Deriving Health Utilities from the Food Allergy Quality of Life Questionnaire – Parent Form (FAQLQ-PF) Using Mapping and Discrete Choice Experiments

Introduction

Peanut allergy is a common, often life-threatening, type 1 hypersensitivity reaction to peanut proteins.¹ People with this allergy produce peanut-specific immunoglobulin E (IgE) antibodies, which cause an allergic response upon direct contact (eg ingestion of peanuts or peanut-containing foods), cross-contact (eg contact with foods processed or handled in exposure to peanuts), or inhalation of traces of peanuts.

The incidence of peanut allergies has been increasing globally,² with current estimates suggesting that over 2% of children and 0.5% of adults suffer from peanut allergies worldwide.^3–5 The methods used to diagnose food allergies have been refined over the past 30 years,⁶ and this may have contributed to the observed increase in peanut allergy prevalence. However, even conservative estimates suggest that the prevalence is at least stable.² This sustained prevalence, together with the early onset of peanut allergy, underscores the long-term health and quality-of-life burden experienced by affected individuals and their families. A systematic review and meta-analysis estimated that the self-reported peanut allergy prevalence in Europe is around 0.4% across all age groups.⁷

While some individuals may outgrow their allergy with age, peanut allergy commonly persists into adulthood.⁸ The early onset and potential lifelong persistence of peanut allergies, coupled with the risks of allergen exposure, affects the quality of life of children and their parents/carers.^9,10 Generic health-related quality of life (HRQoL) instruments, such as the EQ-5D-3L,¹¹ are mandated by health technology assessment (HTA) agencies, such as the UK’s National Institute for Health and Care Excellence.¹² However, these instruments often lack the sensitivity to capture the impact that food allergies may have on children’s mental health and daily activities.¹³

Allergy-specific instruments, such as the Food Allergy Quality of Life Questionnaire–Parent Form (FAQLQ-PF),¹⁴ offer more sensitivity to these issues. The FAQLQ-PF is completed by parents/caregivers and provides insights about the impact of a peanut allergy on the health and wellbeing of their child. However, unlike generic instruments, the FAQLQ does not generate health utilities – the type of HRQoL measure required for economic evaluations by HTA agencies As a result, several studies have explored alternative approaches, including mapping (cross-walk) techniques¹⁵ and discrete choice experiments (DCEs),^16,17 to derive utilities from condition-specific measures in food allergy and other chronic conditions.

Despite this growing body of work, evidence remains limited on the comparative application of mapping and DCE approaches within the same pediatric peanut allergy population, particularly using the FAQLQ-PF. Moreover, few studies have demonstrated how such derived utilities can be applied directly to clinical trial data to inform economic evaluations of emerging peanut allergy interventions. Mapping approaches leverage statistical relationships between condition-specific outcomes and generic utility measures, whereas DCEs enable the direct elicitation of preferences and trade-offs associated with disease-relevant attributes. Employing both methods allows triangulation of health utility estimates and provides a more robust assessment by addressing complementary methodological strengths and limitations.

This is particularly relevant in the context of novel disease-modifying therapies, where improvements in disease-specific quality of life may not be adequately captured by generic instruments alone. More sensitive health utility estimates derived from the FAQLQ-PF may therefore better reflect meaningful treatment benefits, supporting patient-centered decision-making and more accurate economic evaluation of interventions such as epicutaneous immunotherapy.

This study aimed to use mapping and DCE methods to derive health utilities from the FAQLQ-PF. These utilities were applied to existing clinical trial data, generated from an intervention to treat peanut allergy.

Materials and Methods

General Methods

The study comprised three projects. Project 1 developed an algorithm that maps FAQLQ-PF responses to the EQ-5D-3L (henceforth “mapping”). Project 2 utilised a composite DCE with time trade-off (TTO) (henceforth “DCE”). Both Project 1 and Project 2 were conducted via surveys and produced health utilities for the FAQLQ-PF. Project 3 involved applying the derived health utilities to existing clinical trial data. Figure 1 illustrates the overarching methods of the study.

Figure 1 Flowchart of study methods.

Separate groups of respondents were recruited for Project 1 (mapping) and Project 2 (DCE); these are described in more detail below. Respondents in both Projects completed the respective tasks online via a third-party survey platform (Qualtrics LLP). The surveys for Project 1 and 2 were prefaced by an introduction page, which detailed the objectives and instructions for the survey, gave information on information governance and the complaints procedures, and sought participant consent. If a participant declined to participate, the survey was terminated and data deleted. If the participant agreed to continue, the survey proceeded to ask socio-demographic questions.

