Trends in Cost of Care With Tirzepatide in Adults Aged Over 55 Years With Obesity or Overweight Without Diabetes: A Matched Cohort Analysis

ABSTRACT

Aims

Compare healthcare cost trends among older adults with obesity or overweight treated with tirzepatide versus no treatment.

Materials and Methods

Retrospective cohort study using real‐world data from US Komodo Research Dataset, 8 November 2022 to 30 September 2025. Adults > 55 years with obesity, or overweight with ≥ 1 weight‐related comorbidity, without T2 diabetes who initiated tirzepatide for weight management were compared with propensity score‐matched individuals who did not receive any incretin‐based therapies. Changes in all‐cause healthcare costs (excluding tirzepatide; per‐person‐per‐month) between cohorts were assessed at 3–6, 6–12 and 12–18 months using a difference in differences (DiD) framework. Informative censoring and differential follow‐up were addressed with inverse‐probability‐of‐censoring‐weights and pairwise censoring.

Results

Each cohort comprised 15843 individuals. Healthcare cost DiD estimates (95% CI) for the tirzepatide versus control cohorts were −$145 (−266, −25; 12.3% reduction; p = 0.022) during 6–12 months and −$319 (−544, −94; 25.4% reduction; p = 0.015) during 12–18 months; using pairwise censoring, estimates were −$181 (−319, −43; p = 0.021) and −$607 (−1007, −207; p = 0.023), respectively. Cost offsets with tirzepatide were driven by higher medical costs in controls during 12–18 months (−$211 [−$407, −$15]; p = 0.065) and fewer inpatient/emergency department visits with tirzepatide versus controls; incidence rate ratio (95% CI) 0.69 (0.50, 0.94; p = 0.021) during 12–18 months.

Conclusions

Monthly cost offsets for persistent tirzepatide users versus untreated individuals grew over time, offsetting $319 (ranging from $94 to $544) tirzepatide costs by 12–18 months. These results have implications for the $195 monthly net Medicare cost for the Bridge programme, suggesting tirzepatide may generate near‐term cost savings in an older population that stays on treatment.

1. Introduction

Obesity is a significant public health crisis in the US and affects over 30% of adults aged > 55 years [1]. This chronic, relapsing disease is associated with high rates of related comorbidities, such as cardiovascular disease, diabetes, and certain cancers [2, 3, 4, 5]. With an ageing population, the rising burden on the healthcare system will be exacerbated by high rates of obesity and weight‐related comorbidities, where increasing obesity class is associated with increases in healthcare resource utilisation (HCRU) and costs [6, 7, 8], including among adults aged ≥ 65 years [5, 7, 8]. As a result, over the next 10 years, Medicare and Medicaid spending on obesity and related diseases is projected to be $4.1 trillion [9].

New classes of incretin‐based medications—glucagon‐like peptide‐1 receptor agonists (GLP‐1 RA) and dual glucose‐dependent insulinotropic polypeptide (GIP) receptor and GLP‐1 RAs—have emerged as promising therapeutic options for weight management among individuals with obesity, independent of a diabetes diagnosis [10, 11]. Tirzepatide is the only long‐acting dual GIP and GLP‐1 RA indicated for weight management in individuals with obesity, or with overweight and ≥ 1 weight‐related complication. It has shown substantial and sustained reductions in body weight and improvements in cardiometabolic parameters (HbA1c, cholesterol, blood pressure, etc.) in randomised clinical trials, including among individuals aged ≥ 65 years, where tirzepatide use was associated with mean body weight changes of up to −25.9% (in the SURMOUNT‐1 3‐year extension study) and had a comparable safety profile to adults aged < 65 years [11, 12, 13].

Savings from Medicare coverage of weight management medications have been estimated at $175 to $245 billion over the first 10 years [14]. However, Medicare Part D was explicitly prohibited by statute from covering drugs used solely for weight management. In July 2026, the Center for Medicare and Medicaid Services implemented the Medicare GLP‐1 Bridge programme, a short‐term (18 month) demonstration that subsidises access to incretin‐based weight management therapies, including tirzepatide, orforglipron and semaglutide [15, 16]. Under negotiated agreement with manufacturers, these medications are available for eligible Medicare beneficiaries at a price of $245, with a $50 copay ($195 net cost to Medicare), for a 30‐day supply. The Bridge programme is designed as a transitional measure, providing coverage until the end of 2027, prior to potential expansion through the BALANCE model [17]. Policymakers and payers are therefore looking to understand the longer‐term economic impact of weight management medications to inform their decision making. As such, the aim of this study was to compare trends in healthcare costs and HCRU among older adults (aged > 55 years) with overweight or obesity who were treated with tirzepatide compared with matched comparators who were untreated. The study focused on adults aged > 55 years since this population is directly relevant for coverage decisions being made for both private payers and the Medicare programme.

