Heterogeneity in response to GLP-1 receptor agonists in type 2 diabetes in real-world clinical practice: insights from the DPV register – an IMI-SOPHIA study

Glucagon-like peptide-1 receptor agonists (GLP-1 RAs) are a cornerstone in type 2 diabetes management, improving glucose-stimulated insulin secretion, slowing gastric emptying and reducing blood glucose levels [1, 2]. Besides these well-established effects, they promote weight loss, mainly through their action in the brain [1, 2], and are therefore also approved for weight management, even in those with normal glucose regulation. In addition to reducing HbA_1c and body weight, accumulating evidence underscores their role in reducing risk for diabetes complications and even mortality [2]. GLP-1 RAs show promise in other conditions linked to diabetes, including hypertension, hyperlipidaemia, liver steatosis and subclinical inflammation [2].

However, despite their efficacy, individual responses to GLP-1 RAs vary significantly, particularly in terms of glycaemic and weight management [3 –6], and this variability in real-world clinical practice, as well as clinical features linked to this, still represent a critical gap in our knowledge. Gaining insights into the heterogeneity could contribute to precision medicine strategies in diabetes care. Currently, precision medicine in diabetes emphasises patient-specific phenotypic data and distinct diabetes endotypes [7]. Incorporating treatment response variability, such as responses to GLP-1 RAs, could further refine and enhance these personalised treatment approaches. In this study, we analysed HbA_1c and body weight changes in the large Diabetes Patient Follow-up (DPV) dataset, focusing on the first 6 months after initiation of a long-acting GLP-1 RA. Success was clinically defined as a 5% relative body weight reduction and a 5.5 mmol/mol (approx. 0.5%) absolute HbA_1c reduction. The analysis was part of the IMI-Stratification of Obesity Phenotypes to Optimize Future Obesity Therapy (IMI-SOPHIA) project.

The DPV registry is a long-term, prospective registry collecting real-world data on diabetes diagnoses, management and outcomes across four European countries (Germany, Austria, Switzerland and Luxembourg) [8, 9]. It aggregates anonymised data from currently 521 centres on nearly 700,000 individuals with diabetes, allowing multicentre analyses and benchmarking of diabetes care. Data are pseudonymised and transferred biannually to Ulm University, Germany, where incomplete or implausible data are verified with the respective centres [8, 9]. The DPV registry is representative of pediatric diabetes care and adults with diabetes treated in diabetes-specialised practices in the four participating European countries [10]. Race and ethnicity are not well documented in the DPV registry for the participant group included and were therefore not analysed in this study. The DPV initiative and pseudonymised data analysis are approved by the ethics committee of Ulm University, Germany (314/21), and by local ethics boards of participating centres.

The study cohort had a median age at GLP-1RA initiation of 60.0 years (IQR 52.1, 68.1), with a median BMI of 34.9 kg/m² (31.0, 40.0) and a median HbA_1c of 60 mmol/mol (52, 72) (7.7% [6.9, 8.7]). Full baseline characteristics of the study cohort are provided in ESM Table 1. In total, 1890 of 4467 participants changed co-medication within the follow-up period. Aside from a decrease in additional dipeptidyl-peptidase 4 inhibitor (DPP-4i) use from 31.5% to 14.3% of participants, rates for other co-medications were relatively stable, with <5% of participants changing other co-medications (insulin, +1%; metformin, −0.1%; sodium–glucose cotransporter 2 inhibitors [SGLT-2is], +1.5%; sulfonylureas/glinides, −1.8%; thiazolidines, −0.4%).

The likelihood of belonging to the only weight responder group compared with the HbA_1c and weight responder group was similar between sexes (p=0.12) but increased with age (1.31 [1.17, 1.47], p<0.001) and longer diabetes duration (1.28 [1.18, 1.38], p<0.001). Higher baseline HbA_1c reduced the likelihood of belonging to this group (0.52 [0.47, 0.59], p<0.001) (Fig. 3b). The presence of microvascular complications was linked to a higher likelihood of membership of the only weight responder group (Fig. 3e, p=0.02).

When comparing the only HbA_1c responder and only weight responder groups, older people and those with a longer diabetes duration were more likely to belong to the only weight responder group (1.31 [1.17, 1.47] and 1.24 [1.16, 1.33], respectively; both p<0.001). Conversely, higher baseline HbA_1c and eGFR were linked to membership in the only HbA_1c responder group (2.15 [1.94, 2.39], p<0.001, and 1.32 [1.16, 1.51], p=0.001, respectively) (Fig. 3c). Complications and comorbidities had no impact on group membership (Fig. 3f).