Projects 1 (mapping) and 2 (DCE) were conducted in the UK, and the surveys associated with them were written in English. The project, studies and surveys were granted ethical approval by the University of York Health Sciences Research Governance Committee (21 April 2021) and have been performed in accordance with the principles stated in the Declaration of Helsinki.

Information about the instruments and clinical trial data used throughout the study are detailed below, followed by the methods for each project of the study.

Instruments

FAQLQ-PF

The FAQLQ-PF¹⁴ provides a parental perspective of the child’s food-allergy quality of life (FAQL). The FAQLQ-PF is separated into 30 “items” across 3 domains: Food-Related Anxiety (FRA), Social and Dietary Limitations (SDL), and Emotional Impact (EI). An “item” is the individual question or statement that respondents are asked to evaluate. These items are split across 3 sections: Section A is completed by all parents, whilst completion of Section B and C is dependent on the child’s age. In a recent systematic review,¹⁸ the FAQLQ-PF received an A grade in terms of the instrument’s psychometric properties.

A previous psychometric analysis of the FAQLQ-PF¹⁹ derived item discrimination parameters for each of the 30 items (Supplementary Table 1). From this, 6 of the total 30 items were selected from this previous analysis for use in Projects 1 and 2 of this study. We reduced the FAQLQ-PF to 6-items removing all age-specific items (which differ between age categories). These 6 items were selected based on the following criteria: the items associated with the highest discrimination parameter reported in the previous analysis,¹⁹ and to ensure that at least one item was selected for each of the three FAQLQ-PF domains. Items were only selected from Section A to ensure that responses were reflective of all survey respondents. Ultimately, two items were present for each of the three domains:

• Item 1 (“…anxious about food”, FRA).
• Item 4 (“…unfamiliar food”, FRA).
• Item 12 (“…limitations on restaurants”, SDL).
• Item 14 (“…limited in social activities”, SDL).
• Item 7 (“…emotional distress”, EI).
• Item 10 (“…negatively affected”, EI).

In the FAQLQ-PF, “response categories” are the set of options provided for each item that the respondent can choose from to indicate the level of impact/frequency of the issue in question. Each of the 7 response categories are assigned a number (score) from 0 to 6, with higher numbers indicating a worse FAQL (ie a higher impact or burden). The response categories allow respondents to quantify their experiences or perceptions in a standardized way.

Using the clinical trial data (described below), we conducted an item response theory (IRT) analysis of the selected 6 items to determine if there was threshold disordering between the response categories. The Graded Response Model (GRM)²⁰ was applied for the IRT analysis as it is able to capture latent traits or variables that are abstract or not directly measurable (eg an individual’s level of anxiety, perception of fairness, or personality type). Specifically, the GRM calculates the probability of selecting a given response category (or higher). The model generates a discrimination parameter (a) for each item along with a number of threshold parameters (b) for the item’s response categories:

Where, Pm the probability of selecting category n (or higher); a is the discrimination parameter; b is the threshold parameter for category m and theta is the latent trait level (or ability).

In this context, threshold disordering would suggest that the adjacent response categories do not discriminate well between different levels of the latent trait. Evaluating the presence of threshold disordering enables response categories that may need adjustment to be identified.

The results of the item-threshold plots are shown in Supplementary Figure 1. These plots illustrate a considerable degree of overlap (ie. lack of clear discrimination) between the thresholds for each of the 6 items. Based on these results, we decided to collapse and re-number the response categories²¹ as follows:

• 1, previously response options 0 to 2 (Not, Barely, Slightly)
• 2, previously response option 3 (Moderately)
• 3, previously response options 4 to 6 (Quite, Very, Extremely)

The GRM was then re-applied to the amended response options (or scale points) to determine the item-threshold discrimination (Supplementary Figure 2) The reduced threshold disordering demonstrated that the three-response-option approach reflected the FAQLQ-PF and was suitable to be used in Project 1 (mapping).

