. Independent contrary findings
12. Transferability to Germany and Europe
Moderate.
13. Overall methodological judgement
Reasonably conducted dose-response meta-analysis. Probable small inverse association of whole fruit with T2D; subtype signals (positive for cantaloupe and fruit drinks, inverse for grapes/raisins and apples) are hypothesis-generating only.
| Multidimensional assessment | |
|---|---|
| Authenticity of the source | verified |
| Methodological quality | moderate |
| Risk of bias | some concerns |
| Precision of results | moderate |
| Directness | direct |
| Reproducibility | partly replicated |
| Consistency with the overall evidence | mixed |
| Regional transferability | moderate |
| Adequacy of the interpretation (by the archive) | not applicable |
14. Open questions
Data for dates, figs, mango and melons in Middle Eastern populations; whether 100% juice differs from sweetened nectars; mediation by adiposity.
15. Linked fact checks
- Watermelon and melon (cantaloupe) should be avoided; the archive's food page supports this with studies showing that a high dietary glycaemic load raises type 2 diabetes risk. – Studies with mixed or unclear results
- Fruit juices, including canned juices and nectars, are allowed or preferred (mango only as juice; on one third-party site fresh juices are forbidden and only canned juices allowed). – Studies contradicting or limiting the claim
- Sweet fruits such as dates, grapes, figs and mango need not be avoided; conventional diets wrongly restrict them and push bland fruit. – Studies supporting the claim