Cost-effectiveness of the use of the continuous subcutaneous insulin infusion pump versus daily multiple injections in type 1 diabetes adult patients at the Mexican Institute of Social Security

Background To estimate the incremental cost-effectiveness ratio (ICER) of the use of continuous subcutaneous insulin infusion (CSII) therapy versus multiple daily injections (MDI) therapy in adult patients with type 1 diabetes (T1D) at the Mexican Institute of Social Security (IMSS). Methods An analysis was developed using the internationally validated Core Diabetes Model (CDM) with which the incidence and progression of acute and chronic complications and the mortality of T1D was simulated throughout life. The baseline characteristics of the simulated cohorts were obtained from Mexican T1D adult patients aged ≥ 18 years that received care at two national IMSS medical centres in 2016. In the base case, the costs of the complications and treatment of the disease with both therapies were estimated in Mexican currency from the perspective of the institution, using Diagnosis Related Groups for outpatient and inpatient care. Utilities were taken from the international bibliography. In a secondary analysis, indirect costs were included using a human capital approach. The model used a lifetime time horizon, and a discount rate of 5% was applied for health outcomes and costs. A one-way sensitivity analysis was conducted on key variables and patient sub-groups; uncertainty was evaluated using a Cost-Effectiveness Acceptability Curve. Results The average age of the cohort was 32 years, with diabetes duration of 19 years, an average HbA1c of 9.2%; 29% were men. A gain of 0.614 Quality Adjusted Life Years (QALYs) was estimated with the use of CSII therapy. The estimated ICER was MXN$478,020 per QALY in the base case, and MXN$369,593 when indirect costs were considered. The sensitivity analysis showed that, in adult patients with HbA1c > 9.0%, the ICER was MXN$262,237. Conclusions This is the first economic evaluation study that compares CSII therapy versus MDI therapy for T1D adult patients in Mexico. The insulin pump therapy can be considered cost-effective in the context of the IMSS when considering a threshold of three GDPs per capita with 43.9% probability. Results improve substantially when patients have an HbA1c above 9%.


