FormalPara Key Summary Points

Utility values that quantify the effect of acute and long-term complications on quality of life are important for health economic modeling analyses of type 2 diabetes

A literature review was performed to identify utility values associated with type 2 diabetes-related complications (e.g., cardiovascular complications, stroke, and renal disease)

Searches were performed in multiple databases to identify articles published since 2010

Fifty-four articles were identified that included utility values associated with type 2 diabetes-related complications

Introduction

Health state utility values provide a quantitative indication of the extent to which a particular disease, complication or treatment-related side effect influences the quality of life (QoL) of an individual. Utility values quantify QoL on a scale of 0–1, where 1 represents perfect health and 0 represents death. Some utility elicitation methods, such as the EuroQoL 5 Dimensions (EQ-5D), allow negative values, which correspond to states considered to be worse than death [1, 2]. Utility values are key inputs in health economic models that are used to project the long-term clinical and economic outcomes associated with new treatments and upon which payer and reimbursement decisions are often based. It is therefore important that these values accurately represent the QoL decrement associated with specific health states.

Individuals with type 2 diabetes (T2D) are at an elevated risk for a number of acute and long-term complications including cardiovascular disease, stroke, ophthalmic complications, renal disease, diabetic foot, neuropathy and hypoglycemic events, many of which can have a considerable impact on QoL. Poor glycemic control exacerbates the risk for long-term complications, and pharmacologic treatments aimed at improving glycemic control can in turn reduce the risk for long-term complications. However, the duration of clinical trials is not sufficient to capture the effect of new treatments on the incidence of long-term complications. Consequently, long-term health economic modeling analyses represent an important component of the reimbursement decision-making process for payers and policy makers.

The International Society for Pharmacoeconomics and Outcomes Research (ISPOR) good practice recommendations state that the use of up-to-date QoL data is an important aspect of long-term health economic analyses but also note that, in practice, many economic models use outdated utility values that may not adequately capture recent advances in treatment [3]. A previous review published by Beaudet et al. collated utility values for T2D-related complications, published over the period 1995–2012 [4]. The aim of the current review was to update and expand the previously published review by Beaudet et al. to provide a synopsis of utility values that could be utilized in future health economic models of T2D. The scope of the review was expanded to capture utility values elicited by the EQ-5D and other methods including (but not limited to) the Short Form 6 Dimensions (SF-6D), 15D, Health Utilities Index-2 or -3 (HUI-2/3), time-trade-off (TTO) and standard gamble (SG) and also to include utility/disutility values related to treatment-related attributes such as dosing frequency and timing flexibility, injection device-related attributes and unpleasant treatment-related side effects such as nausea or weight gain. The time frame of the review was also updated such that the review was limited to studies published since 2010, thereby capturing contemporary clinical practice relating to the management of people with T2D. Only findings related to T2D-related complications are presented here; summary findings of utilities for treatment-related attributes are presented separately in Part 2 of this review.

Methods

The literature review was performed using the PubMed, Embase and Cochrane Library databases. Search strategies were designed in alignment with recommendations outlined in the UK-based National Institute for Health and Care Excellence (NICE) Decision Support Unit (DSU) Technical Support Document 9 [5]. Search strategies utilized high level Medical Subject Heading (MeSH) terms supplemented with free-text terms, and search syntax was adjusted as required for use across the different databases (full details of the search strategies used are provided in Supplementary Material, Tables 1–3). Supplementary hand searches were also performed to identify pertinent studies presented at major congresses between late 2019 and 2023( specifically the virtual meeting of the American Diabetes Association [ADA], the ISPOR Annual Congress and the 55th annual meeting of the European Association for the Study of Diabetes [EASD] in 2019). Relevant abstracts presented at the 2019 ISPOR meeting have been published; therefore, relevant publications should have been captured within the literature database searches. Studies published only in abstract form prior to 2019 were excluded on the basis that study results were likely to have been subsequently published in full-text form.

