In the US T2D market, prices reflect both branded and generic competition, within and across classes. Despite recommendations, comparative economic analyses usually hold relative prices constant and, equally unrealistic, often use list prices. Even the few analyses that use prices net of manufacturer rebates understate opportunity costs by excluding time-varying intermediary margins. We illustrate the implications of these choices by simulating canagliflozin 300mg vs. sitagliptin 100mg in treating T2D.

In 1993, the US Public Health Service convened the First Panel on Cost-Effectiveness in Health and Medicine, which concluded that decision-makers should consider both health and cost impacts when choosing among competing technologies. An important feature of economic analysis is how it captures heterogeneity (i.e., ‘diversity in character or content’). In particular, the dominant practice of performing economic evaluations at mean values can lead to sub-optimal decisions. Unfortunately, guidelines are vague on methods, key academic underpinnings are spread across the literature, and methods papers focus primarily on the European setting. Our aim was to fill this gap by performing an integrative review of heterogeneity in guidelines and the methodology literature to provide best practice recommendations for the US setting. We performed an integrative review of modeling guidelines and used PubMed to search for “patient heterogeneity” AND “economic evaluation” AND “cost-effectiveness” in the English language. All materials until November 1, 2021 were considered for inclusion independently by study authors (and discrepancies resolved). Methodological studies were subject to forward and backward citation searches. Qualifying studies and guidelines were reviewed and US-focused recommendations developed. Forty-six methodological studies, 31 model taxonomy studies, and 8 best practice recommendations were chosen. Best practice guidelines are vague on methods. Several methodological studies provide more detail, including guidance on how appropriate subgroups should be identified. Most studies focus on patient heterogeneity; less is written on other types of heterogeneity important primarily for the US. For example, insurance coverage heterogeneity and individual risk preferences were identified and incorporated in the broader framework. Properly accounting for heterogeneity is essential for economic analysis and analyses that short shrift its consideration result in sub-optimal decisions. By synthesizing the existing literature and guidelines and tailoring to the US setting, a broader framework including heterogeneity can improve economic analyses.

Based on the renal and cardiovascular benefits demonstrated in the CREDENCE trial, canagliflozin has received a licence extension for the treatment of DKD in Europe. As canagliflozin is the first treatment for DKD in nearly two decades, a new micro-simulation model was developed to estimate the cost effectiveness of canagliflozin added to standard of care (SoC) versus SoC alone. We present results of a cost utility analysis (CUA) for England, conducted in accordance with NICE methods. The model was built with patient-level data from CREDENCE and includes risk equations for start of dialysis, hospitalisation for heart failure (HHF), non-fatal myocardial infarction (MI), non-fatal stroke, and all-cause mortality. Progression of DKD is modelled using evolution equations for estimated glomerular filtration rate (eGFR) and urinary albumin:creatinine ratio (UACR), with patients transitioned to dialysis at an eGFR of 6ml/min/1.73m2 as a fail-safe. The model was loaded with baseline characteristics and treatment effects from CREDENCE. Unit costs and disutility weights were sourced from the literature with costs inflated to £2019. The base case time horizon was 40 years and 500 cohorts of 500 hypothetical patients were simulated. Sensitivity analyses were performed. Costs and QALYs were discounted at 3.5% annually. Canagliflozin plus SoC was associated with gains in life-years and QALYs versus SoC of 2.47 and 1.95 respectively over 40 years, driven largely by reductions in the rates of dialysis start (60%) and HHF (29%), MI (16%), and stroke (17%) events. Substantial cost offsets for dialysis (£12651) and transplant (£8871) lead to cost savings for canagliflozin (£8897 per patient). Extrapolating the CREDENCE trial over 40 years suggests that canagliflozin can substantially reduce rates of renal and cardiovascular outcomes, increase longevity, and save costs in this high unmet need population.

