Longitudinal associations of dietary fiber and its source with 48-week weight loss maintenance, cardiometabolic risk factors and glycemic status under metformin or acarbose treatment: a secondary analysis of the March randomized trial

Type 2 diabetes mellitus (T2DM) prevalence is rising globally, imposing a substantial health burden, particularly in rapidly developing Asian regions. Effective T2DM management requires both pharmacological therapy and optimization of modifiable lifestyle factors, especially diet. While prior work often examined drugs or diet in isolation, synergistic effects remain understudied. The MARCH trial provided an opportunity to investigate interactions between initial monotherapy with acarbose or metformin and dietary factors. Prior analyses from MARCH linked higher carbohydrate intake with higher BMI, HbA1c, postprandial glucose, and insulin AUC. Emerging evidence suggests carbohydrate quality—emphasizing higher fiber and optimal fiber proportion—is more influential than total carbohydrate amount for glycemic control and weight maintenance. This study aimed to assess longitudinal and dose-dependent associations of total and source-specific dietary fiber and the carbohydrate-to-fiber ratio with weight, glycemic, and cardiometabolic outcomes over 48 weeks under metformin or acarbose treatment in newly diagnosed T2DM patients.

Epidemiologic and clinical trials indicate increased dietary fiber intake, especially in populations with low baseline intake, benefits T2DM management, though health effects can vary by fiber source and physicochemical properties. Metformin improves glucose homeostasis via reduced intestinal glucose absorption, inhibition of hepatic gluconeogenesis, and increased peripheral glucose uptake through AMPK activation. Acarbose, an intestinal amylase inhibitor, blunts postprandial glycemic excursions by inhibiting carbohydrate digestion. Dietary fibers may affect drug bioavailability via gastrointestinal interactions. Meta-analyses report fiber improves FBG and HbA1c, and whole grains improve glycemic markers and insulin sensitivity. Mechanisms include delayed nutrient absorption by viscous/insoluble fibers and modulation of the gut microbiota with increased SCFA production. Interactions between gut microbiota and antidiabetic drugs (metformin, acarbose) may mediate therapeutic effects and interindividual responses. Prior findings on fruit intake and lipids are mixed; some studies show no benefit of fruit fiber for T2DM risk and potential sugar-related dyslipidemia concerns. The carbohydrate-to-fiber ratio and broader carbohydrate quality indices (CQI) have been proposed to characterize carbohydrate quality and predict cardiometabolic risk, with suggested cutoffs (e.g., ≤10:1) associated with higher diet quality. However, evidence remains heterogeneous, warranting further research.

Design: Secondary analysis of the MARCH randomized controlled trial, a 48-week, multicenter (11 sites in China) RCT comparing acarbose vs metformin as initial therapy in newly diagnosed T2DM. Participants received 24 weeks of monotherapy followed by 24 weeks with add-on insulin secretagogues if glycemic targets were unmet. Trial registration: ChiCTR-TRC-08000231; ethics approvals obtained; written informed consent provided. Participants: Adults 30–70 years, diagnosed with T2DM within 12 months (1999 WHO criteria), BMI 19–30 kg/m², HbA1c 7–10%, FPG ≤11.1 mmol/L, no oral antidiabetic drugs or short-term (1 month) treatment within 3 months pre-enrollment. Of 1099 screened, 788 randomized (393 metformin, 391 acarbose); 784 started treatment. For this analysis, exclusions included insufficient dietary data (98 acarbose, 107 metformin) and implausible energy intake (10 acarbose, 18 metformin), yielding 551 participants (286 acarbose, 265 metformin). Dietary assessment: 24-h dietary recalls at baseline, week 24, and week 48 across 15 food groups (rice, starch, whole grains, potatoes/tubers, fruits, vegetables, eggs, dairy, meat, fish, nuts, legumes, sugars, alcohol, oils). Trained nutritionists estimated portions relative to reference portions. Nutrient intakes computed using China Food Composition Tables (2002). Exposures: total dietary fiber; fiber from vegetables, fruits, legumes, whole grains; carbohydrate-to-total fiber ratio. Outcomes: Assessed at baseline, weeks 24 and 48. Glycemia: HbA1c, FBG, 2h PPG (2hPPG). Insulin resistance/secretion: fasting insulin, HOMA-IR, HOMA-β, early insulin secretion index (delta30-min insulin/delta30-min glucose), AUC insulin, WBISI (10,000/√([fasting glucose × fasting insulin] × [mean glucose × mean insulin during OGTT])). Hormones: AUC glucagon, GLP-1. Cardiometabolic: BMI, TC, LDL-C, HDL-C, triglycerides, SBP, DBP. Statistical analysis: Dietary measures were skewed and summarized as median (IQR). Friedman’s two-way ANOVA by ranks tested within-group changes over time, with Bonferroni-adjusted post hoc comparisons when P<0.05. Longitudinal associations between dietary measures (continuous) and outcomes were evaluated within each treatment group using mixed-effects linear models and restricted cubic spline (RCS) models with four knots at the 5th, 35th, 65th, and 95th percentiles. Fixed effects: duration of intervention, diabetes duration, sex, age, total energy intake. Random effects: center and individual. Results reported as β coefficients (95% CI) and P-values; nonlinearity tested in RCS. Two-tailed P<0.05 deemed significant.