EQ-5D-3L and EQ-5D-Y-Proxy-1

The EuroQol-5-Dimension-3 Levels (EQ-5D-3L)¹¹ is a generic patient-reported outcome measure widely used to assess HRQoL. The instrument comprises five single-item domains: Mobility, Self-Care, Usual Activities, Pain and Discomfort, Anxiety and Depression. The EQ-5D-3L is scored on a three-point response scale ranging from no problems (one) to severe problems (three). The responses to the five domains are converted to health utilities using a country-specific algorithm (value set) on a scale from 0 (“dead”) to 1 (“perfect health”). The UK health utilities range from −0.594 (a state worse than death) to 1. The EQ-5D-Y-Proxy-1²² is a version of the EQ-5D-3L that is adapted for children and adolescents, assessed through a proxy respondent. Project 1 respondents completed the EQ-5D-Y-Proxy-1, with participants asked to rate the HRQoL of a child if they had a hypothetical peanut allergy.

Trial Data

Data from the PEPITES clinical trial, and the associated Open Label Extension, PEOPLE, were used within this study. PEPITES (NCT02636699) was a Phase 3, multicenter, randomized, double-blind, placebo-controlled trial that assessed the efficacy and safety of an epicutaneous peanut patch (DBV712 250 μg) in 356 children aged 4 to 11 years.²³ The participants were required to have a physician-diagnosed peanut allergy and to have reacted to ≤300 mg peanut protein upon a double-blind, placebo-controlled food challenge.²³ HRQoL data were collected via the FAQLQ-PF, which parents completed at 12, 24 and 36 months. Participants were randomized 2:1 to receive DBV712 250μg epicutaneous peanut patch (DBV712, “DBV Technologies SA, France”) or placebo patch daily for 12 months.²³ The primary outcome was defined as reaching a month-12 eliciting dose (ED) of ≥300 mg (entry baseline ED ≤10 mg) or a month-12 ED of ≥1000 mg (entry baseline ED >10 mg and ≤300 mg).²³ All participants completing the PEPITES study were eligible to enroll in the open-label extension (PEOPLE) trial, designed to evaluate the long-term safety, tolerability and efficacy of DBV712 for ≤5 years treatment (NCT03013517).²⁴ This enabled participants who had received the placebo patch for the duration of PEPITES to receive DBV712 for the rest of the study period. The HRQoL in the PEOPLE trial is more relevant because patients are aware of how much their reactive dose has improved and, therefore, have reduced food-related anxiety and emotional impact (improving overall HRQoL).

Project 1: Mapping

Study Sample and Design

In Project 1 (mapping), parents of children were recruited through two UK-based patient advocacy groups, Allergy UK and Anaphylaxis Campaign. These groups distributed the link to the online survey to potential respondents via e-mailing lists and social media.

The target population for Project 1 was parents of children with a peanut allergy aged between 4 and 11 years. The survey involved parents of children with peanut allergies assessing their child’s FAQL by completing both the FAQLQ-PF and the EQ-5D-Y-Proxy 1. The study aimed to recruit:

1. 100 parents of children aged between 4 and 7 with a peanut allergy.
2. 50 parents of children aged between 8 and 11 with a peanut allergy.

The recruitment criteria for the clinical trials and Project 1 are shown in Supplementary Table 2.

Data Analysis

In Project 1 (mapping), descriptive statistics detailing a selection of study-relevant socio-demographic and allergy-related characteristics (eg sex, age, siblings/children, the severity of peanut allergy) were generated. Continuous variables (eg age) were summarized using mean and standard deviation. Categorical variables (eg sex, allergies, siblings, geographical region) were summarized using frequency counts and percentages. All data were analyzed using the R Studio software package (Version 2023.12.0).

Ordinary least squares (OLS) regression models were applied to map the responses from the six FAQLQ-PF items onto the EQ-5D-3L. The actual EQ-5D-3L¹¹ health utility values were calculated from participants’ responses using the UK tariff. Pearson product moment correlations were used to evaluate the level of association between the actual and mapped health utility values. Model fit was evaluated with R², the root mean squared error (RMSE) and mean absolute error (MAE). The initial mapping algorithm was derived using the Allergy UK data, comprising (n=127); this was subsequently tested using data from the Anaphylaxis Campaign (n=32). Only data from those participants who had complete responses for both instruments were included for the mapping algorithm analysis. The two datasets were then combined, and a Bland-Altman plot was generated to evaluate the mean difference against mean agreement between the mapped and actual EQ-5D-3L values. The percentage of difference scores (actual and mapped EQ-5D-3L) falling outside 2 standard deviation of the difference (2SD) was used to evaluate the level of agreement; <5% was deemed to indicate a good level of agreement.²⁵

Project 2: DCE

Study Sample and Design

For Project 2 (DCE), potential participants were recruited by Qualtrics from online panels maintained by the company. Participants in the DCE received a nominal payment for completing the task in the form of e-vouchers. The DCE aimed to recruit a sample from the general UK population; parents who had a child with a peanut allergy were excluded. The study aimed to recruit:

1. 300 parents of children aged between 4 and 7.
2. 300 parents of children aged between 8 and 11.

