Real-World Effectiveness of Tirzepatide versus Semaglutide on HbA1c and Weight in Patients with Type 2 Diabetes

Abstract

Introduction

To evaluate real-world hemoglobin A1c (HbA1c) and weight change in adults initiating treatment with tirzepatide (dual glucose-dependent insulinotropic polypeptide and glucagon-like peptide-1 receptor agonist [GLP-1 RA]) or injectable semaglutide (GLP-1 RA) indicated for type 2 diabetes (T2D) management.

Methods

This retrospective analysis utilized the Healthcare Integrated Research Database® to identify adults with T2D starting tirzepatide or injectable semaglutide between May 13, 2022 and May 29, 2023. GLP-1 RA naïve and non-naïve cohorts were identified based on the history of GLP-1 RA use within ≤ 6 months of initiation. Propensity score matching balanced 6-month baseline characteristics between groups. HbA1c and weight changes were assessed from initiation to 12 months for matched patients with HbA1c and weight data at both time points.

Results

Both matched naïve cohorts were comprised of 10,702 patients (tirzepatide: 1399 with HbA1c data and 454 with weight data; semaglutide: 1173 with HbA1c data and 432 with weight data). Mean baseline HbA1c and weight were 7.8% and 112.4 kg, respectively, for the tirzepatide group and 7.8% and 110.7 kg for the semaglutide group. Both matched non-naïve cohorts were comprised of 5577 patients (tirzepatide: 792 with HbA1c data and 296 with weight data; semaglutide: 738 with HbA1c data and 224 with weight data). Mean baseline HbA1c and weight were 7.7% and 112.5 kg for tirzepatide, and 7.9% and 108.5 kg for semaglutide. Tirzepatide was associated with greater mean reductions in HbA1c (naïve: – 1.3% vs. – 0.9%; non-naïve: – 0.9% vs. – 0.6%; p < 0.001) and weight (naïve: – 10.2 kg vs. – 6.1 kg; non-naïve: – 7.9 kg vs. – 3.7 kg; p < 0.001) than semaglutide.

Conclusions

Patients with T2D starting tirzepatide had greater HbA1c and weight reductions at 12 months post-initiation than those on injectable semaglutide, regardless of previous GLP-1 RA use, consistent with previous clinical trial results.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13300-025-01794-9.

Plain Language Summary

Type 2 diabetes (T2D) is a serious health issue, often linked to other health conditions like obesity. There is little information on the real-world effects of tirzepatide (a dual glucose-dependent insulinotropic polypeptide and glucagon-like peptide-1 receptor agonist [GLP-1]) compared to semaglutide (a GLP-1 receptor agonist only) in helping people with type 2 diabetes improve their blood sugar control (HbA1c) and lose weight. This study used real-world data from the Healthcare Integrated Research Database (HIRD®), including adults with T2D who initiated either tirzepatide (Mounjaro®) or semaglutide (Ozempic®). Patients with and without past use of GLP-1 receptor agonists were looked at separately. HbA1c and weight were measured at the start of treatment (baseline) and again after 12 months. After 12 months, people who initiated tirzepatide achieved lower HbA1c levels and lost more weight than those who initiated semaglutide. Among GLP-1 naïve patients, tirzepatide users reduced their HbA1c by 1.3% and lost 10.2 kg, while semaglutide users reduced their HbA1c by 0.9% and lost 6.1 kg, on average. Tirzepatide was also more effective at reducing HbA1c and weight than semaglutide among people with prior GLP-1 receptor agonist use. Clinical trials show how well and how safely medicines work in controlled settings. Once available for use in routine medical practice, real-world studies give additional information about how the medicines work in everyday medical care. This analysis compares two approved medicines used to treat T2D in real life and may help doctors and health leaders make decisions to improve care for people with this condition.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13300-025-01794-9.

Key Summary Points

Why carry out the study?

Few real-world studies have compared the effectiveness of tirzepatide with semaglutide in patients with type 2 diabetes.

We used data from an administrative claims database to provide real-world insights on glycated hemoglobin (HbA1c) and weight changes after 12 months for patients with type 2 diabetes who started tirzepatide (Mounjaro®) or injectable semaglutide (Ozempic®).

What was learned from the study?

Patients with type 2 diabetes starting tirzepatide (any dose) had greater HbA1c and weight reductions at 12 months post-initiation than those starting injectable semaglutide (any dose).

The study showed that tirzepatide was more effective than injectable semaglutide in real-world settings, aligning with clinical trial results; this may provide additional information to help clinicians and policymakers aiming to improve treatment strategies for type 2 diabetes.

Introduction

The increased use of real-world evidence has transformed the understanding and evaluation of the safety and efficacy of medical interventions in clinical practice [1]. Real-world evidence bridges the gap between clinical research and practice by drawing on data from everyday health care environments, shedding light on how treatments perform across diverse patient groups and practical settings. With the rapid development of treatments for people with type 2 diabetes (T2D), guidelines increasingly recommend multiple treatment options depending upon patient characteristics and treatment goals [2–7]. Alongside clinical guidelines, real-world evidence studies may provide important clinical insights to help inform clinical decisions, optimize patient outcomes, and improve health care delivery for this patient population.

T2D is a significant global health concern, with an estimated 537 million adults affected in 2021, projected to increase to 783 million by 2045 [8], and is associated with numerous comorbidities and complications (e.g., cardiovascular disease, nephropathy, retinopathy, lung diseases, mental health, and inflammatory conditions) [2, 3, 9–11]. Obesity is a particularly prominent associated comorbidity (63% of patients with T2D) [4], which exacerbates insulin resistance, complicates disease management, and creates a cycle of increasingly poor clinical outcomes if not effectively managed [5].

