Meta-analysiscited by the Tayebat archivefull text checked★ key study A-WHT-03 in-depth review

White rice and type 2 diabetes

White rice consumption and risk of type 2 diabetes: meta-analysis and systematic review.

Overview

Hu EA et al.. BMJ 2012 · PMID 22422870 · DOI 10.1136/bmj.e1454

Each daily serving of white rice was linked to 11% higher diabetes risk; strongest in China/Japan.

Does the study support the specific claim? Partly · Methodological quality: moderate

Original publication

Plain-language summary

Study explained simply

What did the researchers want to find out?: Link white rice and diabetes
How was the study done?: Meta-analysis of 7 cohorts
How many people, animals or samples were studied?: 352,384 adults
What was actually found?: Higher rice intake linked to diabetes
How large was the observed effect?: RR 1.11 per serving/day; RR 1.55 highest vs lowest in Asian cohorts; about 167 extra cases per 100 000 people per year per daily serving (authors' estimate)
What do the results mean in everyday life?: Large daily rice portions may raise diabetes risk
What are the study's limitations?: Only 4 studies; observational; high heterogeneity overall; no protocol; Western results not significant
What does it say about Elawady's specific claim?: Contradicts unlimited rice in the system
Do the results apply to Germany and Europe?: Moderate
What can explicitly not be concluded from this study?: Cannot show causation

Scientific in-depth review

1. Full reference

Authors: Hu EA, Pan A, Malik V, Sun Q
Journal: BMJ
Year: 2012
Identifier: PMID 22422870 DOI 10.1136/bmj.e1454 · PMC3307808
Verification: verified – PubMed E-utilities efetch, 2026-10-25
Retracted: no
Corrections: none
Extent checked: full text
Conflicts of interest and funding: QS supported by NHLBI career development award K99HL098459; funder had no role in design, analysis, writing or submission. ICMJE forms: no support for the submitted work, no relevant financial relationships, no other relationships (full text).

2. Research question

White rice and T2D risk; dose-response

3. Study design and population

Study design: Meta-analysis – Meta-analysis of prospective cohorts
Population: Adults free of self-reported diabetes at baseline in 4 prospective cohort studies (7 cohort analyses): Shanghai Women's Health Study (China), JPHC (Japan; men and women analysed separately), HPFS, NHS, NHS II (USA), Melbourne Collaborative Cohort Study (Australia); ages 26-87 years
Country: China, Japan, USA, Australia
Period: Literature search up to January 2012; cohort follow-up 4 to 22 years
Sample size: 4 articles, 7 cohorts, 352,384 participants, 13,284 cases
Inclusion and exclusion criteria: Included: prospective cohorts, self-reported prevalent diabetes excluded at baseline, RR with 95% CI or SE available/derivable, loss to follow-up <20%. Excluded: animal studies, trials, cross-sectional, case-control, reviews/letters; 1 cohort (Yu et al, Hong Kong) excluded for 31.7% loss to follow-up. 36 full texts assessed, 4 articles included.

4. Methods and endpoints

Exposure / intervention: White rice intake
Comparator: Lowest intake
Endpoints: Incident T2D
Statistical methods: Random-effects pooling of fully adjusted extreme-category RRs (ORs from 2 low-incidence studies treated as RRs); Cochran Q and I2; stratification by ethnicity (Altman method for interaction); Begg funnel plot and Egger test; dose-response by restricted cubic spline (3 knots) and GLST under a fixed-effects model, category midpoints imputed (open top category = cut-off + 25%); 1 serving = 158 g cooked rice; quality score of 15 items (study-own instrument).

