A cost‐effectiveness analysis of behavioural, pharmacological, and surgical obesity treatments in Canada

Abstract

Aims

Effective weight management pharmacotherapies are a new alternative to bariatric surgery or health behaviour intervention (HBI) alone. Comparative cost‐effectiveness evaluations can guide decision‐making. We aimed to evaluate the cost‐effectiveness of sleeve gastrectomy (SG), Roux‐en‐Y gastric bypass (RYGB), semaglutide, tirzepatide, and HBI compared to no treatment in preventing cardiometabolic complications among Canadian adults with class III obesity.

Materials and Methods

We developed a cohort‐based Markov model with a 40‐year time horizon. We examined cardiometabolic complications in 40‐year‐olds with class III obesity (BMI ≥40 kg/m²) without type 2 diabetes or cardiovascular disease at baseline. We compared SG, RYGB, semaglutide 2.4 mg, tirzepatide 15 mg, HBI, and no treatment. We obtained data on treatment effects, probabilities, utilities, and costs from published literature. We expressed effectiveness in quality‐adjusted life years (QALYs) and estimated costs from a Canadian public payer perspective. Outcomes included incremental cost‐effectiveness ratios (ICERs) evaluated at a CAD $50 000/QALY willingness‐to‐pay threshold.

Results

RYGB and HBI were cost‐effective strategies. HBI was cost‐effective versus no treatment (ICER $14 279/QALY). RYGB demonstrated the highest QALYs (20.20) and was the most cost‐effective strategy versus tirzepatide (ICER $44 667/QALY). Semaglutide and SG were strongly dominated due to higher costs and lower effectiveness. Tirzepatide was extendedly dominated by RYGB. Sensitivity analyses confirmed these findings and showed that lower drug prices could improve pharmacotherapy cost‐effectiveness.

Conclusions

RYGB and HBI are cost‐effective for managing class III obesity. While RYGB provided the greatest health gains, access remains limited. Neither pharmacotherapy was cost‐effective at current prices. Lower drug prices could significantly improve pharmacotherapy cost‐effectiveness.

1. INTRODUCTION

Type 2 diabetes (T2D) and cardiovascular disease (CVD) are costly consequences of obesity, with the annual economic impact approaching USD 70 billion in Canada by 2035. ¹ , ² , ³ As a growing range of pharmacotherapies for obesity management continues to emerge and bariatric surgery remains a specialised and limited‐access option, economic analyses are essential for guiding evidence‐based decision making in health systems.

While prevention remains a critical public health mandate, healthcare systems must deliver cost‐effective treatment options. Evidence‐based approaches include dietary and physical activity counselling (health behaviour interventions, HBI) alone or in combination with bariatric surgery or pharmacotherapy. ⁴ , ⁵ Over one‐quarter of Canadian adults lived with obesity (BMI ≥30 kg/m²) in 2016, and half are projected to be living with obesity by 2035. ² , ⁴ , ⁶ The two most common bariatric surgeries in Canada are sleeve gastrectomy (SG) and Roux‐en‐Y gastric bypass (RYGB), associated with 25%–30% net weight loss compared to HBI alone, leading to reductions in T2D and CVD rates. ⁴ , ⁷ More recently, pharmacologic options such as glucagon‐like peptide‐1 receptor agonists (GLP‐1 RA) and dual glucose‐dependent insulinotropic polypeptide (GIP)/GLP‐1 RA have emerged as promising, non‐invasive alternatives. In the STEP trials, semaglutide (GLP‐1 RA) leads to 14.9%–15.2% weight loss among individuals with obesity without T2D. ⁴ , ⁵ , ⁸ , ⁹ , ¹⁰ Weight loss with tirzepatide (GIP/GLP‐1 RA) is 20.9% in a similar population. ¹¹ In these trials, the pharmacotherapy agents were consistently paired with 13 counselling sessions of HBI, which is also effective as a standalone intervention for diabetes prevention and has been successfully implemented in real‐world settings, such as the National Health Service England's Diabetes Prevention Programme. ¹² , ¹³ , ¹⁴

Public and private insurance payers must now decide which obesity treatments to fund, and clinicians must choose which interventions to advocate. ¹⁵ We compared no treatment, the current reality for many Canadians with obesity; optimal HBI (13 counselling sessions), not universally available in Canada but with precedent for implementation, as demonstrated in England; SG and RYGB, the two most common bariatric procedures in Canada though access remains limited; and semaglutide and tirzepatide, agents with limited access for obesity treatment. ¹⁶ , ¹⁷ We included tirzepatide in this analysis to inform imminent reimbursement decisions both nationally and internationally. We evaluated the cost‐effectiveness of these interventions for individuals with class III obesity without baseline T2D or CVD from a Canadian public healthcare payer perspective.

2. METHODS

We adhered to the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) and Canada's Drug Agency guidelines. ¹⁸ , ¹⁹ We did not require Research Ethics Board review as we incorporated publicly available data in our Markov model.