The sample size was chosen based on a commonly applied rule-of-thumb (Johnson and Orme rule): N> 500c/t*a, where c (for main effects) is the highest level for any given attribute (4 in this study), t is the number of choice tasks per respondent (6, see below) and a is the number of alternatives (2). This resulted in a minimum sample size of 208. The sample size of 300 was therefore deemed sufficient both for analysis purposes as well as to ensure representativeness of the UK population.

The Project 2 (DCE) survey comprised a composite DCE and a TTO approach. Each DCE item was presented in two scenarios. The respondents were required to choose the preferred scenario by considering a list of attributes derived from the six FAQLQ-PF items and a final time duration attribute (the TTO element). Each attribute had four levels: “Not at all”, “A little bit”, “Very much”, and “Extremely”. The time duration indicated a time-until-death duration for the health state outlined by the other attributes in the scenarios. The levels for the time variable were 1, 2, 5 and 10 years.

Each hypothetical scenario was worded from a parental perspective (proxy), requiring respondents to imagine that they are a parent of a child whose HRQoL is affected by a peanut allergy. For example, “Imagine that you are the parent of a seven-year-old child, and they are unable to go to parties of their friends due to a severe allergy…”). It was emphasized that the parent must imagine themselves in the allocated scenarios rather than imagining how others would behave in the state.

Respondents were then asked to choose the preferred scenario described. In effect, each respondent was asked to repeatedly pick out the preferred scenario from a set of different competing options presented on each choice set. An orthogonal fractional factorial design was used for the DCE to reduce the cognitive burden on participants. This resulted in a DCE 3 blocks with 6 scenarios.

The survey also included a vignette, which outlined a hypothetical scenario describing a child with a peanut allergy. The most severe levels of the six FAQLQ-PF items were used in the vignette description. Respondents subsequently completed the EQ-5D-Y-Proxy1 ranking this worst-case health state whilst imagining their child experiencing the symptoms described in the vignette. An example DCE vignette is illustrated in Table 1.

Table 1 Example DCE Vignette

Data Analysis

A conditional logit model was used to derive preference weights for Project 2 (DCE). The parameters derived from the DCE provide indirect preference weights, but not health utilities. The results of the vignette provided an anchor to convert the indirect preference weights from the DCE into utilities for the six FAQLQ-PF items, using the three response categories. This approach followed a previously published anchoring method for indirect preference weights:¹⁶

Where V_hs is the sum of the indirect preference weights (for a given combination of response categories per item); V_best is the indirect utility weight for the best health state (=1); V_worst is the indirect utility weight for the worst health state; U_mean is the mean of these two; and U is the utility value.

Project 3: Application of Derived Health Utilities

In Project 3, the health utilities derived from Project 1 (mapping) and Project 2 (DCE) were applied to the trial data (PEPITES and PEOPLE). Mann–Whitney U-tests were employed to identify statistically significant differences (p<0.05) in the mapped utilities between timepoints. Specifically, the mean differences in health utility values (for both the mapping and DCE tasks) were tested over time for trial participants who had been randomized to the intervention arm in PEPITES, and separately for those initially randomized into the placebo arm. In both instances, baseline differences were evaluated relative to the follow-on open-label trial (PEOPLE). Differences between these groups (eg DBV712/DBV712 versus Placebo/DBV712) were also evaluated for each timepoint. Responsiveness to change was assessed through effect sizes (small >0.2, moderate >0.4, large >0.6) for both forms of the derived health utilities (difference in scores between baseline and month 36 divided by the standard deviation at baseline).

