Valuing Maintenance of Visual Function in Retinitis Pigmentosa: A Discrete Choice Experiment with Patients and the General Public

Abstract

Background and Objectives

Retinitis pigmentosa (RP) is a progressive and sight-threatening condition with few treatment options. Patients with RP experience variable progressive vision deterioration that can severely affect functioning. Currently, there is no published literature to indicate the preferences that patients with RP place on potential treatments. This study compared patient preferences to those of a sample from the general public.

Methods

An online, discrete choice experiment (DCE) elicited preferences of a United States (US) sample for maintaining visual function-related activities of daily living (VF-ADL). The data were stratified by sociodemographic and clinical characteristics. Participants completed best-/worst-choice tasks using predefined attributes and varying levels, including time to deterioration in mobility and ability to read associated with worsening visual function.

Results

A total of 150 patients with RP and 301 public participants completed the DCE. By the age of 32–40 years, VF-ADL deteriorated substantially in patients versus public (81% vs 13% reported lower combined mobility/ability to read). Patients with RP placed the greatest value on ability to get around (relative attribute importance [RAI] 51%; standard error [SE] 3.4; 95% confidence interval [CI] 45–58), whereas the public sample preferred ability to get information (RAI 48%; SE 4.7; 95% CI 39–57). Patients with RP placed less weight on extending life expectancy compared with public participants.

Conclusions

Patients with RP demonstrated preferences in VF-ADL and life expectancy associated with deterioration in visual function and potential treatment benefits that differ from those of the public. Considering patient preferences is critical to understand disease burden and estimate the value of therapeutic interventions.

Supplementary Information

The online version contains supplementary material available at 10.1007/s41669-026-00638-x.

Plain Language Summary

A survey of 150 patients with retinitis pigmentosa (RP) and 301 general public participants in the United States demonstrates that patients and the general public have very different perceptions of the burden of RP and the value of treatment outcomes. Patients report a significant burden of RP. For instance, by the age of 32–40 years, visual function (abilities related to vision) had deteriorated substantially in patients versus the general public (81% of patients vs 13% of the general public reported lower combined mobility and/or ability to read). Patients put much more weight on delaying the deterioration of visual function, preferring to delay loss of visual function over extending their life expectancy. Patients also placed greater emphasis on maintenance of mobility, whereas the general public participants highly valued the ability to read. These differences indicate the importance of including preferences from patients with RP in value assessments of potential new treatments.

Supplementary Information

The online version contains supplementary material available at 10.1007/s41669-026-00638-x.

Key Points for Decision Makers

Patients with retinitis pigmentosa (RP) have meaningfully different preferences for treatment outcomes compared with the general public.

Patients with RP placed greater value on delaying visual function deterioration and maintaining the ability to get around compared with participants public who placed greater importance on the ability to get information.

When evaluating RP disease burden and assessing RP treatment outcomes, it is important to consider the patients’ perspectives, as they are meaningfully different from those of the public.

Introduction

Retinitis pigmentosa (RP) is a group of inherited vision disorders characterized by early onset and rapid progression of vision loss due to breakdown of the retinal tissue [1–3]. Worldwide, RP affects approximately 1 in 4000 to 1 in 5000 individuals, and 1 in 4000 in the United States (US) [3, 4]. RP is currently estimated to affect approximately 1.5 million people worldwide [1]. Patients with RP typically experience visual difficulties at night or in poor lighting and in very bright lighting, along with impaired peripheral vision and reduced light to dark adaptation [5]. Such issues significantly impact activities of daily living, including the ability to read, leisure activities, mobility and navigation, and driving, which in turn impairs quality of life [5, 6].

X-linked RP (XLRP) accounts for approximately 10–15% of patients with RP worldwide, within which more than 70% of those affected harbor mutations in the RPGR gene [3, 7]. Patients with XLRP present with early onset of night blindness [3], often within the first or second decade of life, progressing to complete functional blindness between 40 and 50 years of age [7]. This is earlier than other forms of RP; XLRP is considered to have a worse long-term prognosis in comparison with other forms of RP [4]. For instance, patients with XLRP are four times more likely to have visual acuity ≤20/200 (legal blindness in the US) compared with those with autosomal dominant RP [1].

