Glucagon-Like Peptide-1 Receptor Agonist Therapy Effects on Glycemic Control and Weight in a Primary Care Clinic Population

Abstract

Glucagon-like peptide-1 receptor agonist (GLP-1a) medications have been shown in randomized controlled trials (RCTs) to have consistent and impressive effectiveness in lowering hemoglobin A1c (HbA1c) and weight, but limited data exists on the efficacy of GLP-1a medications in clinical practice. We studied the association between GLP-1a therapy and changes in weight and HbA1c in a real-world patient population. In this retrospective cohort study of patients seen in a primary care clinic between 2012-2021, we examined the change in weight and HbA1c over 12 months in a cohort of patients with at least one prescription for a GLP-1a. Within this cohort, treatment was defined as having ≥ 2 GLP-1a prescriptions at a therapeutic dosage separated by ≥ 10 months. The cohort included 693 patients of whom 393 (57%) were treated with GLP-1a therapy. The treatment group had a mean change in BMI of −0.83 kg/m² (±2.88) compared to −0.70 kg/m² (±2.99) in the without GLP-1a group (p=0.57). Treated patients had mean change in HbA1c of −1.00% (±2.07) compared to −0.83% (±1.92) in the without GLP-1a group (p=0.27). For treated and without GLP-1a patients respectively, the proportion of patients with a decrease in BMI was 65% vs. 64% (p=0.86), and the proportion with a decrease in HbA1c was 73% vs. 69% (p=0.28). In clinical practice, GLP-1a therapy was associated with more modest reductions in weight and HbA1c than shown in prior RCTs. As GLP-1a use continues to expand throughout primary care, the real-world impact of this pharmacotherapy will require further evaluation.

Introduction

Obesity and diabetes present an ever-increasing threat to the health of patients in the United States (1, 2). Both conditions are increasingly prevalent and associated with a multitude of negative health outcomes including cardiovascular disease, stroke, cancer, cardiovascular disease, and death (1-5). Glucagon-like peptide-1 receptor agonists (GLP-1a) have emerged in recent years as perhaps the most effective class of medications to manage both diabetes and obesity (6-11). With weight loss and improved glycemic control shown to decrease the risk of negative health outcomes, GLP-1a medications have been incorporated into the guidelines for obesity and diabetes management,(7-9) and their usage has increased dramatically over the past decade (12, 13).

In randomized controlled trials (RCTs), systematic reviews, and meta-analyses, GLP-1a medications have consistently demonstrated statistically significant, and profound, reductions in body weight for patients with and without diabetes (14-21). Clinical trials have also demonstrated that GLP-1a medications significantly reduce glycosylated hemoglobin A1c (HbA1c), with most RCTs demonstrating a −1.0% to −2.0% reduction in HbA1c depending on the specific GLP-1a medication, dosage, and duration of treatment (22, 23). A more limited number of studies have investigated the real-world efficacy of GLP-1a medications in lowering weight and HbA1c (24-30). While results have been variable, these studies suggest GLP-1a medications have a significant, but more modest, impact on weight and HbA1c in practice. Nevertheless, it remains unclear in primary care how the outcomes of patients treated with GLP-1a medications compare to similar patients who are without GLP-1a therapy.

In this study, we aimed to determine the association between prolonged treatment with GLP-1a medications and changes in BMI and HbA1c in a primary care population. We hypothesized that patients treated with therapeutic dosages of GLP-1a medications for one year would demonstrate significantly greater reductions in BMI and HbA1c compared to patients without GLP-1a therapy.

Materials and Methods

We performed a retrospective cohort study of patients seen in an academic primary care clinic with at least one prescription for a GLP-1a medication using electronic health record (EHR) data from a patient-centered medical home between July 2012 (installation of the comprehensive EHR with computerized physician order entry) and December 2021. All data come from the EHR and Enterprise Data Warehouse (EDW) at the Medical University of South Carolina (MUSC). This project was reviewed and approved by the MUSC Institutional Review Board (IRB) as a study of previously existing retrospective data from human subjects with adequate protection for the rights and welfare of the individuals involved.

