Danish value sets for the EORTC QLU-C10D utility instrument

For this study, we followed the methods for the generation of utility weights established by King et al. [24] which were also used in the valuation surveys for other countries [14, 15, 18, 20, 30]. In this cross-sectional valuation survey, participants from the Danish general population were recruited via a third-party provider online panel (https://​surveyengine.​com/​). Participants of the survey received a small payment for their participation. We used quota sampling to obtain representativeness of the sample with regard to sex and age based on United Nations census data for Denmark [40].

All participants provided electronically signed informed consent and the study was approved by the Ethics Committee of the Medical University of Innsbruck (approval number 1079/2023).

The survey comprised the assessment of sociodemographic data, basic clinical data (chronic diseases), the EORTC QLQ-C30, the DCE to establish Danish utility weights, the EQ-5D-5L [34], and the Kessler psychological distress scale [22]. The EQ-5D-5L and the K10 were not evaluated in this paper.

The EORTC QLU-C10D is a utility scoring algorithm that was developed by the MAUCa Consortium and the EORTC based on the EORTC QLQ-C30 [23]. The QLQ-C30 is a 30-item questionnaire for patients with cancer. It assesses functional health (physical functioning, role functioning, social functioning, emotional functioning, cognitive functioning), symptom burden (fatigue, pain, nausea/vomiting, sleep disturbances, dyspnea, appetite loss, constipation, diarrhea, financial impact), and global health/quality of life. For the QLU-C10D, 10 key domains from the QLQ-C30 were selected: physical functioning, role functioning, social functioning, emotional functioning, pain, fatigue, sleep disturbances, appetite loss, nausea, and bowel problems (which merges the constipation and diarrhea scales from the QLQ-C30). As in the QLQ-C30, items in the QLU-C10D have four possible levels ranging from “Not at all” to “Very much” impairment in that domain. With 10 domains with each four levels of severity, the QLU-C10D covers a total of 4^10 = 1,048,576 different health states. Supplementary Table 1 shows the classification system of the QLU-C10D. The robustness of the QLU-C10D methodology and its psychometric properties like test–retest reliability have been demonstrated in past studies [16, 31, 33].

Due to its basis on the QLQ-C30, the QLU-C10D is backward-compatible with the QLQ-C30, meaning that health utilities can be calculated from QLQ-C30 data if utility weights for the respective population are available.

Utilities were elicited by DCE in this study, which was constructed analogously to the one used by King et al. [24]. Out of a set of 960 possible choice sets, participants were randomly assigned to complete 16 choice sets to limit participant burden. For each choice set, participants had to choose one of two scenarios. Scenarios were set up and described based on the 10 domains from the QLU-C10D and also included survival time as an anchor. Survival time varied between 1, 2, 5, or 10 years (wording: “You live in this health state for X years, and then die”). To keep the choice sets comparable and participant burden acceptable, the scenarios of a single choice set only differed in 5 out of the 11 aspects (10 QLU-10D domains and survival time) which was previously found to be an adequate format [33]. The differing aspects were highlighted to increase their visibility for participants. See Supplementary Fig. 1 for an example of the DCE display in English language.

In the DCE section of the survey, randomization occurred at two levels: (i) each participant was assigned 16 choice sets out of the 960 available without duplication; (ii) the assignment of options as Situation A or B within each choice set was randomized to minimize potential ordering biases. It is worth noting that the presentation order of dimensions remained consistent, as prior research indicated that dimension order did not consistently impact the utility weights for the QLU-C10D [31].

Descriptive statistics are reported for the sample characteristics (i.e., demographic information, clinical information). Chi-square tests were used to compare the sample characteristics of gender, age, and education with national census data [39]. To calculate utility weights, we followed an established procedure [2] that was also used in previous studies on the EORTC QLU-C10D, [15, 18, 33]. The utility (U) of an option j in the choice set s for a respondent i can be described by the following formula:TIMEisj is the survival time in option j and X’isj is a set of dummies related to the levels of the corresponding health state. Errors (εisj) are assumed to be independent and identically Gumbel-distributed [2, 24]. DCE data were analyzed using a conditional logit model with cluster sandwich estimators to adjust for correlated observations within respondents. Analogous to previous valuation studies, estimates were converted into utility decrements based on the ratio of health state parameters (β) and the time coefficient (α) to reflect the trade-off between HRQOL and survival. Finally, we ran an additional analysis to correct for non-monotonic levels in the EORTC QLU-C10D utility weights (i.e., for domains where the severity levels did not show an increasing decrement in utility weight) in line with previous valuation studies.

