Dose–Response and Clinical Equivalence of Semaglutide and Tirzepatide for Weight Loss in Type 2 Diabetes: A Model-Based Analysis Authors: Carlos E Builes-Montaño, Andres F Suarez-Rodriguez, Maria A Alzate-Vinasco Published: 2026 ABSTRACT AbstractIntroductionSemaglutide and tirzepatide are highly effective incretin-based therapies for weight reduction in individuals with type 2 diabetes (T2DM). However, direct comparisons across the full range of clinically relevant doses in T2DM are limited. This study aimed to evaluate dose–response relationships and clinical equivalence between semaglutide and tirzepatide using a model-based approach.MethodsArm-level data from phase III randomized controlled trials of semaglutide (SUSTAIN, STEP) and tirzepatide (SURPASS, SURMOUNT-2) in T2DM were analyzed. Dose–response relationships for percent weight change were modeled using generalized additive models and Bayesian hierarchical spline models. Posterior probabilities of clinical equivalence were estimated across prespecified dose pairs using an equivalence margin of ± 2 percentage points. Model-based intensification scenarios were evaluated to estimate incremental benefit associated with treatment switching or dose escalation.ResultsA total of 48 treatment arms (n = 16,524 participants) were included. Both agents demonstrated nonlinear dose–response relationships, with attenuation of incremental effects at higher doses. High probabilities of equivalence were observed for semaglutide 2.4 mg versus tirzepatide 10 mg (99.4%) and semaglutide 7.2 mg versus tirzepatide 15 mg (94.8%). Lower doses of semaglutide were not equivalent to higher doses of tirzepatide. In intensification scenarios, both switching and dose escalation strategies improved weight loss, although the probability of achieving ≥ 2 additional percentage points varied across regimens.ConclusionSemaglutide and tirzepatide demonstrate comparable weight-loss efficacy at specific dose combinations, with equivalence driven by dose rather than drug identity. Model-based analyses provide a quantitative framework to support individualized treatment decisions when direct comparative evidence is unavailable.Supplementary InformationThe online version contains supplementary material available at 10.1007/s13300-026-01886-0.Keywords: Bayesian modeling, Clinical equivalence, Dose–response, Semaglutide, Tirzepatide, Treatment intensification, Type 2 diabetes, Weight loss Key Summary Points Why carry out this study? Direct comparisons between semaglutide and tirzepatide across clinically relevant dose ranges in type 2 diabetes are limited.The clinical equivalence and interchangeability of these agents at different doses remain uncertain.This study evaluated whether dose–response modeling can inform equivalence and treatment decisions. What was learned from the study? Specific dose pairs (e.g., semaglutide 2.4 mg and tirzepatide 10 mg) showed high probability of clinical equivalence for weight loss.Dose escalation and switching strategies provide variable probabilities of achieving clinically meaningful additional weight loss. Introduction Glucagon-like peptide 1 receptor agonists (GLP-1 RAs) are established therapies for type 2 diabetes (T2DM), exerting their effects through glucose-dependent insulin secretion, suppression of glucagon release, delayed gastric emptying, and central regulation of appetite [1]. These mechanisms collectively contribute to improved glycemic control and clinically meaningful weight reduction, positioning GLP-1-based therapies as key agents in the management of cardiometabolic risk in T2DM. Beyond glycemic control and weight loss, GLP-1-based therapies exhibit broader biological effects, including improvements in lipid metabolism, anti-inflammatory activity, reductions in oxidative stress, and favorable effects on vascular function and atherosclerosis, which contribute to cardiovascular risk reduction [2–4]. These pleiotropic effects reinforce their clinical positioning but also highlight the complexity of their pharmacodynamic profile. From a pharmacological perspective, many of these effects are exposure-dependent, supporting the concept that treatment response may vary across dose levels. In particular, pharmacokinetic/pharmacodynamic analyses of semaglutide suggest an approximately linear exposure–response relationship for weight reduction, in contrast to the more saturable response observed for glycemic endpoints. Phase III trials have shown that increasing doses are associated with progressively greater weight loss across patient subgroups [5]. Similarly, tirzepatide, a dual glucose-dependent insulinotropic polypeptide (GIP) and GLP-1 receptor agonist, demonstrates a dose-dependent gradient of efficacy, with increasing weight reduction observed from 5 to 15 mg weekly [6]. At the class level, dose–response meta-analyses further support nonlinear but predictable relationships between dose and weight loss across incretin-based therapies, with higher doses consistently associated with greater efficacy [7]. These findings support the hypothesis that weight loss is largely exposure-driven and may be comparable across agents at specific dose levels. Despite this evidence, a critical gap remains. The only direct head-to-head