The American Diabetes Association (ADA) guidelines recommend T2D treatment to address cardiovascular and renal risk factors, prioritizing concurrent glycemic control and weight management [6, 7]. The American Association of Clinical Endocrinology, ADA, and European Association for the Study of Diabetes recognize that achieving hemoglobin A1c (HbA1c) targets of ≤ 6.5% and weight loss ≥ 10% can improve long-term outcomes [12–15]. Standard treatment approaches include lifestyle modifications, oral antihyperglycemic agents like metformin, and injectable therapies such as insulin and glucagon-like peptide-1 (GLP-1) receptor agonists (RAs).

GLP-1 RA treatments have demonstrated improvements in glycemic control and weight loss [16, 17]. Semaglutide, a GLP-1 RA, is recognized for its efficacy in lowering blood glucose levels and promoting weight loss [18]. Tirzepatide, a dual glucose-dependent insulinotropic polypeptide (GIP) and GLP-1 RA, provides enhanced glycemic control and significant weight reduction through its dual-action mechanism [19, 20]. The SURPASS-2 clinical trial reported greater body weight reduction and superiority of tirzepatide (5, 10, and 15 mg) versus semaglutide (1 mg) in reducing HbA1c in participants with T2D [21]. The SURPASS-2 results underscore tirzepatide’s potential advantages in both glycemic control and weight management over semaglutide in patients with T2D. However, a direct comparison of tirzepatide and semaglutide, in a real-world setting, is useful to help understand their relative effectiveness in relation to that from clinical trials. This is one of the first real-world studies comparing HbA1c and weight change in GLP-1 RA-naïve and non-naïve patients with T2D starting tirzepatide (all approved doses) versus injectable semaglutide (all approved doses) and complements the findings from established clinical trials.

Methods

Data Source

This retrospective, observational cohort analysis utilized the Healthcare Integrated Research Database (HIRD®), a United States (U.S.) based administrative claims database that includes information on beneficiary eligibility, medical and pharmacy claims, as well as electronic health records (EHR) and laboratory data for a subset of all members. Demographic and health care-related characteristics of individuals in the HIRD are sourced from routinely updated real-world data (e.g., health insurance claims). Additionally, HIRD demographic characteristics are closely aligned to the U.S. Census population in terms of sex, age, and regional residence [22, 23]. Information on members’ race and ethnicity was collected from enrollment details, self-reports, EHR data, and algorithm-derived estimations based on names and geography [24]. Assignment of race and ethnicity using these sources was internally validated against self-reports, showing strong concordance.

Study Population

Inclusion criteria included: ≥ 1 pharmacy claim for any dose of T2D-indicated tirzepatide (Mounjaro® only) or injectable semaglutide (Ozempic® only) (index medications; hereafter referred to as tirzepatide and semaglutide) between May 13, 2022, and May 29, 2023, with the date of the first claim serving as the index date. Eligible patients were ≥ 18 years old on the index date and had ≥ 2 medical claims with a diagnosis of T2D (ICD-10-CM E11.xx), in any position, in any place of service, at any time prior to the index date, going back to January 1, 2016. Patients were required to have continuous enrollment in medical and pharmacy benefits during the six-month pre-index period and the 12-month post-index period. Exclusion criteria included any diagnosis in medical claims for gestational diabetes or type 1 diabetes during the 6-month pre-index period, any claims related to bariatric surgery or other obesity-related procedures from January 1, 2016, through 12 months post-index, and patients with pregnancy, labor, or delivery claims from 6 months pre-index to 12 months post-index.

Eligible patients were first stratified according to their index medication, either tirzepatide or semaglutide, excluding anyone prescribed both on the index date. Further stratification was based on prior use of GLP-1 RAs indicated for T2D during the 6-month pre-index period. Those with no previous fills within the 6-month pre-index period were categorized as GLP-1 RA naïve, while those with ≥ 1 fill for a non-index GLP-1 RA were categorized as GLP-1 RA non-naïve. No patients were prescribed tirzepatide during the 6-month pre-index period, as it had not yet been approved for T2D at that time. The study compared GLP-1 RA naïve (henceforth referred to as naïve) patients initiating tirzepatide to those initiating semaglutide, as well as GLP-1 RA non-naïve (henceforth referred to as non-naïve) patients initiating tirzepatide to those initiating semaglutide. No statistical comparisons between the GLP-1 naïve and GLP-1 non-naïve cohorts were conducted.

Variables

This study evaluated demographic characteristics of eligible patients, including age, sex, race, ethnicity, and geographic region, as of the index date. All baseline characteristics, including comorbidities, medication usage, and laboratory values, were analyzed over the 6-month pre-index period. All outcomes, including treatment patterns, were analyzed over the 12-month post-index period. Treatment patterns included dosing and persistence information. A dose change was defined as any increase or decrease from the index dose, while persistence was defined as the absence of a gap between fills (taking into account days’ supply and stockpiling) longer than 45 days [25].