Patients

We identified patients with at least one prescription (both printed and electronically transmitted to an outpatient pharmacy) for a GLP-1a medication during the study period. The date of the first GLP-1a prescription served as the index for follow-up. Cohort inclusion required patients to have both BMI and HbA1c values at the time of the first GLP-1a prescription and BMI and HbA1c values at least 12 months after the first GLP-1a prescription (Figure). The baseline BMI and HbA1c were the values closest in time to the index GLP-1a prescription but were permitted to come from a window of time from 12 months before to 1 month after the first GLP-1a prescription. The follow-up BMI and HbA1c values were those closest in time to 12 months after the GLP-1a script but included results from 12 to 24 months after the index GLP-1a prescription.

Figure.

Study cohort inclusion criteria and treatment definition. GLP1a=Glucagon-like peptide-1 receptor agonist, BMI=Body mass index, HbA1c=Hemoglobin A1c

Outcomes

Our primary outcomes of interest were the absolute changes in BMI and HbA1c from the time of the first GLP-1a prescription to the first BMI and HbA1c values at least 12 months later. Secondary outcomes included the proportion of patients that experienced a reduction in BMI and HbA1c from baseline until follow-up.

Treatment Definition

Our treatment group included patients who had evidence of long-term treatment with a therapeutic dosage of a GLP-1a medication (Supplementary Table S1). To mitigate selection bias, the full primary cohort included only patients with at least one prescription for a GLP-1a. However, treatment was a dichotomous variable (treated / without GLP-1a) based upon prescription patterns identified during the one-year study period. The treatment group included patients who met all the following criteria: (i) having at least 2 prescriptions for the same GLP-1a medication at a therapeutic dosage; (ii) having at least 2 such prescriptions written at least 10 months apart; and (iii) having evidence of continuous GLP-1a treatment (based upon timing of the prescriptions and the number of refills provided) identified on chart review. Patients who did not meet these criteria in the primary cohort were classified as without GLP-1a.

Other Variables

We collected demographic variables for the cohort including age (continuous), sex (female / male), race (Black / Other / White), and marital status (yes/no). Clinical variables included BMI, HbA1c, weight, systolic blood pressure (SBP), diastolic blood pressure (DBP), total cholesterol (TC), low-density lipoprotein (LDL), high-density lipoprotein (HDL), triglycerides, alanine aminotransferase (ALT), and aspartate aminotransferase (AST), all of which were treated as continuous variables and collected at baseline and follow-up (using the date windows specified for BMI and HbA1c collection). We also collected data on other diabetes and weight loss medications prescribed to patients at the time of the index GLP-1a prescription. Comorbidity variables were identified using Elixhauser algorithms for composites of ICD-9 and −10 codes placed during the study period and included hypertension, diabetes mellitus, hyperlipidemia, cardiovascular disease, and chronic kidney disease (31, 32).

Statistical Analysis

Univariate analyses were performed to describe the cohort overall and by GLP-1a treatment status. Continuous variables were compared by GLP-1a treatment status using two-sample t-tests and categorical variables were compared using Chi square tests. Baseline and follow-up clinical variables were calculated and compared by paired t-tests for the treated and without GLP-1a groups. The absolute change in BMI and HbA1c was calculated for each patient in the cohort. The mean changes in BMI and HbA1c were calculated for the overall cohort and by treatment group. The mean changes in BMI and HbA1c were then compared between the treated and without GLP-1a patients using two sample t-tests. We also calculated the proportion of patients experiencing a reduction in BMI and HbA1c between baseline and follow-up for the overall cohort and based on treatment status. The proportion of patients with a negative change in BMI and HbA1c was compared using Chi square tests. We also developed multivariable linear regression models for the outcomes of change in BMI and change in HbA1c. All predictor variables were pre-specified and included GLP-1a treatment, the demographic variables, and the comorbidity predictors previously described. SAS 9.4 (Cary, NC) was used for all statistical analyses. An analysis of covariance (ANCOVA) was also performed with post-intervention outcome measures used as dependent variables and with primary exposure being the GLP-1a treatment group after adjusting for the baseline outcome measure along with other covariates.