Sample: 551 participants included (286 acarbose; 265 metformin). Changes in dietary measures over time: - Acarbose group: Significant time effects for legume fiber (P=0.001) and carbohydrate-to-fiber ratio (P=0.002). Post hoc: legume fiber lower at week 48 vs week 24; carbohydrate-to-fiber ratio higher at week 48 vs week 24 (adjusted P<0.05). - Metformin group: Significant time effects for fruit fiber (P=0.013), vegetable fiber (P=0.002), and carbohydrate-to-fiber ratio (P=0.003). Post hoc: vegetable fiber decreased at week 24; carbohydrate-to-fiber ratio increased at week 48 (adjusted P<0.05). Fruit fiber post hoc not significant after Bonferroni. Longitudinal associations (mixed-effects models): Acarbose group: - Total fiber: positively associated with HOMA-β (β=1.694, P=0.024) and inversely with 2hPPG (β=−0.054, P=0.005). - Whole grain fiber: positively associated with HOMA-β (β=2.988, P=0.006) and WBISI (β=0.089, P=0.048); inversely with 2hPPG (β=−0.070, P=0.014). - Fruit fiber: positively associated with DBP (β=0.653, P=0.010), TC (β=0.091, P=0.007), and LDL-C (β=0.068, P=0.017). - Vegetable fiber: inversely associated with DBP (β=−0.309, P=0.005) but positively with HbA1c (β=0.039, P=0.008). - Legume fiber: inversely associated with 2hPPG (β=−0.097, P=0.033). - Carbohydrate-to-fiber ratio (RCS): non-linear associations with 2hPPG (P non-linear=0.009) and HDL-C (P non-linear=0.001) showing S-shaped patterns (2hPPG decreases slightly to ~25, then rises markedly to ~75; HDL-C slightly increases to ~25 then decreases). Metformin group: - Legume fiber: inversely associated with BMI (β=−0.052, P=0.024), FBG (β=−0.024, P=0.035), and 2hPPG (β=−0.059, P=0.028). - Whole grain fiber: positively associated with DBP (β=0.172, P=0.026). - Carbohydrate-to-fiber ratio (RCS): non-linear associations with insulin AUC (P non-linear=0.005) and BMI (P non-linear=0.009), each S-shaped (slight decline to ~25, increase to ~75; subsequent divergence with continued rise in insulin AUC and sharp BMI decrease). Overall interpretation: - Under acarbose, higher total and whole grain fiber intake relates to improved β-cell function, insulin sensitivity, and postprandial glycemia; legume fiber improves postprandial glycemia. Fruit fiber aligns with higher DBP, TC, LDL-C; vegetable fiber lowers DBP but associates with higher HbA1c. High carbohydrate-to-fiber ratios relate to poorer glycemic control and lower HDL-C. - Under metformin, legume fiber associates with better glycemia and lower BMI (weight loss). Whole grain fiber associates with higher DBP. High carbohydrate-to-fiber ratios show complex, non-linear relations with insulin secretion and BMI.

Findings support that dietary fiber quantity and source modulate clinical outcomes differently depending on whether patients receive acarbose or metformin. Under acarbose, higher total and whole grain fiber appear to enhance β-cell function, insulin sensitivity, and postprandial glycemia, consistent with fiber’s viscosity-related slowing of carbohydrate absorption and potential microbiome-mediated benefits. Legume fiber consistently improved glycemic indices in both groups and supported weight loss under metformin, aligning with evidence that soluble/viscous or hemicellulose-rich fibers promote satiety, reduce energy intake, and beneficially modulate gut microbiota and SCFAs. The non-linear S-shaped relationships between carbohydrate-to-fiber ratio and outcomes underscore the importance of carbohydrate quality beyond quantity, and suggest thresholds where high carbohydrate with low fiber adversely affects HDL-C and postprandial glycemia under acarbose, while complex relationships with insulin AUC and BMI emerge under metformin. Unexpected associations included fruit fiber correlating with higher DBP, TC, and LDL-C under acarbose, possibly reflecting co-ingested simple sugars and low overall fiber intakes, and vegetable fiber correlating positively with HbA1c despite lower DBP. Whole grain fiber’s positive association with DBP under metformin contrasts with some prior literature and may reflect population specifics, intake levels below recommended thresholds, or dietary pattern confounding. Overall, results emphasize tailoring dietary fiber type to pharmacotherapy to optimize glycemic and cardiometabolic outcomes.

Combining antidiabetic pharmacotherapy with targeted dietary fiber strategies may yield additive metabolic benefits in newly diagnosed T2DM. Under acarbose, higher total and whole grain fiber intakes were associated with improved β-cell function, insulin sensitivity, and postprandial glycemia. Legume fiber intake was linked to better glycemic control in both acarbose and metformin groups and to greater weight loss under metformin. Optimizing drug–fiber combinations may outperform drug monotherapy for glycemic control, insulin sensitivity, and weight management. Future research should refine dose–response thresholds for carbohydrate-to-fiber ratios, evaluate gut microbiome mediators, and test tailored dietary prescriptions alongside specific glucose-lowering agents.

Dietary intake was assessed by 24-h recalls, which are susceptible to misreporting and may not reflect usual intake. The cohort comprised 551 Chinese patients newly diagnosed with T2DM, limiting generalizability to other populations. Exclusions due to insufficient dietary data and implausible energy intake raise potential selection bias. The study did not include gut microbiota sequencing, limiting mechanistic insights into microbe–diet–drug interactions. Overall dietary fiber intakes were relatively low, potentially attenuating observable benefits and contributing to some conflicting associations.