Although voretigene neparvovec (Luxturna^®), a retinal gene therapy targeting RPE65, is currently approved for RP in the US, no treatments are specifically approved for XLRP. Only a few management options are available to delay vision loss associated with XLRP, including acetazolamide, vitamin A palmitate, special sunglasses, and retinal implants. However, there are over 100 treatments in development to address this unmet need for RP treatments, including multiple therapies for XLRP [8, 9]. Retinal gene therapies are currently under evaluation for treatment of XLRP through restoring functionality of the RPGR gene, halting or slowing disease progression, and maintaining or possibly improving visual function-related activities of daily living (VF-ADL) [10].

Visual function deterioration in patients with RP is multifaceted and includes changes in both visual acuity and peripheral vision. There is significant variance in the phenotypic presentation of patients with RP, even when they have identical genotypes [3, 4]. Assessing potential new therapies therefore requires the measurement of these multiple aspects, alongside the perceived value of such treatment outcomes. In turn, this begs the question of whose perspective should be used to value treatment outcomes [11].

Because they represent taxpayers or funders in many healthcare systems, and because patients are thought to adapt to their condition and thus undervalue treatment gains, the views of the general public are often considered most relevant to outcome valuation [12–19]. However, it is unclear whether this logic can be extended to RP treatment funding decisions in the US, as the taxpayer is not the funder of many US healthcare systems. Unlike in socially funded healthcare systems, US value assessment framework (US-VAF) reports, which often leverage patient perspectives, are primarily used by insurance providers to inform healthcare coverage decisions in the US [20]. This emphasis on the patient perspective is supported by key US agencies, such as the Patient-Centered Outcomes Research Institute [21], the Academy of Managed Care Pharmacy [22], and the Agency for Healthcare Research and Quality [23].

However, US institutions are not unanimous in their adoption of the patient perspective, and there are still some US decision makers for whom the public perspective informs healthcare decision-making priorities. For instance, in the case of RP treatment funding decisions, the Institute for Clinical and Economic Review (ICER)’s assessment of the effectiveness and value of voretigene neparvovec (Luxturna^®) and sonpiretigene isteparvovec (MCO-010) considered health state value results from studies that considered visual impairment only within the general public, not patient preferences [12, 24].

Whether the approach of identifying utility values based on general public preferences will provide the best health outcomes for the public will depend on whether and how patient and public preferences vary. Differences in patient and public preferences are expected to arise from two sources, which point to different evaluation approaches: (1) patients’ greater insight into the burden of disease may cause them to put greater weight on treatment benefits, and (2) the possibility that patients adapt to their condition may cause them to put less weight on treatment benefits. Comparisons of patient and public preferences conclude that, on average, patients give higher values to their health states than non-patients, but this varies by patient and by valuation method [25].

The primary aim of the current study was to compare preferences for treatment outcomes among patients with RP and the public in the US, specifically focusing on visual function deterioration over the lifespan. A subgroup of patients with XLRP was enrolled and compared with the overall population of patients with RP. Preferences were elicited for activities of daily living impacted by visual function deterioration—the ability to get around (ATGA) and the ability to get information (ATGI).

Methods

Study Design

A 30-min, online, US-specific survey, including a discrete choice experiment (DCE) preference elicitation instrument, was used to measure preferences for visual function maintenance (specified via time to deterioration in selected VF-ADL) and life expectancy. Potential participants were required to complete a screening questionnaire to determine their eligibility. Eligible participants were then required to provide online consent before being presented with the purpose of the study and a series of treatment features. Warm-up tasks with questions were included to check participants’ understanding and engagement with the study. The DCE required participants to complete a set of best-/worst-choice tasks (see Sect. 2.4), choosing between hypothetical treatment effects described by their performance against a set of predefined attributes and levels.

This study was conducted in multiple stages following good research practices and received ethical approval from Ethical & Independent Review Services (#21142), in accordance with the Declaration of Helsinki and the Common Rule. The study confirmed patient consent and was compliant with patient confidentiality and ethics requirements.

