This analysis used the recently published, validated, person-level microsimulation model DEDUCE [18]. The DEDUCE model assigns costs and utilities according to the complications and acute events (diabetic ketoacidosis [DKA], severe hypoglycemic events [SHE], and non-severe hypoglycemic events [NSHE]) experienced by each simulated person living with diabetes, with the incidence and history of complications updated for each 1-year cycle [18]. Complications are modeled with risk equations that use characteristics of simulated persons living with diabetes, including HbA1c, to predict the likelihood of a person developing each complication (myocardial infarction, stroke, congestive heart failure, angina [T1DM only], chronic kidney disease, blindness, and neuropathy [T1DM only]) in each cycle [18].

Analyses were conducted separately for populations of persons living with T1DM and T2DM. For persons living with T1DM, complications are modeled using the Sheffield Type 1 Diabetes Model equations, which were primarily derived from the Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications Study and the Wisconsin Epidemiological Study of Diabetic Retinopathy [19,20,21]. The complications of T2DM are modeled using the Risk Equations for Complications Of type 2 Diabetes (RECODe) risk engine, which was developed using data from the Action to Control Cardiovascular Risk in Diabetes (ACCORD) randomized controlled trial (RCT) [22].

The analyses were conducted from a Canadian private payer perspective. Canada has a universal, publicly funded healthcare system which covers medically necessary hospital and physician services and accounts for around 70% of total health expenditure [17, 23]. In addition, two-thirds of Canadians have private health insurance, which is mostly paid for by employers, unions, or other organizations and makes up around 12% of health spending [17, 23]. Out-of-pocket spending accounts for a further 14% of total health expenditure in Canada [23].

For each analysis, a total of 10 000 people living with diabetes were simulated in Microsoft Excel using the DEDUCE model. Time horizons (the durations over which health outcomes and costs are calculated) of 40 years (T1DM) and 25 years (T2DM) were used in order to capture relevant costs and outcomes during the period of typical private payer coverage. Costs (in 2023 Can$) and utilities were discounted at 1.5% (i.e., future costs and health benefits were weighted less heavily than those that occur in the present), as per current Canadian Drugs Agency (CDA) guidelines [24]. Costs and disutilities were applied per event for acute diabetic events and annually for complications.

No ethical approval was required for this analysis, which is based on previously conducted studies and does not contain any new studies with human participants or animals performed by any of the authors. The clinical trials from which data were taken for this analysis were conducted in accordance with the Declaration of Helsinki, with the relevant consents obtained from all study participants.

The modeled population characteristics are shown in Table 1 [11, 25] and Table 2 [7, 22, 26, 27]. The target population of this model was people ≥ 18 years of age living with T1DM or T2DM (with or without insulin use). The T1DM model was populated using the characteristics of persons living with T1DM—age, sex, baseline clinical parameters (HbA1c level, systolic blood pressure, body mass index, and cholesterol levels), the percentage of people who were smokers, and their history of complications—from the IMPACT RCT and the Finnish registry study FinnDiane [11, 25]. At model entry, the mean age for persons living with T1DM was 23 years. For the T2DM analysis, the modeled persons were assumed to be aged 40 years at model entry; sex and ethnicity data were taken from the ACCORD and IMpact of flash glucose Monitoring in pEople with type 2 Diabetes Inadequately controlled with non-insulin Antihyperglycaemic ThErapy (IMMEDIATE) RCTs and from a real-world study conducted in the USA and Canada [7, 22, 26, 27]. Baseline clinical parameters (HbA1c level, systolic blood pressure, cholesterol levels, serum creatinine, and urine albumin:creatinine ratio), the percentage of people with T2DM who were smokers, the percentage who had cardiovascular disease, and data on medication use for persons living with T2DM were based on the IMMEDIATE and ACCORD RCTs and on US real-world data [7, 22, 26]. T2DM treatment was assumed to be 84% non-insulin, 10% basal insulin, and 6% multiple daily injections of insulin (MDI), based on a US retrospective cohort study [26].