The time horizon of the searches was limited to articles published since 2010, and all searches were performed in March 2020. For inclusion in the review, studies were required to be published in full-text form (except for recent abstracts as outlined above) in English and present utility or disutility values for health states related to acute or long-term T2D-related complications or treatment-related attributes or process characteristics. Complications captured in the review included cardiovascular disease (angina, myocardial infarction [MI], and congestive heart failure [CHF]), stroke, renal disease (albuminuria/proteinuria, end-stage renal disease [ESRD] and renal transplant), ophthalmic complications (retinopathy, macular edema, cataract and severe vision loss/blindness), neuropathy, diabetic foot, amputation, peripheral vascular disease, overweight/obesity, hypoglycemia and fear of hypoglycemia (FoH). Studies that were conducted in mixed populations of patients with type 1 and type 2 diabetes were excluded if results were not presented according to diabetes type. Secondary studies (i.e., studies listing previously published utility values), and discrete choice experiments were also excluded.

Here, reporting of results is limited to acute and long-term complications; utility values for treatment-related attributes and process characteristics are reported in a separate review [6].

This article is based on previously conducted studies and does not contain any new studies with human participants or animals performed by any of the authors.

Results

Literature Searches

Literature searches across the three databases yielded a total of 8566 hits, which included 1383 duplicates, therefore resulting in a total of 7183 unique hits. First-round screening of titles and abstracts was performed by one investigator and identified a total of 241 hits for full-text review (Fig. 1). During second-round screening, a further 176 articles were excluded, leaving a total of 65 articles detailing utilities/disutilities associated with either T2D-related complications or treatment-related attributes for inclusion. A further three articles were identified via bibliographies of included articles. Searches of meeting abstract databases identified one relevant abstract for inclusion. The final review therefore included a total of 69 studies. Of these, a total of 39 presented findings exclusively related to acute or long-term diabetes-related complications, 15 presented findings exclusively related to the influence of treatment-related attributes on QoL, and 15 captured findings on both complications and treatment attributes. In total, 54 articles presented utility/disutility values for T2D-related complications (Table 1); these included a total of 18 studies conducted in Asia [7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24], 13 conducted in Europe [25,26,27,28,29,30,31,32,33,34,35,36,37], 13 in North America [38,39,40,41,42,43,44,45,46,47,48,49,50], 2 in Latin America [51, 52], one in the Middle East [53] and 1 in Africa [54], and 6 were multinational (or setting not stated) [55,56,57,58,59,60]. Nearly all identified studies were conducted in individuals with T2D, although two presented data gathered from general population samples [34, 56].

Fig. 1
figure 1

Summary of literature searches. T2D, type 2 diabetes. Publication/study type refers to articles that were reviews, editorials, letters, case reports, secondary sources of utility values or discrete choice experiments

Full size image
Table 1 Summary of studies presenting utility/disutility values for complications related to T2D
Full size table

T2D with No Complications

Eleven studies were identified that provided baseline utility values for individuals with T2D with no complications, all of which used the EQ-5D-3L or EQ-5D-5L (where studies did not state which version of the EQ-5D was used it was assumed that the EQ-5D-3L was used) (Table 2) [7, 14, 16, 22, 24, 28, 35, 36, 38, 44, 49] with one additionally using the SF-6D [38]. Five of the identified studies were conducted in Asia [7, 14, 16, 22, 24], three in North America [38, 44, 49] and three in European populations [28, 35, 36].

Table 2 Utility values associated with type 2 diabetes and no complications
Full size table

Using the EQ-5D, reported mean baseline values for individuals with T2D with no complications ranged from a low of 0.76 in a community-dwelling sample in Canada aged ≥ 40 years and with HbA1c ≥ 7% [44] to a high of 0.956 in a sample in China with mean (SD) age of 64.9 (9.1) years [14]. One study conducted in Vietnam reported a median baseline value of 1.0, which corresponds to perfect health [16], and is also slightly higher than that previously reported in a general population sample in Vietnam [61]. Differences in baseline characteristics such as age, BMI and duration of disease as well as the source of utility weights may have contributed towards the differences in baseline values between the different study populations.

For economic modeling analyses, the optimal baseline utility values should be as closely representative of the simulated patient population under investigation as possible. In particular, factors such as setting, baseline age and duration of diabetes should be taken into account and, if necessary, adjustments applied to the reference population. The description of the reference population of patients with no complications varied between studies; however, one US-based study provided a comprehensive description. Zhang et al. [49] reported a baseline utility value of 0.92 for no complications with this value referring specifically to individuals that were male, non-Hispanic white, with BMI < 30 kg/m2 and with no risk factors for cardiovascular disease, with income > USD 40,000 per annum and not treated with insulin.