5. Actual results

Results: Highest vs lowest, all 7 comparisons: RR 1.27 (95% CI 1.04 to 1.54), I2=72.2% (Q P=0.001). Asian: RR 1.55 (1.20 to 2.01); Western: RR 1.12 (0.94 to 1.33); P for interaction=0.038. Dose-response (fixed effects): per serving/day RR 1.11 (1.08 to 1.14), P linear trend <0.001, no evidence of non-linearity (P=0.51). Women RR 1.46 (1.16 to 1.83) vs men 1.08 (0.87 to 1.34). Excluding Hodge et al: Western 1.17 (0.97 to 1.42), all 1.33 (1.09 to 1.63). Including Yu et al: Asian 1.45 (1.11 to 1.89), total 1.24 (1.03 to 1.50). Egger P=0.30.
Absolute risks: Authors' extrapolation: using a US middle-aged incidence of 15.2 cases/1000, an estimated 167 additional diabetes cases per 100 000 middle-aged people per year for each serving/day increase in white rice (calculated as background incidence x (RR-1)); authors note this may underestimate the risk difference in Asian populations.
Authors' conclusion: Higher white rice consumption is associated with increased T2D risk, especially in Asian populations.

6. Strengths

Large pooled sample (352 384 participants, 13 284 cases) with 4-22 years follow-up
Fully adjusted estimates used; most cohorts adjusted for BMI, activity, family history, energy intake
Duplicate independent screening and data extraction
Wide intake range (Asian 3-4 servings/day vs Western 1-2 servings/week) enabling dose-response analysis
Sensitivity analyses (excluding the low-quality Hodge study; including the high-attrition Yu study) give similar results
Authors provide an absolute-risk translation

7. Weaknesses

Only 4 articles / 7 comparisons; Egger test underpowered
Substantial overall heterogeneity (I2=72.2%); the single overall estimate mixes very different intake ranges
Observational primary studies: residual confounding, especially by socioeconomic status (authors' stated concern)
FFQ-based exposure with moderate validity (r 0.53-0.66); measurement error
Possible inclusion of undiagnosed diabetes at baseline (reverse causation), judged small by authors
Literature search done by one investigator; no registered protocol; list of excluded full texts not provided
Dose-response used fixed effects and imputed category means (assumptions)
Hodge et al did not separate white from brown rice (approx. 95% white) and scored 7/15 on quality
No Middle Eastern/African cohorts; Western stratum not statistically significant; brown rice substitution could not be meta-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: high
Rationale: Strict AMSTAR 2 rating: critically low confidence (critical items 2 and 7 not met, items 4, 9, 11, 15 only partially met), typical for a 2012 review. Underlying primary evidence is observational (confounding by SES, FFQ error). The direction of the Asian association is consistent with later evidence (e.g. PURE, I-GRN-03), so the low AMSTAR 2 rating concerns review conduct and certainty, not necessarily the direction of the finding. Simplified single-reviewer judgement from the full text; a formal AMSTAR 2/ROBIS assessment requires two independent reviewers and the protocol.

8. Risk of bias
Domain	Judgement	Rationale
Item 1 PICO components	low	Exposure (white rice), population (prospective cohorts), outcome (incident T2D) clearly defined.
Item 2 (critical) protocol registered before review	high	No protocol or registration reported.
Item 3 explanation of included designs	low	Restriction to prospective cohorts explained implicitly by aim (incidence, temporality).
Item 4 (critical) comprehensive literature search	some concerns	Medline and Embase plus reference lists; one investigator searched; no grey literature, trial registries or language statement (partial yes).
Items 5-6 duplicate selection and extraction	low	Two investigators independently screened and extracted; discrepancies resolved with senior author.
Item 7 (critical) list of excluded studies with reasons	high	Only counts by exclusion reason (5, 16, 10, 1); no list of excluded full texts.
Item 8 description of included studies	low	Tables 1-2 describe populations, follow-up, FFQ validity, outcome ascertainment, covariates.
Item 9 (critical) risk of bias of individual studies	some concerns	Study-own 15-item quality score (supplementary table A); not a validated RoB tool; confounding not assessed domain-wise (partial yes).
Item 10 funding of included studies	high	Not reported.
Item 11 (critical) appropriate meta-analytic methods	some concerns	Random effects for extreme categories appropriate; overall pooling despite I2=72%; dose-response under fixed effects with imputed category means; ORs treated as RRs (justified by low incidence).
Item 12 impact of RoB on synthesis	low	Sensitivity analysis excluding the low-quality study and including the high-attrition study.
Item 13 (critical) RoB considered in interpretation	low	Residual confounding by SES, measurement error and undiagnosed diabetes discussed explicitly.
Item 14 heterogeneity explained	low	Ethnicity and sex stratification; heterogeneity discussed by intake level.
Item 15 (critical) publication bias	some concerns	Funnel plot and Egger test (P=0.30) done but uninformative with 7 comparisons.
Item 16 conflicts of interest of review	low	Funding and ICMJE declarations reported; funder had no role.