trial comparing tirzepatide and semaglutide (SURMOUNT-5) was conducted in individuals with obesity without T2DM and evaluated maximum tolerated doses rather than the full range of clinically relevant dosing strategies [8]. As a result, comparative dose equivalence in T2DM populations remains uncertain. This limitation has direct clinical implications, as treatment decisions frequently involve dose escalation or switching between agents in the absence of head-to-head evidence. A model-based dose–response framework offers a quantitative approach to address this gap, enabling indirect comparisons and estimation of pharmacodynamic similarity across therapies. Accordingly, this study aimed to evaluate pharmacodynamic similarity between semaglutide and tirzepatide by comparing their dose–response relationships for weight loss in individuals with T2DM, using a model-based synthesis of phase III randomized clinical trial data. Methods Data Sources and Processing We conducted a structured extraction of arm-level data from phase III randomized controlled trials evaluating semaglutide and tirzepatide in individuals with T2DM. Eligible studies were predefined and restricted to trials within the SUSTAIN, STEP, SURPASS, and SURMOUNT clinical development programs. For semaglutide, included trials comprised STEP 2 [9], STEP UP-DM [10], SUSTAIN-1 through SUSTAIN-11, and SUSTAIN FORTE [11–22]. For tirzepatide, eligible trials included SURPASS-1 through SURPASS-6 [23–28], and SURMOUNT-2 [29]. Extension phases and trials enrolling exclusively individuals without diabetes were excluded. Sex-disaggregated data were not consistently available at the trial arm level and were therefore not included in the analysis. Arm-level data were extracted from primary publications and supplementary materials. Extracted baseline characteristics included mean age, sex distribution, baseline HbA1c, and baseline body weight. Intervention variables included study drug, weekly dose (mg), and treatment duration. Outcomes extracted comprised percent and absolute change in body weight, categorical weight loss thresholds (≥ 5%, ≥ 10%, ≥ 15%, ≥ 20%), and HbA1c change. Adverse events and discontinuation rates were extracted where available; however, these outcomes were not included in the modeling framework because of heterogeneity in reporting and limitations inherent to aggregated data. When multiple time points were reported, data were extracted at the primary efficacy endpoint defined in each trial. When measures of dispersion for continuous outcomes were not directly reported, standard deviations or standard errors were derived from available confidence intervals assuming approximate normality. Treatment arms lacking sufficient information for variance estimation were excluded from analyses requiring uncertainty inputs but retained for descriptive summaries. All doses were harmonized to weekly milligrams. Risk of bias was assessed using the Cochrane Risk of Bias 2 (ROB2) tool across five domains [30]. Baseline Body Weight Comparison Baseline comparability between semaglutide and tirzepatide arms was assessed using pooled estimates of mean baseline body weight. Treatment arms with missing baseline weight or dispersion measures were excluded. Pooled estimates were calculated using two approaches: (i) weighted mean baseline weight with 95% confidence intervals, weighting each arm by sample size; and (ii) random-effects meta-analysis of raw means using inverse-variance weighting. Between-study variance was estimated using the DerSimonian–Laird method, and confidence intervals were adjusted using the Hartung–Knapp approach. Indirect comparisons between drug groups were performed by calculating the difference between pooled random-effects means. Standard errors were derived assuming independence, and 95% confidence intervals were constructed accordingly. Statistical heterogeneity was quantified using the I2 statistic and between-study variance (τ2). Dose–Response Analysis Dose–response relationships between weekly dose and percent weight change were evaluated separately for semaglutide and tirzepatide. As a descriptive step, weight change distributions were summarized across clinically relevant discrete dose levels (semaglutide: 1.0, 2.4, and 7.2 mg; tirzepatide: 5, 10, and 15 mg). Distributions were visualized using boxplots with overlaid jittered arm-level observations. Continuous dose–response relationships were modeled using generalized additive models (GAMs), with percent weight change as the dependent variable and weekly dose (mg) modeled as a smooth function. GAMs were selected to allow flexible nonlinear relationships without imposing a predefined parametric structure [31]. Penalized regression splines were estimated using restricted maximum likelihood (REML). To improve model stability at the lower boundary of the dose range, pseudo-observations corresponding to zero weight change at 0 mg were included for each drug. These pseudo-observations serve to anchor the dose–response relationship at a physiologically plausible baseline (i.e., no treatment effect in the absence of exposure), thereby preventing unrealistic extrapolation near zero dose. Because these points are located outside the range of observed doses, they