2.1. Model structure

We developed a cohort‐state transition Markov model to examine the progression of cardiometabolic complications in a cohort of 40‐year‐olds with class III obesity (BMI ≥40 kg/m²) without baseline T2D or CVD (Figure 1). We compared six treatment strategies: SG, RYGB, semaglutide 2.4 mg, tirzepatide 15 mg, HBI alone, and no treatment. Based on the Diabetes Prevention Program and the STEP 1 and SURMOUNT 1 trials, our HBI incorporated 13 dietician‐led counselling sessions over 1 year. ⁹ , ¹¹ , ²⁰ The model included five health states to reflect key obesity‐related complications: class III obesity (1) without comorbidities, (2) with T2D, (3) with CVD, (4) with CVD and T2D, and (5) death. CVD was a composite of myocardial infarction, unstable angina, stroke, and transient ischemic attack, weighted by Canadian prevalence data. ²¹ Death was an absorbing state and could follow any health state (Table S1 in the Supporting Information). The cohort started in class III obesity without comorbidities, to evaluate the preventive impact of each intervention on the development of T2D and CVD in an at‐risk population. Beginning at the end of Year 1 and at 1‐year intervals thereafter, people could remain in this state or transition to one of the four other health states. Remission of T2D was modelled only for individuals in the T2D health state (State B). We did not allow remission from the combined CVD and T2D health state (State D) as available evidence does not provide sufficient data on remission likelihood among individuals with established CVD. ²² , ²³ We derived annual probabilities of developing T2D or experiencing a CVD event using the QDiabetes and QRISK3 10‐year risk equations, applying a discrete‐time approximation that assumes a constant hazard over the 10‐year period ($p=1-{\left(1-R\right)}^{1/10}$, where R represents the 10‐year cumulative risk). ²⁴ , ²⁵ , ²⁶ , ²⁷ While we used a discrete approximation for simplicity, the result is mathematically equivalent to an exponential conversion under the assumption of a constant hazard. Nonetheless, we acknowledge that the choice may influence transition probabilities particularly at higher risk levels. We estimated all‐cause mortality rates from Statistics Canada, adjusted by health‐state specific relative risks. ²⁸ , ²⁹ , ³⁰ , ³¹ Given obesity's chronic nature, we used a 40‐year horizon, consistent with Canadian life expectancy at 40 years of age. ³²

FIGURE 1.

Markov model diagram. Simulated patients entered the model through their surgical, pharmacotherapy, HBI, or no treatment arm in the initial state of class III obesity (BMI ≥40 kg/m²) without T2D or CVD. From the first cycle onwards, patients either remained in the initial state or transitioned annually to one of four health states: T2D, CVD without T2D, CVD with T2D, or death. Remission from T2D was only permitted for individuals in the T2D only health state (State B). Simulated patients with both CVD and T2D were not eligible for remission. Death was an absorbing state and could occur from any health state. CVD, cardiovascular disease; HBI, health behaviour intervention; T2D, type 2 diabetes.

2.2. Study population

Our cohort excluded individuals with baseline T2D or CVD to reflect a high‐risk but comorbidity‐free subset of the class III obesity population. In a pooled analysis of over 9500 adults with class III obesity, only 15%–22% had diabetes at baseline, indicating that a substantial proportion remain free of major cardiometabolic conditions despite being at elevated risk for future complications. ³³ This supports the plausibility of modelling a prevention‐focused cohort. We derived the cohort's baseline characteristics from publicly available Canadian health surveys. ³⁴ , ³⁵ , ³⁶ We categorized the population into four groups (male smokers, non‐smokers; female smokers, non‐smokers) with a baseline BMI of 40 kg/m² for the base‐case. We attributed mean Canadian height values (175.10 cm males, 162.30 cm females), calculated corresponding weights for the baseline BMI (122.64 kg males, 105.37 kg females), applied an equal sex distribution, and incorporated smoking proportions (14% males, 10% females) from Statistics Canada's Canadian Community Health Survey. ³⁴ , ³⁵ , ³⁶

We used the same approach across treatment arms, setting 1‐year treated BMIs at 38.40 kg/m² for HBI, 34.04 kg/m² for semaglutide, 31.64 kg/m² for tirzepatide, 28.20 kg/m² for SG, and 27.20 kg/m² for RYGB (Table S2). We then calculated corresponding weights for each BMI using the set height values.

2.3. Model inputs

We obtained transition probabilities, costs, and utilities from published literature, government databases, and pharmaceutical manufacturers (Table S3). We applied weight loss estimates from the STEP 1 (semaglutide) and SURMOUNT 1 (tirzepatide) clinical trials, which were conducted in persons with obesity without T2D or CVD. ⁹ , ¹¹ We modelled pharmacotherapy adherence using real‐world discontinuation rates from a 40 000‐person cohort study: 55.9% for tirzepatide and 52.5% for semaglutide. ³⁷ We sourced surgical weight loss estimates from the literature. ³⁸ , ³⁹ Surgical transition probabilities incorporated weight reduction, T2D remission, complications, reoperation, and mortality rates. ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ From Year 2 onward, we modelled ≥10% weight regain from nadir, adjusting transition probabilities and utilities accordingly. ⁴⁴

We used quality‐adjusted life years (QALYs) as the unit of effectiveness. We derived utility scores from published literature, applying an additive approach where the disutilities of obesity, T2D, and CVD were subtracted from the baseline Canadian population utility values. ⁴⁵ , ⁴⁶ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ We calculated treatment‐related utility gains by applying the percent weight loss to utility gains per one‐unit BMI reduction. While we did not model most pharmacotherapy‐related adverse event rates, as they were typically mild or moderate, we incorporated injection‐related disutilities in Year 1. ⁹ , ¹¹ , ⁵¹ We capped utilities at 1.0 to avoid overestimation.