Results

Project 1: Mapping

Participants

There were 159 complete survey responses out of 254 total responses. A summary of the socio-demographic characteristics of the Project 1 (mapping) survey respondents is shown in Supplementary Tables 3–7. The majority of the adults who responded were female, and the majority of the children were male. The median age of the children was 6 years old. Over half of the children had another food allergy in addition to a peanut allergy, and only 12 children had no other allergy-related diseases. Most (59%) children had had their peanut allergy confirmed with a skin test and had been most commonly diagnosed by an allergist at a local hospital (45%). 80% of children had never been admitted to an emergency unit concerning their allergy. The most common peanut-allergy symptoms were “Skin reactions (eg hives)” (70%), “Itching or tingling around the mouth or throat” (50%) and “Swelling of the face, lips, tongue, or throat” (44%).

Mapping the EQ-5D-3L

The Allergy UK dataset (N=127) was used to generate the initial regression weights for the mapping study. The results of the OLS regression are shown in Table 2. The mean health utility for the actual EQ-5D-3L responses was 0.8431 (SD: 0.202, range: 0.193 to 1). The mean mapped utility value was 0.97 (SD: 0.0243, range 0.947 to 1). Figure 2 illustrates the distribution of the actual and mapped EQ-5D-3L with the latter largely falling at the extreme end of scores. The association (Pearson’s product-moment correlation) between the mapped utilities and actual health utility values was 0.199 (p=0.019). The R² was 0.15, RMSE 0.19, and MAE 0.14.

Table 2 Regression (EQ-5D-3L Mapped Utility) Weights for the FAQLQ-PF

Figure 2 Histogram for the EQ-5D-3L utilities and mapped utilities.

The regression weights were fed into the Anaphylaxis Campaign dataset (N=37) and both actual and mapped health utility values were generated. The correlation coefficient (Pearson’s product-moment) was 0.207 (p=0.219). There was a good level of agreement between the mapped and actual EQ-5D-3L values from the combined datasets; only 5.1% of difference scores fell outside the 2SD criterion (Figure 3). The final mapping algorithm is as follows (see also Supplementary Table 1):

Figure 3 Bland-Altman Plot EQ-5D-3L.

M_U(EQ5D) = 0.83 + ((ANX_1j+ANX_2j*-0.0676+ANX_3j*-0.129)+(AFR_1j+AFR_2j*0.131+AFR_3j*0.139)+ (SOCE_1j+SOCE_2j*-.00799+SOCE_3j*-0.0155)+(SOCA_1j+SOCA_2j*-0.0183+SOCA_3j*-0.0533)+(EMO_1j+EMO_2j*-0.0536+EMO_3j*-0.1627)+(AFF_1j+AFF_2j*0.08996+AFF_3j*0.146)).

Project 2: DCE

Participants

Only participants with complete responses to the DCE were retained by the system. There were 767 complete survey responses in total. A summary of the socio-demographic characteristics of the Project 2 (DCE) survey respondents is shown in Supplementary Tables 8–10. There was a relatively even split in the sex of both adults and children completing the survey. The most common allergy-related conditions reported were asthma (17%) and eczema (19%). Most parents had either one (44%) or two (37%) children.

DCE Disutility Values

The disutility values from the DCE are shown in Table 3. The most severe attribute level was associated with the greatest degree of disutility for all six items, with the most severe food-related anxiety (Item 1) having the largest disutility. High disutilities were also shown for emotional distress (Item 7) and limitations on social activities (Item 12).

Table 3 Disutility Values for the FAQLQ-PF (DCETTO)

Project 3: Application of Derived Health Utilities

Finally, the health utilities derived from Projects 1 and 2 were applied to the trial data. A total of 1675 observations were extracted from the PEPITES and PEOPLE trials. A total of N=218 patients’ data were available at baseline (month 0) from the DBV712/DBV712 group, and N=106 for Placebo/DBV712 group.

The Project 1 (mapping)-derived mean health utility values by visit (month) and descriptions for each visit number are shown in Table 4. The mapped EQ-5D-3L health utility values showed a small increase over time for both groups (DBV712/DBV712 versus Placebo/DBV712). This amounted to approximately 0.05 each from baseline to 36 months. The differences from baseline to 12- and 36-months were statistically significant for the DBV712/DBV712 group (p=0.034 and p=0.035, respectively) but not the Placebo/DBV712 group. Larger differences were observed for the Project 2 (DCE) health utilities: 0.12 and 0.09 between baseline and 36 months, respectively, for the two groups. Only the differences from baseline to 12 months (p=0.011), 24 months (p=0.02) and 36 months (p=0.001) were statistically significant for the DBV712/DBV712 group.

