Document resource
WHAT IS ALREADY KNOWN ON THIS TOPIC To date, evidence describing the long-term effectiveness of once-weekly subcutaneous semaglutide in routine clinical practice is limited.WHAT THIS STUDY ADDS The SEmaglutide PRAgmatic (SEPRA) trial evaluated the effectiveness of treatment intensification with once-weekly subcutaneous semaglutide versus alternative treatments when added to up to two oral antidiabetic medications in adult participants with type 2 diabetes (T2D) in routine clinical practice. In SEPRA, greater improvement with semaglutide versus alternative treatment was observed on measures of glycemic control, body weight, and some patient-reported outcomes, with no new safety concerns identified. Treatment intensification, change, or discontinuation occurred less frequently in the semaglutide group than in the alternative treatment group over the 2-year study period.HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY These results complement findings from semaglutide phase III randomized controlled trials, supporting the use of semaglutide as an effective and well-tolerated treatment option for adults with T2D in real-world clinical practice.Introduction An estimated 537 million adults worldwide are living with type 2 diabetes (T2D), and approximately 1.2 million adults in the USA were newly diagnosed with T2D in 2021. 1 2 Guidelines for managing T2D recommend selecting therapies that achieve weight loss and glycemic control goals.3 However, there are few long-term, evidence-based comparative studies to guide treatment selection in routine clinical practice.Semaglutide is a glucagon-like peptide-1 receptor agonist (GLP-1 RA) approved in the USA for once-weekly subcutaneous administration (0.25–2.0 mg) as an adjunct to diet and exercise to improve glycemic control in adults with T2D; to reduce the risk of major cardiovascular events in adults with T2D and established cardiovascular disease; and to reduce the risk of sustained estimated glomerular filtration rate decline, end-stage kidney disease, and cardiovascular death in adults with T2D and chronic kidney disease.4 These indications are derived from results of randomized controlled trials (RCTs), such as the Semaglutide Unabated Sustainability in Treatment of Type 2 Diabetes (SUSTAIN) phase III clinical trial program, the SUSTAIN 6 cardiovascular trial, and the FLOW trial.5–9The SEmaglutide PRAgmatic (SEPRA; NCT03596450) trial is a 2-year, open-label, randomized, pragmatic clinical trial examining the effectiveness of treatment intensification with once-weekly subcutaneous semaglutide compared with alternative treatment in adults with T2D when added to ≤2 oral antidiabetes medications in routine clinical practice.10 The comparison of semaglutide with alternative treatment is relevant as it assesses treatment usage and performance of the randomized intervention in a setting that approximates real-world usage as closely as possible in the context of a clinical trial. Such studies are increasingly important to generate high-quality evidence regarding real-world treatment outcomes while minimizing confounding through randomization.11 12 Here we report 1-year and 2-year results from SEPRA.Methods The trial design and methods were described previously. 10 Briefly, SEPRA is a randomized, open-label, highly pragmatic clinical trial comparing once-weekly subcutaneous semaglutide with alternative treatments excluding semaglutide in real-world, routine clinical practice. SEPRA was evaluated by the research team as scoring 4–5 (on a scale of 1–5, with 5 being very pragmatic) across all domains of the PRagmatic Explanatory Continuum Indicator Summary-2 tool (PRECIS-2).10 13Participants Participants were enrolled at 138 US sites (primary care and endocrinology practices). Patients were approached for potential enrollment by their treating physician during standard care, or through proactive identification and contact within the study physician’s practice. At trial onset, eligibility criteria included the following: adults (aged ≥18 years) with T2D treated with metformin; requirement for further treatment intensification for glycemic control with an additional oral or injectable antidiabetes medication (at the treating physician’s discretion); current member of study-affiliated commercial health plan with pharmacy benefits; and recorded glycated hemoglobin (HbA 1c) value within the past 90 days before randomization. Key exclusion criteria