We based annual costs for tirzepatide (15 mg) and semaglutide (2.4 mg) on the average group sales estimates from Eli Lilly (April 2024) and Canada's Drug Agency, respectively. ⁵² , ⁵³ We sourced SG and RYGB surgical costs and related expenditures (e.g., discharge, follow‐up, reoperations, and complications) from an Ontario government health technology assessment and other literature. ⁴⁵ , ⁵⁴ We calculated HBI costs using Statistics Canada's earnings data for dietitians, adjusted for a 40‐hour workweek and fringe benefits (Table S4). ⁵⁵ We obtained direct medical costs for managing obesity, T2D, and CVD from literature. ⁵⁶ , ⁵⁷ , ⁵⁸ We converted all costs to 2024 Canadian dollars (CAD) using the Bank of Canada's Inflation Calculator. ⁵⁹ We adopted a Canadian public healthcare payer perspective and discounted future costs and utilities at 3% annually (1.5% and 0% in sensitivity analyses per Canada's Drug Agency guidelines). ¹⁹

2.4. Model assumptions

We assumed equal eligibility for all six treatments and did not allow treatment crossover. All strategies incorporated HBI, in accordance with guidelines. ⁴ We assumed that weight loss occurred in the year following treatment and was maintained with continued pharmacotherapy. This assumption is supported by clinical trial data showing that weight loss with semaglutide (STEP 5) and tirzepatide (SURMOUNT 4) is largely sustained or improved over 1–2 years of continuous use. ¹⁰ , ⁶⁰ Discontinuation rates were applied only in Year 1, based on the follow‐up period of the real‐world observational study informing these rates. ³⁷ This reflects the typical timing of adverse effects such as nausea and vomiting that typically occur early in treatment, and is supported by trial data showing high long‐term adherence: in STEP 5, 86.8% of participants remained on semaglutide at 104 weeks, and in SURMOUNT 4, 14.4% discontinued tirzepatide during the first 36 weeks, but only 1.8% discontinued between weeks 36 and 88. ¹⁰ , ⁶⁰ In the absence of long‐term adherence data, individuals who remained on treatment after Year 1 were assumed to continue therapy for the remainder of the model. For those who discontinued pharmacotherapy, we assigned transition probabilities, costs, and utilities equivalent to the HBI group, which reflects a partial regain of weight and treatment effect. ³⁷ , ⁵⁵ , ⁶¹ We applied injection‐related disutilities in Year 1 only, assuming individuals adapted to injections over time. For surgical arms, we modelled short‐ and long‐term weight reduction. ³⁸ , ³⁹ We included a 5‐year surgical wait time based on Canadian data, during which patients followed HBI‐related transition probabilities, costs, and utilities. ⁶² , ⁶³

2.5. Model validation

We cross‐checked model‐predicted cumulative mortality rates and 10‐year cumulative incidence of T2D and CVD in the base‐case (no treatment group) with real‐world estimates from Statistics Canada, QRISK3, and QDiabetes (Tables S5 and S6). ²⁴ , ²⁵ , ²⁶ , ²⁷ , ³¹ Given the lack of long‐term data on the cumulative incidence of T2D and CVD for treated populations (e.g., individuals receiving pharmacotherapy or bariatric surgery), model validation was limited to the no treatment group.

2.6. Statistical analysis

Our primary outcome was the incremental cost‐effectiveness ratio (ICER), expressed in CAD per QALY gained. We conducted a fully incremental analysis, ranking treatments from lowest to highest total QALYs and listing total costs. We then excluded strongly dominated interventions (e.g., more costly and less effective than the next highest‐ranked treatment). Among the remaining treatments, we calculated incremental costs and effects sequentially and divided incremental costs by incremental effects to generate ICERs. We compared these ICERs against a CAD 50 000 per QALY willingness‐to‐pay (WTP) threshold, a commonly used benchmark in Canada. Strategies below this threshold qualified as cost‐effective. ¹⁹ We also calculated the incremental net monetary benefit (INMB) by multiplying incremental QALYs by the CAD 50 000 per QALY WTP threshold and subtracting the corresponding incremental costs.

2.7. Sensitivity analysis

We performed deterministic (one‐way) and probabilistic sensitivity analyses (PSA). In the deterministic sensitivity analysis, for parameters with a reported 95% CI, we used the upper and lower CI limits. Otherwise, we varied costs ±50%, utilities ±25%, and transition probabilities ±25%. We created tornado diagrams to identify influential variables. For the PSA, we assessed the level of parameter uncertainty and variability in data inputs using Monte Carlo simulation with 1000 iterations. We applied gamma distributions to cost parameters (shape (k) = 4; scale (θ) = base‐case value divided by 4), and beta distributions to probabilities and utilities (α₁ and α₂ parameters = 4), then applied a maximum and minimum value equal to ±50% of the base‐case value. We displayed results using the incremental cost‐effectiveness plane and cost‐effectiveness acceptability curves, which illustrate the probability that an intervention provides the greatest net benefit across WTP thresholds.

Finally, through scenario analyses, we explored the impact of several alternative assumptions: eliminating surgical wait times, applying a 20‐year time horizon, modelling an older starting cohort (age 45) with a 10‐year horizon, assuming weight regain to baseline among individuals who discontinue pharmacotherapy (e.g., no sustained treatment effect after discontinuation), incorporating post‐patent expiry generic pricing for pharmacotherapy, and modelling the cardiometabolic benefits of pharmacotherapy. ⁶⁴ For semaglutide and tirzepatide, we assumed generic entry at Year 2 and Year 12 respectively, based on the patent status in Canada. ⁶⁵ , ⁶⁶ We applied a Canadian tiered pricing structure consistent with pan‐Canadian Pharmaceutical Alliance for new generic products entering the market post‐patent expiry (e.g., 75% of brand pricing at listing, declining to 55% after 3 months, 50% with two generics at Year 2, and 35% with three or more generics in Year 3 onwards). ⁶⁷ To model cardiometabolic benefits, we incorporated the 20% relative risk reduction in major adverse cardiovascular events for semaglutide based on the SELECT trial, and applied reduced T2D incidence for both semaglutide and tirzepatide based on the SELECT and SURMOUNT 1 trials, respectively. ⁶⁸ , ⁶⁹ , ⁷⁰ Full details and hazard ratios are available in Table S3. We performed all analyses using Excel, version 16.94.

