A cost-utility analysis of long-acting insulin analogues (detemir, glargine and degludec) for the treatment of adult type 1 diabetes in South Africa

Abstract

Background

Type 1 Diabetes Mellitus (T1DM) is a life-threatening condition that is managed with administered insulin. Intermediate- to long-acting insulin represents the basal insulin constituent of the total insulin used in treating T1DM. In South Africa, intermediate-acting Neutral Protamine Hagedorn (NPH) insulin has been the mainstay basal insulin recommended in the public sector, despite the availability of newer (ultra) long-acting insulin analogues. A cost-utility analysis of the newer long-acting insulin analogues insulins degludec, glargine U100, glargine U300 and detemir in comparison to current practice (NPH insulin) has yet to be performed in the South African public health sector context.

Methods

A cost-utility analysis was carried out utilising Markov modelling. Long-acting insulins degludec, glargine and detemir were compared to NPH insulin in the model. For each comparator, two Markov states were created, one in which no complications occurred and another representing severe nocturnal hypoglycaemic events. Quality-Adjusted Life Years (QALYs) gained per patient year was the health outcome assessed over a one-year time horizon.

Results

NPH insulin was the least costly and least effective; while Determir and Glargine U100 were extended and absolutely dominated respectively. The ICER for Glargine U300 in comparison to NPH was USD 40,104.91 per QALY gained, while Degludec was USD 64,831.20 per QALY gained in comparison to Glargine U300.

Conclusions

The ICERs of long acting insulins were considerably higher than South Africa’s indicative cost-effectiveness threshold. The status quo of NPH insulin in the management of T1DM in adults remains the most cost-effective option for the South African public health sector.

Background

The United Nations Sustainable Development Goal (SDG) target 3.4 aims to reduce premature mortality attributable to non-communicable diseases (NCDs) by one third by 2030, compared to 2015 estimates [1]. One of the four priority NCDs to be targeted by world leaders is diabetes mellitus [2].

There is an appreciable growing global trend in Disability Adjusted Life Years (DALYs) attributed to diabetes, which is also realised in sub-Saharan Africa, where diabetes was responsible for 623 DALYs per 100,000 population in 2017, representing an 8.4% increase from 1990 [3, 4]. In South Africa specifically, it is estimated that premature adult deaths attributed to diabetes have increased by 38% between 1999 and 2006 [5, 6]. One of the major subtypes of diabetes is Type 1 Diabetes Mellitus (T1DM). The estimated burden of T1DM in South Africa as of 2019 is approximately 130,000 for the adult population above 20 years of age, which equates to a prevalence of about 3.7 per 1000 individuals [7].

The disease (T1DM) is hallmarked by high levels of blood glucose secondary to insulin deficiency that is fatal when left untreated. Insulin promotes the uptake of glucose from the vascular space into the intracellular space, where glucose is utilized to fuel the functions of the cell. Under normal physiology, a basal level of insulin circulates in the bloodstream with additional spikes of insulin corresponding to meals. The treatment of T1DM, therefore, aims to imitate physiological levels and responses of insulin in inter-prandial and prandial states. With insulin use, acute and chronic complications of T1DM, including diabetic ketoacidosis and micro- and macrovascular complications, can be averted. Exogenous insulin, however, can result in severe adverse drug events like hypoglycaemia, a medical emergency and potentially life-threatening condition characterised by low blood glucose and associated symptoms. Severe hypoglycaemia in particular is an important adverse drug event. Its significance, however, is complicated by disparities in the definition used in studies of the condition. Definitions range from serologically confirmed hypoglycaemia (< 2 mmol/L) to hypoglycaemia requiring third-party assistance (ranging from non-medically trained individuals to hospital admission) and has subsequently received further research interest [8].

