GLP-1RA Type 1 Diabetes Trials Criticized for Inadequate Hypoglycemia Reporting

A warning from a new clinical critique is putting a spotlight on a potentially dangerous blind spot in research on glucagon-like peptide-1 receptor agonists, or GLP-1RAs, used alongside insulin in people with type 1 diabetes. These drugs have attracted intense interest because they can improve body weight, blood pressure, cholesterol levels and glycated hemoglobin, commonly known as HbA1c. But a letter published in Endocrinology, Diabetes & Metabolism argues that the evidence base remains incomplete unless it reports hypoglycemia—the potentially life-threatening fall in blood glucose—with the same precision as the benefits. The authors say this omission could leave clinicians with an inaccurately reassuring picture of add-on treatment, particularly because people with type 1 diabetes depend on injected or infused insulin to survive. Their concern is not that every GLP-1RA prescription inevitably causes dangerous hypoglycemia, but that trials must show clearly when, how often and how severely low-glucose events occur. Without that information, a modest improvement in HbA1c may be difficult to interpret safely.

The critique responds to a systematic review and meta-analysis that combined results from 21 randomized controlled trials involving 3,417 people with type 1 diabetes. That analysis reported statistically significant improvements in several cardiometabolic measures when GLP-1RAs were added to insulin, including systolic and diastolic blood pressure, low-density lipoprotein cholesterol, C-reactive protein, body weight, body-mass index and HbA1c. The reported weighted mean difference in HbA1c was −0.21 percentage points, a relatively small average reduction but one that may still matter clinically when combined with weight loss or lower insulin requirements. The responding authors commend the breadth of the review and its use of the PRISMA 2020 framework for systematic reviews. However, they say that the absence of a systematic evaluation of hypoglycemia prevents a balanced assessment of the treatment’s benefit-to-risk ratio. For an intervention layered onto insulin, they argue, low blood glucose should be treated as a central safety endpoint rather than a secondary detail.

The biological rationale for the warning lies in how GLP-1RAs alter glucose regulation. These medicines mimic the activity of glucagon-like peptide-1, a hormone released by the intestine after eating. In people with functioning pancreatic beta cells, GLP-1 signaling can enhance insulin secretion when glucose levels are elevated, suppress glucagon release after meals and slow gastric emptying, reducing the speed at which glucose enters the bloodstream. In type 1 diabetes, beta-cell insulin production is severely impaired or absent, so the glucose-lowering effect of a GLP-1RA occurs in a system already controlled by externally administered insulin. Suppressing post-meal glucagon may reduce one of the body’s defenses against falling glucose, while slower gastric emptying can create a mismatch between the timing of food absorption and the action of injected insulin. The drugs may also reduce appetite and food intake, potentially changing carbohydrate consumption without an immediate, carefully adjusted reduction in insulin. These interacting effects make dose, timing, meals, exercise and individual physiology especially important.

Evidence from the ADJUNCT ONE trial illustrates why the issue has become contentious. In that randomized trial, people with type 1 diabetes received liraglutide, a GLP-1RA, in addition to insulin. At the 1.8-milligram dose, symptomatic hypoglycemia occurred at a rate of 16.5 events per patient per year, compared with 12.3 events per patient per year among participants receiving placebo. The higher rate was not the only concern: increased hyperglycemia with ketosis was also observed, leading investigators to conclude that the clinical usefulness of liraglutide in this population was limited. Ketosis occurs when the body begins producing ketone bodies because it cannot use glucose effectively; in type 1 diabetes, substantial ketone accumulation can progress to diabetic ketoacidosis, a medical emergency. The combination of low-glucose events and ketosis is particularly challenging because reducing insulin to prevent hypoglycemia can, if done too aggressively or without adequate monitoring, increase the risk of inadequate insulin delivery and ketone production.