Secondary Analysis

We also performed a post hoc propensity score matching analysis to address the potential bias and confounding inherent in this observational study. From a set of 1,153 clinic patients with no exposure to GLP-1a therapy (including patients never receiving a prescription), at least one HbA1c > 6.5%, and with two pairs of BMI and HbA1c values separated by at least 12 months (to emulate the timing of the primary analysis), we used propensity scores to match a cohort based on all baseline covariates from the primary analysis, baseline HbA1c, baseline BMI, and medication history. We performed the matching strategy looking for the most robust analysis that allowed for statistically similar baseline characteristics between exposure groups (33). We described the propensity score-matched cohort overall and by GLP-1a exposure status, comparing continuous variables with two-sample t-tests and categorical variables with Chi square tests. The absolute change in BMI and HbA1c was calculated for each patient and mean changes in BMI and HbA1c were calculated by GLP-1a exposure group. The mean changes in BMI and HbA1c were compared between GLP-1a treated patients and propensity score-matched patients without GLP-1a therapy using two sample t-tests. For our propensity score-matched analysis, multivariable linear regression models using the outcomes of change in BMI and change in HbA1c were developed using GLP-1a treatment as the primary predictor variable.

Results

The study cohort included 693 patients with a mean age of 56 years and of whom 70% were female, 61% identified as Black, and 95%, 97%, and 35% had hypertension, diabetes, and cardiovascular disease, respectively (Table 1). Of this cohort, 393 patients (57%) had prolonged treatment with a therapeutic dose of a GLP-1a medication. The 393 patients treated with GLP-1a therapy did not significantly differ from the without GLP-1a group by age, sex, race, or marital status. Treated patients had higher mean weight (236.2 vs. 226.4 lbs, p=0.030) and BMI (37.9 vs. 36.4 kg/m², p=0.020) than without GLP-1a patients. A higher proportion of treated patients had diabetes (99.5% vs. 92.7%, p<0.001) and hyperlipidemia (93.6% vs. 87.0%, p=0.003) compared to their without GLP-1a counterparts. A greater proportion of patients in the GLP-1a treatment group were on insulin during the study period (56.5% v. 45.7%, p<0.01). The use of metformin, other oral diabetic drugs, and weight loss medications were similar between the two groups. All other measured variables were similar between groups.

Table 1.

Cohort characteristics overall and by GLP-1a treatment status at baseline.