2. Methods

2.1. Study Design and Data Source

This retrospective, observational, matched‐cohort study used longitudinal, de‐identified data from the Komodo Research Dataset (KRD), a data schema designed for real‐world evidence studies enriched with Komodo‐populated imputed allowed amounts and is derived from Komodo's Healthcare Map. The Komodo Healthcare Map is a nationally representative database of de‐identified, real‐world patient data, comprising the individual healthcare experiences of > 330 million de‐identified patients across the US [18]. The study period was 8 November 2022 through 30 September 2025. This start date allowed inclusion of individuals who initiated tirzepatide on the date of its US regulatory approval for weight management (8 November 2023) and had 12 months of prior coverage.

Each individual's follow‐up was right‐censored at the earliest of: death, disenrolment from medical or pharmacy benefits, pregnancy, initiation of an alternative incretin‐based medication (for treated individuals, any other incretin‐based weight management medication other than tirzepatide; for controls, any incretin‐based weight management medication) or treatment discontinuation (defined as a gap of > 45 days after the index medication's days' supply was exhausted).

Outcomes were assessed using a difference‐in‐differences (DiD) framework, which compares within‐cohort changes in outcomes from a 12‐month pre‐index baseline period to successive follow‐up intervals up to 18 months post‐index between the tirzepatide cohort and the matched control cohort.

2.2. Study Population

This analysis included adults aged > 55 years on the index date with evidence of overweight or obesity during the preceding 12‐month baseline period. The population for this analysis was a pre‐specified subpopulation of the overall cohort of adults aged ≥ 18 years who met the study criteria described below. Obesity was defined as body mass index (BMI) ≥ 30.0 kg/m² or a qualifying obesity diagnosis code. Overweight was defined as BMI 27.0 to −29.9 kg/m² or a qualifying overweight diagnosis code, plus ≥ 1 weight‐related comorbidity (WRC). Individuals were required to be continuously enrolled in medical and pharmacy benefits for ≥ 12 months before the index date, with no enrolment gap > 30 days.

Individuals were excluded based on the presence of any of the following during the 12‐month baseline period: an incretin‐based weight management medication, a diagnosis of diabetes mellitus (identified by ICD‐10‐CM diagnosis codes or a haemoglobin A1c [HbA1c] value ≥ 6.5%); any pharmacy claim for a diabetes medication, excluding metformin but including incretin‐based therapies indicated for diabetes or weight management; pregnancy; cancer diagnosis; bariatric surgery; myocardial infarction or hip or knee replacement within 30 days before the index date; intensive care unit admission within 60 days before the index date; or cardiac surgery within 90 days before the index date. To reduce the influence of outlier baseline cost episodes, individuals with all‐cause baseline healthcare costs ≥ $250 000 USD (i.e., the > 0.01 percentile of costs) were excluded.

Included individuals who had ≥ 1 paid pharmacy claim for a tirzepatide autoinjector (indicated for weight management under brand name Zepbound) between 8 November 2023 and 30 September 2025 were identified as the tirzepatide cohort. Individuals who received tirzepatide vials directly from the manufacturer were not recorded in the dataset and thus were not distinguishable as tirzepatide initiators. The control cohort had no pharmacy claims for tirzepatide or any other incretin‐based weight management medication during the identification period.

2.3. Index Date and Study Periods

For the tirzepatide cohort, the index date was the date of the first pharmacy claim for tirzepatide. For the control cohort, the index date was assigned based on a randomly sampled date from the empirical distribution of index dates of the tirzepatide cohort. Baseline was the 12‐month period prior to index. There were three follow‐up periods: 3–6 months, 6–12 months and 12–18 months from the index date. The 0–3 month period was excluded from the analyses to attenuate the influence of tirzepatide dose titration, which spans approximately 12 weeks of therapy, as any changes observed during this period may have been attributable to baseline risk.

