Cost-effectiveness of advanced hybrid closed loop therapy compared to standard insulin therapy for type 1 diabetes in pregnancy: an economic evaluation of the CRISTAL trial

Abstract

Background: A multicentre, randomised controlled trial (the CRISTAL trial), demonstrated the safety and efficacy of MiniMed™ 780G advanced hybrid closed loop (AHCL) therapy during pregnancy, showing improved glycaemic control overnight, less hypoglycaemia, and improved treatment satisfaction compared to standard insulin therapy (SoC, mainly open-loop insulin pump therapy). This study aimed to assess the cost-effectiveness of AHCL, which has a higher initial cost, compared to SoC in pregnant women with type 1 diabetes (T1D).

Methods: A decision tree model was developed to estimate the cost-effectiveness of AHCL compared to SoC in pregnant women with T1D, covering pregnancy to birth and postpartum hospital discharge (a time horizon of 28 weeks). Total costs per strategy (in 2024 euros, €) were calculated from a healthcare payer perspective. The base-case analysis derived prevalence of pregnancy complications and hospitalisations directly related to diabetes management from the CRISTAL trial. Uncertainty was analysed by exploring multiple scenarios and sensitivity analyses.

Findings: In the base-case analysis, the cost of using AHCL during pregnancy was estimated at €13,988.75 (95% CI: €12,240 to €16,062) compared to €14,221.33 (95% CI: €12,380 to €16,420) for SoC, indicating cost-savings of €232.57 per individual, alongside the demonstrated clinical benefits of AHCL. The primary cost driver was the AHCL device cost. This cost was offset by savings from shorter and less frequent hospital admissions (mainly due to severe hypoglycaemia and dysregulated diabetes) in the AHCL group compared to SoC. In our probabilistic sensitivity analysis, AHCL was dominant in 73% of the simulated cost-effectiveness pairs.

Interpretation: AHCL might be cost-saving compared to SoC for pregnant women with T1D. However, more robust data are needed to assess the potential impact of AHCL therapy on pregnancy and long-term health outcomes.

Funding: Diabetes Liga Research Fund and Medtronic.

Keywords: Advanced hybrid closed loop therapy; Cost-effectiveness; Economic evaluation; Healthcare costs; Pregnancy; Randomised controlled trial; Type 1 diabetes.

Fig. 1

Decision treemodel. The blue square represents the decision point between the two strategies. Green circles are probability nodes, representing uncertain events with branches showing possible outcomes. All probabilities of outcomes emanating from the same node sum to 1. The red triangles are terminal points representing the final outcomes. The vertical line in the decision tree model indicates that the subsequent subtrees follow the same structure and are therefore not shown in detail. The decision tree, read from left to right, first accounts for the possibility of miscarriage. For pregnancies continuing beyond 20 weeks, patients with complicated pregnancies face two potential scenarios: treatment with hospitalisation or treatment without hospitalisation. Delivery can be vaginal or through a C-section, with possibilities for stillbirth and BPI from vaginal deliveries. BPI is used in the decision tree model as a representative outcome of shoulder dystocia, without separately modelling shoulder dystocia, as both outcomes have a relatively small probability of occurring (approximately 1%). N(I)CU admissions include preterm birth, hypoglycaemia, hyperbilirubinemia, and neonatal respiratory distress, as per NICE guidelines. Abbreviations: T1D type 1 diabetes; AHCL advanced hybrid closed loop; BPI brachial plexus injury; N(I)CU neonatal (intensive) care unit.

Fig. 2

Probability distribution of the incremental costin the base-case analysis,based on 10,000Monte Carlo simulations (AHCL therapy vs. SoC). The results of the Monte Carlo simulations reflect the impact of sampled parameter uncertainties on the model results. The means represent averages from 10,000 simulations, each incorporating an assumed standard deviation of 20% for uncertain parameters (as reported in Table 1, Table 2 and 2). The histogram displays the distribution of the cost difference (€) between the AHCL therapy group and the SoC group across 10,000 simulations. The x-axis represents the difference in costs between AHCL therapy and SoC, ranging from approximately −2,300 to 600 euros. The y-axis represents the probability density of the cost differences, with higher bars indicating more frequent outcomes. The dashed line, positioned at −233€, represents the estimated incremental cost difference. The red solid line, positioned at 0 €, marks the point where the cost difference between AHCL and SoC is zero. The area to the left of the red line represents 73% of the simulated outcomes. In these cases, the additional cost of the MiniMed™ 780G device is offset by healthcare cost savings, indicating that AHCL is cost-saving. The area to the right of the red line represents 27% of the simulated outcomes. In these cases, SoC becomes cost-saving compared to AHCL. The distribution is left-skewed, with most of the values left of the red vertical line indicating that in most simulations, AHCL therapy results in cost savings (negative cost differences). Abbreviations: AHCL advanced hybrid closed loop; SoC standard insulin therapy; € euros.