		Overall	GLP-1a Treatment
Variables			Yes	No	p-value
		n=693	n=393	n=300
Demographics
Age, mean (SD)		56.3 (12.7)	56.4 (12.1)	56.3 (13.5)	0.98*
Sex, % (n)					0.73†
	Female	70.0% (485)	69.5% (273)	70.7% (212)	-
Race, % (n)					0.25†
	Black	61.0% (423)	63.1% (248)	58.3% (175)	-
	White	36.5% (253)	35.1% (138)	38.3% (115)	-
	Other	2.5% (17)	1.8% (7)	3.3% (10)	-
Married, % (n)		44.0% (305)	45.8% (180)	41.7% (125)	0.28†
Resides > 50 miles from clinic, % (n)		9.2% (64)	10.2% (40)	8.0% (24)	0.33†
Clinical variables, mean (SD)
Weight, lbs.		231.9 (58.7)	236.2 (59.0)	226.4 (57.9)	0.03*
BMI, kg/m2		37.2 (8.8)	37.9 (9.0)	36.4 (8.5)	0.02*
Systolic BP, mmHg		132.8 (16.2)	133.3 (16.3)	132.2 (16.1)	0.35*
Diastolic BP, mmHg		77.3 (10.7)	77.5 (10.4)	77.1 (11.1)	0.65*
HbA1c, %		8.9% (2.1%)	9.1% (2.0%)	8.8% (2.2%)	0.11*
ALT, IU/L		28.8 (18.9)	28.4 (16.1)	29.2 (22.1)	0.65*
AST, IU/L		24.8 (14.7)	24.6 (13.2)	25.2 (16.4)	0.65*
Total cholesterol, mg/dL		170.5 (50.1)	170.2 (53.8)	170.8 (45.1)	0.89*
LDL, mg/dL		95.0 (43.5)	95.6 (47.2)	94.2 (38.4)	0.71*
HDL, mg/dL		44.4 (13.3)	43.4 (12.3)	45.7 (14.4)	0.06*
Triglycerides, mg/dL		165.3 (134.2)	167.1 (131.7)	163.0 (137.6)	0.74*
Comorbidities, % (n)
Hypertension		95.0% (658)	96.2% (378)	93.3% (280)	0.09†
Diabetes		96.5% (669)	99.5% (391)	92.7% (278)	<0.001†
Hyperlipidemia		90.8% (629)	93.6% (368)	87.0% (261)	0.003†
Cardiovascular disease		35.4% (245)	34.9% (137)	36.0% (108)	0.76†
Hypothyroidism		21.2% (147)	20.6% (81)	22.0% (66)	0.66†
Chronic kidney disease		30.6% (212)	32.1% (126)	28.7% (86)	0.34†
Smoking		8.5% (59)	7.9% (31)	9.3% (28)	0.50†
Medications, % (n)
Metformin		74.8% (518)	77.1% (303)	71.7% (215)	0.10†
Other oral diabetes agent		45.0% (312)	46.8% (184)	42.7% (128)	0.28†
Insulin		51.8% (359)	56.5% (222)	45.7% (137)	<0.01†
Oral weight loss therapy		6.2% (43)	5.9% (23)	6.7% (20)	0.66†

^*Two sample t-test.

^†Chi square test. ALT=alanine aminotransferase. AST=aspartate aminotransferase. BMI=body mass index. BP=blood pressure. GLP-1a=glucagon-like peptide-1 receptor agonist. HbA1c=hemoglobin A1c. HDL=high-density lipoprotein. Lbs=pounds. LDL=low-density lipoprotein. SD=standard deviation.

The GLP-1a treatment group had significant changes in BMI (37.9 vs. 37.1 kg/m², p<0.001), HbA1c (9.1% vs. 8.1%, p<0.001), total cholesterol (170.2 vs. 163.9 mg/dL, p=0.010), LDL (95.6 vs. 89.3, mg/dL, p=0.009), and triglycerides (167.1 vs. 154.5 mg/dL, p=0.041) from baseline to follow-up (Table 2). The without GLP-1a patient group also had significant reductions in BMI (36.4 vs. 35.7 kg/m², p<0.001), HbA1c (8.8% vs. 8.0%, p<0.001), and total cholesterol (170.8 vs. 169.7 mg/dL, p=0.036).

Table 2.

Body mass index, hemoglobin A1c, and other clinical variables at baseline and follow-up, by treatment group.