Study Population

Patients with RP were recruited through patient databases and physician referrals, whereas members of the general public were identified via recruiter databases. For the qualitative pilot testing phase, the target sample size was ten; for the main study phase, the target sample size was 450. For the qualitative phase, a 1:1 ratio of the general public participants to patients with RP was targeted, and for the main study, a 2:1 ratio was targeted. Participants in each phase received financial remuneration (for the qualitative pilot, $125 for participants with RP and $100 for general public participants; for the main study, $50 for participants with RP and $10 for general public participants). Soft recruitment quotas were utilized to ensure a 20% minimum participation level in each age category (≥18–<25, 25–<32, and 32–40 years of age). RP diagnosis was self-reported by patients. Included patients were required to meet the criteria shown in Table 1.

Table 1.

Inclusion and exclusion criteria (main study phase only)

Inclusion criteria	Exclusion criteria
1. Between 18 and 40 years of age	1. Participated in an earlier part of the study
2. US resident	2. Unwilling or unable to provide (online) consent
3. Willing and able to provide (online) consent to participate in the study	3. Significant visual impairment (unable to drive and move around without an aid, even in well-lit conditions, and unable to read, even with the use of aids or adaptations)a
4. Willing and able to access a computer or laptop and complete an online survey
5. Self-reported diagnosis of RP (patients only)
6. Self-reported English speaker

RP retinitis pigmentosa, US United States

^aThis exclusion criterion was added following the pretesting phase of the study

Given the challenges posed by a priori minimum sample size estimation for DCEs [26, 27], no formal sample size determination took place. However, a target sample size of 100 in each target population was set, in line with the typical sample sizes of healthcare DCEs in the literature.

This study sought to characterize the preferences of patients with RP and explore differences in subgroups by enrolling patients with other forms of RP, such as XLRP.

Attribute Selection

The selection of attributes and levels was based on a targeted literature review (TLR) of preference literature (both quantitative and qualitative; see Supplementary Tables 1 and 2), a review of available clinical data for RP treatments, and two stakeholder workshops, which included discussions with 15 experts in the field (12 representatives from the sponsor and three representatives from Evidera, with ophthalmologists and pharmacoeconomists represented). The TLR identified 368 potentially relevant papers, including 74 quantitative and 294 qualitative studies. Following abstract and full-text screening, three quantitative and seven qualitative studies relating to patient preferences were included for review. Of the qualitative publications, two were systematic reviews referencing qualitative data (Supplementary Fig. 1). The TLR was supplemented by identifying endpoints, including patient-reported outcomes (PROs), in two clinical trial protocols for RP provided by the study sponsor.

Attributes were selected through the following process: first, potentially relevant attributes, including commonly used PROs, were identified through a review of the literature. A review of the extracted data identified benefit measures, risk measures, and nonclinical measures (treatment duration, quality of life, etc.) and are summarized in Supplementary Table 3. Second, overlapping concepts were removed to simplify the design. Clinical data and attributes that were identified as important to patients in the TLR were discussed during the stakeholder workshops to finalize the relevant content used for the DCE. Through this process, functional impact was chosen over visual impairment or quality of life impact because it provided participants with more information and because of the overlap of function with quality of life (i.e., visual impairment alone does not necessarily translate to a quality of life impact that would be most critical to value determination). Third, functional impact concepts were mapped to each other, with particular importance assigned to attributes represented in relevant PROs.

Three attributes were included in the DCE: time to deterioration in ATGA, time to deterioration in ATGI, and life expectancy. Time to deterioration in ATGA and ATGI had attribute levels of delays of 10, 20, and 30 years until further deterioration, whereas life expectancy had levels of 70, 75, and 80 years of age (Table 2). Although treatments were not expected to improve life expectancy, it was included as a numeraire, allowing the values of other attributes to be expressed in terms of changes in survival that would give the same value. Participants were provided VF-ADL descriptors for each attribute to inform their selection in the best-/worst-choice tasks (Supplementary Table 4).