Based on renal and cardiovascular benefits demonstrated in the CREDENCE trial, canagliflozin (CANA) received a novel indication to reduce the risk of end-stage kidney disease (ESKD), doubling of serum creatinine, cardiovascular death, and hospitalization for heart failure (HHF) in patients T2D and DKD with albuminuria >300 mg/day. As CANA is the first new pharmacotherapy for this population in nearly two decades, economic stakeholders are unfamiliar with assessing the cost offsets associated with treatment. The purpose of this study was to use an economic microsimulation model to estimate health outcomes and cost offsets associated with CANA versus standard of care (SOC) in this patient population from a US payer perspective. The economic model, CREDEM-DKD, was built with patient-level data from CREDENCE and includes risk equations for the following events: HHF, start of dialysis, non-fatal myocardial infarction (MI), non-fatal stroke, and all-cause mortality. Equations for estimated glomerular filtration rate (eGFR) and urinary albumin:creatinine ratio (UACR) evolution were used to capture DKD progression. The model was loaded with baseline characteristics and treatment effects from CREDENCE. Unit costs were sourced from literature and inflated to $2018. In the base case, health trajectories were simulated over 5 years for 1,000 cohorts of 1,000 hypothetical patients. Sensitivity analyses were performed. In the base case, treatment with CANA versus SOC was associated with 20% fewer dialysis starts; 36%, 16%, and 18% fewer HHF, MI, and stroke events, respectively; and 0.08 life-years gained. The largest estimated 5-year cost-offsets were for dialysis and HHF at about $4,000 and $1,000 per patient, respectively, increasing to approximately $7,700 and $1,500, respectively, in the 10-year simulation. Results of these simulations suggest that for this high unmet need patient population, treatment with CANA will result in fewer renal, cardiovascular, and mortality events and sizable cost offsets.

We estimated the cost-effectiveness of once-weekly semaglutide 1 mg vs. canagliflozin 300 mg in Canada from the payer and societal perspectives. Modeling methods were used to extrapolate benefits observed in the SUSTAIN 8 clinical trial into long-term costs and outcomes, measured by quality-adjusted life-years (QALYs), for patients treated with semaglutide or canagliflozin. The SUSTAIN 8 trial, a 52-week, randomized, double-blind clinical trial, demonstrated significantly greater lowering of HbA1c and body weight for semaglutide 1 mg vs. canagliflozin 300 mg for patients with type 2 diabetes mellitus (T2DM) uncontrolled on metformin. The Swedish Institute for Health Economics diabetes cohort model (IHE-DCM) was used for modelling, and to consider structural uncertainty and robustness of results, the analysis was also run using the microsimulation model ECHO-T2DM. Patient baseline characteristics and treatment effects were sourced from SUSTAIN 8. In the simulation, both agents were discontinued, and insulin therapy initiated when HbA1c exceeded 8.0%. Unit costs (CAD$) and utilities were sourced from the literature. Semaglutide 1 mg was associated with incrementally more QALYs than canagliflozin 300 mg over 40 years (0.38 in IHE-DCM and 0.36 in ECHO-T2DM). The gains came with increased total costs (CAD 8,097 in IHE-DCM and CAD 8,867 in ECHO-T2DM), yielding incremental cost-effectiveness ratios (ICER) of CAD 21,307 and CAD 24,513 per QALY gained, respectively, below the often-cited willingness-to-pay threshold of CAD 50,000 per QALY. Including productivity costs and the full societal perspective lowered the ICERs to CAD 19,303 and CAD 18,962, respectively. Sensitivity analyses around assumptions related to costs, treatment effects, time horizon, and biomarker evolution generally confirmed the results. Using two different models and modeling approaches (cohort and microsimulation), semaglutide 1 mg was found to be cost-effective compared to canagliflozin 300 mg for the treatment of patients with T2D uncontrolled on metformin in Canada.

Microsimulation using risk equations to convert biomarker values into event risks is the norm in T2DM. Though risk factor clustering (whereby individuals with one unfavorable risk factor are likely to have other unfavorable risk factors as well) is common in diabetic populations, accounting for it in empirical applications is rare despite the longstanding example of the Global Diabetes Model (GDM). This absence can potentially bias cost-effectiveness estimates. While the GDM approach is data-intensive, the problem can be addressed simply by allowing correlation of baseline patient characteristics. This study aims to leverage National Health and Nutrition Examination Survey (NHANES) data to estimate correlation coefficients and fill a gap in the literature and to explore bias potential using examples from the US 3rd party payer perspective. Two cohorts of individuals with T2DM—biguanide only and sulfonylurea + biguanide—were identified in the five NHANES cross-sections between 2007 and 2016. Baseline characteristics and correlation coefficients were estimated using stratification weights. Baseline population characteristics were entered into ECHO-T2DM, a validated microsimulation model of T2DM, and 20-year cost-effectiveness simulations were run including and then excluding the correlation coefficients. While hypothetical, the intervention reflects common scenarios. The correlation coefficients spanned from tightly correlated (-0.91 for HDL and triglycerides) to almost uncorrelated (0.01 for HDL and BMI), with some variation by cohort. While estimated cost-effectiveness ratios did not change qualitatively when correlations were added, there were important numerical differences. It should be acknowledged that the NHANES sample sizes are relatively small and that conditioning on HbA1c failure (typical in T2DM modeling) was not possible. Risk factor clustering may be important for modeling the cost-effectiveness of T2DM interventions, correlation in sampling baseline characteristics is easy, and now two sets of correlation coefficients (albeit crude) are available for other researchers.

