Health utilities and willingness to pay in adult patients with coeliac disease in Hungary

Background Coeliac disease (CD) is a life-long food-related disorder with a global prevalence of approximately 1%. Patients with CD often experience reduced health-related quality of life that could improve with a strict adherence to a gluten-free diet (GFD). Objectives To obtain visual analogue scale (VAS), time trade-off (TTO) and willingness-to-pay (WTP) values amongst patients with CD. Methods In 2020–2021, a cross-sectional online survey was conducted amongst 312 adult CD patients in Hungary. Patients completed the Gastrointestinal Symptom Rating Scale (GSRS) and evaluated their current health and three hypothetical health state vignettes defined based on dietary adherence using VAS, conventional 10-year TTO and WTP. Multivariate regressions were used to explore the effect of patients’ demographic and clinical characteristics on utility and WTP values. Results Mean VAS values for current health and ‘CD with strict adherence to GFD’, ‘CD with loose adherence to GFD’ and ‘CD without GFD’ hypothetical health states were 79.69 ± 18.52, 85.36 ± 16.18, 62.44 ± 19.91 and 36.69 ± 25.83, respectively. Corresponding mean TTO utilities were: 0.90 ± 0.19, 0.91 ± 0.20, 0.87 ± 0.23 and 0.76 ± 0.29. Mean annual WTP values for returning to full health were: €845 ± 1077, €648 ± 1002, €862 ± 1135 and €1251 ± 1496. Older age at diagnosis, male sex, more severe gastrointestinal symptoms (GSRS) and having comorbidities were associated with lower VAS and TTO or higher WTP values for current own health (p < 0.05). Conclusion This is the first study to report TTO utilities for CD health states. Strict adherence to the GFD may result in substantial health gains in symptomatic patients. Utilities and WTP results can be used to estimate benefits of GFD in cost-utility and cost–benefit analyses.


Introduction
Coeliac disease (CD) is an immune-mediated systemic disorder activated by the ingestion of gluten in genetically susceptible individuals [1]. It is the most common life-long food-related disorder with a global prevalence of 0.7-1.4% [2,3]. CD can develop at any age and it is approximately 1.5 times more common in females than in males [2,4,5]. The clinical presentation is heterogeneous and may include gastrointestinal problems, signs of malabsorption, as well as extraintestinal symptoms, such as dermatitis herpetiformis, arthritis, neurological symptoms and anaemia [6]. CD is also associated with and increased risk of depression, anxiety, attention deficit hyperactivity disorder and eating disorders [7]. Currently, the only treatment option is a strict, life-long gluten-free diet (GFD) [8]. However, adherence to a GFD may be challenging in everyday life mostly due to the limited availability, high costs and variable quality of gluten-free products, and individual lifestyle aspects (difficulties with dining out and travelling as well as lack of social support) [9][10][11][12].
So far, the EQ-5D has been the most commonly used instrument to derive health utilities in CD patients [15,16,[24][25][26][27]. However, its five dimensions (mobility, self-care, usual activities, pain/discomfort and anxiety/depression), and in particular the pain/discomfort dimension may be insensitive to symptoms, such as diarrhoea and constipation [28]. It is therefore possible that the EQ-5D overestimates the HRQoL of CD patients. This is supported by three earlier EQ-5D studies that reported similar or even higher utilities (representing a better HRQoL) in CD patients than amongst members of the general population [15,16,25]. Another limitation in previous studies is that the vast majority of diagnosed CD patient populations lived gluten free at the time of assessment, and thus, utilities prior to the start of GFD were only assessed retrospectively; however, retrospective assessment may be subject to recall bias [16, 23-25, 27, 29]. As a consequence, no reliable utility estimates are available for CD before treatment. Moreover, none of the existing studies focussed on the impact of different levels of dietary adherence on utilities.
When generic preference-accompanied measures are not feasible to assess utilities, vignette-based methods may be considered. A vignette is a description of a hypothetical health state that is directly valued by using a preference elicitation method [30]. As vignettes are hypothetical, they enable to elicit preferences for any health state. Preferences for the vignettes may be measured in several ways; for example, people may trade off life years (time trade-off, TTO) or risk of death (standard gamble) to improve health. TTO is the most widely used method to directly obtain health utilities [31]; nevertheless, no prior studies used this approach in CD patients. Furthermore, preferences may also be assessed by contingent evaluation, where respondents are asked to reveal the amount they would be willing to pay to improve their health. To date, two studies have employed this method in CD. One study measured WTP for CD screening in parents of children diagnosed with CD in Sweden, and another one surveyed adult patients with CD in Switzerland [32,33]. However, no vignette-based WTP studies have been carried out in CD to date.
This study therefore aims to assess health states by visual analogue scale (VAS), TTO and WTP values in CD patients using a vignette-based study design.