Table 2.

Selected attributes in the DCE survey

Attribute	Attribute definition	Attribute levels
	ATGA was defined as ability to move and drive with glasses or contact lenses, if used
Time to deterioration in ATGA	Deterioration was defined as follows:	10 years 20 years 30 years
	If current state → Then deteriorated state
	Able to get around → Unable to get around in the dark or at night
	Unable to get around in the dark or at night → Unable to get around in poor light
	Unable to get around in poor light → Unable to get around, even in good light
	ATGI was defined as ability to read with glasses or contact lenses, if used
Time to deterioration in ATGI	Deterioration was defined as follows:	10 years 20 years 30 years
	If current state → Then deteriorated state
	Able to read without an aid → Able to read only with an aid
	Able to read only with an aid → Unable to read, even with an aid
Life expectancy	–	70 years of age 75 years of age 80 years of age

ATGA ability to get around, ATGI ability to get information, DCE discrete choice experiment

Experimental Design of the DCE

Participants completed a series of choice tasks; each task involved indicating the best and worst preferred selection for three potential hypothetical treatments (treatments A, B, and C; Fig. 1). A best–worst scaling approach was adopted rather than a more conventional DCE to optimize the number of data points collected within a given sample size. Ahead of the experimental choice task, attributes were introduced one at a time with attribute-specific exercises used to engage participants, and a full practice question was included to familiarize the participants with the format of the DCE choice questions. Prior to the final DCE, qualitative pilot interviews (with five patients and five general public participants) were conducted to explore if the chosen attributes, their levels, and the format of their presentation were complete, meaningful to respondents, and well understood, which informed the final DCE prior to collection of the study data.

Fig. 1.

Example of DCE choice task^a. DCE discrete choice experiment. ^aThe illustrated DCE choice task forms an example of the content that was presented to 20-year-old respondents, with an adequate baseline visual function (note the age timeline starting at 20 years)

The VF-ADL and life expectancy experience on each treatment in the choice task were tailored to patients with RP based on their current age, level of mobility, and ability to read. Patients with RP used a starting point based on their current level of visual impairment. For ATGA, baseline visual function was self-rated by patients with RP and categorized according to mobility (with glasses or contact lenses, if used): (1) able to get around; (2) unable to get around in the dark or at night; (3) unable to get around in poor lighting. Corresponding deterioration states were (1) unable to get around in the dark or at night, (2) unable to get around in poor lighting, and (3) unable to get around, even in good light. For ATGI, baseline visual function was defined according to the ability to read (with glasses or contact lenses, if used): (1) able to read without an aid; (2) able to read only with an aid. Corresponding deterioration states were (1) able to read only with an aid and (2) unable to read, even with an aid.

For the public participants, the starting point/current status was based on a simulated status with estimation of impairment at their current age, as if they had RP. Participants were asked to imagine themselves at a specific baseline level of visual function, based on their age, before the design was applied. Treatment profiles were defined by different times to initial visual function deterioration (10-, 20-, or 30-years’ delay), after which visual function was assumed to continue to deteriorate at a constant rate (linearly) over the remaining life expectancy.

A D-efficient experimental design generated by Ngene software (version 1.3.0; ChoiceMetrics) was used to optimize the combinations of attributes and levels. To minimize the risk of ordering effects, DCE choice tasks, treatment profiles within tasks, and attributes within treatment profiles were randomly ordered [28]. Three nonexperimental choice tasks assessed internal validity, including a repeated choice task to assess consistency of responses, and a dominance test that presented a superior option that all patients were expected to prefer.

Other Endpoints

The impact of visual function impairment on quality of life was assessed using the EuroQol 5-dimension 5-level instrument (EQ-5D-5L).