2.4. Outcomes

The primary outcome was the DiD in mean all‐cause healthcare costs (excluding tirzepatide costs) between the tirzepatide cohort and the control cohort at each of the three follow‐up periods from baseline. Secondary cost outcomes were DiD of all‐cause medical costs and all‐cause pharmacy costs (excluding tirzepatide costs). Tirzepatide costs were excluded from the outcomes, because changing pricing dynamics of tirzepatide are not standard across payers, rebates paid by manufacturers to payers directly are not captured, and these costs may obscure changes in costs due to other causes (e.g., changes in disease status).

All costs were imputed allowed amounts as supplied by or reported in the KRD [19] adjusted to 2025 US dollars using the Medical Care component of the Bureau of Labor Statistics Consumer Price Index. Costs were expressed as per person per month (PPPM), computed at the individual × period level as the period all‐cause cost divided by that individual's person‐month time at risk within the period under the censoring scheme defined in the statistical analysis.

HCRU outcomes were a composite of all‐cause inpatient admissions plus emergency department (ED) visits, and all‐cause outpatient and office visits, both assessed as DiD of visit rates per 1000 person‐months and ratio of incidence rate ratio of visit rates at each time period relative to baseline. When an inpatient admission and an ED visit for the same individual occurred on the same date, the two events were deduplicated to avoid double counting.

2.5. Statistical Analysis

2.5.1. Propensity Score Matching

Propensity score matching was used to reduce confounding from imbalances in baseline characteristics between cohorts. The propensity score was estimated using a generalised linear model with a logit link, taking into account baseline variables. Individuals treated with tirzepatide were matched 1:1 to eligible controls using nearest‐neighbour matching without replacement with a calliper of 0.2 × the standard deviation of the logit propensity score and an exact match requirement on index year‐month.

Covariate balance between the tirzepatide cohort and the control cohort was assessed before and after matching using absolute standardised mean difference (SMD); an absolute SMD < 0.10 was treated as evidence of acceptable balance.

2.5.2. Covariates

Baseline characteristics used in the propensity score matching included demographics (age, sex, race and ethnicity [White, Black, Hispanic/Latino, Asian/Pacific Islander, other, missing/unknown], US census region [Northeast, South, Midwest, West, missing/unknown], payer type, and calendar year of the index date), comorbidities (Charlson Comorbidity Index score, WRC counts, specific WRCs and cardiometabolic conditions), older generation oral weight management medication use and lifestyle interventions, HCRU and costs and BMI at index. Laboratory measurements (HbA1c, lipids, blood pressure and weight) were not included in the propensity model due to substantial missingness but are reported in Table S1.

2.5.3. Outcome Analyses

Baseline demographics, clinical characteristics, and prior HCRU were summarised using descriptive statistics. DiD estimates isolated treatment‐attributable changes by removing time‐invariant differences between the cohorts and secular changes that affect both cohorts equally [20]. This approach not only accounts for changes from baseline within the tirzepatide cohort but also takes into account changes from baseline within the control cohort. Generalized estimating equation (GEE) models for PPPM costs were fit, separately for each follow‐up period compared with baseline, with gamma distribution and log link function to accommodate right skewed cost distributions. Robust sandwich standard errors were clustered at the matched pair, with all observations from both pair members included inside that pair‐level cluster, which accounts for within pair correlation introduced by propensity score matching. The HCRU outcome models used the same follow‐up period time‐band specific approach as cost models, but with Poisson distribution, log link function and the log of person‐month at risk within the band included as an offset to express the outcomes on a rate‐per‐1000 person‐month scale. The relative change in incident rates in tirzepatide arm from baseline compared to relative change in incident rates in control arm from baseline were expressed in incidence rate ratio (IRR), obtained by exponentiating the treatment × period interaction coefficient, has the analogous DiD interpretation transposed to the rate scale. To address residual imbalances remaining between cohorts after matching, pairwise censoring outcome models were adjusted for six covariates that exceeded the SMD ≥ 0.1 threshold: Medicare payer, obstructive sleep apnoea, baseline BMI, the number of baseline WRCs, race (Black and White as separate indicators) and US census region.