Cardiovascular and Cerebrovascular Complications

A total of ten studies that presented disutility values for MI [10, 26, 28, 33, 37, 44, 46, 55, 57, 58], were identified (Table 3) (three studies presented utility values for MI; data not shown [11, 36, 46]). Two further studies presented disutilities for angina [10, 46], and nine studies were identified that presented disutility values for CHF [10, 27, 28, 37, 46, 49, 55, 57, 58]. In studies that utilized the EQ-5D, the reported disutility associated with an MI ranged from − 0.073 in a sample of individuals with T2D attending outpatient clinics in South Korea [10] to − 0.0119 in a sample of patients in Sweden [28]. One study also reported a positive value of + 0.004 in a post hoc analysis of the multinational LEADER trial [58]. However, this value was not an event-related disutility per se but specifically referred to the change in utility value reported in patients who experienced an MI at any point during the 36-month follow-up period rather than a recent MI and may therefore potentially have captured improvements in QoL in patients recovering from an MI. Indeed, in a post hoc analysis of ACCORD trial data, Shao et al. noted that the timing of the event relative to the timing of measurement of QoL was an important determinant of QoL and although the effect of an MI waned over time a degree of long-term impairment persisted [46]. More specifically, using the HUI-3, Shao et al. reported that the mean (SE) disutility associated with MI was − 0.042 (0.016) in the year of the event but − 0.011 (0.006) in subsequent years after the event.

Table 3 Disutility values associated with cardiovascular complications
Full size table

Two studies reported disutility values associated with angina (two studies also reported utility values for patients with angina; data not shown [36, 46]). In a South Korean study that utilized the EQ-5D, Lee et al. reported a mean (SE) disutility of − 0.0266 (0.0114) [10], while for US-based patients, Shao et al., who utilized the HUI-3, reported a disutility of − 0.032 (0.006) in the year of event but − 0.010 (0.021) in subsequent years [46].

A total of nine studies reported disutility values for CHF [10, 27, 28, 37, 46, 49, 55, 57, 58], eight of which used the EQ-5D [10, 27, 28, 37, 49, 55, 57, 58], with the remaining study using the HUI-3 [46] (four studies presented utility values for patients with CHF; data not shown [11, 27, 36, 46]). In studies using the EQ-5D, the mean disutility for CHF ranged from − 0.0821 in Sweden [28] to − 0.037 in a study conducted in Greece [37]. Additionally, a US-based study that used the HUI-3 reported a disutility of − 0.089 in the year of diagnosis and − 0.041 in subsequent years [46].

The interpretation of the QoL impact of stroke was complicated by the heterogeneity in the definition of stroke used between different studies. In two studies, separate disutility values were presented for stroke and transient ischemic attack (TIA) [10, 46] and one study further separated stroke into events with and without hemiplegia [46]. In comparison, other studies either did not state how stroke was defined or grouped TIA and stroke together. A total of 14 studies reported disutility values for stroke [10, 24, 26, 28, 33, 35, 37, 44, 46, 49, 54, 55, 57, 58] (Table 4), and 8 studies presented utility values for stroke [11, 18, 24, 29, 36, 38, 46, 54] (data not shown). Overall, stroke was typically associated with profound deficits in QoL relative to other complications. In studies that used the term stroke, disutility values ranged from − 0.59 to − 0.035 (Table 4). Notably, the two studies that reported the largest disutilities for stroke used the HUI-2 or HUI-3 [46, 54], and another study reporting one of the smallest decrements of − 0.04 utilized the SF-6D [33]. When limited to studies that elicited utility values using the EQ-5D, the disutility associated with stroke ranged from − 0.135 [35] to − 0.035 [49]. Notably, the − 0.035 value reported in a US-based study by Zhang et al. did not include strokes resulting in hemiplegia; for events resulting in hemiplegia, the disutility was − 0.094 [49].