It was not possible to include the productivity impacts of visual function deterioration in the attribute list. This was because the productivity impact associated with each level of visual function deterioration was participant specific and not known to the research team at the time the experiment was designed. For instance, for any given level of deterioration in ATGA, the impact on productivity would be a function of type of employment—it might be expected that deterioration would impact the productivity of someone whose job involved driving more than someone who had a desk-based job. Given this constraint, a novel method was adopted to estimate the effect of productivity impact on preferences for delayed deterioration in visual function. To estimate the contribution to preferences from the productivity and nonproductivity impacts of visual function deterioration, patient- and treatment profile-specific productivity levels were retrospectively estimated and included in the analysis (see Sect. 2.6). To facilitate this, patients with RP were asked to assess the impact of different levels of visual function on productivity, using a 4-level scale: (1) has no effect on my ability to work or study, (2) impacts the quality of my work or study, (3) limits the types of work I am able to do and I might need to consider changing jobs, and (4) completely prevents me from working or studying.

Statistical Analyses

Study results were analyzed according to an a priori statistical analysis plan using R statistical software (version 4.0.5; R Foundation for Statistical Computing, Vienna, Austria).

Descriptive statistics were reported for sociodemographic and clinical characteristics and productivity impact. Sociodemographic, clinical information, and categorical variables (e.g., sex) were summarized by frequency statistics (N [%]). Continuous variables (e.g., age) were summarized using means, standard deviations (SDs), medians, and ranges.

Bivariate analysis investigated how self-reported productivity, employment status, income, and quality of life (EQ-5D-5L) varied with VF-ADL, and how visual function varied by age, type of RP (XLRP or other forms of RP), and severity of RP. Bivariate outcomes were descriptively listed as percentage of sociodemographic variable by age category (≥18–<25, 25–<32, and 32–40 years of age) or RP severity category (less and more severe). Less severe RP included the following states: (1) able to get around/able to read without an aid, (2) able to get around/able to read only with an aid, and (3) unable to get around when very dark or at night/able to read without an aid. More severe RP included the following states: (4) unable to get around when very dark or at night/able to read only with an aid, (5) unable to get around in poor lighting/able to read without an aid, and (6) unable to get around in poor lighting/able to read only with an aid.

Several measures of preference data quality (i.e., dominance [participants chose the superior option as the most preferred treatment] stability, serial nonparticipation, response time, dominated decision-making) were combined into a quality (attentiveness to the task) score, signaling the number of passed tests, leading to a score ranging from 0 (when all quality tests are failed) to 5 (when none of the quality tests are failed). The sensitivity of multinomial logit (MNL) results to the changes in the underlying quality of the choice data was examined by progressively excluding respondents with a lower quality score and re-estimating the MNL model.

The best (most desirable) and worst (least desirable) choices were jointly analyzed following a simultaneous decision-making model, as no information about the timing/sequence of these decisions was recorded. The predictors of the worst decisions were multiplied by −1 before entering the choice model to measure disutility (instead of utility for best decisions).

The best model for choice modeling (the econometric analyses of treatment choices) was identified following a multistep approach (Supplementary Fig. 2). A dummy-coded MNL model, which assumes preference homogeneity, uncorrelated choice, and independence from irrelevant alternatives, was used as the initial/reference model. In this model, all the attributes were treated as categorical variables. Following the estimation of the MNL model, the predictive validity of models with and without interaction effects assessed whether interaction effects between attributes should be added to the reference model. The error-component logit (ECL) model was estimated, and the likelihood ratio (LR) test was used to compare its performance to the MNL model and determine whether the assumption of choice independence at the core of the MNL model was supported by the data. Finally, to test the effects of patient personal characteristics on the results, an interacted multinomial logit (IMNL) model was created (Supplementary Methods 1). Four mixed logit (MXL) models compared levels of ECL and MXL statistical performance (via Bayesian Information Criteria [BIC] and Adjusted McFadden Pseudo-R² [APR]) to determine whether the assumption of preferences homogeneity made by the ECL model was supported by the study data: (1) BMXL 1: All the attributes followed a normal distribution; (2) BMXL 2: All the attributes followed a multivariate-normal distribution; (3) BMXL 3: All the attributes followed a lognormal distribution; and (4) BMXL 4: All the attributes followed a multivariate-lognormal distribution.

