This was a real-world, retrospective, observational cohort study using the Komodo Research Dataset (KRD) from December 15, 2020, through May 30, 2024. Komodo Health obtains de-identified, patient-level claims data from insurance companies as well as clearing houses. These data are then linked with electronic health records using tokenization technology. The dataset, comprising records on > 320 million individuals in the US from 2012 through 2023, included data on insurance enrollment, pharmacy claims, and medical claims with linkage to electronic medical records (EMR). The dataset contains demographic and clinical variables relevant to the present study, such as age, sex, race, height, body weight, body mass index (BMI), laboratory test results, prescription fills, and diagnostic and procedure codes. As semaglutide 2.4 mg received FDA approval for chronic weight management in June 2021, data starting from 2020 were included to establish an appropriate baseline period prior to the earliest possible index outpatient visit for semaglutide 2.4 mg.

This study included adults aged ≥ 18 years with obesity (BMI ≥ 30.0 kg/m2) or overweight (BMI 27.0–29.9 kg/m2) with at least one ORC (e.g., hypertension, T2D, dyslipidemia, obstructive sleep apnea, cardiovascular disease). Patients in the treated cohort initiated once-weekly semaglutide 2.4 mg indicated for chronic weight management at a dose of 0.25, 0.5, 1.0, or 1.7 mg after June 15, 2021, and escalated to and remained on the 2.4 mg maintenance dose during the 12-month follow-up period (proportion of days covered ≥ 80%).

To identify the treatment cohort, the KRD was searched for all patients with ≥ 1 claim for semaglutide 2.4 mg between June 15, 2021 and July 31, 2022. Semaglutide 2.4 mg refers to the formulation of semaglutide indicated for weight management, under the brand name Wegovy®. The first claim for the weight management brand of semaglutide 2.4 mg was the index date for patients in the treated cohort. The index date for patients not treated with obesity medication (non-treated cohort) was the earliest available outpatient visit after June 15, 2021, plus a random number of days based on the time between the last visit and treatment initiation among patients initiating semaglutide. All patients must have had continuous medical and pharmacy enrollment ≥ 6 months prior to the index date (baseline period) and ≥ 12 months after the index date (follow-up period; Fig. 1). Patients were excluded if, in the baseline period, weight ≥ 181 kg, latest BMI < 27.0 kg/m2, or no weight/BMI value was available (Supplemental Table 1 provides additional exclusion criteria details).

Study design graphic. BMI body mass index

This study was conducted according to the guidelines of the Declaration of Helsinki of 1964 and its later amendments and the STROBE guidelines. Komodo Health approved use of the dataset for this study. Because this study used only de-identified data compliant with the requirements of the Health Insurance Portability and Accountability Act, Institutional Review Board review and approval were not needed.

Patient demographic and clinical characteristics, including relevant comorbidities, as well as procedures and medications, were evaluated during the baseline period. Baseline weight characteristics included body weight, BMI, and BMI classification (overweight [27.0–29.9 kg/m2], obesity class I [30.0–34.9 kg/m2], obesity class II [35.0–39.9 kg/m2], and obesity class III [≥ 40 kg/m2]). Baseline cardiometabolic risk factors included SBP, DBP, HbA1c, LDL-C, HDL-C, and triglycerides. The presence of T2D at baseline was determined based on either diagnosis of T2D (International Classification of Diseases, Tenth Revision, Clinical Modification [ICD-10-CM] codes) or HbA1c value ≥ 6.5%. ICD-10-CM codes were used to identify ORCs (dyslipidemia, hypertension, prediabetes, T2D, musculoskeletal pain, psoriasis, urinary incontinence, asthma, polycystic ovary syndrome, obstructive sleep apnea, gastroesophageal reflux disease, HFpEF, and knee osteoarthritis). Number of medications (antihypertensive, antidiabetic, antihyperlipidemic) and health care resource utilization (HCRU; outpatient, inpatient, and emergency department visits) were assessed.

Primary outcomes (percentage change in weight, categorical percentage change in weight, absolute change in weight, and change in BMI) were assessed from baseline (– 30 days to index date) to 12 months (365 days ± 30 days). Secondary outcomes (changes in SBP, DBP, HbA1c, LDL-C, HDL-C, and triglycerides) were assessed from baseline (– 90 days to index date) to 12 months (365 days ± 90 days). If multiple measurements for a given outcome or characteristic were available, the measurement closest to each assessment point (baseline and follow-up, respectively) was used.

Descriptive statistics were used to analyze baseline demographic and clinical characteristics. Cohorts were matched on propensity scores before conducting outcome analyses. Baseline demographics, BMI, BMI class, weight, comorbidities, cardiometabolic medication use, HCRU, and cardiometabolic risk factors were used to develop a logistic regression-based propensity score model. Risk factors were incorporated as an interaction term between the binary data availability flags (indicators) and measurement results to ensure similar distributions of risk factor availability and their respective results in the matched samples. This procedure was used to create the overall study population. The cohorts were matched 1:4 on the propensity score using the nearest neighbor algorithm without replacement and a maximum caliper of 0.1. Standardized mean differences (SMD) were estimated to evaluate the balance of covariates between cohorts. Absolute SMD ≥ 0.1 indicated lack of balance between cohorts. Based on the overall study cohort, eight separate sub-cohorts were created, one for each outcome (weight, BMI, SBP, DBP, HbA1c, LDL-C, HDL-C, triglycerides), with patients with data available to estimate change in the outcomes from baseline (Supplemental Tables 2–9 provide baseline characteristics and sample sizes for each sub-cohort).

Outcome analyses were conducted using the propensity score-matched cohorts and generalized linear models with a Gaussian distribution for continuous outcome variables and binomial distribution for categorical outcome variables. Bivariate and adjusted models were developed. The adjusted models were adjusted for residual unbalanced covariates arising from sub-setting the overall cohort and were at minimum adjusted for the baseline values of the respective risk factors. Outcomes were assessed between baseline and the 12-month follow-up. Estimates were reported with 95% confidence intervals. A two-sided p value  < 0.05 was considered statistically significant. Additionally, outcomes were assessed descriptively for sub-groups defined by BMI class and diabetes status. Analyses were conducted in R (v4.2.1) and MatchIt package [22].