Table 4 Disutility values associated with cerebrovascular complications
Full size table

Renal Disease

Heterogeneity regarding definitions applied was also evident for studies presenting disutility values for renal complications. The literature review was initially designed to capture studies that reported utility/disutility values for the health states of microalbuminuria, macroalbuminuria, end-stage renal disease (ESRD), dialysis and renal transplant. However, few studies were identified that delineated renal disease according to this terminology, with many studies instead utilizing the overarching term of nephropathy. However, the definition of the nephropathy health state was in general poorly defined, and in instances where definitions were provided, these were inconsistent between studies. For example, Grandy et al. defined nephropathy as “self-reported diagnosis of chronic kidney disease, dialysis, ESRD, kidney transplant or proteinuria” [39]. In contrast, Luk et al. applied a more specific definition of nephropathy as “either proteinuria or chronic kidney disease (defined by the Renal Association as estimated glomerular filtration rate < 60 ml/min/1.73 m [2] on at least two occasions 90 days apart with or without markers of kidney damage, which include albuminuria, hematuria, electrolyte disorders due to tubular disorders, renal histologic abnormalities, structural abnormalities evident on imaging or a history of renal transplantation” [62]) [12]. The broad definition applied in analyses such as that presented by Grandy et al. means that the term nephropathy captures both patients with microalbuminuria and ESRD, despite these states representing very different severities of renal disease.

A total of seven studies presented disutility values for the broad health state of nephropathy [10, 12, 16, 37, 39, 50, 58] (all of which used the EQ-5D), and a further four studies presented values for ESRD [44, 46, 49, 57], three of which used the EQ-5D [44, 49, 57] and one used the HUI-3 (Table 5) [46]. The mean disutility for nephropathy ranged from − 0.08 in a sample of patients with good glycemic control (median HbA1c 6.8%) and median duration of diabetes of 7 years based in Vietnam [16] to –0.0044 in patients with a mean age of 57 years based in South Korea [10]. For individuals with ESRD, the mean disutility value in EQ-5D studies ranged from –0.049 [57] to − 0.1018 [44], while the one study that utilized the HUI-3 reported a disutility of − 0.024 [46].

Table 5 Disutility values associated with renal disease
Full size table

Ophthalmic Complications

The review was designed to capture the QoL impact of ophthalmic complications including retinopathy (background and proliferative), macular edema, cataract and severe vision loss/blindness. However, no studies were identified that examined the effect of macular edema on QoL, and only one study that presented a disutility value for cataract was identified (Table 6) [10]. Additionally, none of the retinopathy studies identified distinguished in severity between background and proliferative retinopathy; however, one Chinese study did distinguish between unilateral and bilateral retinopathy [15]. Here, using the EQ-5D-5L, the mean disutilities for unilateral and bilateral retinopathy were –0.013 and − 0.019, respectively [15]. A total of nine further studies reported disutility values for retinopathy elicited using the EQ-5D [10, 14, 16, 24, 26, 28, 37, 39, 58]. These included one US-based longitudinal study that captured the decline in QoL in individuals with retinopathy over a 5-year period. Using the EQ-5D, Grandy et al. reported a decline in EQ-5D of − 0.058 over 5 years, assuming a linear rate of decline over time corresponding to an annual decline of − 0.0116 per year [39]. Excluding the longitudinal study by Grandy et al., the mean disutility associated with retinopathy, elicited using the EQ-5D, ranged from − 0.0578 in a sample of people with T2D aged > 65 years (mean [SD] age 70.3 [9.3] years) based in Germany [26] to − 0.001 in a multinational study by Nauck et al. [58]. A further study utilized the 15D questionnaire and reported a mean disutility for retinopathy of − 0.036 [29]. The same authors also noted that in their analysis both the EQ-5D and SF-6D were relatively insensitive to retinopathy.

Table 6 Disutility values associated with ophthalmic complications
Full size table

Three further studies presented disutility values for health states defined as “impaired vision,” “severe vision loss” and “blindness” (Table 6), findings that collectively suggested a deterioration in QoL with increasing severity of vision loss. In a Norwegian study, visual impairment (severity not defined) was associated with a mean (EQ-5D) disutility of − 0.012 [35]. In a US-based study, severe vision loss (defined as visual acuity of < 20/200 on a Snellen chart) was associated with a mean (HUI-3) disutility of − 0.057 [46], and in a multinational study blindness was associated with a mean (EQ-5D) disutility of − 0.083 [57].