The MXL model was estimated using a Bayesian technique (Monte Carlo Markov Chain [MCMC]) implemented within the Apollo package (version 0.2.7) for R software (version 4.1.2). For the MCMC simulation, 300,000 draws were used prior to convergence; 30,000 draws were used post-convergence; every tenth draw was retained, leaving 3000 effective draws to compute statistics of interest (mean [SD]). For patients with RP, the model was estimated based on 2700 observations (choices) from 150 participants. For the public participants, the model was estimated with 5418 observations from 301 participants. A 2-tailed z test was used to evaluate the significance of model estimates at a conventional 5% level of statistical significance.

Best fitting models were run with and without the addition of a simulated productivity attribute. Leveraging each participant’s estimate of the impact of combinations of visual function deterioration on productivity levels, participant- and treatment profile-specific estimates of productivity impact were estimated for each choice task. To test the impact on preferences of the productivity and nonproductivity impacts of visual function deterioration, this simulated productivity attribute was included in the analysis.

Preference models computed marginal rates of substitution (MRS) and were used to derive how much life expectancy patients with RP would sacrifice to delay deterioration of visual function. Main outputs were reported in years with standard error (SE) and 95% confidence intervals (CIs). The estimated preferences were used to determine the relative importance of treatment attributes, measured using mean (SE and 95% CI) relative attribute importance (RAI) scores.

The RAI score was calculated to measure the importance that patients place on each attribute. The difference in preference weights between the best or most preferred level of an attribute and the worst or least preferred level of the same attribute provided an estimate of the relative importance of that attribute over the range of levels included in the DCE. RAI was given by the following formula:

$${\text{RAI}}_k=100\times \frac{{\widehat{\beta }}_k^{\text{max}}-{\widehat{\beta }}_k^{\text{min}}}{{\sum }_{k\in \left[1;K\right]}\left({\widehat{\beta }}_k^{\text{max}}-{\widehat{\beta }}_k^{\text{min}}\right)}$$

where (${\widehat{\beta }}_k^{\text{max}}-{\widehat{\beta }}_k^{\text{min}}$) corresponds to the largest change in utility obtained with moving from the worst (least desirable) to best (most desirable) level of the $k$th attribute. If the worst level also happens to be the reference category for the dummy coding of the effects, then ${\widehat{\beta }}_k^{\text{min}}=0$.

Results

Study Populations and Sociodemographic and Clinical Characteristics

The DCE survey was launched in November 2021 and completed in January 2022. In total, 150 patients with RP and 301 public participants completed the survey. The attrition of both groups is presented in Supplementary Fig. 3. Patients and public participants had comparable mean (SD) age at 29 (6.7) and 31 (6.2) years, respectively (Fig. 2). However, although similar percentages of patients with RP were aged either 25–<32 years (37%) or 32–40 years (35%) given patient soft target methodology requirements, public participants were predominantly aged 32–40 years (53%). Proportions of patients with RP and public participants aged ≥18–<25 years were comparable (28% and 24%, respectively). Similar proportions of patients and public participants were female (47% vs 48%). In terms of clinical characteristics, most patients received their RP diagnosis ≥6 years before the time of the study (64%; n = 96), had undergone genetic testing for RP (93%; n = 140), had a family history of RP (67%; n = 101), and wore special sunglasses (61%; n = 91). In total, 37 patients (24.7%) had XLRP, and the remaining 113 had other forms of RP.

Fig. 2.

a Sociodemographic characteristics of all study participants and b clinical characteristics of patients with RP. RP retinitis pigmentosa, XLRP X-linked retinitis pigmentosa. ^aOther includes Asian or Asian American, Native Hawaiian or other Pacific Islander, and American Indian or Alaska Native. ^bOther treatment includes cells/tissue replacement, currently on clinical trial testing a new treatment, treatment unknown, and report as other

Over three times as many of the public participants achieved at least a college/university degree as patients with RP (68% vs 22%), and almost three times as many of the public participants as patients were employed full time (63% vs 23%). More patients than the public participants had self-provided health insurance (41% vs 26%) or were covered by Medicaid (17% vs 8%). Full comparisons of sociodemographic and clinical characteristics in patients with RP and public participants are presented in Fig. 2.

