Impact of the timing of metformin administration on glycaemic and glucagon-like peptide-1 responses to intraduodenal glucose infusion in type 2 diabetes: a double-blind, randomised, placebo-controlled, crossover study

Metformin is first-line pharmacotherapy for type 2 diabetes, yet its mechanisms remain incompletely understood. Increasing evidence indicates that much of metformin’s glucose-lowering action occurs within the gastrointestinal tract, including stimulation of GLP-1 secretion, slowing of gastric emptying, and reduced intestinal glucose absorption, rather than solely via systemic exposure. GLP-1 appears to play a key role in metformin’s effects on glucose homeostasis. Interventions that enhance GLP-1 to lower postprandial glycaemia are often more effective when administered before a meal. Although metformin is commonly advised to be taken with meals (primarily to mitigate gastrointestinal side effects), the impact of timing on its postprandial glucose-lowering efficacy is poorly defined. The authors hypothesised that varying the timing of metformin administration prior to a standardised nutrient load would differentially affect glycaemia and be associated with differences in GLP-1 and insulin responses. They conducted a proof-of-concept, double-blind, randomised, placebo-controlled crossover study exposing the small intestine to metformin at defined intervals before a standardised intraduodenal glucose infusion to precisely assess timing effects while controlling for gastric emptying.

Prior mechanistic work shows metformin can reduce fasting glucose via suppression of hepatic glucose production, but substantial in vivo evidence supports gastrointestinal mechanisms irrespective of systemic bioavailability. Enteral metformin is more effective than intravenous or intraportal routes, and a delayed-release formulation with minimal systemic exposure performs similarly to immediate/extended-release. Documented GI effects include stimulation of GLP-1, slowing gastric emptying, reduced intestinal glucose absorption, inhibition of bile acid resorption, and modulation of gut microbiota. Blocking GLP-1 signalling attenuates metformin’s glucose-lowering effects in rodents and humans, highlighting GLP-1’s mediating role. Pre-meal strategies (e.g., whey protein preload) enhance GLP-1 and reduce postprandial glycaemia more than co-ingestion with meals. A small open-label pilot (n=5) suggested that taking 1000 mg metformin 30 min before a meal improved glucose-lowering, enhanced GLP-1, and slowed gastric emptying vs taking it with the meal. However, rigorous evaluation of timing in metformin-treated people with type 2 diabetes has been lacking.

Design: Double-blind, randomised, placebo-controlled, 4-period crossover study conducted at the University of Adelaide Clinical Research Facility. Each participant completed four study visits separated by at least 7 days. Participants: 16 adults with type 2 diabetes on stable metformin monotherapy (500–2000 mg/day for ≥3 months). Characteristics: 16 White; 14 men, 2 women; age 69.9 ± 1.9 years; BMI 28.7 ± 1.0 kg/m²; HbA1c 48.2 ± 1.6 mmol/mol (6.6 ± 0.1%); diabetes duration 10.4 ± 2.6 years. Exclusions: significant GI symptoms, impaired renal/liver function, prior GI surgery, or medications affecting GI function/appetite. Ethics approval obtained (2021/HRE00130); trial registered (ACTRN12621000878875). Pre-visit standardisation: Participants maintained usual metformin dosing except timing adjustments: immediate-release users withheld morning dose; extended-release users withheld evening dose before and morning dose of study day. Standardised evening meal (~19:00; beef lasagne, 2472 kJ) before each visit; then fasting (water allowed) until 08:00 arrival. Avoid vigorous exercise and alcohol for 24 h pre-visit. Procedures: A silicone nasoduodenal catheter inserted and positioned with infusion port 12 cm distal to pylorus; placement verified continuously by transmucosal potential difference (stomach ~−40 mV, duodenum ~0 mV). An IV cannula placed for arterialised venous sampling. Interventions (double-blind, randomised): Intraduodenal (ID) bolus over 2 min of either metformin 1000 mg in 50 mL 0.9% saline or 50 mL 0.9% saline at t = −60, −30, and 0 min per one of four treatments: (1) Met at −60 min, saline at −30 and 0; (2) Met at −30 min, saline at −60 and 0; (3) Met at 0 min, saline at −60 and −30; (4) saline at all timepoints (control). Randomisation generated online; solutions prepared by a research officer independent of procedures to maintain blinding. Glucose challenge: At t = 0, ID infusion of 45 g glucose in 180 mL water over 60 min (3 mL/min; 12.56 kJ/min [3 kcal/min]). Catheter removed at t = 60; monitoring continued to t = 120. Sampling and measures: Blood collected at t = −60 (pre-infusion) and every 30 min from t = −60 to 120 for plasma glucose, total GLP-1, and insulin. Nausea and appetite sensations (hunger, fullness, desire to eat, anticipated meal size) assessed via 100 mm visual analogue scales at the same times. Assays: Plasma glucose via glucose oxidase (YSI 2300 STAT PLUS). Total GLP-1 by RIA (Millipore GLPIT-36HK; sensitivity 3 pmol/L; intra-/inter-assay CVs 7.9%/12.9%). Insulin by ELISA (Mercodia 10-1113; sensitivity 6 pmol/L; intra-/inter-assay CVs 2.1%/11.1%). Statistical analysis: iAUC0–120 for glucose, insulin, and total GLP-1 calculated (trapezoidal rule, baseline-subtracted). Whole-body insulin sensitivity by Matsuda index using fasting and mean 0–120 min glucose/insulin. Fasting values (−60, −30, 0) and iAUC0–120 compared across treatments by one-factor repeated-measures ANOVA. Time-course data analysed by two-factor repeated-measures ANOVA (treatment, time), with Bonferroni-adjusted post hoc tests when interactions significant. Power: n=16 provided ≥90% power (α=0.01) to detect a 20% difference in glucose iAUC0–120 between metformin and control based on prior data. Significance p<0.05. Analyses in GraphPad Prism 9.5. Safety/tolerability: Nausea and appetite VAS monitored; adverse symptoms documented.