3. RESULTS

3.1. Base‐case analysis

Among all interventions, not treating incurred the lowest total discounted costs (CAD 60 251), followed by HBI (CAD 68 643), tirzepatide (CAD 118 206), semaglutide (CAD 120 993), RYGB (CAD 158 352), and SG (CAD 204 948) (Table 1). RYGB yielded the highest total discounted QALYs (20.20), followed by SG (19.71), tirzepatide (19.28), semaglutide (18.94), HBI (18.20), and no treatment (17.61).

TABLE 1.

Base‐case cost‐effectiveness results ^a .

Strategy	Costs, $ b	Incremental costs, $	Effectiveness, QALYs	Incremental effectiveness, QALYs	ICER, $/QALY	INMB, $
3% discount rate c
Comparison across the next most effective strategy d
No treatment	60 251	NA	17.61	NA	NA	NA
Health behaviour intervention	68 643	8392	18.20	0.588	14 279	20 995
Semaglutide	120 993	52 350	18.94	0.744	70 322 (dominated)	−15 128
Tirzepatide	118 206	−2787	19.28	0.340	−8198	19 783
Sleeve gastrectomy	204 948	86 742	19.71	0.426	203 448 (dominated)	−65 424
Roux‐en‐Y gastric bypass	158 352	−46 596	20.20	0.498	−93 621	71 481
Comparison across the next most effective strategy excluding strongly dominated strategies e
No treatment	60 251	NA	17.61	NA	NA	NA
Health behaviour intervention	68 643	8392	18.20	0.588	14 279	20 995
Tirzepatide	118 206	49 563	19.28	1.084	45 708	4665
Roux‐en‐Y gastric bypass	158 352	40 146	20.20	0.924	43 445	6057
Comparison across the next most effective strategy excluding extendedly dominated strategies f
No treatment	60 251	NA	17.61	NA	NA	NA
Health behaviour intervention	68 643	8392	18.20	0.588	14 279	20 995
Roux‐en‐Y gastric bypass	158 352	89 709	20.20	2.008	44 667	10 712
1.5% discount rate c
Comparison across the next most effective strategy d
No treatment	81 861	NA	22.05	NA	NA	NA
Health behaviour intervention	92 176	10 315	22.80	0.752	13 719	27 280
Semaglutide	157 384	65 208	23.77	0.961	67 859 (dominated)	−17 161
Tirzepatide	153 590	−3794	24.20	0.436	−8709	25 579
Sleeve gastrectomy	267 722	114 132	24.72	0.522	218 446 (dominated)	−88 008
Roux‐en‐Y gastric bypass	206 009	−61 713	25.39	0.671	−91 953	95 270
Comparison across the next most effective strategy excluding strongly dominated strategies e
No treatment	81 861	NA	22.05	NA	NA	NA
Health behaviour intervention	92 176	10 315	22.80	0.752	13 719	27 280
Tirzepatide	153 590	61 414	24.20	1.397	43 973	8417
Roux‐en‐Y gastric bypass	206 009	52 419	25.39	1.194	43 916	7262
Comparison across the next most effective strategy excluding extendedly dominated strategies f
No treatment	81 861	NA	22.05	NA	NA	NA
Health behaviour intervention	92 176	10 315	22.80	0.752	13 719	27 280
Roux‐en‐Y gastric bypass	206 009	113 833	25.39	2.590	43 947	15 679
0% discount rate c
Comparison across the next most effective strategy d
No treatment	114 759	NA	28.48	NA	NA	NA
Health behaviour intervention	127 832	13 073	29.47	0.992	13 179	36 525
Semaglutide	211 672	83 841	30.75	1.278	65 585 (dominated)	−19 923
Tirzepatide	206 361	−5311	31.33	0.576	−9218	34 120
Sleeve gastrectomy	360 006	153 645	31.98	0.653	235 310 (dominated)	−120 997
Roux‐en‐Y gastric bypass	276 014	−83 992	32.91	0.926	−90 687	130 301
Comparison across the next most effective strategy excluding strongly dominated strategies e
No treatment	114 759	NA	28.48	NA	NA	NA
Health behaviour intervention	127 832	13 073	29.47	0.992	13 179	36 525
Tirzepatide	206 361	78 529	31.33	1.855	42 345	14 197
Roux‐en‐Y gastric bypass	276 014	69 653	32.91	1.579	44 108	9304

^a Analyses conducted over a 40‐year horizon.

^b All costs are shown in 2024 CAD.

^c Discount rate refers to the annual rate at which costs and utilities were discounted as per health economic guidelines (3%, 1.5%, and 0% rates shown).

^d Each strategy compared with the next most effective strategy.

^e Each strategy compared with the next most effective strategy (excluding any strongly dominated strategies).

^f Each strategy is compared with the next most effective strategy (excluding any extendedly dominated strategies).

Abbreviations: ICER, incremental cost‐effectiveness ratio; INMB, incremental net‐monetary benefit; NA, not applicable; QALYs, quality‐adjusted life year.