Table 4 Utility Values (Mapped EQ-5D-3L and DCE Utilities) by Visit (Month)

The difference between the Project 1 EQ-5D-3L mapped health utilities for DBV712/DBV712 and Placebo/DBV712 was only statistically significant at 36 months (p=0.011). The Project 2 (DCE) health utilities had statistically significant differences at 24 months (p=0.041) and 36 months (p=0.005).

Finally, the effect sizes (ES) for the Project 1 mapped EQ-5D-3L values were −0.26 (95% Confidence Interval, CI: −0.29; −0.24) and −0.29 (95% CI: −0.33; −0.25) in the DBV712/DBV712 and Placebo/DBV712 group, respectively. For the DCE utilities (Project 2) the effect sizes were −0.39 (95% CI: −0.44; −0.35) and −0.34 (95% CI: −0.41; −0.28) for Project 2 (DCE). This indicates a small effect in general for the mapped EQ-3D-3L compared to a moderate effect for the DCE utilities and suggests that the DCE utilities were more responsive to change than the mapped utilities.

Discussion

The aim of the study was to derive health utilities for the FAQLQ-PF which could subsequently be used within health technology assessment (HTA) submissions. This was achieved using two approaches: a direct mapping of the FAQLQ-PF onto the EQ-5D-3L (Project 1), and a DCE (Project 2). The health utilities generated were then applied to clinical trial data for further validation (Project 3).

Project 1 (mapping) allowed health utilities to be generated from the FAQLQ-PF. When applied to trial data, the mapped health utilities showed that there was a statistically significant improvement in HRQoL for patients who had been treated with the intervention during either the preliminary or open-label extension studies. Larger changes were observed in the health utility values generated through Project 2 (DCE). These changes were also statistically significant. The results indicate that the DCE-derived health utilities are more responsive to change than the mapped counterparts, possibly due to a greater effect size (ES) which quantifies the magnitude of an observed effect.

These findings highlight the potential value of preference-elicitation approaches such as DCEs for capturing disease-specific aspects of health-related quality of life that may not be fully represented in generic instruments such as the EQ-5D.²⁶ DCEs can also reduce the cognitive burden on respondents while increasing the amount of relevant information obtained, in turn allowing for more accurate modelling of complex decision-making processes. A potential limitation of the mapping algorithm is the modest model fit (R²).²⁷ This may have resulted from the relatively small sample size, which in turn also restricted the testing of the algorithm to the Anaphylaxis Campaign dataset alone. Future studies could focus on alternative models, eg, Tobit, beta regression to attempt to improve predictive power. The small sample size also meant that additional sensitivity or other cross-validation analyses were not possible to enhance the robustness of the mapping function.

Despite this, there was a high level of agreement between the mapped and actual health utility values, with only around 5% falling outside the threshold. The degree of content overlap between the FAQLQ-PF and EQ-5D-3L may also have had implications for the mapping algorithm. Generic preference-based measures have limitations because they mainly focus on physical dimensions of health, without sufficiently capturing other aspects of emotional and social well-being. Other aspects of well-being have been shown to relate to many aspects of morbidity, mortality and other adverse outcomes in chronic diseases.^9,10 In food allergy, psychosocial domains (particularly allergen related anxiety) have been found to be the most affected by food allergy for both patients and caregivers.^28–30 Higher generalized anxiety scores were also found to be more common among children in households with allergenic food exclusions.³¹

The six FAQLQ-PF items primarily focus on food anxiety, emotional distress, and limitations on social activities; the FAQLQ-PF does not include items such as mobility, pain/discomfort, or self-care. In comparison, the EQ-5D-3L has single-item domains for social activities (“usual activities”) and anxiety and depression. However, the EQ-5D-3L does not include, for example, single-items relevant to unfamiliarity with food. Finally, it is important to note that the seven-scale response category for the FAQLQ-PF was reduced to three. Although the consequent psychometric properties of the FAQLQ-PF were subsequently evaluated, this simplification may have impacted on the results, by reducing sensitivity. Consequently, a pragmatic approach was taken to simplify the response categories. Another potential limitation is the absence of a gold standard for DCE design, presentation format or anchoring method, meaning that EQ-5D-Y valuation studies use a variety of approaches and that there exists no clear consensus on best practice. The final possible limitation to note is that, as with many of other of these types of studies, participant self-selection has the potential to introduce a source of bias in the results. This may be particularly the case, for instance, for the mapping study where participants were recruited from advocacy groups, which should be noted as a possible limitation. This may have also to an extent have affected the online panel; however, the larger sample size for the latter may have mitigated response bias from this group.