were treatment with any medication indicated for diabetes other than metformin in the 30 days before eligibility assessment and pregnancy, breastfeeding, or intention to become pregnant. During the enrollment period (July 2018 through March 2021), several factors caused substantial recruitment challenges, including the COVID-19 pandemic. The eligibility criteria were amended to allow broader recruitment of adults with T2D treated with ≤2 oral glucose-lowering medications (excluding semaglutide) and current members of any health plan with pharmacy benefits.10Interventions Eligible participants were randomized to once-weekly subcutaneous semaglutide or alternative treatment (determined by the treating physician’s choice among commercially available oral or injectable glucose-lowering medications other than semaglutide) as an add-on to ≤2 oral antidiabetic medications. Treating physicians determined the starting dose, escalation regimen, and maintenance dose of the randomized treatment according to local practice and product labeling and continued to manage participants’ diabetes care, including treatment adjustments, based on clinical judgment. During the recruitment period, the maximum approved dose for once-weekly subcutaneous semaglutide was 1.0 mg; therefore, this dose was the highest allowed by protocol. Participants randomized to semaglutide were able to stop or start treatment at any time throughout the study. Participants in the alternative treatment group were afforded similar flexibility but could not switch to semaglutide. Trial products were dispensed by the participants’ pharmacies. Semaglutide or alternative treatment copayments were equalized at a maximum of US$20 per month to minimize the impact of differential out-of-pocket costs between treatments.Study procedures In keeping with the pragmatic design, there were few study-specific evaluations, and no protocol-specified attempts were employed to ensure participants adhered to treatment or completed follow-up assessments. Participants were followed for 2 years irrespective of any changes in antidiabetic treatment. Dedicated trial visits were at randomization, year 1 (±6 weeks), and year 2 (±6 weeks). Clinical data were also collected during routine diabetic care visits.Data from patient-reported outcomes (PROs), including the Diabetes Treatment Satisfaction Questionnaire (DTSQ),14–17 12-Item Short Form version 2 (SF-12),18 and Work Productivity and Activity Impairment: General Health (WPAI:GH) questionnaire,19 were collected at years 1 and 2. Serious adverse events (SAEs) reported by participants during the trial were recorded. AEs not meeting the SAE definition were not required to be collected unless they led to study drug discontinuation.Outcomes The primary endpoint was the proportion of participants achieving HbA 1c <7.0% (53.0 mmol/mol) at year 1. Confirmatory secondary endpoints included change in HbA1c (percentage point) from baseline to years 1 and 2, and proportion of participants achieving HbA1c <7.0% at year 2. Key secondary endpoints and additional derived outcomes for supportive analyses are summarized in online supplemental table S1. Baseline was defined as ≤90 days before randomization (for HbA1c) or ≤4 weeks before randomization (for non-HbA1c secondary endpoints).SP110.1136/bmjdrc-2025-005161.supp1Supplementary dataSupportive secondary endpoints included time to treatment intensification, discontinuation, or change. Treatment intensification was defined as initiating an antidiabetic medication (besides the 1–2 oral antidiabetic medications taken before randomization) in addition to the study drug, or initiation of >1 antidiabetic medication after discontinuing the study drug. Treatment discontinuation was defined as cessation of study drug without initiating another antidiabetic medication for the remainder of the trial. Treatment change was defined as discontinuing the study drug and initiating another antidiabetic medication besides the 1–2 oral antidiabetic medications taken before randomization.Randomization Participants were randomized 1:1 to once-weekly semaglutide or alternative treatment with permuted blocks of size four using centralized web-based allocation via the electronic data capture system. In this open-label study, participants and clinic staff were aware of the treatment assignments, but masking procedures were employed by the study team to minimize bias. The study steering group included the