Two co-primary and five key secondary outcomes were designated for hypothesis testing. The co-primary outcomes were absolute change in HbA1c (%) and weight (kg) from baseline to follow-up. Key secondary endpoints were the percent change in weight from baseline to follow-up, proportion of patients who achieved target follow-up HbA1c levels of < 5.7% or ≤ 6.5%, adherence (measured by ≥ 80% proportion of days covered [PDC]) [26] and cost per responder (mean cost of antihyperglycemic medication divided by the proportion of patients who achieved HbA1c ≤ 6.5%). Additionally, the study included seven exploratory endpoints, including the proportion of patients achieving various alternative follow-up HbA1c targets (< 7%, < 8%), weight reduction benchmarks (≥ 5%, ≥ 10%, ≥ 15%), and a composite outcome, defined by patients achieving a follow-up HbA1c level of ≤ 6.5% simultaneously with a weight reduction of ≥ 10%. Baseline HbA1c and weight values were assessed from 90 days before to 14 days after the index date, while follow-up values were assessed within ± 45 days from the 12 months post-index date. These variables were evaluated among eligible patients with available HbA1c or weight data for respective endpoints. Baseline and follow-up weight or HbA1c data were required for both co-primary endpoints, percent change in weight, and all weight reduction measures. Follow-up HbA1c data were required for the patients who achieved HbA1c targets. Baseline and follow-up weight data and follow-up HbA1c alone were required for the composite outcome.

Analysis

To address potential confounding bias due to non-random treatment allocation, a causal inference design using propensity score (PS) matching [27] was employed. PS was calculated with logistic regression (dependent variable = 1 if patient indexed on tirzepatide, and = 0 if on semaglutide) and used for 1:1 nearest-neighbor matching between tirzepatide and semaglutide initiators. Matching was performed separately for GLP-1 RA naïve and non-naïve patients. Balance was assessed using standardized mean differences (SMDs), reported as absolute values. Although there is no universally agreed upon criterion as to what threshold indicates imbalance, we explored balance using two commonly used thresholds, < 0.1 [28] for key covariates from a clinical perspective (age, sex, HbA1c, weight, antihyperglycemic medication use, and the adapted Diabetes Complications Severity Index (aDCSI) score) and < 0.25 [29] for measures not regarded as key covariates.

After matching, primary and secondary outcomes were compared within the subgroups (naïve and non-naïve) of matched patients with available weight and HbA1c. Leveraging the large sample size available, unpaired t tests were applied to continuous outcomes and chi-square tests to categorical outcomes [30]. Holm’s method was employed to adjust for multiple comparisons, applying a two-sided significance level of α = 0.05 [31]. Holm’s method was applied in a gated manner, first to co-primary outcomes and then, if these were statistically significant, for key secondary outcomes. Inferential testing was performed for exploratory outcome tests; p values were reported but claims of significance were not made [32]. While we confirmed balance in the subgroups of matched patients used to evaluate the primary outcomes, we also conducted ordinary least squares regression analyses to generate doubly robust estimates and address potential residual confounding in the comparison of co-primary outcomes between tirzepatide and semaglutide initiators. Although balance in baseline characteristics was not further confirmed in the subgroups used for secondary outcome analyses, Rassen et al. have demonstrated that a properly specified PS remains valid in large subpopulations [32]. Additional details on PS matching and analysis are included in the supplemental materials.

All analyses were executed using a combination of the Instant Health Data platform provided by Panalgo (Boston, MA), R (version 4.0.2, R Foundation for Statistical Computing, Vienna, Austria), and Statistical Analysis Software (SAS) Enterprise Guide (version 9.4 SAS Institute Inc., Cary, NC, USA).

Role of the Funding Source

This study, including design, data collection, data analysis, data interpretation, and medical writing, was funded by Eli Lilly and Company.

Ethical Considerations

Researchers accessed data in the format of a limited data set for which data use agreements were in place with the covered entities in compliance with the Health Insurance Portability and Accountability Act Privacy Rule. An Institutional Review Board did not review the study since only this limited data set was accessed. As this observational study used previously collected data and did not impose any intervention, the data were de-identified to protect participant privacy. Therefore, a formal Consent to Release Information form was not required. The study was conducted according to the ethical principles that originate in the Declaration of Helsinki and is consistent with Good Pharmacoepidemiology Practices and applicable laws and regulations in the US.

Results

Patient Characteristics

Prior to PS matching, we identified 10,833 GLP-1 RA naïve and 5897 GLP-1 RA non-naïve tirzepatide initiators in the HIRD, along with 52,960 GLP-1 RA naïve and 8437 GLP-1 RA non-naïve semaglutide initiators. Details on patient identification are included in Table S1. Attrition rates were similar across both cohorts for all selection criteria, except when stratified by GLP-1 receptor agonist treatment history. This difference reflects the limited availability of alternatives when most patients initiated semaglutide, whereas patients initiating tirzepatide had greater prior exposure to GLP-1 therapies due to more options being available at that time. Following PS matching, 10,702 naïve tirzepatide and semaglutide initiators and 5577 non-naïve tirzepatide and semaglutide initiators remained. Among naïve patients, there were 1399 tirzepatide and 1173 semaglutide initiators with baseline and follow-up HbA1c data, whereas, respectively, 454 and 432 patients had baseline and follow-up weight measurements. For non-naïve patients, 792 tirzepatide and 738 semaglutide initiators had baseline and follow-up HbA1c data, while 296 and 224, respectively, had baseline and follow-up weight measurements. Adequate balance between treatment groups overall (Table 1) and among subgroups with clinical data (Table S2) was successfully achieved.

Table 1.