Tirzepatide strongly dominated semaglutide, with higher QALYs (19.28 vs. 18.94) at lower costs (CAD 118 206 vs. CAD 120 993), and an ICER of CAD‐8198/QALY. Similarly, RYGB strongly dominated SG, with higher QALYs (20.20 vs. 19.71) and lower costs (CAD 158 352 vs. CAD 204 948). We therefore excluded semaglutide and SG from further analysis.

Among the remaining treatments, RYGB emerged as the most cost‐effective strategy overall, delivering the highest QALY gains with an ICER of CAD 44 667 per QALY compared with HBI. While HBI was less effective, its substantially lower cost resulted in the highest INMB compared to no treatment. Moving from HBI to RYGB produced stepwise QALY gains but at increasing costs, resulting in a lower INMB for RYGB. Tirzepatide was extendedly dominated by RYGB at 3% and 1.5% discount rates, as it had higher ICERs than the most effective option. At a 0% discount rate, tirzepatide was no longer extendedly dominated and remained a potentially cost‐effective alternative.

3.2. Sensitivity analyses

Tornado diagrams illustrated the impact of the 10 most influential variables for ICERs (Figure S1). For tirzepatide versus HBI, tirzepatide's annual cost (CAD 2703 to 8108) was the primary driver; at the lowest cost, the ICER fell below CAD 20 433 per QALY, while at the highest, it rose to CAD 70 982 per QALY (Figure S1A). For RYGB, post‐op healthcare costs (CAD 2955 to 8867) were a key driver in comparisons with both HBI (Figure S1B) and tirzepatide (Figure S1C).

A PSA scatter plot demonstrated that 25.7% of simulations for RYGB versus tirzepatide were cost‐saving, with RYGB being more effective and less costly (Figure 2). This represents a dominant strategy in over one‐quarter of modelled scenarios, indicating a meaningful likelihood of economic dominance under real‐world variability. These simulations fell into the southeast quadrant of the cost‐effectiveness plane, reflecting the ideal scenario for decision makers: greater health benefit at lower cost. At a WTP threshold of CAD 50 000 per QALY, RYGB had a 60% probability of being cost‐effective compared to tirzepatide (Figure 3). For RYGB versus HBI, cost‐savings were rare; most simulations were in the northeast quadrant, indicating higher effectiveness but also higher costs, where value depends more heavily on the WTP threshold (Figure 2). In this comparison, 67% of the simulations indicated RYGB was cost‐effective at a WTP threshold of CAD 50 000 per QALY (Figure 3).

FIGURE 2.

Probabilistic sensitivity analysis depicted as a cost‐effectiveness scatter plot. Each dot represents one simulation, with incremental costs (∆$, CAD) on the y‐axis and incremental QALYs (∆QALYs) on the x‐axis. Blue dots: Roux‐en‐Y Gastric Bypass versus Tirzepatide. Purple dots: Roux‐en‐Y Gastric Bypass versus Health Behaviour Intervention. QALYs, quality adjusted life years.

FIGURE 3.

Probabilistic sensitivity analysis depicted as cost‐effectiveness acceptability curves (CEACs) depicting the proportion of simulations that are cost‐effective across different willingness‐to‐pay thresholds. The CEACs compare Roux‐en‐Y Gastric Bypass to Tirzepatide and Health Behaviour Intervention. Each dot represents one simulation, with proportion cost‐effective (%) on the y‐axis and incremental willingness to pay ($ CAD/QALY) on the x‐axis. Blue line: Roux‐en‐Y Gastric Bypass versus Tirzepatide. Purple line: Roux‐en‐Y Gastric Bypass versus Health Behaviour Intervention. QALYs, quality adjusted life years.

Eliminating surgical wait times slightly improved the cost‐effectiveness of surgical treatments but did not alter intervention rankings (Table 2). Applying a tiered price reduction framework for semaglutide and tirzepatide following patent expiry, pharmacotherapies became more favourable with ICERs of CAD 30 721 per QALY (for tirzepatide vs. semaglutide) and CAD 27 915 per QALY (for semaglutide vs. HBI). In this scenario, both surgical interventions were dominated. At a 47% price reduction, semaglutide became the most cost‐effective strategy, with an annual cost of CAD 2728.67, total costs of CAD 94 655, 18.94 total QALYs and an ICER of CAD 34 956 per QALY. Similarly, tirzepatide became the most cost‐effective strategy at a 12% price reduction, with an annual cost of CAD 4756.75, total costs of CAD 111 629, 19.28 total QALYs, and an ICER of CAD 39 642 per QALY. Over a 20‐year horizon, RYGB and HBI remained cost‐effective while tirzepatide remained extendedly dominated (ICER CAD −6344 per QALY). When we applied trial‐based weight loss values for tirzepatide (relative to placebo), RYGB and HBI continued to be cost‐effective, and tirzepatide was again extendedly dominated with an ICER increasing to CAD 56 772 per QALY versus HBI (Table S7). If individuals who discontinued pharmacotherapy regained weight to baseline within 2 years (rather than transitioning to HBI parameters), it resulted in slightly higher ICERs for semaglutide (ICER CAD 70 322 to CAD 72 952 per QALY) and tirzepatide (ICER CAD −8198 to CAD −7989 per QALY). Nonetheless, overall conclusions remained unchanged, and RYGB continued to be the most cost‐effective strategy. Incorporating the cardiovascular benefits of semaglutide modestly improved its ICER from CAD 70 322 to 67 773 per QALY, but it remained strongly dominated by tirzepatide, and RYGB remained the most cost‐effective strategy. When both semaglutide and tirzepatide arms included reductions in T2D incidence and cardiovascular risk, tirzepatide became the most cost‐effective intervention (Table 2).