Background
In Mexico, type 1 diabetes (T1D) is escalating. In the last decade, the number of new T1D cases rose from 3.4 to 6.2 per 10,000 persons under 19 years of age [1]. This increase in the number of T1D cases represents an additional challenge for health systems to address. Since T1D patients require insulin replacement throughout life [2], they must adhere to a complicated insulin regimen and regular blood glucose monitoring to maintain glycemic control, prevent complications, and preserve life. Insulin-dependent type 2 diabetes (T2D) patients also require continuous, if not daily, delivery of insulin. Therefore, these conditions mandate the delivery of efficient treatment.
The growing number of patients with diabetes, either type 1 or type 2, represents a heavy toll for the Mexican Institute of Social Security (IMSS). This public institution provides medical, social, and economic benefits to almost 63 million people [3] that are formal sector workers and their families. The medical benefits include preventive and curative care, medications, rehabilitation, and orthopedic devices, free at the point-of-care. The economic benefits include work-related risk insurance, non-workrelated disability, life insurance, and old-age pensions, among others [4].
Currently, IMSS provides health care to approximately 3.8 million T1D and T2D adult patients [3], of which approximately 10% have T1D. Given the increase in the number of patients, the expenditures due to diabetes are substantial. In 2016, the expenditure on diabetes, cardiovascular diseases, cervical cancer, breast cancer, and HIV/AIDS reached USD$4.7 billion, of which diabetes expenditures alone accounted for USD$2.5 billion. In total, these five conditions accounted for 32% of IMSS's total health expenditures [3]. Since medical, social, and economic benefits are intertwined, the health and economic consequences of treatment for working-age adult patients with diabetes are particularly relevant to minimize disability leave, early pensions, and premature mortality.
The primary challenge in diabetes treatment is to achieve glycemic control that reduces the risk of acute and chronic complications, improves quality of life, and reduces premature mortality [5]. However, achieving glycemic control has proven to be challenging. Different studies have reported a low rate of glycemic control among T1D adult patients. In Mexico, the mean glycosylated hemoglobin (HbA1C) of T1D cases has been reported as 8.8% [6, 7] and 9.2% [8]. This figure is similar to other regions; for example, the mean HbA1C was estimated as 8.4% across 16 European countries, and 8.9% in Brazil and Colombia [9][10][11].
In the last decade, there have been significant breakthroughs in the development of diabetes technology that improves insulin delivery, such as continuous subcutaneous insulin infusion (CSII) pump therapy. Studies have shown that CSII is an effective alternative to multiple daily injections (MDI) [12][13][14][15], particularly for adult patients with poorly controlled diabetes [16]. Though several developed countries adopted this technology after performing cost-effectiveness studies, its use is not generalized [17]. The estimates indicate that in England and in the USA, 11.7% and 40% of T1D patients, respectively, use an insulin pump [18].
A systematic review of eleven CSII cost-effectiveness studies in eight countries found that in comparison with MDI, the use of CSII therapy was cost-effective and associated with improved life expectancy and quality adjusted life years (QALYs) (0.4-1.1 QALYs in adults). CSIItreated adult patients reached lower HbA1c and lower frequency of hypoglycemic events than MDI-treated adult patients. Although CSII therapy was associated with higher lifetime direct costs due to higher treatment costs when compared with MDI therapy, the costs avoided from reduced diabetes-related complications offset the increase in total expenditures [19].
Health care systems of developing countries such as Mexico have strict budget constraints in providing highquality care to patients with chronic disease; therefore, introducing innovations such as CSII for T1D requires a careful analysis of its economic costs and potential health gains. Analyzing whether CSII therapy can be a costeffective option compared with MDI therapy is pertinent in the IMSS context for several reasons. First, IMSS is the most significant health care provider in the country, covering the privately employed population; thus, the consequences of its decisions have substantial economic and health impacts. Second, results on healthcare costs pertaining to IMSS are at times generalized to other health institutions in Mexico where data is less available, making the availability of results particularly impactful. Finally, it is justifiable to evaluate new treatment options that improve the health outcomes, but there is also a need in IMSS to decelerate the escalating expenditures for diabetes. These considerations justify the present study, the objective of which was to estimate the incremental costeffectiveness ratio of the use of CSII therapy versus MDI therapy in IMSS adult patients.

IQVIA CORE Diabetes Model
We conducted a cost-effectiveness analysis using the validated IQVIA Core Diabetes Model (CDM). The use of CDM allowed for simulation of the incidence and progress of T1D complications and non-specific mortality  [20,21]. Previous publications have described the structure, data inputs, and validation of the CDM [20].

Simulation cohort
The baseline characteristics of the simulation cohort were taken from a cross-sectional study, whose primary objective was to describe the demographic and clinical characteristics of T1D patients affiliated with IMSS [8]. Additional unpublished data is presented here, however the detailed methodology is presented elsewhere [8].
Data was collected from 192 T1D adult patients aged ≥ 18 years who attended follow-up visit(s) at the Endocrinology Department of two IMSS tertiary care hospitals in Mexico City during the year 2016. Specifically, two trained research nurses reviewed all clinical records of the T1D adult patients (n = 192) and conducted telephone interviews to verify and complement the information that was missing in the clinical records.
Data on baseline demographics-sex, age, schooling, occupation, smoking and alcohol consumption, risk factors, pre-existing complications, and patient management (e.g., the percentage of adult patients taking statins) were collected. Table 1 shows the cohort characteristics; 29% were men, the mean age of the cohort was 32 years, mean duration of diabetes was 19 years, the mean BMI was 25.1 kg/m 2 , and HbA1c was 9.2%.

Intervention effects CSII versus MDI
CSII therapy has been associated with improved HbA1c levels and increased body weight, and a decrease in the occurrence of severe hypoglycemic events. The analysis estimated two intervention effects of CSII versus MDI.

