Occurrence of Gastrointestinal Adverse Events Upon GLP-1 Receptor Agonist Initiation With Concomitant Metformin Use: A Post Hoc Analysis of LEADER, STEP 2, SUSTAIN-6, and PIONEER 6

Gastrointestinal (GI) adverse events associate with both metformin and glucagon-like peptide-1 receptor agonists (GLP-1RAs), two first-line agents for the treatment of type 2 diabetes that are frequently prescribed in combination. GLP-1RAs are initiated at a low dose and titrated to minimize the occurrence of GI symptoms. Still, most treatment-related GI adverse events occur during the dose titration period. Few analyses have examined whether concomitant metformin use exacerbates GLP-1RA–associated GI symptoms, and real-world data and meta-analyses of randomized controlled trials are inconsistent (1–3). Regardless, some experts suggest metformin discontinuation or dose reduction during GLP-1RA initiation (4). The purpose of this study was to examine whether concomitant metformin use exacerbates GLP-1RA–induced GI adverse events or increases early GLP-1RA discontinuation. Using data from four clinical trials examining GLP-1RA in type 2 diabetes (Liraglutide Effect and Action in Diabetes: Evaluation of Cardiovascular Outcome Results [LEADER], Semaglutide Treatment Effect in People with Obesity [STEP 2], Trial to Evaluate Cardiovascular and Other Long-Term Outcomes With Semaglutide in Subjects With Type 2 Diabetes [SUSTAIN-6], and Peptide Innovation for Early Diabetes Treatment [PIONEER] 6), we compared the incidence of GI adverse events during the GLP-1RA titration period in participants with and without concomitant metformin use and evaluated whether concomitant metformin use impacted study product discontinuation.

The trial designs for LEADER, STEP 2, SUSTAIN-6, and PIONEER 6 have been previously published (5–8). All studies analyzed were large, randomized, double-blind, placebo-controlled clinical trials assessing the safety and efficacy of GLP-1RA in people with type 2 diabetes and did not require participants to be treated with metformin at baseline. Efficacy outcomes varied by trial.

In this post hoc analysis, we evaluated GI adverse events during GLP-1RA initiation and titration until the maintenance dose was achieved. We chose this period because GI adverse events commonly start during GLP-1RA initiation, continue through titration, and resolve after the maintenance dose is reached (9). We defined the observation window as the time from GLP-1RA initiation until the day of the last titration plus four half-lives of the GLP-1RA (estimated time to achieve final steady state). The observation windows are shown in Table 1. LEADER evaluated liraglutide (half-life: 0.5 days; target dose 1.8 mg). STEP 2 and SUSTAIN-6 evaluated subcutaneous semaglutide (half-life: 7 days; STEP 2: target doses 1.0 and 2.4 mg and SUSTAIN-6: target doses 0.5 and 1.0 mg). PIONEER 6 evaluated oral semaglutide (half-life: 7 days; target dose 14 mg).

The primary end point was the percentage of participants with at least one GI adverse event during the observation window stratified by metformin use at baseline. Repeat events were not analyzed. We also evaluated specific events of clinical interest, including nausea, vomiting, and serious GI adverse events. Finally, we examined the number of participants experiencing GI adverse events who subsequently discontinued the study product at any time. Sensitivity analyses that evaluated a consistent 140-day window for all trials are provided in the Supplementary Material.

Because baseline metformin use was not randomized, statistical comparison between participants with or without metformin within the treatment arms (GLP-1RA and placebo) could not be made. Instead, we used a logistic regression to evaluate the effect of the randomized treatment on GI adverse events within subgroups of metformin users or nonusers. The logistic model used GI adverse events as the binary outcome and GLP-1RA, metformin, and an interaction term as predictors. A Wald test was used to test for significance of the interaction within each trial.

Of 16,996 participants in these four trials, 12,928 (76%) were treated with metformin at baseline. STEP 2 had a higher proportion of metformin users (89.6%) among all trials (LEADER: 74.7%, SUSTAIN-6: 73.2%, and PIONEER 6: 77.3%). Baseline characteristics are provided in Table 2 and were similar between treatment arms within the trials, except for the estimated glomerular filtration rate (eGFR). Participants treated with baseline metformin had a higher eGFR compared with those not treated with metformin in all trials except in STEP 2, where the eGFR was equivalent (metformin users: 95.7 mL/min/1.73 m², metformin nonusers: 92.9 mL/min/1.73 m²). Differences in characteristics among trials, such as duration of type 2 diabetes, age, and presence of chronic kidney disease, reflected the eligibility criteria and the primary outcome of the specific trial.

The primary outcome is shown in Table 3. Across all trials, participants randomized to a GLP-1RA experienced more GI adverse events. However, concomitant metformin use did not increase the percentage of participants who developed any GI adverse event during the observation window. The coefficient for interaction was not significant for any study except LEADER (Table 3). Sensitivity analyses that evaluated a consistent 140-day window for all trials demonstrated the same trends, but the coefficient of variation was no longer significant for the LEADER trial (Supplementary Table 1). Concomitant metformin use was not associated with an increased risk of serious GI adverse events, nausea, or vomiting in any trial (Supplementary Tables 2–4).