Given that treatment discontinuation as a censoring criterion is only applicable to the tirzepatide cohort, it was likely that patients in the untreated cohort would have longer follow‐up. To account for the differential follow‐up time, pairwise censoring was specified as the primary approach, in which follow‐up time for an individual was censored when their matched counterpart follow‐up time was censored due to meeting the earliest of the censoring criteria. This approach was selected because the primary study objective was to compare changes in total costs and HCRU rates with tirzepatide treatment relative to changes observed over the same time period in the control arm. By preserving the symmetric exposure structure introduced through matching, pairwise censoring minimises bias that could arise from differential follow‐up duration, particularly when higher event rates or costs may be observed among individuals with longer follow‐up [21]. A pre‐specified sensitivity analysis relaxed the pairwise censoring scheme by allowing each individual to be censored at their own follow‐up endpoint (defined as a composite of death, medical or pharmacy disenrolment, pregnancy, switch to a non‐index incretin‐based therapy, index medication discontinuation or end of study) with surviving person‐time reweighted using inverse‐probability‐of‐censoring weights (IPCW) [22]. However, post hoc inspection of the results revealed a lower number of patients remaining in the 12–18 months follow‐up time band across treatment arms (as reported in the Results section), raising the likelihood of informative censoring between arms. Given this pattern, IPCW‐adjusted results are presented as the primary analysis, with pairwise censoring results retained as a supporting analysis. The IPCW‐weighted outcome models were not additionally adjusted for the variables used in weight derivation, further adjustment may introduce collinearity and reduce efficiency without providing additional bias reduction.

All analyses were conducted in SAS version 9.4 (SAS Institute, Cary, NC). Propensity score estimation was implemented in R version 4.4 (R Foundation for Statistical Computing, Vienna, Austria). Data extraction and intermediate aggregations were performed in Snowflake (Bozeman, MT).

3. Results

3.1. Demographic and Clinical Characteristics

Following sequential application of the pre‐specified inclusion criteria (Table S2), propensity score matching and derivation of the pre‐specified aged > 55 years cohort, 15 843 individuals were included in each of the treated and control cohorts (Figure 1). Baseline demographics, clinical characteristics, HCRU, and costs were broadly similar in each cohort with most SMDs < 0.10 (Table S1). Mean (SD) age was 64.5 (6.8) years in the tirzepatide cohort and 64.7 (7.0) years in the control cohort, and 68.8% and 67.8% were female, respectively. Most individuals (55.4% in the tirzepatide cohort, 59.5% in the control cohort) had commercial insurance, 38.5% versus 31.9% had Medicare, and 4.4% versus 5.9% were on Medicaid. Mean (SD) BMI was 34.8 kg/m² (4.2 kg/m²) in the tirzepatide cohort and 34.7 kg/m² (4.4 kg/m²) in the control cohort, with 25.1% and 26.0% of individuals, respectively, having class III obesity (≥ 40.0 kg/m²), and the mean (SD) number of WRCs in each cohort was 3.5 (2.2) and 3.3 (2.1). The most common WRCs were dyslipidaemia and hypertension, which affected over 60% of individuals in each group.

FIGURE 1.

Cohort identification and attrition. HbA1c, glycated haemoglobin; ICU, intensive care unit; PS, propensity score; SD, standard deviation.

Of the variables in the propensity score matched sample, six clinically meaningful covariates slightly exceeded the SMD threshold (≥ 0.10). These were Medicare payer type (38.5% in the tirzepatide cohort vs. 31.9% in the control cohort; SMD, 0.14), obstructive sleep apnoea (31.5% vs. 25.9%; SMD, 0.12), the lowest BMI category (27.0–29.9 kg/m² with WRC; 12.2% vs. 16.3%; SMD, 0.12), the number of baseline WRCs (mean, 3.5 vs. 3.3; SMD, 0.10), race (White, 64.6% vs. 60.0%; SMD, 0.10) and US census region (West, 16.5% vs. 13.1%; SMD, 0.10). BMI as a continuous variable was well balanced (mean 34.8 vs. 34.7; SMD, 0.02). A pre‐index parallel‐trends assessment was performed, which supported the parallel‐trends assumption underlying the subsequent DiD specifications (Supplementary Methods; Table S3).

With the individual censoring of study cohorts, of 15843 individuals in each cohort at baseline, in the tirzepatide cohort there were 8645 during 3–6 months follow‐up, 4511 during 6–12 months follow‐up and 1181 during 12–18 months follow‐up; in the control cohort, there were 11101, 7161 and 3004, respectively. In the pairwise censoring approach, in each cohort there were 15843 individuals at baseline, 7623 during 3–6 months follow‐up, 3603 during 6–12 months follow‐up and 793 during 12–18 months follow‐up.