Demographics and baseline clinical characteristics of patients with T2D initiating tirzepatide or semaglutide before and after PS matching

	Before PS matching						After PS matching
	GLP-1 RA naïve			GLP-1 RA non-naïve			GLP-1 RA naïve			GLP-1 RA non-naïve
	Tirzepatide initiators	Semaglutide initiators	SMDa	Tirzepatide initiators	Semaglutide initiators	SMDa	Tirzepatide initiators	Semaglutide initiators	SMDa	Tirzepatide initiators	Semaglutide initiators	SMDa
Number of patients	10,833	52,960	–	5897	8437	–	10,702	10,702	–	5577	5577	–
Age on index (years), continuous
Mean (SD)	53.1 (9.92)	55.0 (10.42)	0.178	54.8 (9.22)	56.2 (9.87)	0.152	53.2 (9.94)	53.2 (9.92)	0.006	54.9 (9.24)	55.0 (9.50)	0.013
Categorical age on index (years), %
18–39	9	7	0.173	6	5	0.180	9	8	0.04	6	6	0
40–49	25	21		21	18		25	25		21	21
50–59	39	38		41	40		39	40		41	41
≥ 60	27	34		32	37		27	27		33	33
Sex, %
Female	56	54	0.040	55	56	0.020	56	55	0.029	55	54	0.02
Raceb, %
White	71	68	0.100	74	70	0.099	71	71	0.06	74	73	0.032
Black	9	12		9	12		9	9		9	9
Asian	3	3		3	3		3	3		3	3
Other	4	5		4	4		4	3		3	3
Undisclosed	13	12		11	11		13	14		11	12
Ethnicityb, %
Hispanic	8	8	0	7	7	0	8	8	0.032	7	7	0
Non-Hispanic	81	81		83	83		81	80		82	82
Undisclosed	12	11		10	10		12	12		10	10
Regiond, %
Midwest	27	25	0.104	29	27	0.071	27	27	0.029	29	26	0.087
Northeast	9	12		10	12		9	9		10	12
South	48	48		46	46		48	49		46	46
West	16	15		15	15		16	15		15	15
Payor type, %
Commercial	92	86	0.194	91	85	0.191	92	92	0.006	91	90	0.013
Medicare Advantage	8	14		9	15		8	8		9	10
Health plan type, %
HMO	16	24	0.201	18	25	0.173	16	15	0.028	18	18	0
PPO	60	54		60	54		60	61		59	59
CDHP	24	22		22	21		24	24		22	22
Quartile of SES index scorec, %
1	22	23	0.028	19	22	0.114	22	23	0.073	20	20	0.067
2	28	27		28	29		28	28		28	28
3	26	26		27	27		26	27		27	27
4	21	21		23	20		21	20		23	22
Missing data	3	3		3	3		3	3		3	3
Index provider specialty, %
Endocrinology	14	11	0.198	26	18	0.233	13	13	0	24	22	0.149
PCP	43	48		34	43		43	43		36	37
Non-physician clinician	36	32		34	32		36	36		34	34
Other/unknown	8	9		6	7		8	8		6	6
Comorbidities, %
Anxiety	18	17	0.034	17	17	0.005	18	18	0.011	17	17	0.003
Depression	15	15	0.006	16	16	0.006	15	15	0.009	16	16	0.017
Dyslipidemia	66	66	0.002	75	71	0.083	66	64	0.044	74	72	0.049
Hypertension	68	69	0.016	73	72	0.013	68	66	0.045	73	72	0.016
Obesity	50	45	0.095	49	45	0.083	49	46	0.078	49	46	0.059
OSA	21	19	0.036	22	21	0.034	21	20	0.03	22	21	0.032
Osteoarthritis	14	13	0.010	14	15	0.027	14	12	0.052	14	13	0.015
aDCSI score, continuous
Mean (SD)	0.80 (1.355)	0.92 (1.463)	0.080	1.10 (1.536)	1.19 (1.631)	0.056	0.80 (1.35)	0.78 (1.36)	0.018	1.10 (1.54)	1.14 (1.57)	0.027
Categorical aDCSI score, %
0	63	60	0.083	53	51	0.072	63	65	0.06	53	52	0.061
1	13	13		16	16		13	13		16	15
2	13	14		16	16		13	12		16	17
3	5	6		7	7		5	4		7	7
≥ 4	6	6		8	10		6	6		8	8
Number of antihyperglycemic agents, %
None	24	18	0.147	0	0	0.000	24	24	0.01	0	0	0
Monotherapy	61	68	0.130	13	12	0.011	62	62	0.005	13	12	0.042
Dual therapy	15	14	0.007	59	60	0.010	15	14	0.019	59	60	0.019
Triple therapy	0	0	0.000	28	28	0.003	0	0	0	28	28	0.011
Antihyperglycemic agents, %
None	24	18	0.147	0	0	0	24	24	0.01	0	0	0
Insulin	19	19	0.008	34	34	0.003	19	18	0.026	33	33	0.002
OAD	72	78	0.139	81	82	0.010	72	72	0.002	81	83	0.046
Metformin	59	66	0.134	64	65	0.036	59	60	0.009	64	67	0.064
SGLT2i	24	22	0.032	42	38	0.091	24	21	0.071	42	39	0.054
Sulfonylureas	15	18	0.086	20	24	0.098	15	16	0.032	20	24	0.1
GLP-1 RA	0	0	0.000	100	100	0.000	0	0	0	100	100	0
Clinical measurements, Mean (SD) or %
Number of patients	3248	17,081	-	1807	2529	-	3226	2901	-	1710	1664	-
HbA1c (%)	7.8 (1.94)	8.0 (1.91)	0.111	7.8 (1.57)	7.9 (1.62)	0.117	7.8 (1.94)	7.8 (1.91)	0.065	7.7 (1.58)	7.9 (1.61)	0.002
Number of patients	1719	8779	-	1000	1470	-	1702	1493	-	913	935	-
Weight (kg)	111.9 (26.81)	108.4 (25.23)	0.132	111.6 (26.06)	106.7 (25.23)	0.191	111.8 (26.62)	110.7 (25.10)	0.057	110.8 (26.24)	109.1 (24.94)	0.003
Number of patients	1587	7901	-	921	1305	-	1571	1354	-	839	850	-
BMI ≥ 30 kg/m2	87	85	0.078	88	82	0.151	87	89	0.050	87	87	0.005