- Plasma glucose: Fasting glucose did not differ across days. ID glucose caused prompt increases (time p<0.001). Significant treatment-by-time interaction (p<0.001). Compared with control, plasma glucose was lower from 30–120 min after Met (−60), and 60–120 min after Met (−30) and Met (0). Compared with Met (0), glucose was lower 30–120 min after Met (−60) and 60–90 min after Met (−30); glucose at 60 min was lower after Met (−60) vs Met (−30) (all p<0.05). Glucose iAUC0–120 differed by treatment (p<0.001): mean ± SEM (mmol/L×min): Met (−60) 439±19.6; Met (−30) 466±20.0; Met (0) 517±22.5; Control 574±24.7. All metformin timings reduced iAUC vs control (p<0.01 each); reduction was greater for −60 and −30 vs 0 (p=0.03 and p=0.001, respectively); no difference between −60 and −30. - GLP-1: Fasting total GLP-1 similar across days. ID glucose increased GLP-1 (time p<0.001), peaking at 60 min. Treatment-by-time interaction p=0.02. GLP-1 higher vs control at 60–120 min after Met (−60) and at 60–90 min after Met (−30); also higher vs Met (0) at 60–90 min after Met (−30) (p<0.05). GLP-1 iAUC0–120 showed a non-significant trend to increase with metformin (p=0.078); means (pmol/L×min): Met (−60) 1879±457; Met (−30) 2025±380; Met (0) 1582±260; Control 1338±241. - Insulin: Fasting insulin similar. ID glucose increased insulin (time p<0.001); peak at 60 min on control and 90 min on metformin days. Treatment-by-time interaction p=0.03: insulin higher at 90 min after Met (−60) and Met (−30) vs control, and at 90–120 min after Met (0) vs control (p<0.05). Insulin iAUC0–120 higher on all metformin days vs control (overall p=0.008); means (pmol/L×min): Met (−60) 20121±4098; Met (−30) 21809±4264; Met (0) 21682±4291; Control 15628±2403; no differences between metformin timings. - Insulin/glucose ratio: Similar at baseline; rose with ID glucose with treatment effect (p=0.025) and treatment-by-time interaction (p<0.001). Higher vs control at 60–90 min after Met (−60), 60–120 min after Met (−30), and 90–120 min after Met (0) (p<0.05). - Insulin sensitivity: Matsuda index did not differ between treatments (p=0.38). - Tolerability: Nausea scores were low and unchanged; appetite VAS measures showed minimal changes without differences between treatments.

Administering metformin 30–60 minutes before a standardized duodenal glucose load produced greater reductions in postprandial glycaemia than administering it at the start of glucose infusion, and this enhancement coincided with greater postglucose increases in GLP-1. The study controlled small intestinal exposure by intraduodenal administration to minimise confounding by variable gastric emptying. Findings align with the concept that pre-exposure of the small intestine to metformin can prime gastrointestinal mechanisms—particularly GLP-1 secretion—to optimize postprandial glucose handling. The GLP-1 augmentation was evident only when metformin preceded the glucose stimulus, suggesting metformin modulates glucose–gut interactions (e.g., by reducing proximal intestinal glucose absorption, increasing distal nutrient exposure and L-cell stimulation). Metformin also enhanced glucose-induced insulin secretion across all metformin timings without altering whole-body insulin sensitivity, indicating that mechanisms beyond GLP-1 may contribute to the insulinotropic effect in this setting. Clinically, the data imply that taking metformin before meals may improve postprandial glycaemic control compared with co-administration at mealtime, challenging routine advice to take metformin with meals and potentially informing strategies to reduce postprandial hyperglycaemia and related complications.

In metformin-treated individuals with well-controlled type 2 diabetes, metformin lowers postprandial glucose more effectively when administered 30–60 minutes before enteral glucose than when given at the start of glucose exposure. This enhanced glucose-lowering is associated with greater GLP-1 responses, while insulin sensitivity remains unchanged and insulin secretion is augmented similarly across metformin timings. These findings suggest that pre-meal administration of metformin could optimize postprandial glycaemic control. Future research should evaluate longer-term clinical outcomes of pre-meal dosing (30–60 minutes before meals) with standard oral formulations, assess effects on mixed meals and other macronutrients, examine different metformin formulations (immediate-, extended-, delayed-release), explore sex-related differences, and further elucidate mechanisms (including intestinal glucose transport, bile acids, and microbiota).

- Formulation/route: Metformin was delivered as a solution directly into the small intestine (immediate exposure), which may limit generalisability to oral extended- or delayed-release formulations. - Model: Intraduodenal administration and glucose infusion provide precise control but are unphysiological compared with oral ingestion; metformin’s known effect to slow gastric emptying could further enhance glucose-lowering with oral dosing. - Test meal: A glucose solution (not mixed nutrients) was used to reduce variability; effects of timing on fat and protein handling remain unknown. - Unmeasured endpoints: Intended assessment of small intestinal glucose absorption using 3-O-methylglucose was not completed; plasma glucagon and metformin concentrations were not measured. - Sex distribution: Predominantly male sample may limit assessment of potential sex-related differences in GLP-1 and glycaemic responses. - Acute study: Findings reflect acute effects; long-term effects of pre-meal metformin timing were not assessed.