3.2. Cost of Care Trajectory Over 18 Months Within Tirzepatide Cohort Versus Control Cohort

During the follow‐up period, all‐cause healthcare costs (excluding tirzepatide costs) PPPM diverged meaningfully between cohorts. Costs in the tirzepatide cohort remained largely unchanged from baseline in the IPCW adjusted analysis (Figure 2A), while costs in the matched control cohort rose over the same period (from $1031 during baseline to $1244 during 12–18 months). This divergence translated into statistically significant DiD estimates (95% CI) during 6–12 months (−$145 [−266, −25]; 12.3% reduction; p = 0.022) and 12–18 months (−$319 [−544, −94]; 25.4% reduction; p = 0.015), with the magnitude of cost offsets increasing over time. Results of the pairwise censoring analysis (Figure 2B) supported the direction of IPCW adjusted analysis approach and the interpretation that there was a sustained cost reduction in the tirzepatide cohort compared with the control cohort, which was greater with longer follow‐up, with significant DiD estimates (95% CI) during 6–12 months (−$181 [−319, −43] PPPM, 15.4% reduction; p = 0.021) and 12–18 months (−$607 [−1007, −207] PPPM, 38.4% reduction; p = 0.023).

FIGURE 2.

All‐cause healthcare costs (excluding tirzepatide costs) PPPM, trajectory among the cohort treated with tirzepatide for weight management and matched controls during the 18‐month post‐index period. (A) Individual censoring with IPCW adjusted approach, (B) pairwise censoring approach. Costs are shown in 2025 USD. In (A) IPCW stabilised weights were derived from two cohort‐stratified Cox proportional‐hazards models fit on the matched cohort: a marginal model with no covariates and an adjusted model conditioning on age, BMI, baseline WRC count, log‐transformed baseline all‐cause cost, sex, payer, race, region and obstructive sleep apnoea. These weights were then applied to the gamma–log GEE specification. In (B) DiD estimates were derived from separate GEE models with gamma distribution with log link fitted for each follow‐up period (3–6, 6–12 and 12–18 months) versus baseline. Estimates within treatment arm at each follow‐up period are adjusted mean derived from covariate‐adjusted GEE models using population‐marginal least‐squares means. Relative reduction defined as 1—the follow up to baseline cost ratio in the tirzepatide cohort divided by the analogous ratio in the control cohort. Error bars indicate 95% CI. DiD p values correspond to the cost ratio scale on which GEE models were fitted; corresponding 95% confidence intervals for the dollar DiD were computed by the delta method, propagating the robust (sandwich) covariance of the model coefficients. Datapoints for each time interval are displayed as offset to aid visualisation of overlapping error bars. Where error bars are not visible, they fall within the range of the symbol depicting the mean. BMI, body mass index; DiD, difference in differences; GEE, Generalized estimating equation; IPCW, inverse probability of censoring weighting; PPPM, per person per month; USD, US dollars; WRC, weight‐related comorbidity.

Medical cost trajectories also diverged between the tirzepatide and control cohorts during follow‐up, consistent with trends in all‐cause healthcare costs (Figure 3). In the IPCW adjusted analysis, medical cost DiD estimates (95% CI) were statistically significant during 6–12 months (−$152 [−254, −50]; p = 0.008) and numerically in the same direction during 12–18 months (−$211 [−407, −15]; p = 0.065) (Figure 3A). Pharmacy costs (excluding tirzepatide) were similar between cohorts through 12 months post‐index and DiD estimates during 12–18 months of −$108 (−207, −9) (Figure 3C). Similar results were seen in the pairwise censoring analysis (Figure 3B,D), except that pharmacy costs rose in the control arm while remaining stable in the tirzepatide cohort, yielding a DiD estimate (95% CI) during 12–18 months of −$278 (−475, −81; p = 0.053).

FIGURE 3.

(A, B) All‐cause medical costs and (C, D) all‐cause pharmacy costs (excluding tirzepatide costs) PPPM, trajectory among the cohort treated with tirzepatide for weight management and matched controls during the 18‐month post‐index period. (A, C) Individual censoring with IPCW approach. (B, D) Pairwise censoring approach. Costs are shown in 2025 USD. In (A) and (C) IPCW stabilised weights were derived from two cohort‐stratified Cox proportional‐hazards models fit on the matched cohort: a marginal model with no covariates and an adjusted model conditioning on age, BMI, baseline WRC count, log‐transformed baseline all‐cause cost, sex, payer, race, region and obstructive sleep apnoea. These weights were then applied to the gamma–log GEE specification. In (B) and (D) DiD estimates were derived from separate GEE models with gamma distribution with log link fitted for each follow‐up period (3–6, 6–12 and 12–18 months) versus baseline. Estimates within treatment arm at each follow‐up period are adjusted mean derived from covariate‐adjusted GEE models using population‐marginal least‐squares means. DiD p values correspond to the cost ratio scale on which GEE models were fitted; corresponding 95% confidence intervals for the dollar DiD were computed by the delta method, propagating the robust (sandwich) covariance of the model coefficients. Datapoints for each time interval are displayed as offset to aid visualisation of overlapping error bars. Where error bars are not visible, they fall within the range of the symbol depicting the mean. BMI, body mass index; DiD, difference in differences; GEE, Generalized estimating equation; IPCW, inverse probability of censoring weighting; PPPM, per person per month; USD, US dollars; WRC, weight‐related comorbidity.