Neuropathy, Foot Ulcer and Amputation

A total of 11 studies reported disutility values for people experiencing neuropathy (Table 7) [12, 14, 19, 26, 29, 35, 37, 39, 46, 49, 51]; however, only 1 disutility study distinguished between painful and non-painful neuropathy [49] (8 studies also reported utility values for the health state of neuropathy, 1 of which distinguished between painful and non-painful neuropathy; data not shown [9, 12,13,14, 18, 24, 29, 46]). For US-based patients, using the EQ-5D, the mean disutility for patients with non-painful neuropathy was –0.039, but for painful neuropathy the mean decrement was as expected, notably more pronounced at –0.105 [49]. Another study identified in the review compared utility (rather than disutility values) and patient characteristics for those with painful versus non-painful neuropathy [9]. Patients with painful neuropathy were slightly older and had longer duration of disease than those with non-painful neuropathy, and the mean (SD) EQ-5D score was 0.8 (0.1) for those with non-painful neuropathy compared with 0.6 (0.3) for those with painful neuropathy [9]. Allied to this, a further US-based study captured the decline in QoL over time in patients with neuropathy, which may at least partially address the issue of progression/increased severity over time. Using the EQ-5D, Grandy et al. reported a decline in QoL of − 0.061 over a period of 5 years, which corresponds to an annual decline of − 0.0122 per year assuming a linear rate of decline [39].

Table 7 Disutility values for neuropathy, foot ulcer and amputation
Full size table

The magnitude of the disutility associated with neuropathy was also influenced by the elicitation method used. For patients in Greece, Kontodimopoulos et al. report a disutility for neuropathy of − 0.117 using the SF-6D, but using the EQ-5D the corresponding value was − 0.247, suggesting different sensitivities to the impact of neuropathy between the two methods used [29].

Diabetic Foot and Amputation

The definitions and terminology used for diabetic foot/foot ulcer health states were heterogenous between studies and in some instances poorly defined, thereby making it challenging to differentiate between the impact of events such as the occurrence of superficial uninfected ulcers that healed without complications and infected/gangrenous ulcer. Additionally, two studies included amputation within the overarching terms of diabetic foot [50, 58]. Reported mean disutility values for diabetic foot/foot ulcer ranged from − 0.016 in a Norwegian study by Solli et al. [35] to − 0.206 in a Greek study by Kontodimopoulos et al. [29], with both of these values elicited using the EQ-5D (Table 7). The broad range of disutility values may be partly attributable to the heterogeneity and differences in severity in terms of the terminology and categorizations used between different studies. Furthermore, as with neuropathy, reported disutility values were influenced by the elicitation method used. Using both the 15D and SF-6D Kontodimopoulos et al. reported a mean disutility of − 0.093, but when the EQ-5D-3L was used the mean utility decrement was notably greater at − 0.206 [29]. Additionally, for amputation specifically, reported disutility values (elicited using the EQ-5D) ranged from − 0.122 [57] to − 0.0631 [44] (Table 7).

Hypoglycemia

Literature searches identified a total of 24 studies that presented either utility or disutility values for patients experiencing hypoglycemia of various severities [11, 12, 17, 20, 21, 24, 25, 30,31,32, 34, 37, 40,41,42,43, 45, 47, 48, 55, 56, 58,59,60] as well as three studies that examined the effect of FoH on QoL [34, 35, 47]. As with other complications, the terminology and categorization of hypoglycemia varied widely between studies. Some studies categorized hypoglycemic events as either non-severe or severe, with non-severe events typically defined as events that could be resolved by the individual and severe events as those requiring third-party assistance. Other investigators used the categorization of mild, moderate, severe and very severe, with mild events typically defined as events causing no interruption of activities, moderate events as those resulting in some interruption of activities, severe events as requiring the assistance of others and very severe events as events that required medical assistance [41, 60]. Some studies employed broader terminology, presenting utility/disutility values for any symptomatic event, and several focused on historical events describing utility values according to hypoglycemia experienced in the previous 1-, 3- or 6-month period. Additionally, a total of four studies distinguished between daytime and nocturnal hypoglycemic events [11, 20, 40, 56]. As with long-term complications, the EQ-5D was the most frequently used utility elicitation method employed to determine the influence of hypoglycemic events on QoL; however, a total of five studies that specifically focused on hypoglycemic events used TTO methodology to determine disutility values [20, 34, 40, 56, 59].

Three studies reported disutility values for hypoglycemic events categorized as non-severe, all three of which included values elicited from both people with T2D and general population samples using TTO methodology [20, 40, 56]. For people with T2D, the disutility for a daytime non-severe event ranged from − 0.0028 in a Canadian study [40] to − 0.0283 in Malaysia (Table 8) [20]. Similarly, for general population samples, the corresponding range was from − 0.004 in a multinational study [56] to − 0.0354 in Malaysia [20]. For people with T2D, two of the three studies reported that a nocturnal non-severe event was associated with a greater decrement compared with an event occurring during the daytime. For example, in Canada, the decrement associated with a daytime non-severe event was − 0.0028, whereas a nocturnal non-severe event was associated with a utility decrement of − 0.0076 [40].

