The type of cost-effectiveness analysis used was a cost-utility analysis. Degludec was compared with glargine U100 in three patient populations:
T1DM using basal-bolus therapy (T1DM)
T2DM using basal oral therapy (T2DMBOT)
T2DM using basal-bolus therapy (T2DMB/B)
Glargine U100 was identified as the most appropriate comparator for the cost-utility analysis. Treatment guidelines in Serbia recommend NPH insulin as the first-line insulin treatment for T1DM and T2DM, with a basal insulin analogue considered if after 6 months on NPH there is persisting hypoglycaemia (< 3.5 mmol/l). Thus, degludec should be considered where a basal insulin analogue is indicated.
The model was a simple, transparent, short-term (1-year time horizon) model developed in Microsoft Excel 2010 (Microsoft Corp., Redmond, WA, US) and has been previously published . The short-term model focuses on the impact of important aspects of insulin therapy, such as hypoglycaemia and dosing, and accommodates the treat-to-target trials required by the Food and Drug Administration (FDA) . In treat-to-target trials patients are treated to the same glycaemic target and no differences in HbA1c are expected; thus, there is no rationale for long-term modelling based on HbA1c differences.
Although cost-effectiveness was analysed in a 1-year setting and is based on data from 1-year clinical trials, the model can be replicated for subsequent years, and the outcomes represent the average annual cost-effectiveness in steady state. As the time horizon was 1-year, no discounting was applied.
The model calculated the costs associated with treatment (insulin, needles and SMBG test strips) and hypoglycaemic events (the cost of treating the event and the use of additional SMBG test strips associated with the event). Health-related quality of life (in the form of QALYs) was calculated by applying a disutility per hypoglycaemic event incurred (Fig. 1).
All costs were estimated from the healthcare payer perspective, RFZO. The threshold was set at 3 × GDP per capita, as although there are alternative approaches to thresholds for cost-effectiveness of interventions , this is an internationally recognised level  and is accepted by the Serbian authorities. According to the World Bank statistics for 2016, the GDP per capita in Serbia was 5348.30 USD. Using the exchange rate of 1 USD = 111.29 RSD, the threshold was set at the level of 1,785,642 RSD.
Data Used in the Model
All costs were estimated from the healthcare payer perspective, RFZO.
Direct Treatment Costs
Costs of insulin, needles and SMGB tests were based on official RFZO prices elicited in November 2017. Other costs of treatment (e.g. use of concomitant medication) or other costs resulting from treatment (e.g. long-term outcomes) were assumed to be equivalent in both treatment groups and were therefore not included.
Cost of Hypoglycaemic Events
Resource use associated with non-severe and severe hypoglycaemic events was derived from the clinical trial data. For non-severe events data were obtained from the patient-completed hypoglycaemia safety questionnaire, and for severe events data were obtained from the serious adverse events case reports .
The use of additional SMBG tests in the week following a non-severe event was also based on the hypoglycaemia safety questionnaire. Data were not collected on the testing pattern following a severe event; therefore, it was conservatively assumed to be similar to a non-severe hypoglycaemic event.
For patients experiencing severe events it was assumed that all patients who were hospitalised used the ambulance service and that all patients experiencing severe events used glucagon to recover.
Patients suffering a hypoglycaemic event received the same treatment regardless of whether they were on degludec or glargine U100. Therefore, any difference in costs of hypoglycaemia between treatments was due to differences in rates of hypoglycaemia and not the cost per event.
The average costs of hypoglycaemic events are calculated by multiplying the unit cost of the services by the share of patients using that treatment/service (Table 1). It was assumed that a non-severe event would be self-managed and not require any additional resource use; thus, costs relate only to additional SMBG testing.
The daily insulin dose for the degludec and glargine U100 treatment groups was based on the end of trial doses captured from the clinical trial data. The meta-analysis of insulin dose from the clinical trials was the source of the glargine U100 dose and degludec/glargine U100 dose ratio . The degludec dose was calculated using the dose ratio to allow for adjustment of covariate factors such as trial, treatment, antidiabetic therapy at screening, age, sex, region and baseline dose (Table 2).
It was assumed that one SMBG test was conducted with every main meal for regimens including bolus insulin (T1DMB/B, T2DMB/B). Three meals per day were assumed.
For SMBG tests related to basal insulin injections, the titration schedule recommended for use with glargine U100  was used to estimate SMBG utilisation for glargine U100. The algorithm recommends seven SMBG tests per week (the dose is adjusted every 3rd day based on the mean of the FPG results over three consecutive days) .
Degludec enables patients to titrate, predict and monitor their blood glucose more efficiently [13, 14]. The recommended titration algorithm for degludec is once-weekly adjustment of dose based on the average of two SMBG measurements from the 2 preceding days for patients with T2DM . Thus, the utilisation of SMBG tests related to degludec was assumed to be two tests per week for patients with T2DM.
Hypoglycaemia Event Rates
Real-world hypoglycaemic event rates from a large-scale questionnaire-based study conducted in seven European countries  were used as the baseline values for severe and non-severe hypoglycaemic events. These rates provide a better estimation of real-life event rates than those from clinical trials , which can be biased in both the selection of patients and the treatment setting. For example, the clinical trials excluded patients with a history of severe hypoglycaemic episodes and anyone considered hypoglycaemia unaware [34,35,36,37,38,39]. The real-world rates were used as the base case event rates in the glargine U100 group (as this is the current treatment on the market); see Table 3. Event rates for the degludec group (Table 3) were calculated using the relative event rates taken from the meta-analysis of hypoglycaemia [20, 21], which were adjusted for trial, type of diabetes, treatment, anti-diabetic therapy at screening, sex and region as fixed factors and age as a continuous covariate. Only rate ratios with a documented statistically significant difference between the treatment arms were used.
Compliance with Ethics Guidelines
This article does not contain any studies with human participants or animals performed by any of the authors.
QALYs were calculated by subtracting a disutility  per hypoglycaemic event experienced from the baseline health utility. Disutilities associated with hypoglycaemic events were obtained from a large-scale TTO study , which reported a disutility of 0.0565 for a severe event (with no significant difference between daytime and nocturnal severe events) and disutilities of 0.0041 and 0.0067 for non-severe daytime and non-severe nocturnal events, respectively (a significant difference in utility was demonstrated for nocturnal vs. daytime non-severe events) . The disutility per hypoglycaemic event was multiplied by the number of events observed in each treatment group.
In addition, the analysis included an estimate of the utility benefit for the option of flexible dosing time with degludec. Boye et al.  reported a utility benefit of 0.006 associated with dosing flexibility; therefore, for degludec an extra utility gain of + 0.006 was applied to the QALY benefit.
One-way and probabilistic sensitivity analyses were conducted to assess the impact of varying key assumptions and outcomes used in the base case analysis (Table 4).