3.3. Changes in HCRU Over 18 Months Within Tirzepatide Cohort Versus Control Cohort

In the IPCW adjusted analysis, the tirzepatide cohort had a more favourable change in rate of inpatient and ED visits relative to baseline compared with the control cohort across all follow‐up time periods, and absolute differences in rates were statistically significant at the 3–6, 6–12 and 12–18 months periods (Figure 4A). IRRs (95% CI) were 0.86 (0.74, 0.99; p = 0.033) during 3–6 months, 0.76 (0.64, 0.89, p = 0.001) during 6–12 months and 0.69 (0.50, 0.94, p = 0.021) during 12–18 months.

FIGURE 4.

Healthcare resource utilisation rate per 1000 person‐months among the cohort treated with tirzepatide for weight management and matched controls during the 18‐month post‐index period. (A, B) Inpatient and ED visits and (C, D) outpatient and office visits. (A, C) Individual censoring with IPCW approach. (B, D) Pairwise censoring approach. In (A, C) IPCW stabilised weights were derived from two cohort‐stratified Cox proportional‐hazards models fit on the matched cohort: a marginal model with no covariates and an adjusted model conditioning on age, BMI, baseline WRC count, log‐transformed baseline all‐cause cost, sex, payer, race, region and obstructive sleep apnoea. These weights were then applied to the Poisson–log GEE specification. In (B) and (D) DiD estimates were derived from separate GEE models with Poisson distribution with log link fitted for each follow‐up period (3–6, 6–12 and 12–18 months) versus baseline. Estimates within treatment arm at each follow‐up period are adjusted mean derived from covariate‐adjusted GEE models using population‐marginal least‐squares means. IRR is IRR in tirzepatide cohort relative to baseline divided by IRR in control cohort relative to baseline. BMI, body mass index; DiD, difference in differences; GEE, Generalized estimating equation; ED, emergency department; IPCW, inverse probability of censoring weighting; IRR, incidence rate ratio; WRC, weight‐related comorbidity.

Pairwise censoring analysis results supported the IPCW analysis and the interpretation that the change in rates of inpatient and ED visits were numerically lower in the tirzepatide cohort relative to baseline compared with the control cohort in the 3–6 months and 6–12 months periods (Figure 4B). IRRs (95% CI) for inpatient and ED visits were 0.87 (0.74, 1.03; p = 0.101) during 3–6 months and 0.83 (0.68, 1.01; p = 0.065) during 6–12 months.

There were no statistically significant changes in rate of all‐cause outpatient and office visits relative to baseline in the tirzepatide cohort compared with the control cohort throughout the post‐index periods in the IPCW adjusted analysis (Figure 4C), with IRR (95% CI) 1.04 (0.99, 1.08, p = 0.141) during 3–6 months, 1.02 (0.96, 1.08, p = 0.593) during 6–12 months and 1.05 (0.94, 1.17, p = 0.380) during 12–18 months. However, numerically, the DiD estimates for absolute differences in rates per 1000 person months were an additional 23.8 visits, 2.9 visits and 6.7 visits within tirzepatide cohort relative to baseline compared to control cohort. The pairwise censoring analysis was consistent with the IPCW analysis (Figure 4D).

4. Discussion

This study assessed real‐world changes in healthcare costs and HCRU following tirzepatide initiation for weight management among older adults (aged > 55 years) with overweight (with ≥ 1 weight‐related comorbidity) or obesity and without diabetes compared with matched untreated controls. In both the IPCW and pairwise censored analyses, all‐cause healthcare costs (excluding tirzepatide drug costs) were significantly lower and continued to diverge over time in the tirzepatide versus control cohorts, primarily driven by rising costs in the control cohort. In analyses adjusting for potential informative censoring, PPPM differences during 6–12 months and 12–18 months were −$145 and −$319 PPPM, respectively, representing reductions of 12.3% and 25.4%, respectively, in the tirzepatide cohort relative to the control cohort. Results for the pairwise censored analysis yielded numerically larger estimated cost differences for the tirzepatide versus control cohorts of −$181 PPPM during 6–12 months and −$607 during 12–18 months. Together, these results suggest a potential economic benefit of persistent tirzepatide treatment among older adults.