We next analysed the associations between baseline characteristics and overall weight or HbA_1c changes across all groups using linear regression. Standardised β-coefficients (with 95% CIs) are presented in ESM Fig. 4. Weight reduction was most strongly associated with sex, with female participants showing greater reductions. Additionally, higher baseline body weight and lower eGFR were associated with greater weight loss (all p<0.05). Higher baseline HbA_1c was associated with greater HbA_1c reductions, while longer diabetes duration was associated with smaller HbA_1c reductions (both p<0.001).

We detected marked heterogeneity in glycaemic and weight responses on GLP-1 RA initiation in adult participants with type 2 diabetes under real-world conditions. As the long-acting (more potent) GLP-1 RAs liraglutide, semaglutide and dulaglutide [1, 2] are becoming the standard in treatment, we focused on these agents. A surprising finding was that only 14% of participants achieved marked reductions in both HbA_1c and body weight, with many responding primarily in one area only. Factors such as sex, age, HbA_1c, BMI and diabetes duration at treatment initiation were linked to the likelihood of response.

It is possible that participants in the group with neither weight loss nor glucose reduction discontinued or only sporadically used the drug, so we focused on those who achieved a reduction in either outcome, as it is more likely that these individuals used the drug as prescribed. Among those achieving reductions, the proportion showing meaningful responses in both areas was lower than in clinical trials, which often involve highly selected, motivated populations that might therefore not be representative of the wide spectrum of individuals with type 2 diabetes in routine clinical practice [3]. In addition, the stricter management in trials may explain some of the discrepancy with findings from real-world settings, a common observation in studies of different drug classes [13]. Our data provide some insights into factors influencing treatment response, but developing robust biomarkers may be necessary for precision medicine approaches at the individual level. Although higher baseline HbA_1c predictably associates with greater glycaemic reductions, the marked heterogeneity in response (as visualised in Fig. 1) highlights that baseline HbA_1c alone does not fully capture treatment variability. This underscores the importance of identifying additional clinical or biological factors that can better predict glycaemic response to GLP-1 RA treatment.

Consistent with most previous trials, participants with higher pretreatment weight were more likely to lose more weight on GLP-1 RA initiation, with women achieving greater weight reductions than men [14 –16]. However, despite assumptions that women experience more side effects from GLP-1 RAs than men [15], they were not over-represented in the non-success group in our analysis. Unexpectedly, we found an association between reduced kidney function at baseline and greater weight reduction, which warrants further investigation.

Individuals with diabetes generally experience less weight reduction with GLP-1 RAs than those without diabetes [2, 6]. We found that higher HbA_1c was associated with a lower likelihood of marked weight reduction, consistent with a recent meta-analysis suggesting that hyperglycaemia may impact weight management [17]. Further research is needed to explore how high glucose levels might reduce the weight loss response to GLP-1 RAs.

Our findings confirm that longer diabetes duration is associated with a lower glycaemic response to GLP-1 RAs [3, 6, 16], probably due to declining beta cell function, as intact insulin secretion is necessary for a robust glycaemic response to this drug class [1, 3]. Genetic factors may also influence the glycaemic response to GLP-1 RAs, although their impact is uncertain, while a genome-wide study did not find a link between genetic variants and weight loss response to this drug class [3], suggesting that different mechanisms are involved in the glycaemic and weight loss effects of GLP-1 RAs.

One strength of our analysis is its focus on individuals newly initiated on GLP-1 RAs, minimising prescription bias. However, the observational nature of our study limits the availability of variables such as treatment adherence or lifestyle factors, which may contribute to response heterogeneity. In future studies, it will be interesting to explore the heterogeneity in response to new co-agonist drugs, which may have even greater glycaemic and weight loss effects [2].

In conclusion, our analysis reveals substantial variability in GLP-1 RA responses in individuals with type 2 diabetes in routine clinical practice, with only a minority of individuals achieving both glycaemic and weight reductions in the first 6 months of treatment. The real-world disparities compared with clinical trials underscore the need for broader studies to better predict response and guide individualised treatment strategies in type 2 diabetes.

Below is the link to the electronic supplementary material. ESM (PDF 524 KB)