In South Africa’s public health care system, Neutral Protamine Hagedorn (NPH) insulin, an intermediate-acting insulin, has been used to replicate basal levels of inter-prandial insulin, and is used in conjunction with rapid-acting insulins in order to achieve physiologically comparable insulin levels in adult T1DM patients. In addition to NPH and rapid-acting insulins, (ultra-) long-acting insulin analogues are also registered for use in South Africa [9], but these drugs are not included in South Africa’s Essential Medicines List nor recommended for use in South Africa’s public health care system [10].

The cost-effectiveness of the newer (ultra-) long-acting insulin analogues has been assessed by many health technology assessment agencies outside of South Africa [11,12,13,14,15] and in individual economic studies [16,17,18,19,20,21,22,23], but no evaluation has been done for the South African adult T1DM population. With this study, we explored the cost-effectiveness of newer long-acting insulins (detemir, glargine and degludec) compared to intermediate-acting NPH insulin in adults with type 1 diabetes mellitus in South Africa.

Methodology

Study design

We conducted a modelled cost-utility analysis (CUA) of long-acting insulins degludec, glargine U100, glargine U300 and detemir in comparison to the status quo of NPH insulin. Our overall study design was based on the South African guidelines for Health Technology Assessment, which are used to guide the selection of medicines for the public health sector [24]. The scope of costs included the full costs associated with the change in insulin from a public sector provider’s perspective, expressed in 2020/21 prices and converted to United States Dollar (USD) using the average South African Rand (ZAR) to USD exchange rate over the same period (16.50 ZAR/USD) [25]. Outcomes were expressed as Quality Adjusted Life Years (QALYs). Analysis was undertaken in Microsoft Excel and TreeAge Pro (2021) [26].

Incremental cost-effectiveness ratios (ICERs) were calculated by ordering each insulin option from least to most costly. After excluding any options with extended or absolute dominance, ICERs were calculated as the difference in costs divided by the difference in outcomes (QALYs) between adjacent options. The resulting ICERs were then compared to the marginal productivity of health spending in the South African public sector which estimates that approximately USD 3012 of marginal spending will avert one DALY [27]. Given the lack of an agreed cost effectiveness threshold (CET) for South Africa, this threshold was used as an indicative CET to interpret QALY based ICERs.

Clinical evidence

A comprehensive literature search for clinical literature was performed to inform clinical parameters and model structure (Appendix 1). Our search identified 209 systematic reviews of which three were selected for further review. Of these three, the National Institute for Health and Care Excellence (NICE) review (2021) [28] was selected to form the basis of the evidence on clinical effectiveness for this current paper, given that it most closely matched our decision problem, it included relevant primary studies, and was judged to be of high quality (see Appendix 2). In evaluating the evidence it was judged that the only significant clinical difference between insulins was rates of severe nocturnal hypoglycaemia, as shown by the summary of primary and secondary endpoints between the three systematic reviews evaluated (Appendix 3).

Modelling approach

Using a Markov modelling approach, and based on the clinical findings of the (NICE) review (2021) [28], we developed models for each alternative insulin with two Markov states: T1DM with no complications and T1DM with severe nocturnal hypoglycaemic events. As mentioned, given the absence of significant differences, we did not model death, adopted a time horizon of one year and therefore did not discount results. Because insulin dosage is weight dependent, we modelled an adult T1DM population with a mean weight of 81.2 kg (SD 17.8) [29].

Our modelling approach therefore required data on rates of severe nocturnal hypoglycaemia, insulin utilisation by weight and associated costs, inpatient admission and costs, and health related quality of life; appropriate to each comparator and Markov state. These data are summarized in Tables 1 and 2. Rates of severe hypoglycaemia for different insulin types and regimens were drawn from the NICE economic model [15], which accompanied the previously mentioned review [28]. These data allowed for the generation of base case, lower and upper values, as well as for an estimation of the proportion of severe hypoglycaemia occurring at night [28].