	GLP-1a treatment
	Yes			No
Variables, mean (SD)	Baseline	Follow-up	p-value	Baseline	Follow-up	p-value
BMI, kg/m2	37.9 (9.0)	37.1 (9.1)	<0.001‡	36.4 (8.5)	35.7 (8.8)	<0.001‡
HbA1c, %	9.1% (2.0%)	8.1% (2.0%)	<0.001‡	8.8% (2.2%)	8.0% (2.3%)	<0.001‡
Weight, lbs	236.2 (59.0)	231.9 (60.0)	<0.001‡	226.4 (57.9)	222.2 (59.1)	<0.001‡
SBP, mmHg	133.3 (16.3)	133.6 (16.0)	0.76‡	132.2 (16.1)	133.3 (17.6)	0.37‡
DBP, mmHg	77.5 (10.4)	77.2 (11.5)	0.59‡	77.1 (11.1)	77.4 (12.4)	0.77‡
ALT, IU/L	28.4 (16.1)	27.8 (16.3)	0.53‡	29.2 (22.1)	27.6 (19.6)	0.01‡
AST, IU/L	24.6 (13.2)	25.4 (18.6)	0.39‡	25.2 (16.4)	23.7 (11.9)	0.04‡
Total cholesterol	170.2 (53.8)	163.9 (42.4)	0.01‡	170.8 (45.1)	169.7 (49.6)	0.04‡
LDL	95.6 (47.2)	89.3 (36.2)	0.009‡	94.2 (38.4)	94.8 (40.1)	0.21‡
HDL	43.4 (12.3)	44.0 (13.0)	0.19‡	45.7 (14.4)	46.2 (16.4)	0.53‡
Triglycerides	167.1 (131.7)	160.2 (103.6)	0.04‡	163.0 (137.6)	156.7 (126.4)	0.31‡

^‡Paired t-test. ALT=alanine aminotransferase. AST=aspartate aminotransferase. BMI=body mass index. DBP= diastolic blood pressure. GLP-1a=glucagon-like peptide 1 receptor agonist. HbA1c=hemoglobin A1c. HDL=high-density lipoprotein. Lbs=pounds. LDL=low-density lipoprotein. SBP=systolic blood pressure. SD=standard deviation.

For the overall cohort, the mean change in BMI was −0.77 kg/m² (± 2.92) and the mean change in HbA1c was −0.93% (± 2.01) from baseline measurement to the end of follow-up (Table 3). The GLP-1a treatment group had a mean change in BMI of −0.83 kg/m² (±2.88) compared to −0.70 kg/m² (±2.99) in the without GLP-1a group (p=0.566). The GLP-1a treatment group had a mean change in HbA1c of −1.00% (±2.07) compared to −0.83% (±1.92) in the without GLP-1a group (p=0.268). Overall, 64% and 71% of cohort patients had reductions in BMI and HbA1c, respectively. There was no significant difference in the proportion of patients with a decrease in BMI (65% vs. 64%, p=0.864) or HbA1c (73% vs. 69%, p=0.277) between treated and without GLP-1a patients.

Table 3.

Mean changes in body mass index and hemoglobin A1c from baseline to the end of follow-up for the overall cohort and by treatment status.

	Overall	GLP-1a Treatment
Outcomes		Yes	No	p-value
Mean change in BMI, kg/m2 (SD)	−0.77 (2.92)	−0.83 (2.88)	−0.70 (2.99)	0.57*
Mean change in HbA1c, % (SD)	−0.93 (2.01)	−1.00 (2.07)	−0.83 (1.92)	0.27*
% of patients with reduction in BMI (n)	64% (446)	65% (254)	64% (192)	0.86†
% of patients with a reduction in HbA1c (n)	71% (493)	73% (286)	69% (207)	0.28†

^*Two sample t-test.

^†Chi square test. GLP-1a=glucagon-like peptide-1 receptor agonist. BMI=body mass index. HbA1c=hemoglobin A1c

The linear regression model for the outcome of change in BMI demonstrated no significant association of GLP-1a treatment (estimate −0.19 kg/m²; 95%CI −0.63, 0.25 kg/m²) with the outcome (Table 4) after adjusting for other demographic and comorbidity covariates. The second multivariable linear regression model showed no significant association between GLP-1a treatment (Estimate −0.16%; 95%CI −0.4%, 0.14%) and change in HbA1c, after adjusting for other potentially confounding covariates. The ANCOVA sensitivity analyses, which evaluated the association of post-intervention BMI and HbA1c measures with GLP-1a exposure after adjusting for baseline BMI and HbA1c, respectively, and other covariates yielded similar results (Table S2, Table S3).