have negligible influence on model estimates within the observed data. Model-based predictions and corresponding 95% confidence intervals were generated across the full dose range. Incremental effects were estimated for prespecified dose intervals by contrasting predicted values at interval boundaries. Posterior Probability of Equivalence To quantify cross-drug dose comparability, posterior probabilities of clinical equivalence were estimated for prespecified semaglutide doses (1.0, 2.4, 7.2 mg/week) and tirzepatide doses (5, 10, 15 mg/week). A Bayesian hierarchical model was fitted using the brms framework, specifying percent weight change as a Gaussian outcome with (i) drug-specific smooth dose effects (splines by drug), (ii) treatment duration included as a covariate to account for variability in follow-up time across trials, and (iii) a study-level random intercept to account for between-trial heterogeneity. Where available, standard errors were incorporated using the se() specification. Models were estimated using Hamiltonian Monte Carlo with conservative tuning parameters (adapt_delta and tree depth), and convergence was assessed using standard diagnostics [32]. Population-level posterior predictions were generated at prespecified dose pairs, holding treatment duration constant at a representative value to ensure comparability across treatments. For each posterior draw, pairwise differences in predicted weight change were computed (Δ = semaglutide − tirzepatide). Clinical equivalence was defined as the posterior probability that the absolute difference fell within a predefined equivalence margin (± δ, expressed in percentage points). The equivalence margin (± 2 percentage points) was selected as a pragmatic threshold for between-treatment differences, rather than an absolute clinical target. While a ≥ 5% weight reduction is commonly considered clinically meaningful at the individual level [33], smaller differences between treatments (e.g., ± 2%) are unlikely to translate into meaningful differences in clinical decision-making and are therefore appropriate for defining equivalence. Posterior probabilities of equivalence were calculated as the proportion of posterior draws satisfying this criterion. Statistical Environment All analyses were conducted using R version 4.5.2 (ucrt build) within RStudio (version 2025.05.0). Data processing and visualization were performed using standard reproducible workflows. Dose–response modeling, meta-analysis, and Bayesian hierarchical models were implemented using validated statistical packages. Bayesian models were estimated using four Markov chains with 4000 iterations per chain, including 2000 warm-up iterations. Convergence was assessed using the potential scale reduction factor ( ) and effective sample size criteria. Additional methodological details, including model specification and sensitivity analyses, are provided in the supplementary material. Ethical Considerations This study was based exclusively on aggregated, arm-level data extracted from previously published clinical trials. No individual participant data were used. Therefore, institutional review board approval and informed consent were not required for the present secondary analysis. According to the information reported in the original trial publications, the included studies received approval from the corresponding ethics committees or institutional review boards, were conducted in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines, and obtained written informed consent from all participants. Results Study Selection and Characteristics A total of 23 phase III randomized controlled trials met the inclusion criteria, contributing 48 treatment arms conducted exclusively in individuals with T2DM (see Table 1 for a summary of trial-level and treatment-arm characteristics, and Supplementary Table S1 for detailed study characteristics). Of these, semaglutide was evaluated in 15 trials (23 treatment arms; n = 9902 participants) and tirzepatide in 8 trials (25 treatment arms; n = 6622 participants). Table 1Characteristics of included trials and treatment armsCharacteristicSemaglutideTirzepatideTrials (n)158Treatment arms (n)2325Participants (n)99026622Dose range (mg/week)0.5–7.25–15Duration, weeks (mean ± SD)55.3 ± 24.553.3 ± 13.0Age, years (mean)54.855.9Female (%)51.149.2Baseline HbA1c (%)7.837.80Characteristics of included phase III randomized controlled trials and treatment armsSummary of trial-level and treatment-arm characteristics for semaglutide and tirzepatide studies included in the analysis. Data are presented as means or ranges across treatment arms. Baseline characteristics reflect aggregated arm-level data and may vary according to trial inclusion criteria Characteristics of included trials and treatment arms Characteristics of included phase III randomized controlled trials and treatment arms Summary of trial-level and treatment-arm characteristics for semaglutide and tirzepatide studies included in the analysis. Data are presented as means or ranges across treatment arms. Baseline characteristics reflect aggregated arm-level data and may vary according to trial inclusion criteria Semaglutide data were derived from the SUSTAIN (1–11, FORTE) and STEP (2, UP-DM) programs, whereas tirzepatide data were obtained from SURPASS (1–6) and SURMOUNT-2 trials. Weekly doses ranged from 0.5 to 7.2 mg for semaglutide and from 5 to 15 mg for tirzepatide. The number of participants per treatment arm ranged from 86 to 2470. Treatment duration ranged from 26 to 104 weeks (mean 55.3 weeks [SD 24.5] for semaglutide and 53.3 weeks [SD 13.0] for tirzepatide). Mean baseline age was 54.8 years for semaglutide and 55.9 years for tirzepatide; the proportion of female participants was 51.1% and 49.2%, respectively. Mean baseline HbA1c was 7.83% and 7.80%, and baseline BMI was broadly comparable across included trials, with mean values of approximately 33.3 kg/m2 for semaglutide and 33.6 kg/m2 for tirzepatide, and an overall range of approximately 31 to 36 kg/m2. Baseline Body Weight A total of 40 treatment arms were included in the baseline weight analysis (21 semaglutide, 19 tirzepatide), comprising 8888 and 6622 participants, respectively. The pooled weighted mean baseline body weight was 94.8 kg (95% CI 94.4–95.3) for semaglutide and 90.6 kg (95% CI 90.1–91.2) for tirzepatide. Random-effects meta-analysis yielded pooled estimates of 95.4 kg (95% CI 92.8–98.0) and 89.0 kg (95% CI 84.6–93.3), respectively. The indirect comparison showed a mean difference of − 6.4 kg (95% CI − 11.2 to − 1.7), with lower baseline weight in tirzepatide arms. Between-study heterogeneity was high for both groups (I2 > 99%). The baseline body weight across treatment arms and pooled estimates are shown in Fig. 1. Variability in baseline weight was observed across studies for both treatment groups, reflecting heterogeneity in trial populations. Despite differences in pooled mean estimates, the range of baseline values overlapped across semaglutide and tirzepatide arms. Baseline body weight across treatment arms and pooled estimates. Baseline body weight (kg) across treatment arms receiving semaglutide and tirzepatide. Points represent mean baseline body weight for each arm, with horizontal lines indicating 95% confidence intervals where available. Pooled estimates for each treatment group were derived using random-effects meta-analysis. Variability across studies reflects differences in trial populations and inclusion criteria Descriptive Weight Change by Dose Across both agents, increasing dose was associated with greater reductions in body weight. The distribution of percent weight change by dose is presented in Fig. 2. Distribution of percent weight change by dose and drug. Distribution of percent weight change across discrete weekly dose levels for semaglutide (1.0, 2.4, and 7.2 mg) and tirzepatide (5, 10, and 15 mg). Boxplots represent median and interquartile range, with individual treatment arms displayed as jittered points. Increasing dose is associated with greater weight reduction for both agents, with increased variability at higher doses For semaglutide, median weight loss increased from approximately 5–6% at 1.0 mg to around 10% at 2.4 mg, with further reductions observed at 7.2 mg. For tirzepatide, a clear gradient was observed from 5 to 15 mg, with the largest reductions at 15 mg. Variability in weight-loss responses increased at higher doses for both drugs. Dose–Response Modeling Semaglutide The semaglutide dose–response curve derived from the GAM is shown in Fig. 3. The model demonstrated a nonlinear, monotonic association between dose and percent weight change, with a steeper slope at lower doses and attenuation at higher doses. Model-based incremental effects are summarized in Table 2. Semaglutide dose–response curve estimated using a generalized additive model. Model-based dose–response relationship between weekly semaglutide dose (mg) and percent weight change, estimated using a generalized additive model (GAM). The solid line represents the predicted mean effect, and the shaded area indicates the 95% confidence interval. The curve demonstrates a nonlinear, monotonic association with attenuation of incremental effects at higher doses Table 2Incremental weight change across semaglutide dose intervalsDose interval (mg/week)Incremental weight change (%)95% CI0–1.0− 5.4− 6.9 to − 3.91.0–2.4− 4.4− 6.0 to − 2.82.4–7.2− 3.1− 5.8 to − 0.5Model-based incremental weight change across semaglutide dose intervalsEstimated incremental percent weight change associated with increasing semaglutide dose across prespecified dose intervals, derived from generalized additive modeling. Values represent differences in predicted mean weight change between interval boundaries, with corresponding 95% confidence intervals Incremental weight change across semaglutide dose intervals Model-based incremental weight change across semaglutide dose intervals Estimated incremental percent weight change associated with increasing semaglutide dose across prespecified dose intervals, derived from generalized additive modeling. Values represent differences in predicted mean weight change between interval boundaries, with corresponding 95% confidence intervals Tirzepatide The tirzepatide dose–response curve is presented in Fig. 4, showing a nonlinear association with the largest incremental effect at lower doses and continued reductions up to 15 mg. Incremental effects are summarized in Table 