Study design and population
An online cross-sectional survey was performed between November 2020 and January 2021 in Hungary. Permission for conducting the study was granted by the Research Ethics Committee of the Corvinus University of Budapest (reference no. KRH/390/2020). A convenience sample of CD patients were recruited through 30 different patient organisations and social media groups. Participation was voluntary and anonymous, no remuneration was provided. The survey was programmed in Qualtrics (Qualtrics 2020, Provo, UT, USA). To be included in the study, participants were required to be aged 18 years or over and to give their informed consent. All questions in the survey were mandatory with the exception of the income question, therefore respondents could not proceed to the next question without answering the previous one.

Questionnaire
The questionnaire consisted of four parts. The first part included questions about CD-related clinical characteristics, including disease duration, comorbidities and adherence to GFD. This section was built on two earlier national surveys involving CD patients in Sweden and the UK [15,16]. The second part consisted of different standardised questionnaires to assess symptoms, HRQoL and well-being. CD-specific symptoms were measured with Gastrointestinal Symptom Rating Scale (GSRS) [34]. The GSRS is widely used in CD patients, including clinical trials and it has showed good validity and responsiveness to both GFD and gluten challenge in this patient population, and additionally, it was already available in Hungarian-language at the time of the data collection [35][36][37]. Regulatory authorities, such as the FDA also have familiarity with the GSRS [38]. The GSRS evaluates common gastrointestinal symptoms experienced by patients, where 15 items combine into five domains: reflux (2 items), abdominal pain (3 items), indigestion (4 1 3 items), diarrhoea (3 items) and constipation (3 items). Each item has seven response options with descriptive anchors ranging from 'no discomfort at all' (= 1) to 'very severe discomfort' (= 7). The total score, calculated by adding up the item scores, may range from 15 to 105, where a higher score represents more health problems. In the third part of the survey, the patients were asked to assess their own health, followed by three hypothetical health state vignettes, which appeared in a randomised order. Both current own health and the hypothetical health state vignettes were valued by VAS and TTO and by contingent evaluation asking WTP. In the last part of the questionnaire, sociodemographic data, including age, sex, employment, place of residence, net household income and education, were collected.

Development of coeliac disease health-state vignettes
The following three hypothetical health state vignettes were developed for this study: (1) CD without GFD, (2) CD with loose adherence to GFD and (3) CD with strict adherence to GFD. The vignettes were presented from a second-person perspective. The vignettes were primarily developed based on existing literature reviews [13,17,[39][40][41]. A recently published model on concepts relevant when assessing health outcomes in CD summarises the signs and symptoms as well as broader HRQoL aspects in CD based on 28 original studies and stakeholder interviews with clinical experts and payers [41]. The model incorporates both gastrointestinal and non-gastrointestinal signs and symptoms of the disease alongside the following six HRQoL aspects: daily activities (e.g. negative impact on career or work), relationships (e.g. family life), social/leisure (e.g. dining out), sleep, dietary burden of GFD (e.g. difficulty adhering to GFD) and psychological impacts (e.g. anxiety or depression). These HRQoL impacts overlap with those covered by the most widely used patient-reported outcome measures in CD. Considering the conceptual model and a thorough review of the item content of CD-specific HRQoL measures and symptom scales, the following six areas were selected to be included in the health state vignettes based on judgement of a patient, a gastroenterologist professor and two health economists experienced in utility assessment: diet, gastrointestinal symptoms, work/ school, physical activities, sleep/fatigue, mood and social life ( Table 1). Comprehensibility of the descriptions was tested in an interview with a CD patient.

Visual analogue scale (VAS)
For the hypothetical health states, a horizontal VAS was used with the endpoints of 'the worst health you can imagine' (= 0), and 'the best health you can imagine' (= 100). To measure patients' own current health, we used the EQ VAS in an earlier section of the questionnaire. Note that the EQ-5D-5L including the descriptive system and EQ VAS were both completed by the patients; however, only the EQ VAS data were used for the present study [42]. The EQ VAS has identical endpoints to the VAS we used for the hypothetical health states, but it is vertically aligned.

Time trade-off (TTO)
The TTO method elicits utility values for imperfect health states by asking patients to make a trade-off between quality and length of life [23]. We opted to use a 10-year time frame, as this is the most commonly used duration in valuation studies in Hungary and beyond [43][44][45][46][47][48][49][50][51]. Patients were asked to imagine living in their current health or in a hypothetical CD-related health state for the next 10 years, followed by death. Then they had to indicate how many life years they would give up in order to regain full health. We used the top-down titration; thus, respondents were offered a predefined list with responses ranging from 10 to 0 years, with the smallest tradable amount of time being 6 months. TTO utilities were computed using the following formula: Suppose, for example, that a patient indicated to give up 2 years, yielding a U = (10-2)/10 = 0.8. There was no worsethan-dead task in this study, therefore utilities ranged from 0 (being dead) to 1 (full health).