TABLE 2.

Scenario analyses.

Strategy	Costs, $ a	Incremental costs, $	Effectiveness, QALYs	Incremental effectiveness, QALYs	ICER, $/QALY	INMB, $
No surgical wait times
No treatment	60 250.91	NA	17.61	NA	NA	NA
Health behaviour intervention	68 643.31	8392	18.20	0.588	14 279	20 995
Semaglutide	120 992.93	52 350	18.94	0.744	70 322 (dominated) b	−15 128
Tirzepatide	118 206.35	−2787	19.28	0.340	−8198 (dominated) c	19 783
Sleeve gastrectomy	204 399.77	86 193	19.90	0.617	139 745 (dominated) b	−55 354
Roux‐en‐Y gastric bypass	157 521.39	−46 878	20.46	0.565	−82 964	75 131
Impact of semaglutide patent expiry d
No treatment	60 250.91	NA	17.61	NA	NA	NA
Health behaviour intervention	68 643.31	8392	18.20	0.588	14 279	20 995
Semaglutide	89 424.16	20 781	18.94	0.744	27 915	16 441
Tirzepatide	118 206.35	28 782	19.28	0.340	84 673 (dominated) b	−11 786
Sleeve gastrectomy	204 947.99	86 742	19.71	0.426	203 448 (dominated) b	−65 424
Roux‐en‐Y gastric bypass	158 352.39	−46 596	20.20	0.498	−93 621 (dominated) c	71 481
Impact of tirzepatide patent expiry e
No treatment	60 250.91	NA	17.61	NA	NA	NA
Health behaviour intervention	68 643.31	8392	18.20	0.588	14 279	20 995
Semaglutide	120 992.93	52 350	18.94	0.744	70 322 (dominated) b	−15 128
Tirzepatide	99 867.03	−21 126	19.28	0.340	−62 149	38 122
Sleeve gastrectomy	204 947.99	105 081	19.71	0.426	246 462 (dominated) b	−83 763
Roux‐en‐Y gastric bypass	158 352.39	−46 596	20.20	0.498	−93 621 (dominated) c	71 481
Impact of semaglutide & tirzepatide patent expiry f
No treatment	60 250.91	NA	17.61	NA	NA	NA
Health behaviour intervention	68 643.31	8392	18.20	0.588	14 279	20 995
Semaglutide	89 424.16	20 781	18.94	0.744	27 915	16 441
Tirzepatide	99 867.03	10 443	19.28	0.340	30 721	6553
Sleeve gastrectomy	204 947.99	105 081	19.71	0.426	246 462 (dominated) b	−83 763
Roux‐en‐Y gastric bypass	158 352.39	−46 596	20.20	0.498	−93 621 (dominated) c	71 481
20‐year horizon g
No treatment	30 782.03	NA	12.51	NA	NA	NA
Health behaviour intervention	37 108.96	6327	12.90	0.394	16 053	13 380
Semaglutide	74 834.88	37 726	13.39	0.487	77 443 (dominated) b	−13 369
Tirzepatide	73400.51	−1434	13.61	0.226	−6344 (dominated) c	12 740
Sleeve gastrectomy	129 605.00	56 204	13.98	0.367	153 210 (dominated) b	−37 862
Roux‐en‐Y gastric bypass	101 401.64	−28 203	14.28	0.295	−95 607	42 953
Subgroup: 45–55 year‐olds h
No treatment	24 642.34	NA	8.59	NA	NA	NA
Health behaviour intervention	28 782.21	4140	8.89	0.295	14 012	10 633
Semaglutide	53 823.81	25 042	9.27	0.384	65 243 (dominated) b	−5851
Tirzepatide	52 556.49	−1267	9.44	0.168	−7532	9680
Sleeve gastrectomy	121 278.25	68 722	9.97	0.528	130 148 (dominated) b	−42 320
Roux‐en‐Y gastric bypass	93 998.09	−27 280	10.27	0.295	−92 477 (dominated) c	42 030
Impact of cardiometabolic risk reduction (CVD and T2D) with pharmacotherapy i
No treatment	60 250.92	NA	17.61	NA	NA	NA
Health behaviour intervention	68 643.31	8392	18.20	0.588	14 279	20 995
Semaglutide	115 049.66	46 406	19.01	0.810	57 276 (dominated) b	−5895
Tirzepatide	113 509.06	−1541	19.32	0.318	−4842	17 450
Sleeve gastrectomy	204 947.99	91 439	19.71	0.382	239 186 (dominated) b	−72 324
Roux‐en‐Y gastric bypass	158 352.39	−46 596	20.20	0.498	−93 621 (dominated) c	71 481

Abbreviations: CVD, cardiovascular disease; ICER, incremental cost‐effectiveness ratio; INMB, incremental net‐monetary benefit; NA, not applicable; QALYs, quality‐adjusted life year.

^a All costs are shown in 2024 CAD.

^b Strongly dominated strategy. A strongly dominated strategy indicates that the strategy results in fewer QALYs gained and higher costs than another strategy, meaning it is strictly inferior and represents an inefficient use of resources.

^c Extendedly dominated strategy. An extendedly dominated strategy is one that is less cost‐effective than a combination of other strategies. This means that a combination of two alternative strategies provides more QALYs at a lower cost per QALY, making the extendedly dominated strategy inefficient.

