GLP-1 vs oral agents: insulin stop in T2D

A target trial emulation published on 14 July 2026 in the Annals of Internal Medicine asked a clinically common question: when a patient with type 2 diabetes is already on basal insulin and you add a second agent, does adding a GLP-1 receptor agonist raise the chances of eventually stopping the insulin, compared to adding an SGLT-2 inhibitor or a DPP-4 inhibitor?^[1] The answer from 8,869 matched Veterans Health Administration patients was: no, not to any meaningful degree.

What the study did

Lipska and colleagues used Veterans Health Administration electronic health record data from 2020 to 2022 to emulate a randomised trial they could not ethically run. They identified adults with type 2 diabetes who were already taking basal insulin and who then added either a GLP-1 receptor agonist, an SGLT-2 inhibitor, or a DPP-4 inhibitor. Participants were propensity-score matched across the three groups, and the primary outcome was insulin discontinuation sustained for at least 90 days during a three-year follow-up^[1].

The study population was 63% aged 65 or older and 93% male, reflecting the demographic composition of the Veterans Health Administration. That skew limits how far the findings generalise to younger patients and to women, which the authors acknowledge as a limitation. The patient profile is also notable for what it is not: it excludes the obese adults without established diabetes who have been the focus of recent large obesity trials such as SURMOUNT-1 for tirzepatide.

What the numbers showed

Over three years, insulin discontinuation rates were similar across all three groups: 16.7% in the GLP-1 receptor agonist group, 17.9% in the SGLT-2 inhibitor group, and 17.1% in the DPP-4 inhibitor group^[1]. The differences between groups were small and consistent with chance. The authors concluded that GLP-1 receptor agonist addition did not increase the likelihood of stopping insulin compared to the other two drug classes.

Put plainly: about one in six patients in each group managed to discontinue insulin within three years, regardless of which add-on agent they received. The drug class did not shift that rate. For context on what GLP-1 receptor agonist doses are used in clinical trials, see the semaglutide clinical overview on this site.

Why this finding matters in the clinic

GLP-1 receptor agonists, including semaglutide and liraglutide, are established treatments for type 2 diabetes that produce meaningful weight loss and, in some trials, reduce cardiovascular events. In clinical practice, there is a widespread expectation that adding a GLP-1 agonist to basal insulin will, over time, allow the insulin to be tapered or stopped. This study tests that expectation in a real-world, older, predominantly male population and finds it is not reliably borne out^[1].

The finding does not argue against GLP-1 receptor agonists. The drugs still offer glycaemic improvement, weight reduction, and cardiovascular benefits in people with established heart disease, as covered in the ACP obesity guideline explainer on this site. What this study questions is the specific narrative that adding a GLP-1 agonist routinely enables patients to come off insulin. In this population, it did not happen at a higher rate than with cheaper alternatives^[2].

What a target trial emulation can and cannot tell us

A target trial emulation uses observational data to approximate a randomised controlled trial. It is not a randomised trial. Residual confounding is possible: physicians who chose to add a GLP-1 agonist may have done so in patients whose diabetes they judged more tractable, or more insulin-resistant, or better suited to weight loss. The propensity matching reduces but does not eliminate those differences. The study also relies on medication records rather than clinical review to determine discontinuation, so it may miss patients who nominally stayed on insulin but drastically reduced their dose. For background on how semaglutide and related GLP-1 drugs are prescribed and titrated in clinical practice, the semaglutide overview covers the approved indications and dose ranges.

Within those caveats, the design is well-suited to the question: a randomised trial of insulin discontinuation as an outcome would be difficult to conduct and fund, so the emulation fills a gap in the evidence.

The comparison drugs: SGLT-2 and DPP-4 inhibitors

SGLT-2 inhibitors (such as empagliflozin and dapagliflozin) lower blood glucose by blocking kidney glucose reabsorption. They carry their own cardiovascular and renal outcome data and can reduce insulin requirements indirectly by lowering blood glucose load. DPP-4 inhibitors (such as sitagliptin and saxagliptin) enhance the incretin effect by preventing GLP-1 degradation, producing more modest glycaemic lowering. Both classes are generally less expensive than GLP-1 receptor agonists in the US and many European markets^[2].

The parity in insulin discontinuation rates across all three classes does not mean the drugs are equivalent on every outcome. It means that on this specific endpoint, in this specific population, the advantage of GLP-1 agonists that many clinicians and patients expect does not appear in the data. For readers interested in how the classes compare on weight and cardiovascular outcomes, the tirzepatide vs semaglutide comparison page covers the SURPASS-2 trial head-to-head data.

What we do not yet know

Three gaps stand out. First, the study population is largely older men. Women, younger patients, and those with shorter diabetes duration might respond differently. Second, three years may not be long enough: some patients taper insulin slowly over five or more years, and the follow-up window may have cut off before that outcome could be observed. Third, the study does not report on glycaemic control, weight, or cardiovascular events in the three groups. It is focused tightly on insulin discontinuation as an outcome, and readers should not extrapolate its conclusions to the broader clinical value of semaglutide or other GLP-1 receptor agonists. For the broader semaglutide evidence base, including SUSTAIN, STEP, and SELECT trial data, see the semaglutide peptide page on this site.

**Medical disclaimer:** This article is for educational and journalistic purposes only and does not constitute medical advice. Always consult a qualified healthcare professional before making any changes to diabetes medication, including insulin. PeptideMethods.com does not sell, distribute, or facilitate the sale of any peptide or pharmaceutical product.