Willingness to pay (WTP)
WTP measures the maximum amount of money an individual would be willing to pay to be free from their own symptoms or those described in the vignettes [52]. In our questionnaire, monthly WTP values were recorded in a closed question format with an open-ended 'other' response option. Sixteen predefined monthly amounts (in HUF) were offered to patients based on a previous survey: none; 500; 1,000; 2,000; 4,000; 6,000; 8,000; 10,000; 15,000; 20,000; 25,000; 30,000; 45,000; 60,000; 80,000 and 100,000 [53].

Statistical analysis
To ensure homogeneity and high quality of data, respondents that (1) have not been diagnosed with CD by a physician or (2) filled in the questionnaire in less than eight minutes were excluded. Socio-demographic and clinical characteristics of the patients were analysed using descriptive statistics. The difference in Patient's answer 10 years

Sample characteristics
Out of 734 individuals that opened the questionnaire, 455 (62.0%) finished it. Of these, 143 respondents were excluded based on the exclusion criteria (Fig. 1). Furthermore, one patient that reported not following GFD was decided to be excluded to ensure homogeneity of the sample. Thus, data of 312 physician-diagnosed CD patients were included in the analyses. Socio-demographic and clinical characteristics of patients are presented in Table 2

VAS, TTO and WTP values
The distribution of VAS, TTO and WTP values is depicted in Fig. 2

Predictors of current own health VAS, TTO and WTP values
The multivariate regressions showed that patients with higher income had significantly higher VAS values (p = 0.006) (  Table 4). Similarly to VAS, a one-point increase in GSRS resulted in a 0.004 decrease in TTO utilities (p = 0.005). These three variables explained 8.6% of the variance of TTO utilities.
The amount patients were willing to pay decreased by 70.45% in females compared to males (p = 0.022). Patients' WTP increased by 6.8% with + 1 GSRS score (p = 0.002), by 3.9% with a 1-year increase in age at diagnosis (p = 0.047) and by 251.6% in case of concomitant gastroesophageal reflux disease (p = 0.012) ( Table 5). A 1% increase in the household's per capita net annual income was associated with an 1.33% increase in the willingness-to-pay amount, on average (p < 0.001).

Predictors of VAS, TTO and WTP values for hypothetical CD health states
Both the 'CD without GFD' and the 'CD with loose adherence to GFD' hypothetical health states were associated with significantly lower VAS and TTO and higher WTP valuations compared to the 'CD with strict adherence to GFD' health state (p < 0.05) (Tables 4, 5). A + 1 GSRS score, indicating worse HRQoL was associated with an, on average, 0.168-point decrease in the VAS values and with a 4.9% increase in the WTP amount (p < 0.05). Having rheumatic or inflammatory bowel disease decreased the patients' VAS valuations and having higher income resulted in higher TTO and WTP values (p < 0.05). Furthermore, every 1-year increase in the patients' age increased the TTO utility by 0.002 (p = 0.045). These variables explained 49.0%, 8.7% and 10.9% of the overall variation in VAS, TTO and WTP values, respectively.

