The mediating role of obesity on the prospective association between urinary sucrose and diabetes incidence in a sub-cohort of the EPIC-Norfolk

Type 2 diabetes (T2D) develops through genetic and environmental factors, with lifestyle and diet playing major roles. While sugar-sweetened beverages show a clear positive association with T2D incidence, evidence for total sugars or sucrose as nutrients is inconsistent, with inverse or null associations reported. A prior meta-analysis suggested an inverse association between sucrose intake and incident T2D, though evidence quality was low. Self-reported dietary assessments are susceptible to bias, notably underreporting among overweight/obese individuals. Objective biomarkers (e.g., urinary sucrose/fructose) may improve assessment; previous EPIC-Norfolk work found urinary sucrose positively associated with overweight/obesity, while self-reported sugars showed inverse/null associations. It remains unclear whether any sucrose–diabetes relationship operates mainly via obesity or includes direct pathways. This study aimed to examine, prospectively, the associations of sucrose intake assessed by urinary biomarker, FFQ, and 7DD with incident diabetes and to quantify mediation by BMI or waist circumference (WC).

Evidence strongly links sugar-sweetened beverages to higher T2D risk, but associations between nutrient-level sugar intakes (total sugars, sucrose) and T2D are mixed, with several cohorts reporting inverse or null findings. A meta-analysis indicated an inverse association for sucrose with T2D incidence, albeit with low certainty. Self-report-based assessments are vulnerable to measurement error and underreporting, especially in overweight individuals. Urinary sugar biomarkers have shown positive associations with adiposity, contrasting with self-reports. Calibration studies using 24-h urinary sucrose and fructose (24uSF) have been developed and validated to estimate unbiased sugar intake; in some cohorts, biomarker-calibrated sugars attenuated inverse associations seen with self-reports. Prior ecological mediation analysis suggested BMI substantially mediates the sugar–diabetes relationship, but individual-level mediation evidence was lacking. Spot urine biomarkers (used here) have been applied but are less validated than 24uSF.

Design and population: Prospective analysis within EPIC-Norfolk. Of 6000 with baseline spot urine, exclusions included prevalent diabetes (n=159), missing specific gravity (n=167), implausible energy intake (women <500 or >3500 kcal/day; men <800 or >4000 kcal/day; n=321), missing covariates (n=65), and urinary sucrose values outside 5–500 µM (n=2292) for the main analysis. Final main sample: n=2996; sensitivity sample with imputation for values <5 µM (n=2246) and >500 µM (n=46): n=5288. Exposure assessment: Dietary sucrose via baseline 7-day diet diary (7DD) and 130-item semi-quantitative FFQ. Urinary sucrose measured from baseline spot urine (with specific gravity adjustment) alongside urea, creatinine, glucose, and fructose using established methods; specific gravity by reagent strips. Outcome: Incident diabetes ascertained via self-reports confirmed by external sources (general practice registers, hospital registers/admissions, mortality records). Covariates: Age, sex, total energy intake (from FFQ/7DD), education level, smoking status, physical activity (validated questionnaire), family history of diabetes; BMI or WC (measured by trained nurses) considered mediators; height included when adjusting for WC. Statistical analysis: Cox proportional hazards to estimate HRs (95% CI) for urinary and dietary sucrose with incident diabetes. Proportional hazards assessed via Schoenfeld residuals and Kaplan–Meier plots. Exposures modeled as log-transformed continuous variables (urinary sucrose per 100 µM; dietary sucrose per 50 g/d) and quartiles (Q1 reference). Models: Model 1 adjusted for age and sex; Model 2 additionally for total energy, education, smoking, physical activity, family history of diabetes (+height for WC). Additional models further adjusted for BMI or WC. Linear trends tested using quartile medians. Nonlinearity assessed with smoothing splines (3 df). Mediation analysis per VanderWeele’s counterfactual framework using the mediator package in R: estimated pure direct effect (PDE), total indirect effect (TIE), total effect, and proportion mediated (PM) with BMI or WC as mediators. Preconditions checked: exposure–outcome association (Cox models), exposure–mediator (linear/logistic regression; BMI and WC also categorized as obese vs non-obese), and mediator–outcome association. Sensitivity analyses: imputed urinary sucrose values outside detection limits; mediation using BMI/WC from second health check (2HC) after 3 years (n=1728) to address temporality; time-varying causal effects examined over follow-up. Analyses conducted in R 4.1.0.