Patients with more severe RP (n = 79) had higher prevalence of unemployment (19% vs 8%), sick leave/short-term disability (8% vs 0%), long-term disability (10% vs 0%), and self-identification as homemakers (16% vs 7%) versus patients with less severe RP (n = 71). Mean (SD) EQ-5D-5L utility scores were higher in patients with less severe RP (0.91 [0.16]) compared with those who had more severe RP (0.66 [0.34]). Notably, the mean EQ-5D utility score was 0.87 (0.2) for the public participants. With respect to the individual domains of the EQ-5D-5L, there were consistently lower percentages of patients with more severe RP versus less severe RP reporting “no problems with” mobility (56% vs 87%), self-care (72% vs 93%), usual activities (38% vs 73%), pain/discomfort (42% vs 80%), and anxiety/depression (30% vs 72%).

Although there were no differences between patients and the public participants in age brackets of ≥18–<25 and 25–<32 years, by 32–40 years of age, only 4% of patients with RP were still mobile compared with 81% of the public participants (Supplementary Fig. 4a). The proportion of patients with RP who had full mobility and could read without an aid decreased from 60% in the group ≥18–<25 years of age to 4% in the group 32–40 years of age, whereas in the public participants, these functions were reduced by a lesser extent at these ages (86% in the group ≥18–<25 years of age; 70% in the group 32–40 years of age; Fig. 3a). Similarly, by 32–40 years of age, visual function deterioration limited the type of work and may have required a job change in 45% of patients with RP compared with 4% of public participants (Fig. 3b). In total, 81% of patients 32–40 years of age versus 24% of the public participants 32–40 years of age required an aid beyond their regular reading glasses to be able to read (Supplementary Fig. 4b).

Fig. 3.

a Relationship between age and mobility and ability to read^a and b ability to work or study.^b ATGA ability to get around, ATGI ability to get information, RP retinitis pigmentosa. ^aMobility and ability to read higher and lower functioning levels were classified by the attribute definition provided in Table 2 (with higher functioning defined as “able to get around” and “able to read without an aid” and lower functioning defined as “unable to get around in the dark or at night,” “unable to get around in poor light,” and “able to read only with an aid”). Patients with a mix of lower functioning and high functioning in ATGA and ATGI were not included in either category, and percentages do not add up to 100% because only high- and low-functioning patients were included. The completed reports of individual functioning level in mobility and ability to read are provided in Supplementary Fig. 4. ^bImpact of RP on productivity was measured as “has no effect on my ability to work or study,” “impacts the quality of my work or study,” “limits the types of work I am able to do and I might need to consider changing jobs,” and “completely prevents me from working or studying.” Patients who responded “impacts the quality of my work or study” were not included in either category, and percentages do not add up to 100%

Discrete Choice Analysis: Main Model Selection, Data Quality, and Preference Results

The BMXL 1 model was found to have the best fit and was selected as the reference model for both public participants (APR 40.6%) and patients with RP (APR 46.9%) samples. This suggests that there was a high level of heterogeneity in both patients and the public. Interactions between attributes and 13 separate patient characteristics were added to the model to test how preferences varied with observable characteristics. Although three of these characteristics (total yearly income, racial background, and ethnic background) improved the model fit, these improvements were not statistically significant (Supplementary Table 5). No characteristics of the public participants improved model fit per the IMNL model.

Data quality was high for patients and moderate for the public, with a median (interquartile range) response time of 4 (3–5) min in patients and 5 (4–7) min in the public. Most participants (>69%) had high health literacy or high numeracy (Supplementary Table 6). Choice data quality was high for patients with RP and moderate for the public participants. Overall, 100% of patients with RP and 97% of the public participants did not always select only A, only B, or only C in their choices. In addition, 91% of patients and 67% of the public participants passed the dominance test. In total, 75% of patients and 63% of the public participants provided consistent answers in the repeated choice task. MNL-based sensitivity analysis further demonstrated that preference estimates were not sensitive to respondent performance on these validation metrics. As a consequence, models specified on data from all participants are presented below.