Novo Nordisk study team, as well as the Carelon Research study statistician, who was blinded to treatment group assignment until the 1-year database lock.Statistical methods Power and sample size determination are described in the online supplemental file 1.The full analysis set (FAS) included all randomized participants. Missing data were imputed for some analyses; details of the imputation methodology are presented in the online supplemental file 1.Two estimands were defined, each involving missing data imputation. The primary “intention-to-treat” estimand evaluated the effectiveness of randomized treatment irrespective of adherence. Data for participants without observations at a study visit were imputed separately by treatment group and “on or off” study drug status using data from all participants with observations at the study visit. The secondary estimand evaluated the effectiveness of randomized treatment as though all participants were adherent. Data for participants without observations and/or no longer taking the study drug at a study visit were imputed separately by treatment group based on data from all participants with observations at the study visit who were still taking the study drug.Analyses of the primary and confirmatory secondary endpoints were based on the estimands and tested under multiplicity control via a hierarchical testing scheme in the order given in online supplemental table S1. Proportions of participants with HbA1c<7.0% at years 1 and 2 were analyzed using a logistic regression model with a logit link function, treatment as a categorical effect, and baseline HbA1c as covariate. From the model, the estimated OR for semaglutide versus alternative treatment was presented, with 95% CI and two-sided p value. Changes in HbA1c (percentage point) at years 1 and 2 were analyzed using analysis of covariance with treatment and baseline HbA1c as independent variables. From the model, the estimated mean difference in change from baseline to endpoint (estimated treatment difference (ETD)) for semaglutide versus alternative treatment was calculated with 95% CI and two-sided p value. These analyses were complemented by additional analyses without imputation of missing data. A complete case analysis was based only on participants with observations at the study visit, whether participants discontinued study drug or not. A complete case on-study drug analysis was based only on participants with observations at the study visit still taking study drug using the same statistical tests as the estimands.Analyses of supportive secondary endpoints; additional derived outcome measures; treatment intensification, change, or discontinuation; PROs; and subgroup data are described in the online supplemental file 1. Safety analyses were descriptive. SEPRA was conducted in accordance with the Declaration of Helsinki20 and the International Conference on Harmonisation Good Clinical Practice Guidelines.21Results Participant flow During the recruitment period, 1308 participants were screened and 1278 were randomized to once-weekly subcutaneous semaglutide (n=644) or alternative treatment (n=634) ( figure 1). Enrollment rates were lower than projected (original target, 2250; revised target, 1387).10 Among other factors, the COVID-19 pandemic negatively impacted recruitment and affected follow-up.10Figure 1Participant disposition (Consolidated Standards of Reporting Trials diagram). aOther reasons for early trial termination include: change in insurance, treating physician, or residence, financial restrictions, lost contact due to COVID-19, and the treating physician withdrew the patient. bIncluding 406 (63.0%) who attended within a ±6 week window around the year 1 time point. cHbA1c years 1 and 2 assessments include all values collected within a ±10 week window around the years 1 and 2 visit dates. dIncluding 368 (57.1%) who attended within a ±6 week window around the year 2 time point. eIncluding 428 (67.5%) who attended within a ±6 week window around the year 1 time point. fIncluding 351 (55.4%) who attended within a ±6 week window around the year 2 time point. FAS, full analysis set; HbA1c, glycated hemoglobin.23 participants (semaglutide, n=10; alternative treatment, n=13) did not initiate study drug. A total of 368 (28.8%) participants ended trial participation before year 2, including 125 participants who terminated before or at year 1. The most common reasons for early termination included