Canagliflozin is a sodium glucose co-transporter 2 inhibitor which reduces HbA1c, blood pressure, and weight loss, and has a low intrinsic risk for hypoglycaemia. In the CANagliflozin cardioVascular Assessment Study (CANVAS), canagliflozin significantly reduced the risk of major atherosclerotic cardiovascular events (MACE) by 14% over 3.6 years in 10,142 T2DM patients with high risk of developing or with established CVD versus placebo. There were also numerical differences in favour of canagliflozin for non-fatal myocardial infarction (MI) and stroke, CV death, and hospitalisation for heart failure (HHF). Given this, we used CANVAS data to estimate CV events and deaths avoided and associated cost offsets for canagliflozin vs. standard of care (SoC) from the perspective of the NHS in England. It was assumed that 50.5% of the T2DM population in England met the inclusion/exclusion criteria of CANVAS based on a US database analysis. The rates observed in CANVAS were applied to the eligible population to estimate the CVD events avoided for canagliflozin over SoC over 5 years. Unit costs, obtained from the literature (primarily UKPDS), were multiplied by the number of events to estimate cost offsets for canagliflozin over SoC. The impact of including drug costs and adverse events as part of the analysis was tested in the sensitivity analysis. It was estimated that the use of canagliflozin could avoid 42,000 MACE events, 16,000 CV deaths and 24,000 all-cause deaths over 5 years. Reductions of 17,000 MIs, 8,000 strokes, and 28,000 HHFs were associated with cost offsets of £360 million (£130, £62 and £168 million, respectively) over this period. Using canagliflozin in T2DM patients with high risk of or with established CVD can improve health in England by controlling HbA1c and reducing CV events, while allowing substantial near-term cost offsets.

When modeling the cost-effectiveness of treating chronic diseases, analysis usually emphasizes the features of the drugs under comparison. The long time horizons necessary to capture the full costs and benefits of intervention means that cost-effectiveness is determined not only by the properties of these comparators, but also by those of rescue therapies. Unfortunately, the effects of rescue treatment are often neglected, and arguments are frequently proffered that they affect both arms equally. When there are differences in durability, however, this can seriously impact results. For example, a treatment that is more durable will generally be found to be more cost-effective when rescue therapy is expensive or ineffective (or both) than when rescue therapy is cheap or effective. Multiple sclerosis is a chronic disease and the recommended treatment sequence has changed considerably in the last decade. We examine the impact of assumptions regarding the modeled sequence of subsequent treatments on estimated cost-effectiveness of natalizumab versus fingolimod in the second-line treatment of relapsing-remitting multiple sclerosis. In particular, we evaluate how the inclusion of a hypothetical expensive, but ineffective rescue therapy affects estimated cost-effectiveness. We used a micro-simulation model that was initially developed to review the cost-effectiveness of multiple sclerosis treatments currently available in Sweden, with a 20-year time horizon and a societal perspective. Costs were sourced from Swedish publications and price lists, and QALYs were based on estimates from a Swedish patient survey. By altering the rescue therapy, estimated cost-effectiveness of natalizumab versus fingolimod switches from around SEK 690,000 per QALY gained to cost-saving (i.e., dominating) versus fingolimod. The treatment sequences modeled in economic simulation can greatly impact estimated cost-effectiveness. It is important to model the sequence of rescue therapies properly in considering chronic and progressive diseases. Good practice guidelines should be amended to include this more explicitly.