aDCSI Adapted Diabetes Complications Severity Index, BMI body mass index, CDHP consumer-driven health plan, GLP-1 RA glucagon-like peptide-1 receptor agonist, HbA1c glycated hemoglobin, HMO health maintenance organization, OAD oral antihyperglycemic drugs, OSA obstructive sleep apnea, PCP primary care provider, PPO preferred provider organization, PS propensity score, SD standard deviation, SES socioeconomic status, SGLT2i sodium-glucose cotransporter-2 inhibitors, SMD standardized mean difference

^aSMD reported as absolute values

^bRace/ethnicity variables come from member-self attestation during enrollment, EHR data, and imputation algorithm

^cSES index was derived from the 2020 American Community Survey

^dCensus region of the members’ residence on patient claims index date

The SES index is a composite measure based on seven factors (unemployment rate, poverty rate, median household income, median home value, proportion of not having a high school degree, proportion with a college degree, proportion of households that average one or more persons per room). The SES index score is reported in quartiles, with “4” indicating a patient is in the top 25% and “1” indicating a patient is in the bottom 25% using US national thresholds

Across all matched cohorts, baseline dyslipidemia, hypertension, and obesity were highly prevalent (> 45%), with higher prevalence in non-naïve patients than naïve patients. Across the naïve cohorts, 76% of tirzepatide and semaglutide initiators had ≥ 1 fill for an antihyperglycemic medication (other than a GLP-1 RA) over the 6-month pre-index period. Monotherapy was most common (62% in both naïve cohorts), and metformin alone was the most prevalent monotherapy treatment regimen (27% tirzepatide; 28% semaglutide). Across the non-naïve cohorts, all initiators had ≥ 1 fill for an antihyperglycemic medication (i.e., baseline GLP-1 RA use, by design) over the 6-month pre-index period. Dual therapy was most common (59% tirzepatide; 60% semaglutide), and metformin in combination with a GLP-1 RA was the most prevalent dual treatment regimen (17% tirzepatide; 18% semaglutide). Dulaglutide was the GLP-1 RA most patients were exposed to prior to index (72% tirzepatide; 82% semaglutide).

Tirzepatide was most commonly initiated at doses of 2.5 mg (59% for naïve, 37% for non-naïve) or 5.0 mg (31% for naïve, 44% for non-naïve), while for semaglutide the most commonly initiated doses were 0.25 mg or 0.5 mg (87% for naïve, 57% for non-naïve) or 1.0 mg (12% for naïve, 31% for non-naïve; Table 2). More than half of the patients increased their dose over the 12-month post-index period. Most patients remained persistent to their index treatment (i.e., tirzepatide or semaglutide) across both cohorts; however, the proportion of persistent patients and the duration of persistence trended higher in the tirzepatide than in the semaglutide groups.

Table 2.

GLP-1 RA treatment patterns in patients with T2D initiating tirzepatide or semaglutide after PS matching over a 12-month follow-up period

	GLP-1 RA naïve		GLP-1 RA non-naïve
	Tirzepatide initiators	Semaglutide initiators	Tirzepatide initiators	Semaglutide initiators
Number of patients	10,702	10,702	5577	5577
Persistent, %	62	47	68	49
Persistent days
Mean (SD)	271.0 (132.86)	238.6 (135.40)	289.3 (123.29)	247.9 (132.63)
Index tirzepatide dose in mg, %
2.5	59	–	37	–
5	31	–	44	–
7.5	6	–	10	–
10	3	–	6	–
12.5	1	–	1	–
15	1	–	1	–
Index semaglutide dose in mg, %
0.5	–	87	–	57
1	–	12	–	31
2	–	2	–	12
Dose change from index dosea, %
Any change	79	54	83	57
Dose increase	78	54	81	55
Dose decrease	32	12	33	15

GLP-1 RA glucagon-like peptide-1 receptor agonist, PS propensity score, SD standard deviation

^aIncreases or decreases in dose throughout the 12-month period were not mutually exclusive per patient

Patient Outcomes

Both primary (Fig. 1) and four of the five key secondary outcomes among the naïve population and all five among the non-naïve (Table 3) were statistically significant by Holm in favor of tirzepatide. All exploratory outcomes trended in favor of tirzepatide (Supplementary Table S3).

Fig. 1.

Absolute changes in HbA1c and weight from baseline to end of 12-month follow-up among patients with T2D initiating tirzepatide or semaglutide after PS matching. HbA1c hemoglobin A1c, N number of patients, PS propensity score, SD standard deviation

Table 3.