lost to follow-up, withdrawal of consent, and “other,” including COVID-19–related issues, change in insurance, or other financial concerns.Overall, 892 (69.8%) and 748 (58.5%) participants had HbA1c data at years 1 and 2, respectively, and 910 (71.2%) participants completed the trial.Baseline characteristics and treatment at randomization Baseline characteristics of the overall population were reported previously 10 and balanced between treatment groups (online supplemental table S2).The initial doses of semaglutide chosen by physicians at randomization were 0.25 mg (n=450 (69.9%)), 0.5 mg (n=135 (21.0%)), 0.75 mg (n=1 (0.2%)), and 1.0 mg (n=48 (7.5%)) (online supplemental table S3). Data were missing for 10 (1.6%) participants. The alternative treatments chosen by physicians at randomization included GLP-1 RAs (n=452 (71.3%)), sodium-dependent glucose cotransporter (SGLT-2) inhibitors (n=98 (15.5%)), dipeptidyl peptidase 4 inhibitors (n=48 (7.6%)), sulfonylureas (n=13 (2.1%)), metformin (n=13 (2.1%)), insulin (n=10 (1.6%)), thiazolidinediones (n=3 (0.5%)), and other medications (n=2 (0.3%)) (online supplemental table S4). The most commonly used alternative GLP-1 RA was dulaglutide (n=371 (58.5%)); the initial doses of dulaglutide chosen by physicians at randomization were 0.5 mg (n=3 (0.8%)), 0.75 mg (n=298 (80.3%)), and 1.5 mg (n=70 (18.9%)) (online supplemental table S5). Data were missing or unknown for 6 (0.9%) participants.Glycemic control Superior HbA 1c reduction was observed in the semaglutide group compared with alternative treatment (online supplemental table S6). Based on the primary estimand, the proportions of participants achieving HbA1c <7.0% (53.0 mmol/mol) for semaglutide and alternative treatment were 53.1% and 45.5% at year 1 (primary endpoint; OR (95% CI): 1.36 (1.03 to 1.79); p=0.033) and 49.9% and 38.9% at year 2 (OR (95% CI): 1.56 (1.13 to 2.16); p=0.007) (figure 2A). The mean HbA1c decreases for semaglutide and alternative treatment (percentage point) were −1.35% and −1.16% at year 1 (ETD (95% CI): −0.20% (−0.39% to 0.00%); p=0.046) and −1.27% and −0.96% at year 2 (ETD (95% CI): −0.31% (−0.57% to −0.05%); p=0.018) (figure 2B). Treatment effects based on the secondary estimand were similar (online supplemental table S7). Complete case and complete case on-study drug analyses for the primary and confirmatory second endpoints supported these results (online supplemental table S7).Figure 2(A) Percentages of participants who achieved HbA1c <7.0% (53.0 mmol/mol) at year 1 (primary endpoint) and year 2, and (B) changes from baseline in HbA1c at year 1 and year 2. All data include imputed data for participants without HbA1c values at years 1 and 2. ETD, estimated treatment difference; HbA1c, glycated hemoglobin.HbA1c reductions were similar in participants taking semaglutide 0.25 mg, 0.5 mg and 1.0 mg (online supplemental table S8); mean (SD) decreases in HbA1c from baseline to year 1 were −1.45% (1.56%), −1.49% (1.59%), and −1.59% (1.57%), respectively, and from baseline to year 2 were −1.51% (1.66%), −1.49% (1.68%), and −1.51% (1.73%), respectively. Results from supportive glycemic control-related endpoints were consistent with the primary and secondary endpoints (online supplemental table S9). In descriptive analyses based on the FAS, HbA1c reduction with semaglutide was greater compared with dulaglutide (taken by 58.5% of the alternative treatment group) and SGLT-2 inhibitors (taken by 15.5% of participants in the alternative treatment group); mean (SD) decreases in HbA1c from baseline to year 1 were −1.46% (1.67%), −1.38% (1.72%), and −0.41% (1.24%), respectively, and from baseline to year 2 were −1.45% (1.72%), −1.14% (1.84%), and −0.60% (1.01%), respectively (table 1).Table 1Descriptive summary of HbA1c data* in patients taking semaglutide, or any alternative treatment, including dulaglutide and SGLT-2 inhibitors (full analysis set)Semaglutide(n=644)Any alternative treatment(n=634)Any alternative treatmentDulaglutide(n=371)SGLT-2 inhibitors(n=98)HbA1c at year 1 N (missing)430 (0)462 (0)250 (0)57 (0) <7.0, n (%)244 (56.7)226 (48.9)134 (53.6)22 (38.6)HbA1c at year 2 N (missing)374 (0)374 (0)198 (0)37 (0) <7.0, n (%)209 (55.9)162 (43.3)102 (51.5)13 (35.1)Change in HbA1c from baseline to year 1 N (missing)430 (0)461 (1)250 (0)57 (0) Mean (SD)−1.46 (1.67)−1.14 (1.71)−1.38 (1.72)−0.41 (1.24) Median (range)−1.3 (−11.2, 4.7)−0.9 (−9.7, 4.4)−1.1 (−6.8, 4.4)−0.5 (−5.1, 2.8)Change in