Health economic simulation models of chronic diseases often use surrogate clinical covariates to extrapolate from short-term clinical trials to long-term outcomes of interest. Often, these treatment effects are parameterized for each agent as change from baseline value (e.g., improvement in biomarkers such as blood glucose in diabetes). As most modeling guidelines recognize the importance of capturing parameter (sometimes called 2ndorder) uncertainty, this poses a modeling challenge as most models where treatment effects are expressed as changes relative to baseline values fail to capture the correlation correctly. In particular, the nature of the evidence on treatment effects coming from comparative clinical studies (or meta-analyses of such studies) imposes a correlation structure on these treatment effects and ignoring these correlations leads to biased estimates of the uncertainty, such as incorrect confidence intervals around incremental cost-effectiveness ratios and incorrect probabilities of cost-effectiveness. The aim of this study is to describe the statistical problem and advocate a simple solution. Correct modelling of the correlations, by either parametrizing the inputs using correlation coefficients, or by mathematically decoupling the incremental effect of one treatment from the change relative to the baseline value applicable to the comparator treatment, overcomes this mistake. This allows the parameter uncertainty to propagate correctly to uncertainty in the final estimates of cost-effectiveness. In models that rely on covariates in order to generate accurate estimates of uncertainty, modelers should honor the correlation structure of treatment effects derived from comparative sources of evidence such as randomized controlled trials or meta-analyses.

OBJECTIVE—To evaluate the efficacy and safety of add-on insulin glargine versus rosiglitazone in insulin-naïve patients with type 2 diabetes inadequately controlled on dual oral therapy with sulfonylurea plus metformin. RESEARCH DESIGN AND METHODS—In this 24-week multicenter, randomized, open-label, parallel trial, 217 patients (HbA1c [A1C] 7.5–11%, BMI >25 kg/m2) on ≥50% of maximal-dose sulfonylurea and metformin received add-on insulin glargine 10 units/day or rosiglitazone 4 mg/day. Insulin glargine was forced-titrated to target fasting plasma glucose (FPG) ≤5.5–6.7 mmol/l (≤100–120 mg/dl), and rosiglitazone was increased to 8 mg/day any time after 6 weeks if FPG was >5.5 mmol/l. RESULTS—A1C improvements from baseline were similar in both groups (−1.7 vs. −1.5% for insulin glargine vs. rosiglitazone, respectively); however, when baseline A1C was >9.5%, the reduction of A1C with insulin glargine was greater than with rosiglitazone (P < 0.05). Insulin glargine yielded better FPG values than rosiglitazone (−3.6 ± 0.23 vs. −2.6 ± 0.22 mmol/l; P = 0.001). Insulin glargine final dose per day was 38 ± 26 IU vs. 7.1 ± 2 mg for rosiglitazone. Confirmed hypoglycemic events at plasma glucose <3.9 mmol/l (<70 mg/dl) were slightly greater for the insulin glargine group (n = 57) than for the rosiglitazone group (n = 47) (P = 0.0528). The calculated average rate per patient-year of a confirmed hypoglycemic event (<70 mg/dl), after adjusting for BMI, was 7.7 (95% CI 5.4–10.8) and 3.4 (2.3–5.0) for the insulin glargine and rosiglitazone groups, respectively (P = 0.0073). More patients in the insulin glargine group had confirmed nocturnal hypoglycemia of <3.9 mmol/l (P = 0.02) and <2.8 mmol/l (P < 0.05) than in the rosiglitazone group. Effects on total cholesterol, LDL cholesterol, and triglyceride levels from baseline to end point with insulin glargine (−4.4, −1.4, and −19.0%, respectively) contrasted with those of rosiglitazone (+10.1, +13.1, and +4.6%, respectively; P < 0.002). HDL cholesterol was unchanged with insulin glargine but increased with rosiglitazone by 4.4% (P < 0.05). Insulin glargine had less weight gain than rosiglitazone (1.6 ± 0.4 vs. 3.0 ± 0.4 kg; P = 0.02), fewer adverse events (7 vs. 29%; P = 0.0001), and no peripheral edema (0 vs. 12.5%). Insulin glargine saved $235/patient over 24 weeks compared with rosiglitazone. CONCLUSIONS—Low-dose insulin glargine combined with a sulfonylurea and metformin resulted in similar A1C improvements except for greater reductions in A1C when baseline was ≥9.5% compared with add-on maximum-dose rosiglitazone. Further, insulin glargine was associated with more hypoglycemia but less weight gain, no edema, and salutary lipid changes at a lower cost of therapy.