Intervention costs
The perspective for the analyses considered IMSS as a single-payer healthcare system. The base case analysis was based on direct medical costs. Medtronic provided information on the cost of the CSII pump in Mexico (MiniMed ™ Paradigm ™ Veo ™ insulin pump) and associated supplies. Costs were averaged over 4 years for a yearly cost, based on the four-year warranty that came with the pump. It was assumed that the pump was replaced every 4 years. The costs of MDI and other medicines and supplies were based upon the prices of consolidated IMSS 2016 tenders [23] Resource use for treatment that was identical between the intervention arms were not included in the analysis. sources: (i) IMSS diagnosis-related groups for ambulatory [24] and hospital care [25], and (ii) 2016-unit costs by level of IMSS medical care [26]. Table 2 shows cost information. All costs were expressed in 2016 Mexican pesos (MXN$) and United States Dollars (USD$). In 2016, the exchange rate of MXN peso to USA dollar was 18.390 (http://www. anter ior.banxi co.org.mx/porta l-merca do-cambi ario/). In the secondary analysis, indirect costs were included using a human capital approach, considering disability leave due to diabetes complications. Costs were calculated based on the duration of the disability as determined by the CORE Diabetes model and daily salary. It was assumed that patients received their first salary at 18 and retired at the age of 60 according to IMSS regulation. Mean daily salary was taken from IMSS (Table 3).

Utility values
The analyses of the quality of life and health state utilities related to the different diabetes complications were obtained from Beaudet et al. [30]. When multiple complications occurred, the lowest utility value was used in the model. The majority of events are including using a oneoff subtraction, while patients who experienced events with long lasting impact were assigned a new health state value baseline. For specific events, a one-off disutility was subtracted in addition to the resulting health state value, such as for severe hypoglycemic events (Table 4).

Time horizon and discounting
The analysis applied the discount rate of 5% per year to costs and clinical outcomes [31]. The simulations used a patient lifetime time horizon, in which the health economic analysis continued until 100% mortality was reached. The simulation time horizon was set to 70 years with the goal of capturing the remainder of a lifetime of a patient with T1D in Mexico.

Sensitivity analyses
To explore the robustness of the base-case findings and determine the key drivers of cost-effectiveness, we performed a series of one-way sensitivity analyses. First, we analyzed the influence of the intervention effects on HbA1c by changing the mean baseline value of − 1.2 to − 0.6%. Second, the impact of the discount rates for future costs and benefits was assessed by applying a discount rate of 3%, and 7% per annum for economic outcomes and 0% and 7% for clinical outcomes, respectively. The applied discount rates were selected based on recommendations by the Mexican Economic Evaluation Guideline [31]. Third, we performed the sensitivity analysis in two subgroups: (1) uncontrolled HbA1c at baseline (> 9%) and controlled HbA1c ≤ 9%; (2) with T1D duration of ≤ 10 years and > 10 years.
The uncertainty surrounding the estimates of costeffectiveness was further assessed through a cost-effectiveness acceptability curve, using a willingness to pay threshold of one and three GDPs per capita (2016), or

Quality adjusted life years and life expectancy
The base case results indicate that in T1D IMSS adult patients, the total lifetime direct costs associated with CSII were projected to be MXN In the secondary case where indirect costs were considered, the use of CSII presented an ICER of MXN$369,593 (USD$20,097.50) per QALY (Table 7). Indirect costs accounted for an increase in MXN$427,884 (USD$23,267) and MXN$494,480 (USD$26,889) for the lifetime costs of CSII and MDI, respectively. Table 6 shows the number of years free of complications for T1D adult patients treated with CSII compared with MDI, demonstrating the delay in complications associated with CSII treatment. While the delay to any complication was limited to 0.39 years (4.68 months), the majority of severe complications were delayed by over 2 years. The most substantial delay was observed in the occurrence of the first ulcer (3.08 years), followed by gross proteinuria (2.82 years), and neuropathy (2.71 years). Heart and renal complications were delayed by approximately 2.5 years; specifically, end age renal disease (2.62 years), congestive health failure (2.57 years), angina (2.46 years), and myocardial infarction (2.46 years).
The uncertainty of the results was tested via a cost-effectiveness acceptability curve. When the willingness to pay was set at three GDPs per capita, or MXN$459,000/QALY, CSII therapy had a 43.9% probability of being cost-effective in the established cohort. There was a 0% probability of being cost-effective when the threshold was established as one GDP per capita, or MXN$153,000 (Fig. 1).