^d Semaglutide patent expiry in 2026 with tiered price reduction applied in year 2.

^e Tirzepatide patent expiry in 2036 with tiered price reduction applied in year 12.

^f Combined tiered price reductions applied in year 12 for tirzepatide and in year 2 for semaglutide.

^g Surgery assumed to occur in year 5, reflecting average surgical wait times in Canada.

^h No surgical wait times are assumed in this scenario.

ⁱ Cardiometabolic benefits to pharmacotherapy arms: a 20% CVD risk reduction for semaglutide (applied to individuals with CVD), and reduced T2D incidence for semaglutide (73% risk reduction) and tirzepatide (93% risk reduction) (applied to those without baseline T2D), starting in year 1 and sustained for duration of treatment.

4. DISCUSSION

In our economic evaluation, RYGB and HBI were cost‐effective over a 40‐year horizon for Canadians with class III obesity and no baseline cardiometabolic comorbidities. While tirzepatide dominated semaglutide by providing greater QALY gains at lower costs, it was extendedly dominated by RYGB at 3% and 1.5% discount rates. Over a shorter 20‐year horizon, cost‐effectiveness was maintained for RYGB and HBI, while neither tirzepatide nor semaglutide met the CAD 50 000 per QALY threshold. These findings suggest that expanded access to HBI and RYGB for class III obesity management may represent cost‐effective strategies for the Canadian healthcare system.

Despite the growing enthusiasm for incretin mimetics, high drug prices remain a major barrier, even though these therapies consistently deliver greater QALY gains than other pharmacotherapies. ⁵⁷ , ⁷¹ , ⁷² Several analyses report that these agents, when combined with HBI, are not cost‐effective at current prices. ⁷¹ , ⁷² One systematic review found wide ICER variability for semaglutide, ranging from £14 827 to USD 122 549 per QALY compared to HBI, largely attributable to differences in drug costs. ⁷¹ Another Canadian study found semaglutide to be cost‐effective compared to HBI (ICER CAD 29 014 per QALY). ⁵⁷ This differs from our findings, likely due to their use of a societal perspective that included productivity gains and higher HBI costs derived from averaging four weight loss programmes. In contrast, we applied a payer perspective and modelled 13 structured counselling sessions. An analysis by Canada's Drug Agency estimated a substantially higher ICER for semaglutide (CAD 204 928 per QALY), further illustrating how drug cost and modelling assumptions can dramatically affect outcomes. ⁵² In our own scenario analysis, semaglutide became the most cost‐effective strategy if its price dropped by 47%, while tirzepatide became the most cost‐effective with a 12% price reduction. Further, applying a tiered price reduction at 12 years for tirzepatide and 2 years for semaglutide, consistent with expected pricing following patent expiry, would markedly enhance their cost‐effectiveness.

In our analyses, HBI delivered modest health benefits at a reasonable cost and was consistently cost‐effective. We modelled 13 counselling sessions—an approach not currently available in the Canadian public system. However, this reflects the protocols used in clinical trials (e.g., STEP 1 and SURMOUNT 1) as well as successful international programmes such as the National Health Service England's Diabetes Prevention Programme, a publicly funded initiative that includes 13 sessions and has shown to effectively prevent T2D. ¹² , ¹³ , ¹⁴ Our modelling decisions reflect a forward‐looking policy perspective: although 13 counselling sessions are not yet standard in the Canadian healthcare system, our findings aim to inform programme design decisions and eventual reimbursement. Our results support the integration of such structured programmes in Canada both for their direct health benefits and as a foundation to optimise the impact of pharmacotherapies, which are consistently combined with HBI in clinical trials. ⁹ , ¹¹

Our analyses incorporated both bariatric surgery and pharmacotherapy, allowing comparisons. Most studies have examined these modalities separately. Among the limited prior comparative studies, Saumoy and colleagues in the U.S. determined that bariatric surgery is cost‐effective for class III obesity, with semaglutide only becoming cost‐effective at a substantially lower annual cost (USD 1879), consistent with our findings on the impact of pricing. ⁷³ Similarly, Haseeb and colleagues found that endoscopic sleeve gastroplasty (ESG) was cost‐effective compared to semaglutide in individuals with class II obesity over a five‐year horizon, with semaglutide requiring an annual price reduction to USD 3591 to achieve cost‐effectiveness. ⁷⁴ In Asiabar and colleagues' systematic review, an American study found semaglutide was not cost‐effective compared to surgical options such as SG, RYGB, or ESG. ⁷¹ In our model, RYGB was not only cost‐effective at a CAD 50 000 per QALY threshold in 60% of simulations against tirzepatide, but in 25% of these simulations, it was cost‐saving. RYGB also dominated the more commonly performed SG, likely due to its greater long‐term weight loss, higher T2D remission rates, and fewer reoperations. Although RYGB entails higher upfront costs and surgical risks, it delivers sustained weight loss that may offset long‐term healthcare expenditures by delaying or preventing T2D and CVD. ³⁹ That said, cost offset evidence remains mixed. Some studies report similar total expenditures between surgical and non‐surgical management, and across surgical procedures. ⁵⁴ , ⁷⁵ , ⁷⁶ However, many of these studies evaluate short‐term horizons (≤5.5 years), which may not capture the full scope of long‐term cost offsets. Our 40‐year horizon captures broader cost offsets, favouring surgery. Importantly, although our model assumes equal access and uptake, real‐world preferences may differ. In a qualitative study of patients with obesity‐related complications, 27% preferred bariatric surgery and 30% preferred pharmacotherapy, with many citing concerns about invasiveness, surgical complications, long wait times, and costs as barriers to surgery. ⁷⁷