Table 4.

Adjusted linear regression models for the outcomes of change in body mass index (BMI) (Model 1) and change in HbA1c (Model 2) with GLP-1a treatment as the primary predictor variable.

	Model 1		Model 2
	Outcome = Change in BMI, kg/m2		Outcome = Change in HbA1c, %
Predictors	Estimate (95% CI)	p-value	Estimate (95% CI)	p-value
GLP-1a treatment	−0.19 (−0.63, 0.25)	0.40	−0.16 (−0.46, 0.14)	0.29
Age	−0.03 (−0.05, −0.01)	0.003	−0.01 (−0.03, 0.00)	0.03
Black	0.13 (−0.35, 0.61)	0.59	−0.38 (−0.71, −0.05)	0.02
Male	0.08 (−0.44, 0.59)	0.77	−0.34 (−0.70, 0.01)	0.06
Unmarried	0.01 (−0.46, 0.48)	0.98	−0.14 (−0.46, 0.19)	0.40
Smoking	0.22 (−0.54, 0.99)	0.57	−0.84 (−1.37, −0.31)	0.002
Resides ≥ 50 mi. away	−0.25 (−1.00, 0.49)	0.50	0.03 (−0.48, 0.55)	0.89
Hypertension	−0.11 (−1.17, 0.96)	0.84	−0.07 (−0.80, 0.66)	0.85
Diabetes	0.10 (−1.13, 1.32)	0.88	−0.70 (−1.55, 0.14)	0.10
Hyperlipidemia	0.64 (−0.20, 1.48)	0.13	0.60 (0.02, 1.17)	0.04
CAD	0.04 (−0.43, 0.51)	0.86	0.11 (−0.21, 0.44)	0.49
Hypothyroidism	−1.30 (−1.85, −0.76)	<0.001	−0.03 (−0.41, 0.34)	0.86
CKD	0.005 (−0.49, 0.50)	0.98	0.20 (−0.14, 0.54)	0.25

GLP-1a=glucagon-like peptide-1 receptor agonist. BMI=body mass index. HbA1c=hemoglobin A1c. CAD=coronary artery disease. CI=confidence interval. CKD=chronic kidney disease. Mi.=miles.

For our secondary analysis, we developed a propensity score cohort, matched by demographic, comorbidity, medication, and baseline HbA1c and BMI variables. This propensity score-matched cohort included 740 patients, 393 of whom were exposed to GLP-1a therapy and 347 patients were unexposed (Table S4). Baseline mean BMI (37.9 kg/m² vs. 37.6 kg/m², p=0.636), baseline mean HbA1c (9.1% vs. 8.9%, p=0.347), and all other measured variables were similar between GLP-1a exposure groups. The GLP-1a treatment group had a reduction in HbA1c of −1.00% which was significantly greater than the cohort without GLP-1a therapy (−0.65%, p=0.014) (Table 5). There was no significant difference in change in BMI between the GLP-1a treatment group and the group unexposed to GLP-1a therapy (−0.83 v. −0.82 kg/m², p=0.983). Multivariable linear regression models for the outcomes of change in BMI and HbA1c in the propensity score-matched cohort, after adjusting for other covariates, showed that GLP-1a exposure was associated with a significant reduction in HbA1c (−0.38%; p=0.009) compared to unexposed patients, while there was no significant association between GLP-1a exposure and change in BMI (−0.02; p=0.911) (Table S5).

Table 5.

Mean changes in body mass index (BMI) and hemoglobin A1c (HbA1c) from baseline to the end of follow-up for a propensity score-matched cohort of primary care patients with and without treatment with GLP-1a therapy.