Table 1 Model inputs: rates of severe nocturnal hypoglycaemia per 100 person years [15]

Table 2 Model inputs: insulin utilisation, inpatient admission, unit costs and health related quality of life

In terms of insulin utilisation, the total recommended daily dose ranges between 0.5 and 1 unit/kg body weight, of which 33–50% consists of the intermediate-acting bedtime basal dose and the remainder is divided into three pre-meal short-acting insulin doses. Although patient and clinician practice may differ, we followed relevant South African Standard Treatment Guidelines complemented where necessary by the recommendations from the U.S. Food and Drug Administration [10, 30,31,32,33]. The analysis only includes once daily regimens as this is recommended in the South African public health sector. Unit costs of insulin were derived from the South African Master Health Product List (MHPL) of October 2020 which lists public sector prices [34]. Because not all insulins are available in the public sector, we also used prices from the Medicines Price Registry (MPR) of December 2020 which provides private sector prices. The cheapest formulation per unit of insulin (10 ml (100 U/ml) injection formulation was selected as the cost input [35]. Where the injection formulation was not available or does not exist, the next cheapest formulation was used, namely the cost per 5 pack of 3 ml (100 U/ml) cartridges of insulin. The insulin cost per patient per year was thus calculated as a function of weight, total daily dose of insulin, proportion of total daily dose of insulin that is basal insulin, cost per unit of insulin and the number of days in a year.

Given that the scope of costs for this study includes any potential changes in provider costs owing to a change in insulin, we included admissions for severe nocturnal hypoglycaemic events based on a study of T1DM patients in Spain, Germany and the United Kingdom (UK) [36]. We assumed utilisation rates from the UK cohort as these were lowest. The cost per inpatient day was taken from the District Health Barometer (2019/2020) which summarizes these costs across all public sector hospitals in South Africa [37] (see Table 2).

Finally, the Health-Related Quality of Life (HRQoL) of T1DM patients with no complications (0.844) or with severe nocturnal hypoglycaemic events (0.77) was derived from a study by Evans et al. (2013). When expressed as disutilities, these values were similar to that found in other available literature [38,39,40].

Sensitivity analysis

Probabilistic sensitivity analyses with 10,000 iterations were performed using the ranges on certain model inputs as shown in Tables 2 and 3. In addition, deterministic sensitivity analysis included a threshold analysis on insulin prices (solving for cost neutrality) as well as a 40% price reduction on private sector medicine prices to approximate public sector prices, as recommended by HTA guidance for South Africa [24].

Table 3 Cost-utility results (USD, 2020–2021 prices)

Results

Cost-utility analysis

Base case

Base case economic evaluation results are presented in Table 3. NPH insulin was the least costly and least effective; while Determir and Glargine U100 were extended and absolutely dominated respectively. The ICER for Glargine U300 in comparison to NPH was USD 40,104.91 per QALY gained, while Degludec was USD 64,831.20 per QALY gained in comparison to Glargine U300. These ICERs are higher than a South African indicative CET of USD 3012.00 per DALY averted.

Sensitivity analysis

Results of the probabilistic sensitivity analyses are presented as scatter plots in Figs. 1 and 2. These are provided for Glargine U300 in comparison to NPH and for Degludec in comparison to Glargine U300 given that these are the only undominated alternatives. As shown, all model runs indicate positive incremental costs; however, there are many model runs indicating negative incremental effects, indicating the uncertainty regarding whether the long acting insulins are more effective than NPH. In terms of deterministic sensitivity analysis, the results of the 40% price reduction on private sector prices is shown in Table 4. Insulin glargine U300 remained the most cost-effective alternative to NPH with an ICER of USD 20,898.47 per QALY gained. Insulin degludec followed, with an ICER of USD 238,172.30 per QALY gained. Insulin detemir was extended dominated whilst Glargine U100 remained absolutely dominated, as expected, as its price was not affected by the price reduction since it was drawn from the Master Health Product List.