Table 8 Reported disutility values for hypoglycemic events
Full size table

A further two studies (both of which used the EQ-5D) categorized events as either mild or moderate and reported disutility values of − 0.009 [41] and − 0.018 [31] for mild events and − 0.055 [41] and − 0.043 [31] for moderate events, respectively. A total of 13 studies examined the effect of severe hypoglycemic events on QoL [11, 20, 24, 31, 34, 37, 40, 41, 43, 46, 56, 58, 60], including 8 that reported disutility values for severe events [20, 24, 31, 37, 40, 41, 56, 58]. In people with T2D, reported mean disutility values for a severe hypoglycemic event ranged from − 0.008 in China [24] (determined using the EQ-5D-3L) to − 0.1938 (determined using TTO methods) in a cross-sectional study conducted in Malaysia [20]. Three studies distinguished between severe events occurring during the day and nocturnal severe events [20, 40, 56], and in people with T2D nocturnal events were consistently associated with a greater disutility than daytime events.

Additionally, in a US-based study presenting disutility values elicited using the EQ-5D, Marrett et al. [41] distinguished between severe events requiring third-party assistance and very severe events requiring medical assistance; the mean disutilities associated with severe and very severe events were − 0.131 and − 0.208, respectively.

Three further studies quantified the influence of FoH on QoL; of these, two were conducted in people with T2D in Norway [35] and the USA [47], respectively, and the third was conducted in a UK-based general population sample [34]. Both studies conducted in people with T2D used the EQ-5D, and in the US, Shi et al. [47] reported that any FoH was associated with a disutility of − 0.003. In Norway, Solli et al. [35] reported that a large FoH was associated with a disutility of − 0.078 compared with a small FoH.

Discussion

It is well established that diabetes-related complications can have a profound effect on QoL. Here, a synopsis of recently published disutility values that reflect current management/treatment practices is provided, which can be used in turn to inform future health economic models and analyses of novel interventions. New treatments for T2D generally provide incremental benefits in terms of glycemic control and/or adverse event rates, in particular hypoglycemia, relative to the standard of care. Glycemic control is a key determinant of the risk for long-term complications [63]. As such, for economic models to project valid outcomes such as quality-adjusted life expectancy for diabetes interventions, it is important that the most appropriate disutility values are used to best reflect the impact of individual complications on QoL.

When selecting utility/disutility values to inform economic models, there are a number of issues that warrant consideration by model developers. These include the elicitation method used and whether utility/disutility values selected are from a population that aligns with the simulated patient cohort under investigation. The choice of elicitation method is in some instances influenced by national guidelines. For example, for economic analyses performed in Sweden or Denmark, direct methods (e.g., TTO or SG) are preferred, whereas in other jurisdictions (e.g., England, Scotland) guidelines advocate the use of utility values elicited using the EQ-5D [64]. Indeed, in national guidelines where a specific generic multi-attribute utility instrument is recommended, this is most commonly the EQ-5D, although several countries do not stipulate which instrument should be used, instead providing examples of acceptable methods [65]. In terms of population, it may also be desirable to source disutility values that are generalizable to the simulated model cohort in terms of location, baseline demographics and disease characteristics. For example, values derived from individuals with newly diagnosed T2D based in Europe may not be appropriate when modeling new interventions in a population with long-standing disease in a country in Asia such as China or Singapore. The statistical approach used and the parameters adjusted for should also be considered when selecting utility values. In some instances, disutilities were calculated simply by subtracting the mean utility value of patients with a particular complication from the mean value for those without, while in other analyses disutilities were calculated using multivariate regression models and adjusted for baseline demographics and disease characteristics. The factors controlled for should be considered when selecting sources of disutility values. For example, women were consistently shown to have lower QoL relative to men [14, 24, 28, 33, 46, 49, 53], so it is important to consider whether values have been adjusted according to gender and other baseline demographics and disease characteristics.