	Overall	GLP-1a Treatment
		Yes	No	p-value
Outcomes	n=740	n=393	n=347
Mean change in BMI, kg/m2 (SD)	−0.82 (2.9)	−0.83 (2.9)	−0.82 (3.0)	0.983*
Mean change in HbA1c, % (SD)	−0.83% (2.0%)	−1.0% (2.1%)	−0.65% (1.9%)	0.014*

^*Two sample t-test. GLP-1a=glucagon-like peptide-1 receptor agonist. BMI=body mass index. HbA1c=hemoglobin A1c.

Discussion

In this retrospective cohort study of a primary care clinic population, we examined the association between GLP-1a treatment and changes in BMI and HbA1c in the real-world setting. Despite improvements in BMI (−0.83 kg/m²) and HbA1c (−1.00%) in patients treated with GLP-1a therapy, there was no significant difference in the degree of change when compared to without GLP-1a patients. A similar proportion of patients treated with GLP-1a therapy experienced a reduction in BMI (65% vs. 64%) and HbA1c (73% vs. 69%) during follow-up compared to without GLP-1a patients.

Prior RCTs have consistently demonstrated significant weight loss with GLP-1a therapy. In a phase-2 RCT, over 52 weeks daily injectable semaglutide at 0.05 – 0.4 mg led to a dose-dependent reduction in body weight ranging from −6.0% to −13.8%, versus −7.8% for liraglutide 3.0 mg daily and −2.3% in the placebo group (18). In the STEP-1 trial, over 68 weeks, patients without diabetes on semaglutide 2.4 mg had a −14.9% loss of body weight compared to −2.4% in the placebo group, and in the STEP-2 trial, patients with diabetes on semaglutide 2.4 mg had a −9.6% body weight change versus −7.0% for semaglutide 1.0 mg and −3.4% with placebo over 68 weeks (19-21). Only a few studies have evaluated GLP-1a efficacy in clinical practice. A 2023 retrospective cohort study of 2405 patients found that over 72 weeks GLP-1a medications were associated with just a 2.2% reduction in body weight, though this study excluded higher dosages of GLP1-a medications used in obesity treatment (24). Nonetheless, these results are consistent with our findings; in our GLP-1a treatment group, treated patients had a −0.83 kg/m² reduction in BMI or −2.2% reduction over 12 months.

In terms of GLP-1a impacts on HbA1c, most RCTs have found a −1.0% to −2.0% reduction in HbA1c with GLP-1a therapy (22, 23). A few studies have looked at the effects of GLP-1a medications on HbA1c in real-world clinical practice (25-30). The TROPHIES observational prospective study of 2005 patients found that over 12 months dulaglutide reduced HbA1c −1.18% and liraglutide reduced HbA1c −1.15% (26). A systematic review of real-world studies of injectable semaglutide found a wide range of HbA1c reduction from −0.3% to −3.4%, while a retrospective 12-month study of 11,211 patients on liraglutide, dulaglutide, or exenatide found an HbA1c reduction of −0.77% to −1.00% (25, 30). Prior results are comparable to our findings, with our GLP-1a treatment group having a −1.00% reduction in HbA1c over 12 months.

To our knowledge, our study is among the first to investigate the impact of GLP-1a treatment on HbA1c and BMI with comparison to a similar group of patients without GLP-1a therapy in real-world practice. Based on prior clinical trial data, it is generally accepted that GLP-1a therapy will offer patients superior HbA1c lowering effect and BMI reduction than alternative treatments available (10, 11, 34). However, in our study of a primary care clinic population, we found no significant difference in HbA1c or BMI change between patients treated with GLP-1a therapy and without GLP-1a patients. One presumable cause of this rests in the inherent differences between the stringent RCT environment as compared to a real-world primary care clinic population. In the latter, weight loss and glycemic control exist in the background of broader complex care goals which are impacted by socioeconomic factors and competing priorities. For example, in the ambulatory setting, it’s common for patients receiving ongoing prescriptions for therapeutic dose GLP-1a medications to suffer treatment interruptions due to shifts in insurance coverage, increases in out-of-pocket expense, or supply-chain issues comprising medication availability. It may also be the case that the decision to prescribe a GLP-1a medication in practice serves as an indicator that a patient is engaged and motivated to manage their disease, regardless of their ability to obtain or tolerate a GLP-1a medication. That patients without GLP-1a therapy experienced a similar effect to treated patients may reflect that these patients are successfully utilizing alternative medications and lifestyle changes when a GLP-1a is not an option.