- Sample: n=2996 (mean age 60.6±9.5 years; 52.5% women); follow-up 11.2±2.9 years; incident diabetes cases n=97. Men had higher urinary sucrose; higher quartiles associated with older age, higher adiposity, lower education, higher energy intake, and more inactivity. - Dietary sucrose (Model 2, per 50 g/d): 7DD HR 0.63 (95% CI 0.43, 0.91); FFQ HR 0.81 (95% CI 0.46, 1.42). Quartiles generally showed inverse but imprecise trends; associations attenuated after adjusting for BMI or WC. - Urinary sucrose: Continuous per 100 µM HR 1.14 (95% CI 0.95, 1.36); Q4 vs Q1 HR 1.36 (95% CI 0.77, 2.41). Associations attenuated with BMI or WC adjustment (per 100 µM: HR 1.08 (0.90, 1.30) with BMI; 1.07 (0.90, 1.29) with WC). Spline suggested rising hazard with higher urinary sucrose. - Mediation (main analysis, per 100 µM urinary sucrose): BMI mediator—PDE HR 1.11 (0.92, 1.35), TIE HR 1.02 (1.00, 1.04), total effect HR 1.13 (0.94, 1.37), PM 16%. WC mediator—PDE 1.12 (0.91, 1.39), TIE 1.03 (1.01, 1.06), total effect 1.16 (0.94, 1.42), PM 22%. - Sensitivity (imputed values; n=5288): FFQ per 50 g/d HR 0.66 (0.44, 0.98); 7DD per 50 g/d HR 0.71 (0.52, 0.96); urinary sucrose per 100 µM HR 1.00 (0.92, 1.08); Q4 vs Q1 HR 1.38 (0.79, 2.42). With 2HC mediators: BMI PM 10% (TIE 1.02 (0.99, 1.06)); WC PM 12% (TIE 1.03 (1.00, 1.07)). Time-varying analyses indicated stable PDE, TIE, and total effects over follow-up.

Self-reported sucrose intake (FFQ, 7DD) showed inverse or imprecise inverse associations with incident diabetes, while urinary sucrose biomarker indicated a tendency toward a positive association, aligning with prior biomarker-based studies linking urinary sugars to adiposity and glycemia. The mediation analyses suggest that part of the urinary sucrose–diabetes association operates via adiposity (BMI or WC), though indirect effects were small and direct effects imprecise, implying potential additional mechanisms beyond obesity. Measurement error and underreporting likely explain inverse associations seen with self-reported intakes; biomarker-calibrated approaches in other cohorts have attenuated such inverse findings. While 24-h urinary sugars are validated as predictive biomarkers, spot urine (used here) lacks full validation, which may contribute to inconsistencies (e.g., u-shaped associations with imputed extremes). Biologically, high sugar intake may increase energy intake leading to weight gain (indirect pathway), and might also influence liver fat, insulin resistance, and glycemic load-related responses (potential direct pathways). Overall, findings support a positive relationship between objectively assessed sucrose exposure and diabetes risk, partially mediated by adiposity, while highlighting the limitations of self-reported dietary assessment.

Sucrose exposure measured by an objective urinary biomarker pointed to increased diabetes incidence, with BMI and WC partially mediating this association. The study underscores potential bias in self-reported sugar intake assessments and suggests additional mechanisms may contribute beyond obesity. Larger, well-powered studies using validated objective sucrose biomarkers (preferably 24-h collections or validated spot measures), repeated measures, and robust mediation frameworks are needed to refine estimates and elucidate causal pathways.

- Limited number of incident diabetes cases (n=97) leading to wide confidence intervals and imprecision. - Potential competing risk: high number of deaths during follow-up (n=1358); death treated as censoring may violate independence from diabetes incidence. - Biomarker measurement constraints: 42.5% of urinary sucrose values below and 0.9% above the 5–500 µM detection range required imputation; associations with imputed values appeared u-shaped, raising concerns about true value representation below detection limits. - Single baseline spot urine measurement; spot urine sucrose is less validated than 24-h collections for reflecting usual intake. - Possible residual confounding despite adjustments; self-reported intake remains prone to underreporting and measurement error. - Generalizability limited to similar populations and measurement settings.