Estimated Cost-effectiveness of Medical Therapy, Sleeve Gastrectomy, and Gastric Bypass in Patients With Severe Obesity and Type 2 Diabetes

Key Points Question Compared with medical therapy, is bariatric surgery with sleeve gastrectomy or Roux-en-Y gastric bypass (RYGB) associated with cost-effective weight reduction in patients with severe obesity and varying type 2 diabetes (T2D) severity? Findings In this economic evaluation using simulated patient cohorts, RYGB was projected to be the preferred strategy in the overall population with T2D at 5 years (probability preferred, 83.0%). The cost-effectiveness of RYGB was highest in those with mild-to-moderate T2D at baseline. Meaning These findings suggest that RYGB is projected to be cost-effective in patients with severe obesity and T2D, regardless of T2D severity.


Individualized Metabolic Surgery (IMS) Score Calculation
The following equation, developed by Aminian et al (2017) 1 , was used to calculate an Individualized Metabolic Surgery (IMS) Score for each modeled individual in our simulation. It considers number of diabetes medication, insulin use, duration of type 2 diabetes mellitus (T2DM), and glycemic control (HbA1c) before surgical intervention (preop). This score was then used to divide patients into three different levels of type 2 diabetes mellitus (T2DM) severity: mild (IMS score ≤ 25), moderate (25 < IMS score ≤ 95), and severe (IMS score > 95). Patients with mild T2DM are more likely to have good glycemic control and less likely to be on insulin before surgery compared to other severity levels. Patients with higher severity of T2DM are more likely to be on insulin and have a high number of diabetes medication with poor glycemic control as well as a long duration of T2DM history. Where the following are indicator variables:

Healthcare Costs
As in prior analyses, we estimated the total healthcare expenditures of United States (U.S.) adults from U.S. Medical Expenditure Panel Survey (MEPS) using a two-part model. [2][3][4] First, we used a survey-weighted multivariable logistic regression model to estimate the probability of non-zero total annual healthcare expenditures. Second, among individuals with non-zero total healthcare expenditures, we used a survey-weighted multivariable generalized linear model with a log link and gamma distribution to estimate annual healthcare costs. We included the same covariates in the generalized linear model as in the logistic regression.
As MEPS does not include institutionalized individuals receiving long-term care, we separately estimated the mean, weighted cost of long-term care, stratified by age and sex. 2 We divided the number of U.S. adults using long-term care in 2013-2014 (i.e., using adult day service centers or residing in nursing homes or assisted living facilities), stratified by age (i.e., <65, 65-74, 75-84, ≥85) and sex, by the number of U.S. adults in the same age and sex strata in the 2010 U.S. Census. 5,6 We multiplied the proportion of all U.S. adults using each type of long-term care service by published annual long-term care cost estimates from the US Department of Health and Human Services and the Genworth Cost of Care Survey, to estimate the mean, weighted cost of long-term care. 7,8 We used the combined two-part model to estimate mean costs and standard errors for age (18-34, 35-44, 45-54, 55-64, 65-74, 75-84, and ≥85 years), sex, T2DM (yes or no), and BMI (18-24.9, 25-29.9, 30-34.9, 35-39.9, and ≥40 kg/m2) stratified groups. Total background healthcare costs (including long-term care costs) for age-and sex-groups are shown in eTable 1. Total background healthcare costs represent all non-diabetes, nonoverweight/obesity, and non-surgery related healthcare costs. The age-specific costs of having diabetes or overweight/obesity are shown in eTable 2. During each year of the simulation, total healthcare costs are calculated by adding the total background healthcare costs, costs of diabetes, costs of overweight/obesity, and costs of surgery and surgical complications.

Net Monetary Benefit (NMB)
Net monetary benefit (NMB) is the monetary value of an intervention for a given cost-effectiveness (CE) threshold. Incremental net monetary benefit (INMB) is calculated as the incremental effectiveness (in qualityadjusted life years, QALYs) of a strategy multiplied by the WTP and then subtracted by its incremental costs.
indicates the intervention is cost-effective compared to the reference strategy for the given CE threshold, while an INMB ≤0 indicates that the reference strategy is preferred. Additionally, the greater the INMB, the more cost-effective the strategy is compared with the reference strategy. When more than two interventions are considered simultaneously, the strategy with the highest INMB is considered the preferred strategy (i.e., the most cost-effective). The