Fig. 3

Cost-effectiveness planesdemonstrating the probability ofcost-effectivenessfor AHCL compared to SoC in thebase-caseanalysis, based on 10,000 Monte Carlo simulations. The y-axis represents the incremental cost of AHCL compared to SoC (€), while the x-axis represents the incremental effects in minutes. In panel a, the incremental effects are defined as the difference in improved overnight pregnancy-specific time in range (TIRp) of AHCL compared to SoC, with positive values indicating increased TIRp (overnight). In panel b, the incremental effects are defined as the difference in reduced pregnancy-specific time below range (TBRp) of AHCL compared to SoC, with negative values indicating a reduction in TBRp. In panel a, 0.02% of the cost–effect pairs fall in the north-west (NW) quadrant (less effective, more costly); 0.07% fall in the south-west (SW) quadrant (less effective, less costly); 72.79% fall in the south-east (SE) quadrant (more effective, less costly); and 27.12% fall in the north-east (NE) quadrant (more effective, more costly). In panel b, where the x-axis measures reductions in TBRp (improvements are negative), 27.07% of the cost–effect pairs fall in the NW quadrant (more effective, more costly); 72.79% fall in the SW quadrant (more effective, less costly); 0.07% fall in the SE quadrant (less effective, less costly); and 0.07% fall in the NE quadrant (less effective, more costly). Both TBRp and overnight TIRp were assumed to be normally distributed. Abbreviations: € euros; TIRp pregnancy-specific time in range; TBRp pregnancy-specific time below range.

Fig. 4

Tornado diagram showing the cost difference of AHCL therapy compared to SoCin the base-case analysis. This tornado diagram illustrates the sensitivity of the cost difference between AHCL and SoC to variations in key model parameters (univariate sensitivity analysis). The x-axis represents the cost difference in euros (€), with negative values indicating cost savings for AHCL and positive values indicating higher costs for AHCL. Blue bars indicate the decrease in the parameter value, red bars indicate an increase. The extent of each bar shows the impact on the cost difference when the parameter is varied within its specified range. Probability and cost parameters were varied for the MiniMed™ 780G device cost, miscarriages, complicated versus non-complicated pregnancies, and treatment with versus without hospitalisation. Since the same values were consistently applied to both the AHCL and SoC groups, parameters associated with perinatal outcomes did not alter the incremental results in the univariate sensitivity analysis and were therefore excluded from this diagram. Abbreviations: AHCL advanced hybrid closed loop; SoC standard insulin therapy; € euros.

Fig. 5

Multivariate sensitivity analysis ontheMiniMed™ 780G cost in euros (€) andthenumber of hospitalisation days in the SoC group. The figure illustrates how changes to the MiniMed™ 780G cost in euros (x-axis) and the hospitalisation days in the SoC group (y-axis) affect the estimated cost difference. The orange area represents the combinations of the MiniMed™ 780G cost and hospitalisation days where SoC is the cost-saving choice, while the blue area indicates combinations where AHCL therapy is cost-saving. The intersection between the areas represents the combinations where the costs of AHCL and SoC are equal. At the observed average number of hospitalisation days in the SoC group (11.5 days) and the current cost of the MiniMed™ 780G (€5,096 on the Belgian market), AHCL therapy is more economical since this combination falls in the blue area, as indicated by the white circle. Given the current MiniMed™ 780G cost of €5,096, the threshold number of hospitalisation days is approximately 8.4 days. If the hospitalisation duration falls below this value, SoC becomes the cost-saving strategy, as indicated by the dark circle, which lies on the threshold. All combinations of fewer than 8.4 days fall below the dark circle and within the orange area.