There are important limitations to our investigation related to study design. This was a single center study with baseline patient characteristics that may not be generalizable to all primary care populations. We were also unable to account for behavioral variables, including lifestyle factors, alcohol use, income, and insurance status which could have potentially impacted our treated and without GLP-1a patient groups differently. Regarding the retrospective nature of our study, the inclusion of all patients with a GLP-1a prescription in the primary study cohort was intended to minimize differences between treated and without GLP-1a patients, though some baseline differences (known and unknown) were present. Our without GLP-1a group had a significantly lower baseline BMI and a lower incidence of diabetes than the GLP-1a treatment group. However, if anything, this might be expected to bias the results towards showing a significant difference between the treated and without GLP-1a groups. Furthermore, adjusted linear regression models and an ANCOVA analysis supported our primary findings even with adjustment for differences in baseline outcome measures and potentially confounding covariates. The GLP-1a treated group was also prescribed insulin at a somewhat higher rate at baseline, but it is difficult to predict what impact this may have had on outcomes. We also compared the GLP-1a treatment group to a propensity matched group of patients unexposed to GLP-1a therapy (including patients never prescribed a GLP-1a). This comparison showed no significant difference in BMI change, though there was a significant difference in HbA1c reduction between the GLP-1a treatment group and the group without GLP-1a therapy. It is possible that by examining a group who never had a GLP-1a prescription, we may have selected for a group of patients where a higher HbA1c target was deemed acceptable.

The criteria we used to define treatment also presents potential limitations. While using only patients with a GLP-1a prescription for our primary analysis allowed for the identification of treated and without GLP-1a patient groups with similar baseline characteristics and treatment goals, this method has the potential to misclassify patients. To mitigate this, we used strict chart review criteria which required treated patients to have therapeutic dose prescriptions for the same GLP-1a medication separated by > 10 months. This was felt to be a strong predictor that a patient was able to obtain and tolerate a ramp-up of the medication and received an ongoing prescription for the duration of the study period.

In conclusion, our findings suggest that GLP-1a treatment led to significant reductions in BMI and HbA1c, though these changes did not differ significantly from patients without GLP-1a therapy. Further study is needed to identify why GLP-1a medications may not be having the expected impact on HbA1c and BMI in clinical practice. Specifically, future investigations would be helpful to assess if specific patient factors predict more significant HbA1c and BMI lowering effect, whether specific GLP-1a drugs or dosages are more efficacious in clinical practice, and how medication cost and availability impact treatment plans.

Summary.

What is already known on this topic?

Glucagon-like peptide-1 receptor agonists (GLP-1a) have consistently demonstrated significant weight loss and glucose reduction in clinical trials and offer hope for radically transforming the care of patients with obesity and diabetes in primary care.

What this study adds?

In this study of real-world GLP-1a use, though patients prescribed GLP-1as had significant weight loss and reduction in hemoglobin A1c (HbA1c), changes in body mass index and HbA1c did not differ from patients prescribed GLP-1as that did not consistently use the therapy.

How this study might affect research, practice, or policy?

As GLP-1a prescribing and use increases for the management of the ongoing obesity epidemic, rigorous evaluation of the therapy’s impact in real-world primary care, compared to its effectiveness in clinical trials, is needed.