Table 2 (continued)
DRGs were converted from 2014 to 2016 costs using a health care specific inflation rate of 0.0865, as reported by INEGI. Ψ Costs were estimated as equivalent to 10% of the cost of the acute event where the acute event is estimated as 90% of the total DRG cost, as proposed by Reynales-Shigematsu et al. [27] a Cost of the insulin pump and annual supplies for insertion (Quick Serter) and infusion (Quick Set), reservoirs, lancets, test strips and insulin. Insulin costs were calculated considering (1) average weight of the baseline cohort of 65.45 kg, (2) daily doses 0.1 IU/kg, (3) daily requirement 6.545 IU, (4) annual requirement 2388.925 IU, (5) unit cost of $2.12 per IU of lispro insulin b Calculated considering (1) insulin types, daily doses obtained from the clinical records of the baseline cohort and (2) unit cost according to the type of insulin c Weighted cost of AMI of patients with discharge due to improvement (56%) and mortality (44%), weight based on patient distribution of GRDs  One-way deterministic sensitivity analysis and subgroup analysis

Discussion
This study is the first cost-effectiveness analysis comparing CSII with MDI using a simulation model, based on a cohort of T1D adult patients treated in a public healthcare institution of Mexico. The main findings indicate that insulin pump therapy can be considered cost-effective in the context of the IMSS, when considering a threshold of three GDPs per capita, with 43.9% probability. Results improve substantially when patients have an HbA1c above 9%. Moreover, the analysis suggests that CSII improves the quantity and quality of life by delaying the onset of complications and increasing life expectancy. The results of this study are consistent with other studies in developed countries that have assessed the efficacy and ICER of insulin pump therapy. Previous studies in the United States, Canada, and Australia performed in adults and child/young adult patients with T1D demonstrated that CSII improved life expectancy and quality of life due to a reduced incidence of diabetes-related complications and represented good value for money [32,33]. A systematic review of the clinical effectiveness and cost-effectiveness of CSII therapy for T1D patients concluded that CSII provides advantages over MDI therapy such as better control of glucose levels as reflected by HbA1c levels. Additional advantages were fewer hypoglycemia events and improved quality of life gains [34]. In our study, treatment with CSII in T1D adult patients was associated with an increase of 0.614 QALYs when compared to treatment with MDI, reflecting these tendencies. Similarly, the model reflects the delay of onset for diabetes-related complications that impacted both the clinical and economic results.
The total lifetime costs estimated in this analysis were higher for CSII therapy (MXN$1,404,173) than MDI therapy (MXN$1,110,573), largely due to the differences in the treatment costs of each arm. The cost of treatment represents approximately 49% of the total cost of CSII, and only 25% of the total cost of MDI. However, the use of CSII therapy in comparison to MDI therapy led to a delay in the onset of diabetes complications of up to 3 years. Delaying complications has clinical and economic significance; it has clinical importance as the delay indicates a better management of the disease and an improvement in quality of life. It has economic significance as the overall costs of CSII were partially offset by savings associated with the delay and overall reduction in diabetes-related complications as the progression of the disease is slowed. As demonstrated in the results, CSII is less costly for all expenses associated with diabetes complications, at times by up to 89% in the case of hypoglycemia. The relatively modest increase in absolute costs associated with CSII was also due in part to the survival paradox. Specifically, the life expectancy of an adult patient was 2.5 years longer in the CSII arm (undiscounted), resulting in additional annual treatment costs. Furthermore, this analysis took a conservative approach to the life-span of the insulin pump, assuming that the pump would be continually replaced at the end of the 4-year warranty period.