Fig. 1

Probabilistic sensitivity analysis results Glargine U300 vs NPH (USD, 2021 prices)

Fig. 2

Probabilistic sensitivity analysis results Degludec vs Glargine U300 (USD, 2021 prices)

Table 4 Cost-effectiveness analysis results with 40% reduction in single exit price applied (USD, 2020–2021 prices)

The price threshold analysis (see Fig. 3) shows that significant price reductions for the long acting insulins would be needed to achieve cost neutrality to NPH.

Fig. 3

Price threshold analysis at which insulin acquisition price results in cost neutrality

A cost-breakdown analysis shows that the pharmaceutical costs of NPH insulin once-daily are about USD 18 per patient per year, compared to insulin degludec, which costs about USD 232 (see Fig. 4). The costs incurred by the health system (i.e. inpatient care to manage adverse events) are about USD 14 for NPH insulin, which is higher than for insulin degludec, at about USD 6 per patient per year (Fig. 4). This is due to the relatively lower risk of severe hypoglycaemia assumed for insulin degludec in the model, which results in an estimated 0.05 fewer severe hypoglycaemic episodes per patient per year.

Fig. 4

Cost breakdown of each insulin per patient per year (USD, 2020–2021 prices)

Discussion

This study aimed to inform the use of intermediate- and long-acting acting insulins in the management of adults with T1DM in South Africa by means of cost-utility analysis modelling.

After review of the available evidence, it was concluded that the insulins under review have few clinically significant differences. The evidence included, derived from systematic review of best available evidence, suggests no significant difference in key outcomes, including all-cause mortality and HbA1c control. Differences in reductions in harm, particularly hypoglycaemia, however, were noted in the NICE NMA [28].

The findings from this study indicate that the long-acting insulin analogues detemir, glargine U100, glargine U300 and degludec are not cost-effective compared to NPH insulin, with insulin detemir being extended dominated and insulin glargine U100 being absolutely dominated. Insulin glargine U300, followed by insulin degludec, proved to be the next cost-effective options in both the main and sensitivity analyses. The cost breakdown suggests that the primary driver of costs for the insulins under review are pharmaceutical costs, as opposed to hospitalisation costs. The point of cost-neutrality sits well below the current pricing of the long-acting insulin analogues.

Therefore, the recommendations from this analysis favour NPH insulin over the newer long-acting insulin analogues. This recommendation is in line with a recent clinical practice guideline recommendation made by the World Health Organization, that draws its evidence from a review by Tricco et al. [41, 42].

Disparities in recommendations from other economic evaluations, however, do exist and depend on study setting, insulin pricing, clinical effectiveness assumptions and the cost-effectiveness threshold used. For example, the recommendation of this study differs with the NICE NG17 clinical practice guideline recommendations for the United Kingdom that was based on evidence generated using many of the clinical efficacy and economic parameters utilised in this analysis [15]. A German economic evaluation found insulin glargine to be the most cost-effective option [21]. In addition, some studies comparing insulins degludec to glargine favoured insulin degludec, which was also not replicated in this review [11, 16].

In this economic evaluation, limitations arise from the input variables identified, some of which were drawn from small studies conducted in a setting not necessarily similar to South Africa. This includes the estimates of hospitalization following severe nocturnal hypoglycaemic events, which are likely overestimates of hospitalization rates in South Africa. The exclusion of long term sequalae of diabetes such as HbA1c and all-cause mortality, on the grounds of available evidence selected, is a major limitation as slight differences in these outcomes may drastically impact cost-effectiveness. Furthermore, given the lack of an agreed-upon CET for South Africa, the ICERs are interpreted in relation to an indicative CET. However, given the large difference in ICERs in comparison to CET, the overall interpretation of cost-effectiveness findings is unlikely to change.