Only a few previous studies have published health utilities associated with peanut allergy. A study by Gallop et al (2022) used a combined survey and structured interview approach with adolescent children affected by peanut allergy and their parents/caregivers.³² This study reported a mean utility value of 0.799 for the child’s current HRQoL, calculated from the parent/caregiver proxy-ratings.³² This was virtually identical (0.80) to the combined dataset from Project 1 of this study. However, it should be noted that the proxy-ratings for participants using the FAQLQ-PF were younger in the Gallop et al study (mean age 6 years)³² Furthermore, the increase in mapped health utilities demonstrated in the trial data (0.05) broadly mirror the 0.057 increase in health utilities associated with the development of tolerance (to 6–8 peanuts) reported by Gallop et al.³²

To strengthen the generalizability of the proposed algorithms, future research should validate both the mapping and DCE-derived utilities in larger and more diverse populations, including children with other food allergies and allergic conditions. Cross-cultural validation across different healthcare systems and preference contexts will also be important, as health state valuations may vary by country and cultural setting, particularly in paediatric populations. Further work could explore the adaptation of the FAQLQ-PF utility algorithms to related instruments or allergy types, such as tree nut or multi-food allergies, to assess their broader applicability.

In addition, the incorporation of child self-reported quality-of-life measures alongside caregiver-reported outcomes may help refine health utility estimates and better capture age-specific perceptions of disease burden, particularly among older children and adolescents.

Peanut allergy has a significant detrimental impact on the HRQoL of affected children.^29,33 Technologies aimed at desensitizing children to peanut exposure require economic evaluations for HTA approval in certain jurisdictions. HTA agencies may require economic evaluations to be completed using health utilities. However, few studies have published utility data in this field. Therefore, from a policy and HTA perspective, the availability of condition-specific utilities derived from the FAQLQ-PF has the potential to improve the sensitivity of cost-effectiveness models in pediatric peanut allergy, particularly where generic instruments may underestimate quality-of-life gains associated with reduced anxiety, dietary restrictions, and risk of accidental exposure. These improvements in utility measurement may, in turn, influence reimbursement and coverage decisions for pediatric immunotherapies by providing a more accurate representation of patient-relevant benefits in economic evaluations.

As disease-modifying treatments for food allergy continue to emerge, the use of more responsive, disease-specific health utility estimates may support more informed appraisal of long-term value and affordability within HTA frameworks, complementing traditional clinical endpoints and strengthening the evidence base for decision-making.

To our knowledge, this is amongst the first studies to generate health utility algorithms derived from a condition-specific patient-reported outcome measure in paediatric food allergy in the absence of direct utility data. This is important in building the consensus needed on best practice for evaluation generally and specifically in food allergy to ensure optimal intervention results.

Conclusion

Two complementary methods – mapping and discrete choice experiment (DCE) approaches – were applied in order to generate health utilities from the FAQLQ-PF. Both methods produced valid and plausible health utility estimates that were consistent with the study findings, with the DCE-derived utilities demonstrating greater responsiveness to changes in health-related quality of life. Although further research is warranted with larger samples, this is an important first step in producing disease-specific health utilities from the FAQLQ-PF for cost-effectiveness evaluations.

The availability of FAQLQ-PF–derived utilities has practical implications for multiple stakeholders. For health economists and HTA bodies, these utilities enable more sensitive cost-utility analyses that better reflect the lived burden of pediatric peanut allergy and the benefits of emerging disease-modifying therapies. For clinicians and decision-makers, improved utility measurement may support more patient-centered appraisal of interventions by capturing quality-of-life changes that generic instruments may overlook. In this context, the application of the derived utilities to clinical trial data illustrates their potential value in the economic evaluation of interventions such as epicutaneous immunotherapy.

Future research should focus on validating these utility algorithms in larger cohorts, assessing their performance across different cultural and healthcare settings, and exploring their applicability to other food allergies. Additional work incorporating child self-reported outcomes alongside caregiver-reported measures may further enhance the relevance and robustness of these utilities. Beyond peanut allergy, the methodological framework presented here may also serve as a template for deriving utilities from other disease-specific quality-of-life instruments, supporting broader advances in health economic evaluation for paediatric chronic conditions.