These findings have implications for the Medicare GLP‐1 Bridge programme, under which tirzepatide is available to eligible beneficiaries at a negotiated net price of $245 per month with $50 co‐pay (i.e., $195 per month cost to Medicare) [23]. The reduction in healthcare costs (95% CI) of $319 (−544, −94) per month observed during the 12–18 months following tirzepatide initiation suggests that downstream economic benefits may offset the net drug cost among persistent tirzepatide users. Consistent findings with an alternative pairwise censoring approach demonstrate cost offsets (95% CI) of $607 (−1007, −207) per month during 12–18 months. Both approaches demonstrated cost offsets with tirzepatide, further supporting the robustness of these results. These findings indicate that coverage for tirzepatide for weight management may have economic benefits for Medicare in the near term. Moreover, there may also be potential for higher cost savings at longer time horizons, since, for example, estimates indicate that weight management medications could save Medicare $175 to $245 billion in the first 10 years of coverage increasing to up to $846 billion after 30 years [14].

The rising costs in the control cohort were driven by increases across both healthcare cost components and higher rates of inpatient and ED visits from 3 to 18 months. In contrast, healthcare costs and HCRU in the tirzepatide cohort remained broadly stable across time periods, with the exception of higher rates of outpatient and office visits. This may reflect increased engagement with routine care following tirzepatide treatment initiation. The increasing cost in the control cohort could be explained by rising costs over time incurred by weight‐related multimorbidity in these individuals, as both cohorts had an average of ≥ 3 WRCs, with over 40% having ≥ 4 WRCs at baseline. Additionally, up to 21% of individuals had annual ED visits during the baseline period, representing a high cost burden in this population. The observed increase in healthcare costs in the control cohort is consistent with a large cohort study that followed adults (aged ≥ 18 years) for 8 years from their first diagnosis of obesity and observed around a 10% growth in all‐cause cost of care per year, from $504 and $5237 at baseline for individuals with class I obesity and class III obesity, respectively, to $9291 and $12 472 at Year 8 [4]. Healthcare costs have also been reported to increase with number of WRCs, with one study demonstrating costs from $4893 to $6926 in Years 1–8 in individuals without WRCs and from $15 585 to $20 349 in those with ≥ 3 WRCs [24]. Similar results have been seen in a UK population with obesity [25]. Together, these results are consistent with the growth in costs observed in the untreated cohort and highlight the benefit of tirzepatide in keeping costs flat.

This is the first study to assess real‐world cost offsets in older adults treated with tirzepatide. However, cost differences between the tirzepatide and control cohorts are consistent with estimates from a previous study in the Medicare setting showing a projected reduction in cost of care with reductions in BMI [26]. Thorpe et al. reported that a 15%–20% BMI reduction was associated with an estimated annual reduction in spending of $3511 to $4513, respectively, among adults in Medicare with obesity and ≥ 1 comorbid condition [26]. Real‐world medical cost reduction for individuals in the KRD with obesity and ≥ 2 WRCs treated with semaglutide have been estimated at $323 per month compared with matched controls [27]. The latter study included individuals with type 2 diabetes, however the observed medical cost reductions are consistent with the findings in medical costs offsets with tirzepatide in this study. Some recent literature has found no significant differences in costs following treatment for weight management [28, 29, 30]. However, as incretin‐based medications have only recently been approved for weight management, these studies have tended to focus on older generation incretin‐based medications (e.g., dulaglutide, liraglutide or semaglutide for a diabetes indication, which has a lower dose than for weight management). They also may have included individuals who were not persistent on the medication [28].