Discussion
This study aimed to provide VAS, TTO and WTP values in adult patients with CD. To our knowledge, this is the first study to report TTO utilities in CD patients. The health state 'CD with strict adherence to GFD' had lower utilities than expected, possibly reflecting the psychosocial aspects of the disease, for example, the patients being prevented from eating with their peers. Older age at diagnosis, male sex, more severe gastrointestinal symptoms and having comorbidities were associated with lower VAS and TTO or higher WTP values for current own health. Compared to the TTO and WTP values, a relatively higher proportion of the variance of VAS values was explained by sociodemographic and clinical predictors. A possible explanation for this observation is that our study did not collect data on some likely predictors of TTO, including cultural values, self-esteem, marital status, having children, religious beliefs and attitudes towards life [54][55][56]. Similarly WTP values may have various additional predictors, such as sociodemographic characteristics and perceived threats and benefits of treatment [57]. It is worthwhile to reconcile our findings with those of previous studies with CD patients. The mean GSRS score in our study (28.3) was in the range of other patient populations on GFD from different countries (21.0-30.4) [58,59]. Females reported significantly more problems on GSRS compared to males, which is in line with the existing literature suggesting that women with CD experience a greater deterioration in their HRQoL than men [60][61][62][63][64]. Despite the higher average GSRS score in our study, women were willing to trade less life years in the TTO and pay less in order to regain full health. Older age at diagnosis was associated  [15,26,27]. Mean EQ-5D-3L utilities in CD patients before the diagnosis (assessed retrospectively) were 0.56 and 0.65 in two UK in 2006 and 2015, whilst after diagnosis these improved to 0.84 and 0.85 [15,27]. Similarly, in Poland, the pre-(assessed retrospectively) and post-diagnosis mean EQ-5D-5L utilities were 0.79 and 0.94 [26]. These results are comparable to our findings, whereby the hypothetical health state of 'CD without GFD' had a mean TTO utility of 0.76 and the 'CD with strict adherence to GFD' health state had a mean TTO utility of 0.91. Preference measurement also helps to shed light on the HRQoL burden associated with a disease. In addition, utilities may be used to calculate quality-adjusted life years (QALYs) in cost-utility analyses, whilst WTP values may be used in cost-benefit analyses of GFD and possible new treatments in the future [70]. Preferences may be derived from the general public or patients. In most European countries, a societal perspective is recommended in the context of economic evaluations in healthcare [71][72][73]. However, there is a growing amount of literature arguing that utilities based on both patient and general population preferences ought to be considered in economic analyses [74][75][76].
CD without a strict adherence to GFD might result in a sizeable QALY loss at a societal level. Preference-accompanied measures, such as the EQ-5D, may not be able to fully capture the health impact of CD [28], therefore vignettebased methods might be superior to indirect utility assessment in this patient population. We found that the TTO method discriminated well between health states according to dietary adherence. In cost-utility analyses of GFD and new treatments for CD, directly elicited utilities may be recommended to be used to calculate QALYs. Over 10 years, untreated CD may cause a loss of between 1.3 (with loose adherence to GFD, calculated as 10*(1-0.87)) and 2.4 (CD without GFD, calculated as 10*(1-0.76)) QALY per patient. These findings highlight the large health gains associated with GFD and may be considered when quantifying effectiveness of programmes to support CD patients' access to gluten-free food products. The WTP results also provide insights for the industry to invest in research and development in alternative treatment methods for CD, given that 69-88% of patients were willing to pay in our WTP tasks. Furthermore, the WTP data from this study will be useful for cost-benefit analyses of GFD.
Many countries offer various forms of reimbursement for CD patients, such as tax reduction (Hungary, Canada, the US, the Netherlands and Portugal), cash transfer (Italy, Argentina, Uruguay, Finland, Greece, France, Norway, Belgium and Slovenia only for children), food provision (some provinces in Argentina and Spain), prescription for gluten-free food (New Zealand, Ireland, the UK) and subsidy (Northern Ireland, Scotland, the Czech Republic) to reduce the individual financial burden of GFD [77,78]. However, other countries or regions provide no coverage of GFD products at all (e.g. Germany, some provinces in Spain, Mexico) [78]. Our findings may contribute new evidence for relevant national health and social policy programmes affecting the access to gluten-free products. This study has a few limitations that should be noted. First, the questionnaire was administered online and relied on self-reported clinical data that were not verified by physicians. Secondly, selection bias might have occurred as the majority of our study population were from the middleand high-income social groups, with college or university degrees, and females were somewhat overrepresented in the study population. It is also possible that patients voluntarily filling in such a questionnaire may differ in their clinical characteristics, and thus, the sample may not be representative of the whole population of CD patients in Hungary. In the valuation tasks, three different hypothetical health states were valued; however, the clinical manifestation may vary widely and other CD health states with atypical symptoms could have also been selected. Given that the majority of patients respond to GFD, we only included hypothetical health states that improve after following the GFD. However, a substantial minority may develop persistent or recurrent symptoms even after following GFD [1]. Furthermore, only symptomatic hypothetical health states were considered and there were no health states describing silent celiac disease due to the limited evidence of the natural history thereof [79]. Albeit, several studies found that up to 15% of patients with positive serologic test (without histological confirmation) develop symptoms after 10 to 45 years [80][81][82]. An additional limitation of the study is that we did not decompose the overall higher utility for the GFD into its components, i.e. the possible utility loss associated with following the GFD (e.g. difficulties of keeping the diet) and the utility gain as a result of the improved HRQoL due to the diet. Finally, TTO utilities in this study may be somewhat upward biased due to the top-down titration approach, and we used the EQ VAS to assess VAS values for current health, but a horizontal VAS for the three hypothetical states that might not be entirely equivalent [83].
To conclude, this study provides a better understanding of the burden of CD by reporting VAS, TTO and WTP values for patients' current health and three hypothetical GFDrelated health states. Utilities and WTP results from the present study may be useful in economic evaluations examining and comparing the value of GFD, screening strategies, subsidies and new alternative treatment options in the future.