Africa and South Africa’s diabetic population have demonstrated varying utilisation rates of insulin, making it difficult for the model to accurately reflect diabetic care. Regarding dosing and frequency, large multicentre study based on a German/Austrian database concluded that, on average, NPH insulin was used 1.9 times a day for T1DM patients. The study further identified that the mean adjusted daily NPH dose was 0.36 IU/kg, lower than the recommended dosing of 0.5–1 IU/kg outlined by the standard treatment guidelines [10, 43]. However, a smaller study, by Mbanya et al., conducted across the African continent, which included 49 patients with T1DM in South Africa, found that on average, across all countries, total daily dose of basal and postprandial insulin amounted to 0.78 IU/kg [44]. In a South African study by Sehloho and van Zyl, the authors found that the mean basal and postprandial insulin dose was 59.8 (standard deviation of 36.7) IU/day [45].

Furthermore, patient practices may vary in the insulin regimen used. Despite the recommendation by the South African EML that the basal bolus regimen be employed in the management of T1DM, a small South African study by Sehloho and van Zyl found that a mere 15.2% of participants used the basal bolus regimen whereas the remainder used the alternative, namely the biphasic regimen [45]. This result was not replicated in the study by Mbanya et al., where the authors found that roughly 33% of T1DM included in the study utilized the biphasic insulin regimen. The alternative treatment regimen, the biphasic regimen, consists of twice daily “pre-mixed” or biphasic insulin, which is a mixture of intermediate- and short-acting insulin.

Although the literature suggests that the recommended treatment regimen, dosing and frequency is most often not utilized by patients, similar total amounts of basal insulin are likely to be utilised across both the recommended and alternative treatment regimens. There may be significant differences between the two regimens with regards to clinical outcomes, which should be a topic for further research. Furthermore, research investigating long-term outcomes of diabetic control between different long-acting insulins in a local context is needed to fully inform clinical practice.

Conclusion

This review has highlighted the need for robust clinical effectiveness evidence in economic models. This review recommends intermediate-acting NPH insulin, the status quo, over long-acting insulins as the basal insulin in the management of adult T1DM patients in South Africa receiving care in the public health sector.

Appendix 1: Literature search

Description of systematic reviews selected for inclusion

From the database search conducted on 8 March 2022, 208 systematic reviews were identified for screening. One additional systematic review was identified through checking reference lists of eligible reviews and clinical guidelines. After duplicate removal, and title and abstract review, 39 systematic reviews were selected for full text review. Of these, 8 eligible systematic reviews were identified for further study composition comparison, from which three systematic reviews were selected for inclusion in this report.

The three studies were selected as they most closely aligned with the review question, included relevant primary studies, were published recently (in past two years), and were conducted using transparent methods. The three studies included were a Cochrane review by Hemmingsen et al., a review conducted by NICE to inform a clinical guideline recommendation, and a review conducted by Martinson et al. [5,6,7]. The quality of these three systematic reviews were appraised (in duplicate) using the AMSTAR tool and were found to be of good quality (Fig. 5).

Fig. 5

PRISMA flow diagram

Appendix 2: Study composition analysis

Description of systematic reviews

We identified three up to date systematic reviews and meta-analyses of good quality that compared the efficacy and safety of insulin glargine, insulin detemir, insulin degludec, and NPH. An overview of the three systematic reviews is presented in Table 5, and the characteristics of all studies included in each reported in Table 6.

Table 5 Overview of systematic reviews

Table 6 Change in HbA1c as reported in the Cochrane evidence review

Appendix 3: Literature review findings and interpretation

All three reviews identified little difference between the clinical effectiveness of long-acting insulin analogues detemir, glargine and degludec compared to one another and compared to NPH insulin.

These insulins were shown to be noninferior with regards to primary outcomes HbA1c and mortality risk.

HbA1c

All three reviews found no significant differences in HbA1c control between long-acting insulin analogues and NPH insulin (Tables 7, 8).

Table 7 Change in HbA1c as reported in the evidence review by Martin et al.

Table 8 Change in HbA1c as reported in the NICE evidence review

Mortality

The Cochrane review was the only review to report on differences in mortality. No significant differences were found (Table 9).

Table 9 Cochrane review all-cause mortality results

Quality of life

Health-related quality of life was shown to lack sufficient data or not reported on.