Key secondary outcomes in patients with T2D initiating tirzepatide or semaglutide after PS matching over 12-month follow-up

	GLP-1 RA naïve			GLP-1 RA non-naïve
	Tirzepatide initiators	Semaglutide Initiators	p value	Tirzepatide initiators	Semaglutide initiators	p value
Number of patients	10,702	10,702	–	5577	5577	–
Percent change in weight from baseline to follow-upa
Valid N	454	432	–	296	224	–
Mean (SD)	– 8.9 (10.00)	– 5.4 (7.01)	< 0.001	– 7.1 (9.39)	– 3.3 (5.87)	< 0.001
Follow-up HbA1c ≤ 6.5%
Valid N	2125	1909	–	1173	1130	–
n (%)	1418 (67%)	1029 (54%)	< 0.001	570 (49%)	368 (33%)	< 0.001
Follow-up HbA1c < 5.7%
Valid N	2125	1909	–	1173	1130	–
n (%)	646 (30%)	337 (18%)	< 0.001	195 (17%)	69 (6%)	< 0.001
Adherent at ≥ 80% PDC
Valid N	10,702	10,702	–	5577	5577	–
n (%)	6413 (60%)	4832 (45%)	< 0.001	3697 (66%)	2685 (48%)	< 0.001
Antihyperglycemic medication cost per responder (HbA1c ≤ 6.5%)b
Valid N	2125	1909	–	1173	1130	–
Point estimate	$20,639	$21,392	0.24	$36,855	$50,689	< 0.001

GLP-1 RA glucagon-like peptide-1 receptor agonist, HbA1c hemoglobin A1c, n number of patients, N total number of patients, PDC proportion of days covered, PS propensity score, SD standard deviation

^aPatients who gained ≥ 50 kg from baseline to follow-up were checked, and their weight was manually revised as needed. Two patients had their weight mistakenly reported in kg instead of lbs. These entries were manually corrected

^bTotal annual antihyperglycemic medication cost calculated as the sum of insulin, metformin, thiazolidinediones, sulfonylureas, dipeptidyl peptidase 4inhibitors, sodium-glucose cotransporter-2 inhibitors, meglitinides, alpha-glucosidase inhibitors, and GLP-1 RAs

Co-Primary Outcomes

In the matched naïve cohorts, 13% of tirzepatide (N = 1399) and 11% of semaglutide initiators (N = 1173) had baseline and follow-up HbA1c data. Both tirzepatide and semaglutide initiators had a mean baseline HbA1c of 7.8%, with a mean reduction of 1.3% and 0.9% for tirzepatide and semaglutide, respectively (Fig. 1A). The additional 0.4% decrease among tirzepatide initiators was significant (p < 0.001; Supplementary Table S4). For weight change in the matched naïve cohorts, 4% of tirzepatide (N = 454) and 4% of semaglutide (N = 432) patients had baseline and follow-up data. Tirzepatide and semaglutide initiators had a mean baseline weight of 112.4 kg and 110.7 kg, respectively, with a mean reduction of 10.2 kg for tirzepatide initiators and 6.1 kg for semaglutide initiators (Fig. 1B). The additional 4.1 kg weight reduction among naïve tirzepatide initiators was significant (p < 0.001; Supplementary Table S4). The additional significant decreases in HbA1c and weight remained in favor of tirzepatide after further regression adjustments to address potential confounding from remaining imbalances in baseline characteristics in these subgroups (Supplementary Table S4).

In the matched non-naïve cohorts, 14% of tirzepatide (N = 792) and 13% of semaglutide initiators (N = 738) had baseline and follow-up HbA1c data. Tirzepatide and semaglutide initiators had a mean baseline HbA1c of 7.7% and 7.9%, respectively, with a mean absolute reduction of 0.9% for tirzepatide and 0.6% for semaglutide (Fig. 1A). The additional 0.3% decrease among tirzepatide initiators was significant (p < 0.001; Supplementary Table S4). In the matched non-naïve cohorts, 5% of tirzepatide (N = 296) and 4% of semaglutide (N = 224) had baseline and follow-up weight data. Tirzepatide and semaglutide initiators had a mean baseline weight of 112.5 kg and 108.5 kg, respectively, with a mean absolute reduction of 7.9 kg for tirzepatide and a reduction of 3.7 kg for semaglutide (Fig. 1B). The additional 4.2 kg weight reduction among naïve tirzepatide initiators was significant (p < 0.001; Supplementary Table S4). The additional significant decreases in HbA1c and weight remained in favor of tirzepatide after further regression adjustments (Supplementary Table S4).

Key Secondary Outcomes

Among the naïve cohorts, most secondary outcomes were in favor of tirzepatide (Table 3). In the matched naïve cohorts, 20% of tirzepatide (N = 2125) and 18% of semaglutide (N = 1909) patients had follow-up HbA1c data. Among tirzepatide initiators, 67% (N = 1418) reached HbA1c levels ≤ 6.5%, and 30% (N = 646) reached < 5.7% at follow-up. For semaglutide initiators, 54% (N = 1029) achieved HbA1c ≤ 6.5%, and 18% (N = 337) reached < 5.7%. In the matched naïve cohorts with baseline and follow-up weight data, tirzepatide initiators had a mean percent weight reduction of 8.9%, while semaglutide initiators had a mean percent reduction of 5.4%. Among the tirzepatide cohort, 60% achieved adherence (i.e., ≥ 80% PDC), compared to 45% for the semaglutide cohort. Tirzepatide averaged $20,639 annual cost of antihyperglycemic medications per patient who achieved HbA1c ≤ 6.5%, while semaglutide resulted in an average of $21,392, the only formal comparison that did not yield a significant result (p = 0.24). All other differences were statistically significant in favor of tirzepatide.