Full-text review (25 October 2026): Added overall pooled RR 1.27 (1.04-1.54, I2 72.2%), sex and sensitivity analyses, the authors' absolute-risk estimate (167/100 000/year per serving/day), detailed inclusion criteria and methods. AMSTAR 2 itemised: no protocol, no excluded-study list, single searcher, own quality score -> overall bias raised from some_concerns to high (critically low confidence by strict AMSTAR 2). Archive derivation refined: 11% and 55% are different estimates.

9. Assessment of the statistical interpretation

Relative risks from observational cohorts. The 11% per serving/day estimate is a fixed-effects dose-response across all populations; the 55% excess applies to highest vs lowest categories in Asian cohorts (approx. 3-4 servings/day vs much lower), not to a single serving. The overall extreme-category estimate (RR 1.27) is heterogeneous (I2 72%). Western stratum CI includes 1. Absolute excess about 167 cases/100 000/year per serving/day under US incidence (authors' model).

10. Relevance to the specific claim

Contradicts C-GRN-05.

11. Independent contrary findings

I-GRN-03 (PURE) confirms a modest association.

12. Transferability to Germany and Europe

European intakes low; relevance moderate.

13. Overall methodological judgement

Moderately consistent evidence that high white rice intake (several servings/day, as in Asian staple diets) is associated with higher T2D risk; weak or uncertain at Western intake levels. Review conduct has methodological gaps by AMSTAR 2 standards; causality not established.

Multidimensional assessment
Multidimensional assessment
Authenticity of the source	verified
Methodological quality	moderate
Risk of bias	high
Precision of results	moderate
Directness	direct
Reproducibility	replicated
Consistency with the overall evidence	consistent
Regional transferability	moderate
Adequacy of the interpretation (by the archive)	taken out of context

14. Open questions

Middle East/Egypt-specific data.

15. Linked fact checks

Rice in all forms (including white rice) is allowed without restriction; 'rice is sugar' and two spoons of rice a day sustain life. – Studies contradicting or limiting the claim
Diets rich in rapidly absorbed carbohydrates (high glycaemic index/load) and poor in cereal fibre raise type 2 diabetes risk by about 50%, and each 10 g/day of cereal fibre lowers risk by 25%; refined flour therefore promotes diabetes. – Studies with mixed or unclear results

Use by the Tayebat archive

Note: according to the archive, the studies were compiled by the archive operators. Whether Dr. Elawady himself cited this study is not established.

What the archive says about the study: Each additional daily serving of white rice raised T2D risk by 11%, up to 55% in Asian populations.
A. What was actually studied?: White rice and T2D.
B. What did the authors find?: Highest vs lowest white rice intake: Asian RR 1.55 (1.20-2.01), Western RR 1.12 (0.94-1.33, not significant); per serving/day RR 1.11 (1.08-1.14) across all populations; overall extreme-category RR 1.27 (1.04-1.54) with high heterogeneity.
C. What conclusions do the authors themselves draw?: Higher white rice intake associated with T2D.
D. What does the archive derive from it?: Cited under the white-flour rule as general support against refined carbohydrates.
E. Is this derivation scientifically justified?: Numbers are correctly reproduced in substance, but the archive's phrasing merges two different estimates: 11% is per daily serving (dose-response, all populations) while 55% is highest vs lowest category in Asian cohorts eating 3-4 servings/day; it is not a ceiling of the per-serving effect. 'Raised' implies causation the observational data cannot establish. The study concerns white rice, not white flour, so it supports the refined-carbohydrate rule only by analogy, and it contradicts the system's unrestricted allowance of rice.
Adequacy of the interpretation (by the archive): taken out of context