3. Tirzepatide dose–response curve estimated using a generalized additive model. Model-based dose–response relationship between weekly tirzepatide dose (mg) and percent weight change, estimated using a generalized additive model (GAM). The solid line represents the predicted mean effect, and the shaded area indicates the 95% confidence interval. The curve shows a nonlinear association with continued incremental weight reduction across the evaluated dose range Table 3Incremental weight change across tirzepatide dose intervalsDose interval (mg/week)Incremental weight change (%)95% CI0–5− 7.1− 8.8 to − 5.35–10− 2.5− 3.9 to − 1.010–15− 2.2− 3.6 to − 0.7Model-based incremental weight change across tirzepatide dose intervalsEstimated incremental percent weight change associated with increasing tirzepatide dose across prespecified dose intervals, derived from generalized additive modeling. Values represent differences in predicted mean weight change between interval boundaries, with corresponding 95% confidence intervals Incremental weight change across tirzepatide dose intervals Model-based incremental weight change across tirzepatide dose intervals Estimated incremental percent weight change associated with increasing tirzepatide dose across prespecified dose intervals, derived from generalized additive modeling. Values represent differences in predicted mean weight change between interval boundaries, with corresponding 95% confidence intervals Posterior Probability of Equivalence Posterior probabilities of equivalence across dose pairs are shown in Fig. 5. Probabilities varied substantially depending on the dose combination under the primary equivalence margin (± 2 percentage points). Posterior probability of clinical equivalence across dose pairs. Posterior probabilities of clinical equivalence across prespecified semaglutide and tirzepatide dose pairs using an equivalence margin of ± 2 percentage points for percent weight change. Values within each cell represent the posterior probability of equivalence, the estimated mean difference in percent weight change (Δ = semaglutide − tirzepatide), and the corresponding 95% credible interval. Positive values indicate greater estimated weight reduction with semaglutide, whereas negative values indicate greater estimated weight reduction with tirzepatide. Color intensity reflects the probability of equivalence The highest probabilities of equivalence were observed for semaglutide 2.4 mg vs tirzepatide 10 mg (99.4%) and semaglutide 7.2 mg vs tirzepatide 15 mg (94.8%). In contrast, comparisons involving semaglutide 1.0 mg vs tirzepatide 10–15 mg showed near-zero equivalence probability. Detailed results are presented in Table 4. Table 4Posterior probability of equivalence across dose pairsSemaglutide (mg)Tirzepatide (mg)Probability (%)Δ (p.p.)95% CrI1.0552.7− 2.0− 2.8 to − 1.11.0100.0− 3.9− 4.7 to − 3.01.0150.0− 5.9− 6.7 to − 5.02.4577.2+ 1.5+ 0.3 to + 2.82.41099.4− 0.4− 1.6 to + 0.92.41525.9− 2.4− 3.6 to − 1.27.250.2+ 4.5+ 2.8 to + 6.27.21023.6+ 2.6+ 0.9 to + 4.37.21594.8+ 0.6− 1.1 to + 2.3Posterior probability of equivalence for percent weight loss across prespecified dose pairsPosterior probabilities of clinical equivalence in percent weight loss for predefined semaglutide and tirzepatide dose combinations. Δ represents the posterior mean difference in percent weight loss (semaglutide − tirzepatide), expressed in percentage points. Credible intervals (CrI) correspond to 95% posterior intervals Posterior probability of equivalence across dose pairs Posterior probability of equivalence for percent weight loss across prespecified dose pairs Posterior probabilities of clinical equivalence in percent weight loss for predefined semaglutide and tirzepatide dose combinations. Δ represents the posterior mean difference in percent weight loss (semaglutide − tirzepatide), expressed in percentage points. Credible intervals (CrI) correspond to 95% posterior intervals Summary of Dose–Response Patterns Both agents demonstrated nonlinear dose–response relationships, with larger incremental effects at lower doses and attenuation at higher doses. Tirzepatide showed larger initial incremental effects, whereas semaglutide maintained incremental weight reduction across an extended dose range, as illustrated in Figs. 3 and 4. Sensitivity Analyses Across Equivalence Margins Sensitivity analyses evaluating wider equivalence margins are shown in Figs. 6 and 7. As expected, posterior probabilities of equivalence increased with wider margins across all dose pairs. The relative ranking of dose combinations remained consistent, with the highest probabilities observed for semaglutide 2.4 mg vs tirzepatide 10 mg and semaglutide 7.2 mg vs tirzepatide 15 mg. Posterior probability of clinical equivalence across dose pairs (equivalence margin ± 5 percentage points). Posterior probabilities of clinical equivalence across prespecified semaglutide and tirzepatide dose pairs using an equivalence margin of ± 5 percentage points for percent weight change. Values within each cell represent the posterior probability of equivalence, the estimated mean difference in percent weight change (Δ = semaglutide − tirzepatide), and the corresponding 95% credible interval. Positive values indicate