Among the non-naïve cohorts, all secondary outcomes were in favor of tirzepatide (Table 3). In the matched non-naïve cohorts, 21% of tirzepatide (N = 1173) and 20% of semaglutide (N = 1130) patients had follow-up HbA1c data. Among tirzepatide initiators, 49% (N = 570) reached HbA1c levels ≤ 6.5%, and 17% (N = 195) reached < 5.7%. For semaglutide initiators, 33% (N = 368) achieved HbA1c ≤ 6.5%, and 6% (N = 69) reached < 5.7%. In the matched non-naïve cohorts with baseline and follow-up weight data, tirzepatide initiators had a mean percent weight reduction of 7.1% while semaglutide initiators had a mean percent reduction of 3.3%. The tirzepatide cohort achieved 66% adherence, compared to 48% for semaglutide. Tirzepatide averaged $36,855 annual cost of antihyperglycemic medications per patient who achieved HbA1c ≤ 6.5% while semaglutide averaged an annual cost of $50,689. All differences were statistically significant in favor of tirzepatide.

Discussion

This is one of the first studies utilizing real-world data (RWD) to examine differences in HbA1c and weight outcomes among GLP-1 RA naïve and non-naïve patients with T2D initiating either tirzepatide or injectable semaglutide.

Comparison with Clinical Trials

Our findings align with previous clinical trials that report blood sugar control and weight reduction with both tirzepatide and semaglutide [21, 33]. This study’s results are aligned with the head-to-head SURPASS-2 phase 3 clinical trial in naïve participants, though with some differences [21]. Reductions in HbA1c were less pronounced than those in SURPASS-2, where mean changes from baseline in HbA1c were: – 2.01% for 5 mg, – 2.24% for 10 mg, – 2.30% for 15 mg tirzepatide, and – 1.86% for 1 mg semaglutide. This study observed – 1.3% in naïve tirzepatide initiators, and – 0.9% for naïve semaglutide. Despite this, the difference in HbA1c reduction between the treatments was similar across both studies: – 0.15%, – 0.39%, and – 0.45% for the respective tirzepatide doses versus semaglutide in the clinical trial, and – 0.4% for all doses in the naïve cohort in this study. In the SWITCH-2 study, participants who switched to tirzepatide 5 mg from GLP-1 RAs experienced an additional HbA1c reduction at week 12 with tirzepatide (0.43%) versus semaglutide (0.36%), also aligned with the results from the current analyses [33]. The weight loss results from the current analyses are similarly aligned with clinical trials, however less pronounced as SURPASS-2 reported a mean weight change from baseline of – 7.6 kg, – 9.3 kg, and – 11.2 kg for tirzepatide doses of 5 mg, 10 mg, and 15 mg, respectively and – 5.7 kg with 1 mg semaglutide [21]. In the current study, the reductions observed were – 10.2 kg for naïve tirzepatide, – 7.9 kg for non-naïve tirzepatide, – 6.1 kg for naïve semaglutide, and – 3.7 kg for non-naïve semaglutide.

These differences may be due to SURPASS-2’s stringent criteria, including only treatment-naïve patients with baseline metformin usage. This group typically responds well to weight and glucose control interventions, unlike the diverse participants in our real-world study. The current population also had lower HbA1c and higher body weight at baseline compared to SURPASS-2, likely due to the more stringent inclusion/exclusion criteria of the clinical trial. Additionally, the controlled setting of clinical trials includes factors like strict regimen adherence and frequent monitoring, which differ from real-world conditions where variability in adherence is common due to behavior, health care access, and socioeconomic factors. Patients may not reach or maintain high doses of medication, potentially leading to the underestimation of treatment benefits in real-world scenarios [34, 35].

Finally, the SURPASS-2 trial analyzed each tirzepatide dose separately (5 mg, 10 mg, and 15 mg), providing a detailed dose–response relationship. Conversely, our study observed a pooled cohort across various dosages, possibly diluting specific dose-related effects. Moreover, the introduction of the 2-mg semaglutide dose in real-world settings may introduce variations in patients’ outcomes, thereby potentially affecting comparative efficacy.

The relationship between sex and weight loss among individuals with T2D using incretin-based therapies has been reported previously. In a recent post hoc analysis of the SURPASS 1–5 tirzepatide clinical trials [36] female participants achieved the composite endpoint (HbA1c < 6.5%, and weight loss target of ≥ 10%) at higher rates than male participants across the five trials. These observations have been hypothesized to be due to a higher medication exposure among women, i.e., increased plasma concentration of the treatment due to their lower mean body weight. The current analyses utilized PS matching of the cohorts to minimize any differences in sex distribution between study cohorts; a subgroup analysis by sex was outside the study’s scope.

Comparison with Prior Real-World Evidence

The exploration of tirzepatide versus semaglutide effectiveness has been informed by a few other real-world studies. Rodriguez et al. examined the effects of tirzepatide and semaglutide on weight change in GLP-1 RA naïve patients with overweight or obesity [37]. After 12 months, PS-matched patients on tirzepatide showed a 15.3% bodyweight reduction, while semaglutide patients showed an 8.3% reduction. Adjusting for residual confounding, they observed a – 6.9% difference, exceeding our naïve cohort’s finding of – 3.5%. However, both studies favored tirzepatide over semaglutide. The discrepancy in weight loss between the two studies is likely due to patient demographics. All patients in our study had a diagnosis of T2D, compared to 50% in the Rodriquez study. GLP-1 RAs tend to result in more significant weight loss for individuals without diabetes [38–42], likely contributing to the greater weight loss and highlighting the importance of considering patient demographics in evaluating GLP-1 RAs.