HbA1c from baseline to year 2 N (missing)374 (0)373 (1)198 (0)37 (0) Mean (SD)−1.45 (1.72)−0.98 (1.77)−1.14 (1.84)−0.60 (1.01) Median (range)−1.2 (−8.4, 3.5)−0.8 (−7.3, 6.3)−1.0 (−7.3, 5.7)−0.6 (−3.9, 1.7)*HbA1c assessments include all values collected per protocol including a ±10 week window around the year 1 and year 2 visit dates.HbA1c, glycated hemoglobin; SGLT-2, sodium-glucose transport protein 2.Weight loss and blood pressure Reduction in body weight was greater for semaglutide than for alternative treatment. Using the primary estimand, the percentage changes from baseline in body weight for semaglutide and alternative treatment were −3.57% and −1.91% at year 1 (ETD (95% CI): −1.65% (−2.92% to −0.39%); p=0.010) and −3.84% and −2.36% at year 2 (ETD (95% CI): −1.48% (−3.62% to 0.66%); p=0.175) ( figure 3A). Treatment effects based on the secondary estimand were similar, but the ETDs were statistically significant at year 1 and year 2.Figure 3Changes from baseline to year 1 and year 2 in (A) body weight, (B) SBP, and (C) DBP. All data include imputed data for participants without HbA1c values at years 1 and 2. DBP, diastolic blood pressure; ETD, estimated treatment difference; HbA1c, glycated hemoglobin; SBP, systolic blood pressure.Systolic blood pressure (SBP) and diastolic blood pressure (DBP) decreased from baseline to years 1 and 2 in both treatment groups. For both the primary and secondary estimands, the decrease in SBP was numerically higher for semaglutide compared with alternative treatment at years 1 and 2, but the ETDs were not statistically significant (figure 3B). Decreases in DBP at years 1 and 2 were comparatively modest, and the ETDs were not statistically significant (figure 3C).Treatment intensification, change, and discontinuation Treatment intensification occurred less frequently with semaglutide than with alternative treatment from baseline to year 2 (88/634 (13.9%) vs 118/621 (19.0%)) ( table 2). Among participants still in the trial and taking semaglutide at the year 2 visit, treatment intensification occurred more frequently with higher semaglutide doses (post hoc analysis; 0.25 mg: 9/113 (8.0%); 0.5 mg: 16/146 (11.0%); and 1.0 mg: 41/181 (22.7%)) (online supplemental table S10). Treatment change also occurred less frequently with semaglutide than with alternative treatment from baseline to year 2 (27/634 (4.3%) vs 60/621 (9.7%)), as did permanent treatment discontinuation (194/634 (30.6%) vs 207/621 (33.3%)).Table 2Treatment changes in glucose-lowering medication from baseline to year 1 and from baseline to year 2 (full analysis set)Participants, n (%)Semaglutide(n=634)Alternative treatment(n=621)Any treatment intensification* Baseline to year 161 (9.6)82 (13.2) Baseline to year 288 (13.9)118 (19.0)Added another antidiabetic medication in addition to study drug Baseline to year 157 (9.0)72 (11.6) Baseline to year 282 (12.9)101 (16.3)Added another antidiabetic medication after discontinuing study drug Baseline to year 16 (0.9)10 (1.6) Baseline to year 28 (1.3)19 (3.1)Treatment change† Baseline to year 120 (3.2)43 (6.9) Baseline to year 227 (4.3)60 (9.7)Early treatment discontinuation‡ Baseline to year 160 (9.5)54 (8.7) Baseline to year 2194 (30.6)207 (33.3)23 participants (semaglutide, n=10; alternative treatment, n=13) who did not initiate study drug following randomization were excluded from the analysis.*Treatment intensification was defined as initiation of an antidiabetic medication (besides the up to two oral antidiabetic medications taken before randomization) in addition to the study drug, or initiation of more than one antidiabetic medication after discontinuation of the study drug.†Treatment change was defined as discontinuation of study drug and initiation of another antidiabetic medication (besides the up to two oral antidiabetic medications taken before randomization).