Model developers may also need to consider the best way in which to address data gaps in terms of the need to utilize disutility values from multiple different sources. Research across different therapy areas has consistently shown that, for particular health states, utility values elicited using direct methods are consistently higher than values for the same state elicited using indirect methods [66, 67], which may potentially lead to the introduction of bias in an analysis. Allied to this, for indirect methods consistency in terms of the source of preference weights used should also be considered owing to potential differences in cultural norms and healthcare provision between different settings [46]. Two studies identified in the current review showed that the source of preference weights can have a considerable influence on baseline utility values [28, 57]. In particular, in a Swedish analysis, Kiadaliri et al. [28] noted that when using a Swedish tariff, the mean utility value for individuals with no complications was 0.89, but when using a UK tariff the corresponding value was considerably lower at 0.79. Similarly, when using the UK tariff, the disutilities associated with MI, heart failure and stroke were approximately two-fold greater than with the Swedish tariff [28].

A few studies identified in the review documented the phenomenon of diminishing marginal disutility specifically relating to cardiovascular events. Diminishing marginal disutility has previously been demonstrated with hypoglycemic events [68] and refers to instances where subsequent events are judged to have a lesser effect on QoL than first events. Briggs et al. observed diminishing marginal disutility with a composite endpoint of major cardiovascular event [55]. Here, a first major cardiovascular event was associated with a disutility of − 0.05, but the decrement associated with a subsequent event was less at − 0.038. Similar findings were reported by Kiadaliri et al. in terms of the effect of first and subsequent MIs [28]. Shao et al. also demonstrated that while the QoL impact of events such as MIs wanes over time, an MI within the previous year was associated with a disutility of − 0.042 but the disutility associated with a history of MI was − 0.011 [46]. However, with any retrospective analysis of events, it is possible that the interval between the event and utility elicitation may influence the QoL finding owing to response shift or recall bias, the potential impact of which should be considered when interpreting results [69].

Several diabetes-related complications such as retinopathy or renal disease may be classified according to severity. However, few studies identified in the review captured differing levels of severity for complications, although the notable exception to this was hypoglycemia, where events were frequently classified as non-severe and severe, or mild, moderate, severe and very severe. For retinopathy, no studies were identified that distinguished between background retinopathy and proliferative retinopathy. The decrement associated with vision loss/blindness, which may occur as a result of proliferative retinopathy, was however captured by several investigators. Additionally, only two studies distinguished between painful and non-painful neuropathy [9, 49]. However, a longitudinal US-based study by Grandy et al. [39] captured the decline in EQ-5Dindex score over a time period of 5 years for patients with various complications including retinopathy and neuropathy. Annualizing this decrement may therefore represent an alternative method of modeling progression/increasing severity of complications over time.

For most complications, the EQ-5D was the most frequently used elicitation method, with more recently published studies tending to use the EQ-5D-5L rather than EQ-5D-3L. Two studies also directly compared values derived from the EQ-5D-3L and EQ-5D-5L in people with T2D, with both concluding that the 5L version showed greater sensitivity and discriminative ability as well as a reduced ceiling effect relative to the 3L [13, 22]. Although the EQ-5D was frequently used for long-term complications, three studies that focused exclusively on hypoglycemic events used TTO methodology [20, 40, 56], with the rationale for this being the greater sensitivity of TTO methods relative to generic instruments such as the EQ-5D. This increased sensitivity may be particularly beneficial for complications such as hypoglycemia where events are distinguished by severity as well as the time of day/night at which events occur.

As with all literature reviews, there are limitations associated with the present review. No formal assessment of study quality or critical evaluation was performed; as a result, no ranking of the data is identified. This was a deliberate approach as different modeling and country-specific approaches may lead to differential priorities for the selection of utilities in a health economic analysis. Little was reported on the management of diabetes-related complications in the studies identified in this review. Different approaches to managing these conditions could lead to different utility scores being reported by patients across studies, likely contributing to between-country variation in outcomes reported. Demographic and socio-economic factors regarding utility scores were not reported in this review, which may also influence the selection of the most appropriate utility scores for health economic analysis. The application of inclusion and exclusion criteria was limited in the present review in an effort to align the work with that of Beaudet et al. and to minimize the risk of introducing bias through study selection [4]. However, this approach also has the potential to capture studies with smaller populations and/or limited quality and imply an equal weight to their results.

The ISPOR good practice recommendations note that economic modeling analyses should utilize up-to-date utility values as these reflect contemporary clinical practices and recent advances in treatment that may influence patients’ QoL [3]. The findings presented here provide a synopsis of recently published disutility values for major complications in T2D diabetes than can be used to inform future economic modeling analyses.