greater estimated weight reduction with semaglutide, whereas negative values indicate greater estimated weight reduction with tirzepatide. Color intensity reflects the probability of equivalence Posterior probability of clinical equivalence across dose pairs (equivalence margin ± 4 percentage points). Posterior probabilities of clinical equivalence across prespecified semaglutide and tirzepatide dose pairs using an equivalence margin of ± 8 percentage points for percent weight change. Values within each cell represent the posterior probability of equivalence, the estimated mean difference in percent weight change (Δ = semaglutide − tirzepatide), and the corresponding 95% credible interval. Positive values indicate greater estimated weight reduction with semaglutide, whereas negative values indicate greater estimated weight reduction with tirzepatide. Color intensity reflects the probability of equivalence Sensitivity Analysis Excluding the Semaglutide 7.2 mg Trial In sensitivity analyses excluding the semaglutide 7.2 mg arm, the overall dose–response relationship remained consistent across the observed dose range. Estimated weight-loss effects for semaglutide 1.0 mg and 2.4 mg were materially unchanged, and comparisons involving tirzepatide were unaffected. As expected, estimates for semaglutide 7.2 mg were not generated in this analysis. The corresponding sensitivity analysis is shown in Supplementary Fig. S1. Model-Based Intensification Scenarios Model-based intensification scenarios are presented in Table 5, summarizing posterior estimates of incremental percent weight loss associated with switching or up-titration strategies relative to a current regimen. Table 5Model-based intensification scenarios (incremental benefit vs current regimen)Current regimenNew regimenΔ (p.p.)95% CrI (p.p.)P (new better)P (≥ 2 p.p. additional loss)Semaglutide 1.0 mgSemaglutide 2.4 mg− 3.5− 4.4 to − 2.6100.0%99.8%Semaglutide 1.0 mgTirzepatide 5 mg− 2.0− 2.8 to − 1.1100.0%47.3%Tirzepatide 10 mgTirzepatide 15 mg− 2.0− 2.6 to − 1.5100.0%54.4%Tirzepatide 10 mgSemaglutide 7.2 mg− 2.6− 4.3 to − 0.999.9%76.3%Tirzepatide 5 mgTirzepatide 10 mg− 1.9− 2.4 to − 1.3100.0%35.3%Tirzepatide 5 mgSemaglutide 2.4 mg− 1.5− 2.8 to − 0.399.2%22.8%Semaglutide 2.4 mgSemaglutide 7.2 mg− 3.0− 4.3 to − 1.7100.0%93.4%Semaglutide 2.4 mgTirzepatide 15 mg− 2.4− 3.6 to − 1.2100.0%74.1%Δ is defined as (new regimen − current regimen) in percentage points (p.p.) of weight loss; negative values indicate greater weight loss with the new regimenProbabilities are derived from posterior draws of the Bayesian hierarchical spline model and represent population-level predictions at a representative treatment duration Model-based intensification scenarios (incremental benefit vs current regimen) Δ is defined as (new regimen − current regimen) in percentage points (p.p.) of weight loss; negative values indicate greater weight loss with the new regimen Probabilities are derived from posterior draws of the Bayesian hierarchical spline model and represent population-level predictions at a representative treatment duration Across scenarios, both switching and dose escalation strategies were associated with higher expected weight loss compared with baseline regimens, with probabilities of benefit (P[new better]) approaching 100% in most comparisons. The magnitude of incremental benefit and the probability of achieving ≥ 2 additional percentage points of weight loss varied across treatment strategies and baseline regimens. Detailed estimates, including posterior mean differences and 95% credible intervals, are reported in Table 5. Discussion In this model-based analysis of phase III randomized trials in individuals with T2DM, we observed that semaglutide and tirzepatide exhibit broadly comparable pharmacodynamic profiles for weight reduction when evaluated across clinically relevant dose ranges. Dose–response relationships for both agents were nonlinear, with greater incremental effects at lower doses and attenuation at higher doses, although the shape and extent of these relationships differed between drugs. Across the evaluated dose pairs, high posterior probabilities of clinical equivalence were observed for semaglutide 2.4 mg versus tirzepatide 10 mg and semaglutide 7.2 mg versus tirzepatide 15 mg, suggesting that similar weight-loss outcomes can be achieved with these dose combinations. In contrast, lower doses of semaglutide were not equivalent to higher doses of tirzepatide, reflecting clinically meaningful differences in expected efficacy. Findings involving semaglutide 7.2 mg should be interpreted cautiously, as this dose was informed by a single trial. Although inclusion of this dose allows exploration of the upper range of semaglutide exposure, corresponding estimates—particularly comparisons with higher doses of tirzepatide—should be considered exploratory until additional data become available. From a clinical perspective, these findings highlight that both agents follow predictable dose–response patterns, but with important practical implications. Tirzepatide demonstrated larger initial incremental effects at lower doses, whereas semaglutide showed continued incremental benefit across an extended dose range. This distinction may be