‡Treatment discontinuation was defined as discontinuation of study drug without initiation of another antidiabetic medication for the remainder of the participant’s time in the trial.Time to first study drug discontinuation was similar for semaglutide and alternative treatment (mean (SD): 1.4 (0.9) years vs 1.5 (0.7) years), but the incidence rate (per year of exposure) of first study drug discontinuation was higher (0.22 vs 0.16; ETD (95% CI): 1.33 (1.08 to 1.65); p=0.008). The difference between groups in time to first study drug discontinuation was more pronounced in men than women (p=0.040), but no differences were noted based on age, baseline HbA1c, BMI, or duration of T2D (online supplemental table S11). Mean (SD) time to first treatment intensification or change was 1.6 (0.7) and 1.4 (0.7) years for semaglutide and alternative treatment, with no subgroup differences noted (online supplemental table S11). The incidence rate of treatment intensification or change was lower for semaglutide than for alternative treatment (0.12 vs 0.20; ETD (95% CI): 0.61 (0.47 to 0.77); p<0.001).Patient-reported outcomes PRO results (complete case analysis) indicated greater improvement with semaglutide versus alternative treatment on several measures ( figure 4). On the DTSQ, there was a significantly greater decrease in perceived hyperglycemia with semaglutide at years 1 (ETD (95% CI): −0.25 (−0.49 to –0.01); p=0.040) and 2 (ETD (95% CI): −0.38 (−0.64 to –0.12); p=0.004), and a significantly larger increase in total treatment satisfaction at year 2 (ETD (95% CI): 0.95 (0.09 to 1.81); p=0.001) compared with alternative treatment. On the SF-12, there was a larger mental component summary score increase (indicating improvement) with semaglutide at years 1 (ETD (95% CI): 1.26 (0.16 to 2.36); p=0.032) and 2 (ETD (95% CI): 2.26 (1.00 to 3.53); p=0.002) compared with alternative treatment. Results on other measures, including the WPAI:GH (work and activity impairment), were highly variable and did not show statistically significant differences between treatment groups. Complete on-study case analyses for PROs are summarized in online supplemental table S12.Figure 4PROs at year 1 and year 2 (complete case analysis). aMean treatment difference (semaglutide–alternative treatment). DTSQ, Diabetes Treatment Satisfaction Questionnaire; ETD, estimated treatment difference; PRO, patient-reported outcome; SF-12 v2, 12-Item Short Form Survey version 2; WPAI:GH, Work Productivity and Activity Impairment Questionnaire: General Health.Safety The safety profile was consistent with prior analyses of semaglutide in T2D. No new safety concerns were identified. There were more non-serious AEs leading to discontinuation with semaglutide (6.6%) than with alternative treatment (4.7%) ( online supplemental table S13), primarily involving gastrointestinal AEs (online supplemental table S14). All SAEs and fatal outcomes were generally balanced between groups. Additional safety data are reported in online supplemental tables S13–S16.Discussion Results of the randomized, open-label, highly pragmatic SEPRA trial demonstrate the effectiveness of once-weekly subcutaneous semaglutide compared with alternative treatment in glycemic control and weight loss, complementing results of phase III RCTs and observational studies and supporting the integration of semaglutide into clinical practice.Mean HbA1c reductions among participants treated with semaglutide were −1.35% and −1.49% (primary and secondary estimands, respectively) from baseline to year 1, and −1.27% and −0.96% from baseline to year 2. In the SUSTAIN clinical trial program (SUSTAIN 1–7; 30–104 weeks), the magnitude of change in HbA1c from baseline ranged from −1.4% to −1.8% for semaglutide 1.0 mg and −1.1% to −1.5% for semaglutide 0.5 mg.5 Compared with participants in the SUSTAIN program, participants in SEPRA were slightly older and heavier, with a shorter duration of T2D and wider range of baseline HbA1c, and fewer were Hispanic; additionally, participants in SEPRA had higher rates of hypertension, ischemic heart disease and stroke, and chronic heart failure.5 8 22–26 In a meta-analysis of real-world studies, once-weekly subcutaneous semaglutide reduced HbA1c by −1.1% (26 to ~30 weeks follow-up).27 Glycemic control (HbA1c <7.0%) was maintained over time to a greater extent with semaglutide (year 1: 53.1%; year 2: 49.9%) compared with alternative treatment (45.5%–38.9%). The glycemic control observed with semaglutide compared with the physician’s choice of alternative glucose-lowering medications is notable, as others have observed waning effectiveness of long-term diabetes treatment.28 Although previous studies that used fixed-dose escalation or randomization to specific doses of semaglutide have demonstrated dose-dependent responses in HbA1c,8 26 there were minimal differences in glycemic outcomes by semaglutide dose in SEPRA, which may have been a result of this pragmatic study not being designed to assess the effect of dose on HbA1c. Specifically, similar HbA1c results