In Canada, the most commonly utilized oral third-line agent for patients with T2DM inadequately controlled on metformin (MET) and a sulfonylurea (SU) is sitagliptin (SITA). Canagliflozin (CANA), a novel agent that inhibits sodium glucose co-transporter 2 (SGLT2), has demonstrated HbA1c lowering, as well as improvements in weight and systolic blood pressure (SBP). The objective of this analysis was to evaluate the cost-effectiveness of CANA 100 and 300 mg versus SITA 100 mg in patients inadequately controlled on MET + SU in the Canadian setting. In accordance with the CADTH guidelines for economic evaluations, cost-utility analysis using ECHO-T2DM, a validated economic model, was done to simulate lifetime outcomes and costs of using CANA versus SITA in the third-line setting. Patient characteristics and treatment effects were sourced from a head-to-head study for the comparison of CANA 300 mg to SITA 100 mg. In the absence of a direct comparison of CANA 100 mg versus SITA 100 mg, relative treatment effects for this simulation were obtained from an indirect comparison via Bayesian network meta-analysis (NMA), with baseline patient characteristics sourced from a pooled analysis of two CANA trials (patients on background therapy of MET + SU) that contributed to the NMA. ECHO-T2DM was populated with Canadian costs and utility estimates relevant to the Canadian population. Using CANA 300 and 100 mg resulted in mean quality-adjusted life year (QALY) gains of 0.08 and 0.04, respectively, and lower costs of $2,035 and $981, respectively, compared to SITA over 40 years in patients failing to meet glycemic control on MET + SU. Therefore, CANA “dominated” SITA. CANA used as a third-line agent added on to MET + SU background therapy may result in better quality of life outcomes and lower costs when compared to SITA (the most common third-line agent in Canada).

To estimate the cost-effectiveness of using CANA versus DAPA or EMPA, three agents that inhibit sodium glucose co-transporter 2 (SGLT2), as monotherapy from the UK NHS perspective. The validated ECHO-T2DM model was used to estimate 40-year outcomes and costs associated with using CANA 100 or 300mg versus DAPA 10mg or EMPA 25mg. Data from a 26-week network meta-analysis (NMA) performed to support a NICE multiple technology assessment were used to populate the model with treatment effects for HbA1c, blood pressure, weight and rates of hypoglycaemic events (hypoglycaemia data for EMPA were not possible to report from the NMA). Changes in lipids and rates of adverse events (AEs) associated with SGLT2 inhibition (i.e., urinary tract infections, genital mycotic infections) were sourced from a CANA monotherapy trial; values for DAPA and EMPA were assumed the same as CANA 100mg (as was the hypoglycaemia rate for EMPA). Sensitivity analyses were also performed. In the base case, CANA 100mg dominated DAPA and EMPA with quality-adjusted life-year (QALY) gains of 0.033 and 0.015 and lower total costs of £69 and £3. CANA 300mg versus DAPA provided an estimated QALY gain of 0.075 and increased cost of £709, resulting in an incremental cost-effectiveness ratio (ICER) of £9,429. Versus EMPA, the ICER was slightly higher (£13,491), but still below the generally accepted threshold in the UK, with a QALY gain of 0.056 and an increased cost of £761. Sensitivity analyses supported these base case findings. Through an insulin-independent mechanism of action, agents that inhibit SGLT2 improve glucose levels, blood pressure, and weight, with a low inherent risk of hypoglycaemia. These results suggest that both CANA 100 and 300mg are likely to be cost-effective monotherapy options versus DAPA and EMPA in the UK.

People aged ≥65 years with T2DM contribute significantly to the increasing rate of health care utilization. CANA, an agent that inhibits sodium glucose co-transporter 2 (SGLT2), and SAXA, a dipeptidyl peptidase-4 inhibitor, have provided meaningful HbA1c reductions when used as monotherapy and as add-on to other antihyperglycemic agents in older patients. This analysis estimates the cost-effectiveness of CANA 100 or 300 mg versus SAXA 5 mg in patients with T2DM aged ≥65 years in the Canadian setting. ECHO-T2DM was used to simulate outcomes associated with using CANA versus SAXA as an add-on therapy in patients with T2DM aged ≥65 years. As head-to-head data were unavailable, an indirect comparison (IC) was performed using published data on SAXA 5 mg and results from a post hoc analysis of CANA data where possible (HbA1c and weight). IC estimates were calculated for those inadequately controlled on a mix of different background therapies (lifestyle intervention alone or combination with metformin, metformin plus sulfonylurea, or metformin plus pioglitazone). For other biomarkers (ie, cholesterol, systolic blood pressure) and adverse event rates, SAXA 5 mg values were assumed to be equal to those of placebo in the post hoc analysis. The post hoc dataset was also the source of the background patient characteristics. Costs and benefits were discounted at 5% and assessed from the Canadian perspective. Sensitivity analyses were performed. Both CANA 100 and 300 mg were dominant compared to SAXA 5 mg (lower net cost and greater quality-adjusted life-years [QALYs]). CANA 100 and 300 mg reduced costs (–$375 and –$771, respectively) and improved QALYs (0.033 and 0.057, respectively) over 40 years. Sensitivity analyses support these findings. These results suggest that using CANA in older individuals is cost-effective versus SAXA in Canada.