The ADJUNCT TWO trial provided a similar signal at a lower liraglutide dose. Symptomatic hypoglycemia occurred at 21.3 events per patient per year with 1.2 milligrams of liraglutide, compared with 16.6 events per patient per year with placebo. Event rates expressed per patient-year account for different follow-up durations and estimate how many episodes would occur if one participant were observed for a full year, but they do not by themselves reveal whether the events were mild, nocturnal, recurrent or severe enough to require outside assistance. That distinction matters because “hypoglycemia” can cover a wide range of experiences, from a measurable glucose reduction with few symptoms to confusion, seizure, loss of consciousness or the need for emergency treatment. The authors therefore emphasize that insulin doses may need careful adjustment whenever a GLP-1RA is introduced, increased or stopped. A treatment that improves average glucose can still worsen short-term glucose instability if the insulin regimen, meal pattern or counter-regulatory responses are not considered.

The critique also points to real-world safety signals identified in the US Food and Drug Administration’s Adverse Event Reporting System, or FAERS, across several GLP-1RA medicines. Such databases can detect patterns that deserve investigation, but they cannot establish incidence or prove that a drug caused an event. Reports may be influenced by publicity, prescribing volume, patient characteristics and incomplete information. Even so, the authors say these signals add to the reason for demanding more consistent trial reporting. A previous meta-analysis of GLP-1RAs as add-on therapy did not find a statistically significant increase in severe hypoglycemia across 13 studies. But a non-significant pooled result is not equivalent to evidence that no risk exists. Trials may differ in their definitions, glucose-monitoring methods, insulin-adjustment protocols, participant populations and thresholds for recording events. If severe episodes are rare, studies may also lack the statistical power to detect meaningful differences. Combining unlike measurements can make a genuine safety signal appear smaller or disappear altogether.

The authors propose that future systematic reviews should define hypoglycemia in advance as a primary safety outcome and divide it into clinically informative categories. They recommend separating mild, moderate and severe episodes, distinguishing nocturnal from daytime events, and analyzing results by GLP-1RA agent and dose. They also support use of the International Hypoglycemia Study Group’s three-level classification, which is designed to standardize reporting across clinical trials. Level 1 corresponds to a glucose concentration of 3.9 millimoles per liter or less, equivalent to 70 milligrams per deciliter, and signals a threshold at which action may be needed. Level 2 is glucose below 3.0 millimoles per liter, or 54 milligrams per deciliter, a level considered clinically significant. Level 3 does not depend on a specific glucose measurement: it describes severe cognitive impairment in which a person needs help from another individual to recover. This framework has been endorsed by major diabetes organizations and the European Medicines Agency, making it a practical common language for comparing studies.

Standardized reporting could also help researchers answer questions that average event counts conceal. Are low-glucose episodes concentrated after meals, overnight or during exercise? Do they occur primarily during dose escalation, when appetite and carbohydrate intake are changing? Are some agents or doses associated with more events than others? Does continuous glucose monitoring identify substantially more level 1 or level 2 episodes than conventional finger-stick testing? And do people with reduced awareness of hypoglycemia experience fewer symptoms before glucose reaches dangerous levels? These details could influence how doctors select patients, adjust basal and mealtime insulin, counsel about food and physical activity, and monitor treatment after initiation. They could also clarify whether a reduction in HbA1c reflects genuinely improved glucose control or an average produced by alternating lower glucose with episodes of hyperglycemia. For patients, the difference is crucial: a lower HbA1c is not automatically safer if it comes at the cost of more severe glucose excursions.

The debate arrives as GLP-1RAs continue to reshape diabetes and obesity medicine, creating pressure to test their usefulness in populations beyond their most established indications. In type 1 diabetes, however, these drugs are not replacing insulin; they are being considered as supplementary therapy, potentially targeting weight, insulin resistance and cardiometabolic risk. That makes the safety context different from treatment in type 2 diabetes, where the body may still produce some insulin and the risk of severe hypoglycemia can be lower when GLP-1RAs are used without insulin or insulin-releasing drugs. The letter’s central message is therefore one of evidence quality rather than a blanket rejection of the drug class. The authors report no funding or conflicts of interest and generated no new data; their work is a critical appraisal of published evidence. They argue that GLP-1RAs can only be recommended confidently for type 1 diabetes when their metabolic advantages are presented alongside a transparent, standardized account of hypoglycemia. Until then, clinicians and patients may be seeing only half of the treatment equation.