at various doses of semaglutide may have occurred because patients who achieved HbA1c targets or experienced adverse events at a low dose would not necessarily increase their dose, while those who did not achieve targets may increase their dose; therefore, dose response was not formally assessed, since the randomization scheme was based on the treatment that was used and not the specific dose that was administered as was done previously.Average weight loss with semaglutide in SEPRA (year 1: −3.57% and −4.0 kg; year 2: −3.84% and −4.44 kg) was consistent with SUSTAIN 1–7 (semaglutide 1.0 mg: −4.9 kg to −6.5 kg; semaglutide 0.5 mg: −3.5 kg to −4.6 kg) based on the doses used in SEPRA.5 Average weight loss of −4.9 kg was reported in other real-world studies (over 26 to ~30 weeks).27 The ETD for average weight loss for semaglutide versus alternative treatment was only significantly different for years 1 and 2 in the second estimand that assumed all participants were adherent, whereas the ETD for the primary estimand was only significant at year 1. These results indicate that adherence to the study drug may influence the degree of weight loss at year 2, although differences in dose escalation patterns, tolerability issues, or treatment discontinuation may have also contributed to the lack of difference in weight loss between treatment groups for the primary estimand. Additionally, in this pragmatic study, practice patterns varied, and patients and practitioners prioritized achieving HbA1c levels near 7% as opposed to maximizing the semaglutide dose to achieve greater weight loss.PROs showed a tendency toward improvement from baseline to year 2, suggesting greater treatment satisfaction and fewer problems with perceived hyperglycemia in participants receiving semaglutide versus alternative treatment. This suggests that semaglutide may be associated with improvements in quality of life in individuals with T2D in a real-world setting. However, these results may be influenced by the open-label nature of the trial and differential loss to follow-up of less satisfied participants.The benefits of semaglutide versus alternative treatment were observed even though only 27.5% of participants were taking semaglutide 1.0 mg, the highest dose approved for this indication at the time, at year 2. Some participants may have taken lower doses because they achieved their HbA1c target; experienced intolerable side effects, treatment inertia, and lack of attention to the titration regimen; missed follow-up appointments; or had difficulty with access due to medication shortages or insurance changes, among other possibilities. Prospective, single-arm, non-interventional, real-world studies conducted outside the US have reported a wide range (11%–71%) of participants taking semaglutide 1.0 mg.29–37 Treatment intensification occurred more frequently with higher doses of semaglutide, possibly because some participants did not achieve their glycemic target even with the highest dose of semaglutide; however, treatment intensification was less frequent with semaglutide overall compared with alternative treatment. Current guidance recommends that treatments selected at initiation or intensification should meet glycemic and weight targets established with patients through shared decision-making.3A large number and variety of therapeutics were taken by participants in the alternative treatment group, but a majority of participants took GLP-1 RAs (71.3%), particularly dulaglutide (58.5%), or SGLT-2 inhibitors (15.5%). Although this study was not powered to compare outcomes by treatment, and alternative treatments were at the discretion of the treating physician, descriptive analyses demonstrated a relatively modest but consistent superiority in glycemic control between semaglutide and dulaglutide, and a considerably larger difference between semaglutide and SGLT-2 inhibitors. Similarly, reductions in HbA1c were greater with semaglutide at year 1 and year 2 compared with dulaglutide and SGLT-2 inhibitors. These findings are consistent with previous studies showing greater efficacy of semaglutide compared with these other therapeutics.26 38 Additionally, head-to-head studies that assessed glucose metabolism or mean change in HbA1c in patients who switched from dulaglutide or liraglutide to semaglutide compared with patients who remained on initial treatment demonstrated improved outcomes when switching to semaglutide, which is