Of particular interest for the IMSS context is the secondary case analysis that included indirect costs. The IMSS is a social security institute whose benefits to affiliates include work-related risk insurance, non-workrelated disability, life insurance, and old-age pensions, among others [4]. As such, disability leave payments are among IMSS´ expenditures. While these expenditures could be defined as a direct healthcare cost, IMSS handles them separately as indirect costs. At the same time, the Mexican Economic Evaluation Guideline also considers disability-related expenditures to be indirect costs [31].  When including disability leave payments associated with diabetes-related complications, this analysis estimated an ICER of MXN$369,593 per QALY-a 20% reduction versus the base case. These results do not consider pension payments for early retirement due to diabetes that would result in a further reduction of the estimated ICER. An analysis of a retrospective cohort of 34,014 Mexican workers with permanent occupational disability caused by T1D and T2D during the years 2000-2013 at IMSS found that the mean age for permanent occupational disability was 51.6 years, and that the expenditure on diabetes-related pensions almost doubled (USD$58.28 to USD$111.62 million) [35]. As such, the results of this analysis may be considered a conservative estimate. Additionally, as several other Mexican institutes have similar payment responsibilities, these results may equally apply.
Finally, this study included a sensitivity analysis that comprise easily defined sub-groups, with the hopes of identifying highly cost-effective populations for a more efficient use of resources. Mexico faces constraints on the health care system and an overall low level of spending on the population versus the OECD average, of 5.8 vs 9.0 respectively in 2016 [36]. To reinforce efficiencies in spending, the Ministry of Health uses a Health Technology Assessment process to evaluate new innovations for  inclusion into the National Formulary [37]. At an institution level a budget impact analysis is also typically recommended for potentially high impact technologies. A budget impact model allows IMSS to allocate adequate resources in the procurement process and ensure future funding for the new health technology. As such, the pursuit for effective diabetes treatment represents an important challenge in Mexico due to the increasing incidence and prevalence of the disease, and consequent high budget impact. The results presented in the sensitivity analysis demonstrate that the use of CSII is most cost-effective in the patient group with uncontrolled diabetes, defined by HbA1c > 9%. By focusing on these patients, the baseline ICER can be reduced by ~ 45%, to MXN$262,237/QALY. While this is a significant improvement, it is important to note that the mean level of HbA1c of IMSS patients was reported as 9.2. As such, while CSII´s clinical impact and cost-effectiveness would be dramatically improved, the budget impact would remain significant in this patient sub-group.
Sensitivity analyses in other studies have also shown that altering the improvement in HbA1c levels increases the incremental cost-effectiveness ratios [38]. In this study, it was estimated that patients would see an improvement in the HbA1c value by − 1.2% versus the general minimum difference regarded as clinically significant at 0.5%. As such this represents a significant clinical benefit and highlights the importance in identifying patient groups with high probabilities of improvement in order to further increase cost-effectiveness. Cost-effectiveness was also sensitive to variations in the discount rates, and duration of the disease.
Regarding strengths, we performed a model-based economic study that was developed using the best available evidence for the context under analysis. The basis of the analysis was the validated CDM that has proven methodological consistency across different studies evaluating the cost-effectiveness of CSII versus MDI in several countries [39]. Equally important, the baseline patient characteristics of the cohort under analysis were taken from Mexican T1D patients treated in IMSS. Together with the use of IMSS direct and indirect costs, the results presented in this analysis reflect the reality of an institution that provides care to over 60% of the Mexican population.
A potential limitation of this analysis is that several inputs came from sources outside Mexico and Latin America, as these data were not available for Mexican T1D patients. Of importance, data on the dis-utilities of complications was based on a non-Mexican population. As such, the Life-Years Gained results may be of equal importance in the interpretation of results for some audiences. To address this limitation, future studies on diabetes management utilities for T1D patients in Mexico will be relevant.

Conclusions
This study is the first economic evaluation comparing CSII therapy versus MDI for T1D adult patients in Mexico. Insulin pump therapy can be considered cost-effective in the context of the IMSS when adult patients have an HbA1c > 9%. The results provide evidence to policymakers for the purpose of making decisions regarding resource allocation for the supply of health services for T1D adult patients. Thus, it is possible to conclude that the health gains from CSII therapy are substantial enough relative to the additional costs that introducing this technology represents.