In a recent review of DCE studies in health state valuation by Wang et al. [41], the majority of studies (48/65, 74%) used conditional logit models. In our study, we also relied on a conditional logit models because its parameterization is more suited our purpose (for public decision-making) compared to mixed logit models that focus on the distribution of individual preferences. Analyses were run using SPSS version 27.0 (IBM Corp., Armonk, NY, USA).

We recorded the total time it took respondents to complete the survey and the number of participants that only chose the A or only chose the B option in the DCE. We also computed an additional model in which we excluded the fastest 10% of participants and compared that to the model with all participants.

Between 04.08.2021 and 24.08.2021, 1744 participants initiated the survey. Of those, 269 (15.4%) participants were excluded due to the quota sampling, 417 (23.9%) were excluded due to timing out (being idle mid-survey for too long and not progressing further in the survey), and 57 (3.3%) were excluded because they did not finish the survey (actively closed the survey before completion). The final sample consisted of 1001 participants.

Socio-demographic and data on chronic diseases are shown in Table 1. For gender and age, the sample was representative of the Danish general population (both p > 0.975). Regarding education, the sample showed significantly higher education levels compared to the general population (p < 0.001) and most participants had upper secondary (47.8%) or tertiary education (41.9%). Most participants (59.9%) reported being married or in a steady relationship. About one-third (35.5%) of participants reported having at least one chronic disease. The most commonly reported diseases were arthritis or rheumatism (11.0%), respiratory diseases (10.4%), and diabetes (8.7%).

The DCE analyses without correction for monotonicity are shown in Fig. 1 and Table 2. The analyses with correction for monotonicity are shown in Fig. 2 and Table 2. We report the utility decrements for each domain and severity level. For the level “not at all” (i.e., no impairment), the decrement is, by definition, 0. Actual utilities corresponding to individual health states can be calculated by subtracting the decrements from Table 2 from 1. For example, the utility for a health state where all domains would have a severity level of “quite a bit” (33333 33333), based on the utility decrements with correction for monotonicity, would be calculated as follows:

The utility score for the worst possible health state (44444 44444), or PITS state (i.e., worst attainable health), is calculated as follows:

The largest utility decrements for the severity level “very much” were observed for the domains physical functioning (− 0.224), pain (− 0.160), role functioning (− 0.136), and emotional functioning (− 0.124). Notably, the cancer-specific domains nausea (− 0.101) and bowel problems (− 0.102) also had decrements around − 0.1. The smallest utility decrements for the severity level “very much” were observed for the domains lack of appetite (− 0.024), sleep disorders (− 0.057), and fatigue (− 0.064). The overall rank order of domains for the other severity levels was similar.

Non-monotonicity of levels was observed for the domains sleep disorders, lack of appetite, and bowel problems. However, none of the deviations from monotonicity were statistically significant. In the model adjusted for monotonicity, the levels showing non-monotonicity were set to be equal (e.g., sleep disorders level 3 = level 4).

The mean time to complete the survey was 16.0 min with a standard deviation of 10.0 min (median 13.6, inter-quartile range 9.5–19.7 min). Excluding the fastest 10% of respondents (≤ 7.03 min) from the analysis did not change the utility decrements and results were almost identical with the analysis that included all respondents (complete data available upon request from the authors).

Few respondents (1.7% of 1001) chose either only the A or only the B option in the DCE, which is comparable to our previous studies [36].

The Danish healthcare system combines universal protection and equal access to healthcare for all citizens, providing high-quality services free of charge. It promotes, among other things, preventive examinations for some diseases, especially to detect cancer in early stages and provide better treatments to patients. The total socioeconomic cost of Danish cancer patients is approximately DKK 12 billion (or roughly € 1.6 billion) over a 5-year period, including direct healthcare costs, and lost income for patients and potential partners [25]. These costs are expected to increase due to the increased incidence of cancer, and therefore, the EORTC QLU-C10D may facilitate cost-utility analysis of cancer interventions in different healthcare settings. Further instructions on how to use the EORTC QLU-C10D can be found in the EORTC user manual by Gamper et al. (https://​qol.​eortc.​org/​manual/​eortc-qlu-c10d-manual/​).

The EORTC QLU-C10D is a relatively new multi-attribute utility instrument (MAUI) and is a promising candidate measure that is cancer-specific. Compared to generic MAUIs like the EQ-5D which offer the advantage of making comparisons between certain condition groups, specific MAUIs might more effectively capture aspects that are particular to a certain condition or disease. A 2020 review of health utility instruments that are used and recommended by national health technology assessment (HTA) guidelines found that generic MAUIs (meaning not focused on a specific disease) like the EQ-5D are still preferred in half of the reviewed guidelines [21]. However, the review also found that about half of the national guidelines do not express a preference for a specific MAUI but do recommend calculating utilities using national preference weights. This highlights the necessity for national utility weights for different countries. In this context, two Danish value sets are available for the EQ-5D-3L [43] and the EQ-5D-5L [19] but so far, no dataset was available for the EORTC QLU-C10D. One considerable advantage of the EORTC QLU-C10D is that it can be used in any cancer trial that assesses the EORTC QLQ-C30, which is the most commonly used quality of life questionnaire in published and registered clinical trials by a large margin [17]. In other words, the EORTC QLU-C10D health utilities can be calculated using patients' EORTC QLQ-C30 scores, even retrospectively for trials that have already been closed. We strongly encourage future analyses of trials using EORTC QLQ-C30 data based on the growing body of utility weights available for the EORTC QLU-C10D [15, 18, 20, 24, 30, 32, 36].