Finally, we examined rates of premature study product discontinuation in participants who experienced any GI adverse event during GLP-1RA initiation and titration. As shown in Table 4, concomitant metformin use did not associate with increased discontinuation of the study product at any point during the trial. Instead, in both randomized arms, GLP-1RA and placebo, a numerically higher percentage of participants who were not treated with concomitant metformin discontinued the study product. The same trend in discontinuation was observed during the GLP-1RA initiation and titration observation window (Supplementary Table 5).

GI adverse events are the most prevalent adverse effects associated with GLP-1RAs. Although largely temporary and mild-to-moderate in severity, GI adverse events lead to discontinuation of GLP-1RAs both in clinical trial programs and in the real world (5–8,10). Factors that modify the risk or severity of GI symptoms are not well established, but practical ways to mitigate GI adverse events and prevent GLP-1RA discontinuation are needed given the glycemic, weight loss, and cardiovascular benefits of GLP-1RAs that have made them a potential first-line therapy for type 2 diabetes and obesity (11–13). Although the mechanisms for GI intolerance are different, reasonable clinical concern exists that metformin use may exacerbate GLP-1RA–induced GI symptoms, as metformin alone can cause abdominal pain, cramps, and diarrhea (14). Whether metformin should be paused to promote successful GLP-1RA initiation remains an important clinical question. In this post hoc analysis of four large GLP-1RA trials in type 2 diabetes, we demonstrate that concomitant metformin use during GLP-1RA initiation does not associate with increased risk or severity of GI adverse events or premature discontinuation of study product. We did not evaluate simultaneous initiation of metformin and GLP-1RA.

Interestingly, in every study, the percentage of participants experiencing any GI adverse event was numerically higher in participants without concomitant metformin treatment, irrespective of randomization to GLP-1RA or placebo. We also found that of those who experienced GI adverse events, a higher percentage of participants without concomitant metformin treatment discontinued the study product in both treatment arms across all four studies. These data are consistent with real-world evidence from the U.K. suggesting metformin use associates with lower odds of GI adverse events or drug discontinuation among patients treated with GLP-1RA (2).

The reasons for this finding are likely multifactorial but may be due to increased baseline or susceptibility to GI symptoms. In LEADER, SUSTAIN-6, and PIONEER 6, people who were not treated with metformin had lower kidney function. As such, people without concomitant metformin treatment may reflect a sicker population with a greater symptom burden. Nonetheless, similar findings were seen in STEP 2, where the eGFR was equivalent between metformin users and nonusers. As metformin was considered the single first-line pharmacological therapy for type 2 diabetes when all four studies were enrolling (15), it is conceivable that participants who did not tolerate treatment with metformin may be more susceptible to GI adverse events independent of the study product. Techniques for mitigating GLP-1RA–induced GI symptoms, such as in-depth counseling and slow titration (4,9), may be necessary in individuals with a low eGFR or those who did not tolerate metformin.

In all studies, we compared active treatment to placebo within the two metformin subgroups (metformin users and metformin nonusers) and demonstrated no statistical difference in three trials. Although there were few GI adverse events within the short observation window of liraglutide (16 days) and a numerically higher percentage of GI adverse events in metformin nonusers, the analysis suggested a statistically significant increase in GI adverse events in participants treated with metformin. The marked difference in pharmacokinetics between liraglutide and semaglutide could account for this difference. However, it is more likely that the statistical significance of this finding is due to the small sample size resulting from the method of adverse event collection. Accordingly, statistical significance was not observed in the sensitivity analysis that included more adverse events by extending the observation to 140 days. Consistently, although a prior meta-analysis suggested statistically significant increased GI-adverse events in people treated with concomitant metformin, this was largely mediated by short-acting GLP-1RAs (exenatide, lixisenatide), as no statistically significant increase in GI adverse events was seen with longer-acting GLP-1RAs like liraglutide (1).

Taken together, these data indicate that concomitant metformin use does not exacerbate GLP-1RA–induced GI symptoms and suggest that metformin discontinuation or dose reduction is unnecessary. Instead, although counseling about GI symptoms is necessary in all patients, increased counseling or slowed titration may be necessary in metformin nonusers.

Funding. K.R.K. is supported by the University of North Carolina Department of Medicine and School of Medicine Physician Scientist Training Program. K.K.B.C. is partly funded by an Innovation Fund Denmark grant.

Duality of Interest. K.R.K. has received personal compensation from Novo Nordisk for consultation. I.L. received research funding (paid to institution) from Novo Nordisk, Sanofi, Merck, Pfizer, Mylan, and Boehringer Ingelheim, and received advisory/consulting fees and/or other support from Novo Nordisk, Eli Lilly, Sanofi, AstraZeneca, Boehringer Ingelheim, Janssen, Intercept, Intarcia, TargetPharma, Merck, Pfizer, Novartis, GI Dynamics, Mylan, Mannkind, Valeritas, Carmot, Zealand Pharma, Shionogi, Mediflix, and Bayer. E.F. is a former employee of Novo Nordisk. K.K.B.C., S.O., and T.J.A. are employees of Novo Nordisk. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. K.R.K., K.K.B.C., E.F., and I.L. conceptualized the study design. K.R.K. and I.L. drafted the manuscript. K.K.B.C., E.F., and S.O. conducted the analyses. T.A. contributed to data quality assurance and design and provided analysis expertise. All authors contributed to manuscript review and editing and clinical subject matter expertise. I.L. contributed to regular oversight and communications. E.F. and I.L. are guarantors of this work and, as such, had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at 83rd Scientific Sessions of the American Diabetes Association, San Diego, CA, 23–26 June 2023.