To assess the cost-effectiveness of CANA versus SITA in patients with T2DM inadequately controlled with metformin and sulfonylurea from the perspective of the Brazilian private healthcare system. The validated Economics and Health Outcomes Model of T2DM (ECHO-T2DM) was used to estimate the cost-effectiveness of CANA 100 and 300 mg versus SITA 100 mg added to metformin and sulfonylurea over a 20-year horizon. Patient characteristics were obtained from a pooled analysis of two CANA trials as add-on to metformin and sulfonylurea (DIA3002 and DIA3015). Efficacy and adverse event inputs were sourced from DIA3002 for CANA 100 mg and from DIA3002/DIA3015 for CANA 300 mg and SITA. Pharmaceutical costs were sourced from list prices; hospitalizations and resource use were from a medical claims database. Outcomes and costs were discounted at 5%. Sensitivity analyses were conducted that varied parameters relevant to the Brazilian setting, including using data from Latin American patients in CANA trials. CANA 100 and 300 mg were associated with QALY gains of 0.09 and 0.21 and mean cost increases of R$2,403 and R$2,947 relative to SITA. Non-medication cost offsets were seen with CANA 100 and 300 mg versus SITA (0.3% and 2.0%). CANA 100 mg was cost-effective per WHO criteria (<3 times the gross domestic product [GDP] per capita) and CANA 300 mg was very cost-effective (<1 times the GDP per capita) based on GDP per capita (R$26,082), with incremental cost-effectiveness ratios of R$27,755 and R$13,904 per QALY gained, respectively. The cost-effectiveness of CANA versus SITA was robust to different specifications in the sensitivity analyses. These Resultssuggest that adding CANA 100 or 300 mg versus SITA in patients with T2DM inadequately controlled on metformin and sulfonylurea would be a more efficient use of healthcare resources in Brazil.

T2DM is a chronic, progressive disease and proper economic evaluation of alternative treatment interventions requires economic modeling over long time horizons. As currently available treatments cannot halt disease progression, most patients eventually require therapy intensification to meet HbA1c goals. This analysis explores the impact of commonly used intensification assumptions on cost-effectiveness estimates using simulations of canagliflozin (CANA) versus maximally-titrated glimepiride (GLIM) in patients with uncontrolled HbA1c on metformin in the US. ECHO-T2DM, a validated micro-simulation model, was used to simulate 30-year outcomes and costs associated with using CANA 100 or 300mg versus GLIM as add-on to metformin. Patient characteristics, treatment effects, and adverse event rates were sourced from a previously reported head-to-head trial. Health utilities and unit costs were sourced from the literature. Two types of treatment intensification triggers were modeled: when HbA1c exceeds a target threshold, and after a fixed amount of time. Treatment was intensified first by adding basal insulin and then prandial insulin, both titrated to maintain HbA1c control (up to pre-specified maximum doses). A simulation with no intensification was also performed. Incremental cost-effectiveness ratios (ICERs) for treatment strategies starting with CANA 100 and 300mg versus GLIM with insulin rescue at HbA1c >7.0% were $29,032 and $22,106, respectively, largely driven by CANA’s ability to keep HbA1c controlled longer, thus delaying insulin initiation. Using a fixed (and equal) time on CANA and GLIM (5 and 10 years) favored GLIM by eliminating this benefit, yielding higher ICERs. The extreme case of no rescue therapy artificially inflated complication costs in both arms, since HbA1c drifts unabated upwards. Assumptions about treatment intensification matter. Unrealistic assumptions like fixing time on agents or omitting intensification had large effects, as ICERs depend on how downstream treatment choices are modeled. Consumers of T2DM economic evaluations should therefore consider these assumptions carefully.