As part of the ongoing efforts to provide utility weights for the EORTC QLU-C10D, we calculated national utility weights for Denmark. The Danish Medicines Council guidelines for methods to assess new pharmaceutical states that QALYs should be used to summarize findings from health economic analyses [9] based on national weights. Sensitivity analyses are recommended if both a generic MAUI (like the EQ-5D) and disease-specific MAUI like the EORTC QLU-C10D are available [9]. With the foundation of the Danish Health Technology Council in 2021 [7], we expect a growing importance of HTA in Denmark in the near future.

Generally, utility decrements for the EORTC QLU-C10D and the rank order of domains were similar to other European countries. Several of the findings merit in-depth discussion. In our sample, the utility decrement for physical functioning was the largest, which was also the domain with the largest decrement for many other countries; the decrement was comparable to Austria [15], Australia [24], and smaller when compared to Italy and Poland [15]. Following physical functioning, the domain with the second largest decrement was pain, which is a pattern that was, for example, also observed in, for example, Austria [15], Germany [20], Spain [14], France [30], and the United Kingdom [32].

Compared to many other countries however [14, 15], the domain with the third largest decrement in the Danish utility weights was not role functioning but emotional functioning. In general, the decrement for emotional functioning was large with − 0.124, indicating that emotional outcomes are more highly valued or considered to have a strong impact on health in the Danish population. In international comparison, only France [30], the United Kingdom [32], and Australia [24] show larger decrements for emotional functioning. Most other countries have a lower decrement for this domain [15, 18, 20, 36]. This might also be influenced by the translation of the domain, which was “Du føler dig deprimeret” and is a strong wording considered a paronym for depression in a more clinical or diagnostic sense compared to “feeling down.” However, similar wordings were also used in other countries that did not show large decrements for this domain, like the USA (“You feel depressed”) [36] or Italy (“Soffre di depressione”) [15]. Social functioning was also valued similarly high as emotional functioning at − 0.104 for the highest level, with the decrements for the lower two levels being slightly larger than the corresponding levels for emotional functioning.

Some of the domains that are considered cancer-specific like bowel problems (diarrhea, constipation) and nausea had decrements around − 0.1. These are larger than in other countries like Germany [20], Austria, Italy, and Poland [15] and similar to the Netherlands [18] and Australia [24]. Three other domains fatigue, lack of appetite, and sleep disturbances were associated with smaller utility decrements. This was also observed in other countries [15, 18, 20]. One potential explanation is that, while people from the general population do experience occasional fatigue-like symptoms or lack of appetite or sleep disturbances, they cannot relate to severe, cancer or therapy-induced fatigue or lack of appetite [36]. It is possible that patients with cancer who have actually experienced those symptoms as a consequence of antineoplastic therapy would assign a different weight to these domains. Evaluations to derive comparable utility weights from a patient population are currently ongoing.

The worst possible health state (PITS state) with the highest level of impairments in all domains had a negative value of − 0.096, indicating that it was perceived as worse than being dead in our population. This is similar to the value of the PITS state in Australia (− 0.096) [24] and the United Kingdom (− 0.083) [32].

One strength of our study is the large sample that was representative of the Danish general population with regard to age and gender. Moreover, we followed an established methodology to calculate the utility weights [23]. This enables cross-country comparisons of utility weights as preferences have been shown to differ between countries [35]. One limitation of our study is that the sample had a higher proportion of highly educated people compared to the general population which was also found in the evaluations of other countries [20, 36] and it is unknown if different health preferences exist in different subpopulations (e.g., in groups with different education levels). Another limitation is that we do not have sociodemographic information on respondents that did not complete the survey and who were excluded from the analysis as those respondents mostly did not complete the questions on sociodemographic information.

Non-monotonicity was found for some of the domains with smaller weights. However, none of these deviations from monotonicity were statistically significant and all could be corrected in the analysis. Finally, another possible limitation of this study is the potential cognitive burden of participants due to the length of the questionnaires included in the survey, since about a third of those who began the survey did not complete it and were excluded from the analysis.