Our study has several limitations. First, we assumed sustained weight loss over 40 years with semaglutide and tirzepatide, despite limited long‐term data. We addressed this by incorporating real‐world discontinuation rates (>50%) from a retrospective cohort study to reflect evidence for treatment attrition. ³⁷ However, since the study only reported short‐term follow‐up, discontinuation was applied in Year 1 only, with adherence assumed thereafter. ³⁷ Second, our model focused on T2D and CVD, potentially underestimating QALY gains from treating other obesity‐related conditions such as hypertension, metabolic dysfunction‐associated steatotic liver disease, osteoarthritis, and cancer. While these conditions may be impacted by all interventions, the magnitude and nature of benefit likely vary by comorbidity. For instance, RYGB and SG may provide greater improvements for knee osteoarthritis due to more substantial and sustained weight loss. ⁷⁸ In a separate cost‐effectiveness analysis, RYGB provided 1.36 additional QALYs over usual care, primarily from mobility and pain improvements, effects that were not captured in our model. ⁷⁸ Third, we modelled a cohort without baseline T2D or CVD to evaluate how each intervention prevents future cardiometabolic complications; therefore, our findings apply to individuals with class III obesity without established comorbidities. Fourth, we adopted a public healthcare payer perspective, which is in alignment with Canada's Drug Agency's guidelines on economic evaluations. ¹⁹ However, this perspective excludes societal costs (e.g., productivity losses and caregiver burden) which may underestimate the full economic value of obesity treatments. Fifth, tirzepatide's cost was based on its T2D indication, as it had not yet been approved for obesity in Canada at the time of model development. However, it has been approved for this indication by Health Canada in May 2025, aligning with earlier approvals in the United States and the United Kingdom, and Health Canada has accepted a submission for its use in obesity. ¹⁶ , ¹⁷ , ⁷⁹ U.S. pricing is comparable between indications (USD 1079.77 per fill for diabetes and USD 1086.37 per fill for obesity). ⁸⁰ , ⁸¹ Given its recent approval in Canada, established regulatory status internationally, and substantial clinical benefits, inclusion of tirzepatide in our model is both timely and relevant to inform upcoming reimbursement and policy decisions in Canada and internationally. We addressed cost uncertainty by varying tirzepatide costs by ±50% in sensitivity analyses. Finally, in the base‐case, we modelled treatment benefits of semaglutide and tirzepatide solely based on weight reduction. We acknowledge that both agents may have cardioprotective effects not mediated by weight loss, and that to date only semaglutide has evidence for CVD impact. ⁷⁰ However, to capture additional cardiometabolic effects, we conducted a scenario analysis incorporating reduced cardiovascular events for semaglutide and reduced T2D incidence for both agents. These assumptions improved ICERs, with tirzepatide becoming the most cost‐effective strategy when reduced T2D incidence was included.

5. CONCLUSION

This economic evaluation found RYGB to be the most cost‐effective intervention for class III obesity in Canada, providing the highest QALY gains. HBI was also cost‐effective, offering modest QALY gains at low cost, even with 13 counselling sessions—the standard in clinical trials. Neither semaglutide nor tirzepatide was cost‐effective over a 40‐year horizon; however, substantial price reductions for incretin mimetics would significantly enhance their cost‐effectiveness. These findings, based on a high‐risk but comorbidity‐free population, support efforts to reduce bariatric surgery wait times and expand access to evidence‐based HBI for Canadians with class III obesity.

AUTHOR CONTRIBUTIONS

Concept and design: Nisha Gupta, Allie Babyak, Areni Chorbajian, Jeromie Ballreich, Kaberi Dasgupta. Acquisition, analysis or interpretation of data: All authors. Drafting of the manuscript: Nisha Gupta. Critical revision of the manuscript for important intellectual content: All authors. Statistical analysis: Nisha Gupta, Allie Babyak, Areni Chorbajian, Jeromie Ballreich. Obtained funding: Nisha Gupta under the supervision of Kaberi Dasgupta. Administrative, technical, or material support: Nisha Gupta, Jeromie Ballreich. Supervision: Jeromie Ballreich, Kaberi Dasgupta.

FUNDING INFORMATION

Nisha Gupta received support through grant BF10‐342730 Post‐diplôme professionnel (Fellowship) from the Fonds de recherche du Québec—Santé (Dr Gupta). The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

CONFLICT OF INTEREST STATEMENT

Dr. Dasgupta has received operating grants for other studies from the Canadian Institutes of Health Research, UK Medical Research Council, Diabetes Canada, and the Lawson Foundation. Dr. Tardio has received a speaker's honorarium from Bausch Health, Eli Lilly, NovoNordisk, Amgen, Boehringer‐Ingelheim, outside the submitted work. No other disclosures were reported.

PEER REVIEW

The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer-review/10.1111/dom.16627.

Supporting information

Appendix S1. Supporting Information.

Gupta N, Babyak A, Chorbajian A, Tardio V, Ballreich J, Dasgupta K. A cost‐effectiveness analysis of behavioural, pharmacological, and surgical obesity treatments in Canada. Diabetes Obes Metab. 2025;27(10):5748‐5760. doi: 10.1111/dom.16627

DATA AVAILABILITY STATEMENT

Data sharing not applicable to this article as no datasets were generated or analysed during the current study.