Meta-analysisindependent cross-checkfull text checked I-FRU-02 in-depth review
Sugary drinks, juice and diabetes
Consumption of sugar sweetened beverages, artificially sweetened beverages, and fruit juice and incidence of type 2 diabetes: systematic review, meta-analysis, and estimation of population attributable fraction.
Overview
Imamura F et al.. BMJ 2015 · PMID 26199070 · DOI 10.1136/bmj.h3576
Sugary drinks were linked to more type 2 diabetes. Fruit juice showed at most a small, unstable link (low-quality evidence), so juice is not a proven healthy alternative.
Does the study support the specific claim? No · Methodological quality: high
Plain-language summary
Study explained simply
- What did the researchers want to find out?
Beverages and diabetes
- How was the study done?
Meta-analysis
- How many people, animals or samples were studied?
17 cohorts
- What was actually found?
Sugary drinks +13% per serving/day after adiposity adjustment; juice +7%, but null in studies with objective diagnosis
- How large was the observed effect?
7% per serving/day
- What do the results mean in everyday life?
Juice not health drink
- What are the study's limitations?
Observational; juice result unstable and rated low quality by the authors
- What does it say about Elawady's specific claim?
Limits juice allowance
- Do the results apply to Germany and Europe?
High
- What can explicitly not be concluded from this study?
Cannot show that juice causes diabetes, only that it does not appear protective
Scientific in-depth review
1. Full reference
- Authors
- Imamura F et al.
- Journal
- BMJ
- Year
- 2015
- Identifier
- PMID 26199070 DOI 10.1136/bmj.h3576
- Verification
- verified – PubMed efetch 2026-10-25
- Retracted
- no
- Corrections
- none
- Extent checked
- full text
- Conflicts of interest and funding
Funded by MRC Epidemiology Unit core support (MC_UU_12015/5); one author supported by an AHA fellowship; sponsor had no role. ICMJE: no support for the submitted work, no relevant financial relationships. No included study or publication was industry-funded.
2. Research question
Beverages and T2D
3. Study design and population
- Study design
- Meta-analysis – Systematic review and random-effects meta-analysis of prospective cohorts, with population attributable fraction (PAF) estimation in US and UK national surveys
- Population
Adults aged >=18 years free of diabetes in 17 prospective cohorts (USA, Finland, Japan, France, Singapore, 8 EPIC countries); PAF in NHANES 2009-10 (n=4729, representing 189.1 million) and UK NDNS 2008-12 (n=1932, representing 44.7 million)
- Country
USA (majority), Finland, Japan, Singapore, France, 8 European countries (EPIC-InterAct); PAF for USA and UK
- Period
Searches 31 May 2013, updated 10 February 2014; cohort baselines 1966-2002; median follow-up 3.4-21.1 years
- Sample size
17 cohorts; 38,253 cases over 10,126,754 person-years; fruit juice analysed in fewer cohorts
- Inclusion and exclusion criteria
Prospective design, beverage intake and incident T2D, adults >=18 years free of diabetes, mean follow-up >=2 years; no time or language restriction. 1937 records, 33 full texts, 21 articles of 16 cohorts plus unpublished data from 1 cohort.
4. Methods and endpoints
- Exposure / intervention
Sugar-sweetened beverages (incl. sugar-sweetened fruit juice/drinks), artificially sweetened beverages, 100% fruit juice (or juice assessed separately from fruit drinks), per serving/day (median 250 mL)
- Comparator
Per one serving/day lower intake (dose-response)
- Endpoints
T2D
- Statistical methods
Random-effects meta-analysis of most-adjusted RRs with and without adiposity adjustment and crude; calibration for within-person dietary variation and self-reported diabetes misclassification; meta-regression; Egger test, contour-enhanced funnel plots, trim-and-fill; influence and sensitivity analyses; simulation of residual confounding by adiposity; GRADE; PAF by Cochrane algorithm assuming causality.
5. Actual results
- Results
Per serving/day, before / after adiposity adjustment: SSB RR 1.18 (1.09-1.28; I2 89%) / 1.13 (1.06-1.21; I2 79%); ASB 1.25 (1.18-1.33) / 1.08 (1.02-1.15), after publication-bias correction 1.22 (0.98-1.52); fruit juice 1.05 (0.99-1.11; I2 58%) / 1.07 (1.01-1.14; I2 51%), crude 0.97 (0.90-1.06). Fruit juice null in studies with objective T2D ascertainment (RR 1.08, 0.97-1.20; P heterogeneity 0.008) and in studies with repeated diet measures (0.98, 0.86-1.11); significance lost after excluding any single supporting study; Egger P=0.03. GRADE: SSB moderate; ASB and fruit juice low quality.
- Absolute risks
10-year absolute T2D event rates 11.0% (USA, 20.9 million events) and 5.8% (UK, 2.6 million). Assuming causality and adiposity-independent effect: 1.8 million US events attributable to SSB (PAF 8.7%, 3.9-12.9%) and 79,000 UK events (PAF 3.6%, 1.7-5.6%); adiposity-unadjusted: 2.6 million (11.9%, 7.4-16.5%) and 126,000 (4.9%, 3.0-7.2%). No absolute estimates for fruit juice.