consistent with the descriptive analyses in this study demonstrating improved glycemic control with semaglutide versus dulaglutide.39 40A total of 368 (28.8%) participants terminated SEPRA early, most commonly due to loss to follow-up (8.3%), withdrawal of consent (5.8%), or other reasons (7.9%), including issues related to COVID-19, change in insurance, or other financial concerns. Reasons for early termination were generally balanced across treatment groups. A limitation inherent in pragmatic clinical trials conducted in routine clinical practice, including SEPRA, is a lack of study participation-related incentives,13 which are common in traditional RCTs but generally antithetical to a hands-off pragmatic approach.11 13 Aspects of the SEPRA study design were implemented to minimize potential causes of missing data, such as equalizing copays and facilitating data collection by recording HbA1c data within a protocol-specified window. In addition to dropout, a certain amount of missing data was anticipated for other reasons, such as the known rate of 22%–30% non-adherence to newly prescribed antidiabetic medications reported in real-world studies.41–44 However, treatment adherence was high at years 1 (semaglutide: 81.5%; alternative treatment: 81.2%) and 2 (semaglutide: 68.3%; alternative treatment: 62.6%) (figure 1). Post hoc exploration of SEPRA participants with and without HbA1c data at year 1 found no meaningful difference in baseline demographic, clinical, and diabetes characteristics.From a statistical perspective, a selection of predefined analytic approaches was used to decrease uncertainty due to missingness by using both imputed and observed HbA1c data. Initial sample size calculations used an estimate of 25% missing data for the primary and confirmatory HbA1c endpoints at year 1, which were slightly less than the observed proportion of 30%, but a conservative approach nonetheless, considering the use of imputed HbA1c data in the primary endpoint analysis. The observed proportion of 41% missing data at year 2 was slightly lower than the estimated 44% for the confirmatory HbA1c endpoints. The protocol-defined, prespecified analytic approach of using multiple imputation rather than repeated measures modeling was chosen based on the pragmatic study design because there were minimal required visits planned at years 1 and 2 post randomization. Additional HbA1c data from routine visits were to be collected if they occurred, but this varied by participant. The analyses using imputed data were supplemented with complete case analysis and complete case on-study analyses of observed data. All analytic approaches consistently demonstrated the superior effectiveness of semaglutide in achieving glycemic control at year 2.A key strength of SEPRA is the highly pragmatic design, scoring 4–5 (on a scale of 1–5, with 5 being very pragmatic) across all PRECIS-2 domains,13 which bridges RCTs and observational study real-world evidence.10 This stands in contrast to the randomized trials on which general practice guidelines for the management of T2D are based, which were judged to be more explanatory than pragmatic when evaluated with PRECIS-2.45 Consistent with the pragmatic design, the trial population was more inclusive than traditional RCTs, and the follow-up period was longer than in other real-world studies (generally 6 months to 1 year).35 46 47 Regarding limitations, the COVID-19 pandemic caused recruitment challenges, and the estimated population size for power calculation was not met (target, 1387 participants; actual, 1278 participants).10 Copay equalization and supply shortages during the study may have impacted the pragmatic nature of SEPRA. Additional limitations include the open-label design, comparator heterogeneity, reliance on multiple-imputation analyses, and lack of comparison by dose of comparator GLP-1 RAs. Despite these limitations, the final design afforded an opportunity to robustly assess real-world evidence.In conclusion, results of SEPRA confirm that semaglutide is an appropriate choice for treatment intensification among individuals with T2D who are receiving 1–2 antidiabetic medications in the USA. This trial complements findings from phase III and observational trials and supports the use of semaglutide as an effective and well-tolerated treatment option for glycemic control and weight management in real-world clinical practice.SP210.1136/bmjdrc-2025-005161.supp2Supplementary data