relevant when considering treatment initiation versus later intensification. Tirzepatide may provide more rapid early weight reduction at lower doses, whereas semaglutide may allow continued titration with incremental benefit at higher doses, particularly in patients who tolerate therapy well and require additional weight loss. Importantly, the observed attenuation of incremental benefit at higher doses for both agents suggests diminishing returns with dose escalation, reinforcing the need to balance expected benefit against treatment burden and patient-specific factors. One of the key clinical questions addressed by this analysis is whether doses of semaglutide and tirzepatide can be considered interchangeable in practice. Our findings suggest that equivalence is dose-dependent rather than drug-specific. For example, semaglutide 2.4 mg appears pharmacodynamically comparable to tirzepatide 10 mg, while semaglutide 7.2 mg approximates tirzepatide 15 mg. These relationships may inform clinical decision-making in situations where treatment switching is required as a result of availability, cost, or tolerability considerations. However, equivalence should not be interpreted as therapeutic interchangeability in all contexts. Differences in tolerability profiles, patient preference, access, and prior treatment response remain critical determinants of treatment selection. The scenario-based analyses provide a more direct translation of model outputs into clinical decision-making. In patients receiving semaglutide 1.0 mg, both escalation to semaglutide 2.4 mg and switching to tirzepatide 5 mg were associated with high probability of improved weight loss. However, escalation within semaglutide was associated with a substantially higher probability of achieving a clinically meaningful incremental benefit (≥ 2 percentage points), suggesting that dose optimization within the same agent may be preferable when tolerated. Conversely, in patients receiving tirzepatide 10 mg, both escalation to 15 mg and switching to semaglutide 7.2 mg were associated with further weight reduction. Notably, switching to semaglutide 7.2 mg was associated with a larger expected incremental benefit and a higher probability of achieving clinically meaningful additional weight loss. This suggests that, at higher dose ranges, switching between agents may provide incremental benefit beyond within-drug dose escalation in selected patients. These findings illustrate that optimal intensification strategies are context-dependent and should consider both the magnitude and probability of clinically meaningful benefit. These results extend existing evidence from randomized trials and network meta-analyses by providing a unified, dose-continuous framework for comparing semaglutide and tirzepatide in individuals with T2DM [7]. Prior studies have demonstrated superior weight-loss efficacy with higher doses of both agents, but direct comparisons have been limited and often restricted to specific dose combinations or populations without diabetes. By integrating data across trials and modeling continuous dose–response relationships, the present analysis provides a more granular understanding of how these agents compare across the full range of clinically relevant doses. This approach complements, rather than replaces, head-to-head trial data, and is particularly useful in contexts where direct evidence is lacking. This study applies a model-based framework combining generalized additive models and Bayesian hierarchical modeling to estimate dose–response relationships and quantify clinical equivalence [34]. A key strength of this approach is the ability to estimate probabilistic equivalence across multiple dose combinations, rather than relying on pairwise comparisons. The use of posterior probabilities allows direct quantification of uncertainty and facilitates clinically interpretable outputs, such as the probability of achieving a predefined threshold of additional weight loss. However, equivalence estimates depend on the prespecified margin (± 2 percentage points) [35, 36], which, while clinically plausible, remains a subjective choice. Sensitivity analyses using wider margins demonstrated consistent relative patterns, supporting the robustness of the findings. Several limitations should be considered. First, this analysis was based on aggregated arm-level data rather than individual participant data, which limited the ability to adjust for patient-level heterogeneity, including comorbidities and concomitant therapies. As a result, differences in background glucose-lowering treatments, weight-modifying therapies, and clinical characteristics across trials may have influenced observed weight-loss responses. Second, variability in study design, patient populations, and clinical management across trials represents an additional source of heterogeneity. Although a study-level random intercept was included and treatment duration was incorporated as a covariate, residual confounding cannot be excluded. Third, the assumption of comparability across trials may not fully account for differences in adherence, treatment intensification strategies, or background care, which could affect the magnitude of treatment effects. Fourth, the use