Chuang et al. reported that tirzepatide treatment was associated with lower risk of all-cause mortality, adverse cardiovascular events, acute kidney injury, and adverse kidney events compared to GLP-1 RAs in patients with T2D [43]. They observed a body weight loss difference of – 2.9 kg between tirzepatide and GLP-1 RAs, although the specific GLP-1 RA was not identified, limiting direct comparison with our study. While specific mortality, cardiovascular, and renal outcomes were not the focus in the current study, the comparable changes in absolute weight observed for both naïve and non-naïve cohorts suggest similar effects for patients with these comorbidities.

Clinical Significance and Future Directions

The differences in HbA1c and weight reduction suggest that tirzepatide’s dual agonism may offer improved glycemic control and weight loss benefits for patients with T2D compared to a single agonist like semaglutide, potentially reducing diabetes-related complications. Clinicians should consider these differences, especially for patients struggling with both weight management and glycemic control, to optimize individualized treatment. Future research should explore tirzepatide and semaglutide’s long-term effects on cardiovascular outcomes, mortality, and quality of life across diverse populations. Studying subgroup (e.g., age, gender, and baseline weight) responses could enhance personalized medicine by identifying patients likely to benefit most from each treatment.

Strengths and Limitations

Our study provides insights into the performance of these treatments beyond controlled clinical trials. The study’s inclusion of a diverse patient cohort, including those with multiple comorbidities and varying GLP-1 RA doses, reflects the complex health scenarios encountered in routine clinical practice and enhances the generalizability of its findings. Furthermore, patients were categorized by previous (within 6 months) GLP-1 RA use to include both populations and account for the potential confounding of outcomes. Large sample sizes and various robustness analyses strengthen the study's conclusions. Additionally, previous comparator trials did not include semaglutide at its highest dose of 2 mg, which was included in the current analysis.

The study also has limitations, including a 12-month follow-up, which may not reflect the complete clinical impact of T2D a chronic condition requiring long-term treatment. Administrative claims data, which are primarily used for billing, may contain diagnostic or treatment inaccuracies due to coding errors or undocumented medication use and exclude free samples and over-the-counter drugs. For instance, a 6-month washout period is clinically sufficient to capture prior GLP-1 RA exposure, yet out-of-pocket purchased medications (including compounded GLP-1 RAs) are unaccounted for. Other relevant factors, such as provider preferences, health plan details, and specific patient characteristics (e.g., social needs, family history), are unavailable, potentially leading to unmeasured confounding. The study’s focus on commercially insured patients may limit generalizability, as such individuals may differ from those with public or no insurance. Exclusion of groups of patients with gestational diabetes, type 1 diabetes, specific surgeries, or pregnancy may further limit generalizability. Continuous enrollment requirements might introduce immortal time bias as outcomes for patients with less than 12 months of follow-up are not included, and selection bias might arise from including only patients with available laboratory results and EHRs. However, we expect these biases to affect both treatment arms similarly. Additionally, analyses of the baseline demographics of patients with baseline and follow-up HbA1c and weight measurements, presented in Table S2, indicate minor imbalances between groups after PS matching; regression models fitted as a sensitivity check verified the original results [29].

Conclusions

The real-world evidence presented in this study supports findings from clinical trials such as SURPASS-2, emphasizing tirzepatide’s treatment effect over injectable semaglutide in reducing HbA1c and weight for patients with T2D. These outcomes highlight tirzepatide’s potential to deliver significant clinical benefits in managing T2D, contributing to improved long-term health outcomes. The study’s insights, consistent with those from previous clinical trials, are important additional information for clinicians and policymakers aiming to optimize therapy for patients with T2D and enhance diabetes management strategies.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contribution

All authors (Meredith M. Hoog, Carlos Vallarino, Juan M. Maldonado, Michael Grabner, Chia-Chen Teng, Kendra Terrell, and Emma L. Richard) contributed to the study conception, design, and methodology. Data analysis was performed by Chia-Chen Teng. The first draft of the manuscript was written by Emma L. Richard and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript. Emma L. Richard is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Funding

Sponsorship for this study and the journal’s Rapid Service Fee was funded by Eli Lilly and Company.

Data Availability

The data that support the findings of this study are available from Carelon Research, but restrictions apply to the availability of these data to external entities, and therefore they are not publicly available. Data may be made available from the authors upon reasonable request and with permission of Carelon Research.

Declarations

Conflict of Interest

Meredith M. Hoog, Carlos Vallarino, Juan M. Maldonado, and Kendra Terrell are employees and stockholders of Eli Lilly and Company. Michael Grabner, Chia-Chen Teng, and Emma L. Richard are employees of Carelon Research, which received funding from Eli Lilly and Company for the conduct of this study as well as for manuscript writing. Michael Grabner, Chia-Chen Teng, and Emma L. Richardare stockholders of Elevance Health.

Ethical Approval

Researchers accessed data in the format of a limited data set for which data use agreements were in place with the covered entities in compliance with the Health Insurance Portability and Accountability Act Privacy Rule. An Institutional Review Board did not review the study since only this limited data set was accessed. As this observational study used previously collected data and did not impose any intervention, the data were de-identified to protect participant privacy. Therefore, a formal Consent to Release Information form was not required. The study was conducted according to the ethical principles that originate in the Declaration of Helsinki and are consistent with Good Pharmacoepidemiology Practices and applicable laws and regulations in the US.