Key strengths of this study include use of a large multipayor dataset with a nationally representative sample. The study design incorporated a well‐matched untreated control group, allowing for a robust temporal comparison of cost trajectories using a DiD approach. As mentioned above, healthcare costs associated with obesity have been shown to increase over time by up to 42% [4, 24, 25], so comparison with a control group provided essential context for interpreting cost changes observed with treatment with tirzepatide. Furthermore, propensity score matching and IPCW adjusted analysis enhanced the validity of group comparisons by minimising bias and accounting for baseline differences, informative censoring and differential follow‐up. Additionally, the consistency of findings with the alternative censoring adjustment approach further strengthened confidence in the robustness of the results. Future research should assess differences between demographic subgroups, for example, stratified by race/ethnicity or geographic region, and evaluate the intention‐to‐treat (ITT) estimand. Given that the time period in this study includes a period of supply shortage during 2024 (36% of included individuals had an index date in 2024), which may have affected treatment persistence [31], an ITT analysis was not conducted in this study, as the results may not be generalisable to current real‐world tirzepatide use where supply shortages have been resolved.

4.1. Limitations

This study has several limitations. The post‐index follow‐up period is constrained by the approval of tirzepatide for weight management in November 2023, with over 60% of the study cohort starting tirzepatide in year 2025; fewer than 10% of individuals remained uncensored during the 12–18 month band reducing precision and generalisability for the estimates and preventing further subgroup analysis (e.g., by census region or race/ethnicity) in this time period. However, analysis with the IPCW adjustment helped to address any informative censoring between cohorts during this period and provides greater confidence in the range of cost offsets observed in the real‐world in this population.

The findings of this study apply only to this age/eligibility‐selected subgroup and cannot be extrapolated to younger or broader tirzepatide‐treated populations. For anyone who discontinued tirzepatide, their follow‐up time was right censored from the analysis; consequently, findings are generalisable only to those who stay on treatment with tirzepatide and cannot be extrapolated post‐discontinuation. Furthermore, individuals receiving tirzepatide through Medicare prior to coverage expansions in July 2026 may represent a higher‐risk population than those who have access after July 2026; however it is important to note that under the Medicare GLP‐1 Bridge programme, only a select Medicare population aged ≥ 65 years with specific BMI classes and with certain pre‐existing conditions receive weight management therapy at the subsidised rate [23].

Additionally, obesity medications obtained via cash‐pay, compounding or directly from the manufacturer were not captured, which may result in misclassification of some untreated individuals and underestimate the true cost differences. Despite propensity score matching and covariate adjustment, residual confounding from unmeasured factors, including socioeconomic status, lifestyle behaviours and health literacy, cannot be excluded. Finally, as with all claims‐based analyses, study results may be subject to coding errors, data entry errors or misclassification and a pharmacy claim does not confirm medication use.

4.2. Conclusion

In this real‐world study of older individuals with overweight (with ≥ 1 weight‐related comorbidity) or obesity and without type 2 diabetes, persistent treatment with tirzepatide for weight management was associated with cost offsets compared with untreated controls, suggesting potential economic benefits of sustained weight management with tirzepatide in this population. Cost offsets grew progressively throughout the follow‐up period. In the context of the Medicare GLP‐1 Bridge programme, where tirzepatide has a net monthly cost of $195 to Medicare, these findings suggest that tirzepatide coverage may partially offset the drug cost by 12 months and may result in cost savings by 18 months among those who stay on the tirzepatide therapy for weight management.

Author Contributions

All authors contributed to the conception and design of the study and the analysis and interpretation of the data. All authors contributed to the drafting and critical revision of the manuscript for important intellectual content and approved the final manuscript for publication.

Funding

This work was supported by Eli Lilly and Company.

Conflicts of Interest

Navneet Upadhyay, Benjamin Behrend and Emily Ruth Hankosky: employees and shareholders of Eli Lilly and Company. Ali Bonakdar and Keshab Subedi: employees of Cencora, which received payment from Eli Lilly and Company for analytical services for this study. Soutrik Banerjee: employee of Aixial Group, which received payment from Eli Lilly and Company for analytical services for this study.

Supporting information

Data S1: Supplementary Methods.

Table S1: Baseline characteristics of study population.

Table S2: Sequential application of inclusion criteria, propensity score matching and complete‐pair filter.

Table S3: Pre‐index parallel‐trend assessment for all‐cause healthcare cost PPPM (excluding tirzepatide costs).

Upadhyay N., Bonakdar A., Subedi K., Banerjee S., Behrend B., and Hankosky E. R., “Trends in Cost of Care With Tirzepatide in Adults Aged Over 55 Years With Obesity or Overweight Without Diabetes: A Matched Cohort Analysis,” Diabetes, Obesity and Metabolism 28, no. 11 (2026): 10418–10427, 10.1111/dom.71250.

Data Availability Statement

All data generated or analysed during the study were derived from the privately held Komodo Research Data and are not publicly available.