of pseudo-observations at zero dose, while methodologically justified to stabilize model behavior at the lower boundary, represents a modeling assumption that may influence extrapolation outside the observed dose range. Fifth, although adverse events and discontinuation rates were extracted, these outcomes were not formally analyzed because of substantial heterogeneity in reporting across trials, including differences in definitions, timepoints, and outcome structure (event-based vs patient-based). In addition, the use of aggregated data precluded adjustment for exposure time. Therefore, these data are presented descriptively in the Supplementary Material (Table S2). Finally, this analysis focused exclusively on weight loss and did not incorporate glycemic outcomes, cardiovascular endpoints, safety, or tolerability. While weight reduction is a key therapeutic target in T2DM, these additional outcomes are critical for a comprehensive assessment of treatment benefit. Therefore, the present findings should be interpreted within the context of a weight-focused modeling framework and integrated with broader clinical evidence when informing treatment decisions. Conclusions In individuals with T2DM, semaglutide and tirzepatide demonstrate comparable weight-loss efficacy at specific dose combinations, with equivalence driven by dose rather than drug identity. Model-based analyses suggest that both within-drug dose escalation and between-drug switching can yield clinically meaningful incremental benefits, depending on the treatment context. These findings provide a quantitative framework to support individualized treatment decisions [37, 38], particularly in settings where direct comparative evidence is unavailable. However, as this analysis is based on aggregated data and focused on weight outcomes, the results should be interpreted as complementary to clinical judgment and broader outcome considerations. Future studies incorporating individual-level data and additional clinical endpoints are warranted to further refine these estimates. Supplementary Information Below is the link to the electronic supplementary material. Supplementary file1 (PDF 287 KB) Acknowledgments Medical Writing/Editorial Assistance Festina Lente S.A.S. provided methodological support, medical writing, and editorial assistance for manuscript preparation. No specific external funding was received for this assistance beyond the funding disclosed for the underlying research project. Artificial intelligence tools (OpenAI, Claude, and Grammarly) were used to assist with language editing and refinement of analytical scripts. All outputs were critically reviewed, validated, and edited by the authors, who take full responsibility for the content of the manuscript. No specific funding was received for the use of these artificial intelligence tools. Author Contributions Carlos E. Builes-Montaño contributed to the study conception and design, statistical analysis, interpretation of results, drafting of the manuscript, and methodological supervision. Andres F. Suarez-Rodriguez contributed to data interpretation, critical revision of the manuscript, and clinical contextualization of the findings. Maria A. Alzate-Vinasco contributed to interpretation of results, critical manuscript revision, and scientific discussion of the clinical implications. All authors reviewed and approved the final version of the manuscript. Funding This work was derived from a research project funded by Novo Nordisk. No additional external funding was received for the present analysis. The sponsor had no role in the design, analysis, interpretation, or writing of this manuscript. The journal’s Rapid Service Fee was funded by Novo Nordisk. Data Availability The datasets analyzed during the current study are derived from published clinical trials and are available in the public domain. Declarations Conflict of Interest The authors declare the following conflicts of interest: Carlos E. Builes-Montaño has received consulting and speaking fees from Sanofi, Novo Nordisk, Novartis, Recordati Rare Diseases, Janssen, Chiesi, Abbott, and Boehringer Ingelheim, and is a shareholder of Festina Lente. Andres F. Suarez-Rodriguez and Maria A. Alzate-Vinasco are employees of Novo Nordisk, the sponsor of the research project from which this work was derived. These disclosures are not necessarily related to the submitted work. Ethical Approval This study was based exclusively on aggregated, arm-level data extracted from previously published clinical trials. No individual participant data were used. Therefore, institutional review board approval and informed consent were not required for the present secondary analysis. According to the information reported in the original trial publications, the included studies received approval from the corresponding ethics committees or institutional review boards, were conducted in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines, and obtained written informed consent from all participants. References Associated Data Supplementary Materials Supplementary file1 (PDF 287 KB) Data Availability Statement The datasets analyzed during the current study are derived from published clinical trials and are available in the public domain.