The effect of FreeStyle Libre was modeled as an immediate absolute reduction in HbA1c which persisted over the model time horizon (Table 3) [13, 14, 27,28,29,30,31,32,33,34,35,36,37]. HbA1c was reduced by 0.42% and 0.59% for insulin-using persons living with T1DM and T2DM, respectively, based on a meta-analysis of clinical trials and real-world observational studies [28]. For persons living with T2DM not using insulin, a reduction of 0.3% versus SMBG was applied, based on the 16-week Canadian RCT IMMEDIATE [27]; this reduction versus SMBG is consistent with the results seen at week 24 in a Japanese RCT [10].

For people living with T1DM and for those with T2DM treated with insulin, the rates of DKA and SHE with SMBG and FreeStyle Libre were taken from a French real-world study, Real-World Evidence of FreeStyle Libre: Analysis of the SNDS Database in France (RELIEF) (Table 3) [13, 29]. For the T2DM non-insulin group, rates of DKA and SHE for SMBG were based on Danish real-world evidence [30] and a meta-analysis of population-based studies [31], respectively, and were assumed to be reduced with FreeStyle Libre to the same degree as in the T2DM basal insulin group in RELIEF [13, 29]. As in previous FreeStyle Libre models [38, 39], in the absence of specific data, mortality due to DKA and SHE was assumed to be the same with SMBG and FreeStyle Libre [32,33,34]. The rate of NSHE was based on published studies [14, 31, 35] and was conservatively assumed not to be reduced with FreeStyle Libre among persons living with T2DM who were not using insulin.

Work absenteeism was modeled using real-world data (Table 3) [36, 37]. The number of days absent from work due to T1DM or insulin-treated T2DM was taken from FLARE-NL4, a prospective registry study conducted in the Netherlands [36]. Work absence due to T2DM not treated with insulin was derived by applying a 0.85 multiplier for non-insulin vs. insulin, derived from a US real-world study of treatment absence due to T2DM, to data for insulin-treated T2DM from FLARE-NL4 [36, 37].

Long-term absence was modeled based on persons living with diabetes not returning to work after diabetes complications (Table 4) [40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55]. The proportion of persons living with diabetes not able to return to work due to blindness was inferred from the gap between labor force participation among blind vs. non-blind Canadians (46.2% vs. 80.2%) [40]. For congestive heart failure, the proportion not returning to work was assumed to be the same as that observed in a German study of 220 people hospitalized with left heart failure and viral myocarditis [41]. The corresponding proportion for myocardial infarction was taken from the international (USA, Spain, and Australia) Variation In Recovery: Role of Gender on Outcomes of young acute myocardial infarction patients (VIRGO) study, which included data for 1680 people aged 18–55 years hospitalized with myocardial infarction [42]. The ability of people with renal failure to work was taken from a Canadian study which included all employed Canadians aged 40–64 years who had end-stage renal disease [43]. The proportions of persons returning to work after a stroke or amputation were taken from literature reviews [44, 45].

Annual costs were taken from Canadian sources and included FreeStyle Libre and SMBG acquisition costs, the costs of treating diabetes complications, and the costs of absenteeism and long-term absence (Table 4) [46,47,48,49,50,51,52,53,54,55]. For each complication, costs were derived from Canadian real-world data identified in a targeted literature search and incorporated outpatient medical costs, rehabilitation, and medication costs [46,47,48,49,50,51]. Hospitalization costs, which would not be relevant to Canadian private payers, were not included. Similarly, acute diabetic events, which typically require transitory hospital treatment but no ongoing outpatient care or medication, were assumed not to have any cost to private payers. Absenteeism costs and the costs of long-term absence were based on the Statistics Canada average wage of Can$31.90 per hour [52]. The cost of long-term absence was conservatively applied only for 1 year after the occurrence of the relevant complication.

Health outcomes were measured as quality-adjusted life years (QALYs). In the absence of Canada-specific data, baseline health utility values (0.839 for T1DM and 0.785 for T2DM) were taken from two UK prospective studies (Supplementary Material, Table S1) [25, 35, 56,57,58,59,60,61,62,63,64].

Disutilities associated with complications, DKA, SHEs, and NSHEs were taken from the literature (Supplementary Material, Table S1) [25, 35, 56,57,58,