- Authors' conclusion
SSB consumption was associated with T2D independently of adiposity; positive associations for ASB and fruit juice likely involve bias, yet neither is a suitable healthy alternative to SSB; under the assumption of causality SSB may relate to a substantial number of new diabetes cases.
6. Strengths
- Four databases, no language restriction, duplicate full-text review and extraction
- Bias domains assessed per Cochrane non-randomised tool; GRADE rating per beverage
- Estimates with and without adiposity adjustment and crude estimates reported
- Calibration for measurement error, extensive sensitivity, influence and publication-bias analyses
- PAF translated into absolute numbers for USA and UK
7. Weaknesses
- Observational cohorts: residual confounding and reverse causation (health-conscious or high-risk people changing beverages)
- Protocol not registered; title/abstract screening by one author
- High heterogeneity for SSB and ASB
- Fruit juice association unstable: null with objective outcome ascertainment or repeated diet measures, crude RR 0.97, Egger P=0.03, lost in influence analysis; graded low quality
- Dose-response partly derived from categorical estimates and ORs
- Few Asian studies, none from the Middle East/Africa; PAF assumes causality
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
Review conduct is good (approximately moderate confidence by AMSTAR 2; unregistered protocol, single-author title screening). The remaining concern is the observational primary evidence, which the authors grade moderate for SSB and low for ASB and fruit juice. Simplified single-reviewer judgement from the full text; a formal AMSTAR 2/ROBIS assessment requires two independent reviewers.
| Domain | Judgement | Rationale |
|---|---|---|
| Item 1 PICO | low | Population, beverages, outcome and design clearly defined. |
| Item 2 (critical) protocol | some concerns | Protocol existed and was available on request but was not registered (partial yes). |
| Item 3 design selection explained | low | Prospective design with >=2 years follow-up justified. |
| Item 4 (critical) comprehensive search | low | PubMed, Embase, Ovid, Web of Knowledge plus hand searches; no time or language restriction; authors contacted for unpublished estimates. |
| Items 5-6 duplicate selection and extraction | some concerns | Title/abstract screening by one author; full-text review and extraction in duplicate. |
| Item 7 (critical) excluded studies justified | some concerns | Flow chart and supplementary table S1 on non-eligible cohorts; full list of excluded full texts not visible in main text. |
| Item 8 description of included studies | low | Table 1 details cohorts, follow-up, dietary and outcome assessment. |
| Item 9 (critical) RoB of individual studies | low | Seven bias domains assessed following Cochrane non-randomised tool; six cohorts flagged for potential bias. |
| Item 10 funding of included studies | low | Funding extracted; no included study industry-funded. |
| Item 11 (critical) meta-analytic methods | low | Prespecified random effects, adiposity-stratified models, calibration, non-linearity tested. |
| Item 12 impact of RoB on synthesis | low | Sensitivity analysis excluding high-risk studies; meta-regression on bias indicators. |
| Item 13 (critical) RoB in interpretation | low | GRADE ratings; explicit statement that ASB and juice findings likely biased. |
| Item 14 heterogeneity | low | Explored by meta-regression; juice heterogeneity explained by outcome ascertainment and repeated diet measures. |
| Item 15 (critical) publication bias | low | Egger, contour-enhanced funnel and trim-and-fill with corrected estimates. |
| Item 16 review COI | low | Funding and ICMJE declarations reported. |
Full-text review (25 October 2026): Added crude and calibrated estimates, juice sensitivity results (objective ascertainment RR 1.08, 0.97-1.20; repeated diet 0.98; Egger P=0.03; influence analysis), GRADE ratings (juice low quality), PAF absolute numbers, funding details. Bias overall unchanged (some_concerns) with itemised AMSTAR 2. Juice evidence reframed as 'no benefit shown, harm uncertain'.
9. Assessment of the statistical interpretation
Relative risks per serving/day from observational cohorts. For fruit juice, the adjusted RR 1.07 (1.01-1.14) is small, borderline, unstable and graded low quality; the crude estimate is 0.97 and objectively ascertained studies are null. The juice data show no evidence of benefit for diabetes prevention rather than robust evidence of harm. PAF figures for SSB assume causality.
10. Relevance to the specific claim
Contradicts juice preference (C-FRU-03).
11. Independent contrary findings
12. Transferability to Germany and Europe
High.
13. Overall methodological judgement
Well-conducted meta-analysis. Moderate-quality evidence links SSB with T2D independently of adiposity. For fruit juice the evidence is low quality: it does not support juice as protective, nor does it establish harm.
| Multidimensional assessment | |
|---|---|
| Authenticity of the source | verified |
| Methodological quality | high |
| Risk of bias | some concerns |
| Precision of results | moderate |
| Directness | direct |
| Reproducibility | partly replicated |
| Consistency with the overall evidence | mixed |
| Regional transferability | high |
| Adequacy of the interpretation (by the archive) | not applicable |
14. Open questions
Effect of 100% fruit juice in populations with high juice intake and in the Middle East; trials of substituting juice for whole fruit or water.
15. Linked fact checks
- 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