Patients with RP placed greater weight on ATGA (RAI = 51.4% [SE 3.4; 95% CI 44.7–57.8]) over ATGI (RAI = 39.2% [SE 3.1; 95% CI 32.9–45.2]), whereas the public participants expressed the opposite preference (RAI for ATGA = 23.0% [SE 3.9; 95% CI 15.3–30.7] vs RAI for ATGI = 48.1% [SE 4.7; 95% CI 38.9–57.3]) (Fig. 4).

Fig. 4.

RAI (%) of patients with RP versus public participants. ATGA ability to get around, ATGI ability to get information, RAI relative attribute importance, RP retinitis pigmentosa, SE standard error

Public participants placed more weight on increases in life expectancy than did patients (Fig. 4). It was possible to use these preference data to estimate that, for a public sample, delaying deterioration in ATGA and ATGI by 10 years (from 10 to 20 years) would provide the equivalent utility of an additional 6.7-year and 14.0-year improvement in life expectancy, respectively (survival equivalence, p < 0.001). Patients with RP attached a positive value to increases in life expectancy. However, given very strong preferences for delaying deterioration relative to the public, this preference was insufficiently large in too small a sample to measure with statistical certainty. As such, it was not possible to estimate the survival equivalence of delaying deteriorations in visual function for patients.

Discussion

This study is the first to measure preferences for visual function deterioration in both patients with RP and the public. Unsurprisingly, both groups had strong preferences for delaying the deterioration of visual function, but there were interesting differences between patient and public preferences that point to the public’s lack of appreciation of the impact of deterioration in visual function. First, whereas public participants placed greater value on maintaining ATGI, patients placed greater value on maintaining ATGA. This suggests that either the public overestimates the impact of deterioration in ATGI and/or underestimates the impact of deterioration in ATGA, or patients are able to adapt to the impact of visual function deterioration on ATGI.

Second, the public significantly underestimates the extent of impact of visual function deterioration. Specifically, patients placed a much greater value on maintaining their visual function than the public, as measured by a difference in the groups’ willingness to trade off delayed deterioration for longer life expectancy—the public was willing to trade off delays in visual function deterioration for extra life expectancy, whereas the patients were not. The importance that patients attribute to delaying visual function deterioration was so strong that they were not willing to trade off delaying visual function deterioration even for an extra 10 years of survival. Given these results, it was not possible to estimate this trade off with sufficient statistical certainty. That was not the case for public participants, for whom the utility gained by maintaining ATGA and ATGI for 10 years was the equivalent of an additional 6.7 years and 14 years, respectively, of improvement in life expectancy.

Third, the public may place a greater weight on the productivity impacts of visual function deterioration, possibly because a higher proportion of the public sample was working full time, although these findings were not statistically significant and will require further testing.

Fourth, although there was a high level of heterogeneity in preferences within both the patient and public samples, it was not possible to attribute this heterogeneity to any participant characteristics. For instance, no relationship was identified between age or disease severity and patient preference for delaying deterioration, suggesting that patients quickly obtain an understanding of the nature and extent of the impact of visual function deterioration.

The study also provided evidence of the burden of RP. Patients with more severe RP had higher prevalence of unemployment, short- or long-term disability leave, and self-identification as homemakers, as well as lower self-reported quality of life. These impacts could already be seen at a young age: for the 18–25-year-old age group, 86% of the public were higher functioning in ATGA and ATGI, but only 60% of patients in this age group showed similar functionality. Further, patients’ visual function deteriorated swiftly: only 11% of patients in the 25–32-year-old age group were higher functioning compared with 76% of the public. The employment impact of RP was not yet manifest in the 18–25-year-old group but emerged in the 25–32-year-old group, among whom 25% reported visual function resulting in a limit in the type of work they were able to do.

These findings have implications for the valuation of treatments for RP. Arguments for basing treatment valuation on public preferences are based partly on the concern that patients adapt to their conditions and thus “undervalue” treatments. Arguments for basing treatment valuation on patient preferences point to the public not understanding the impact of a disease. The differences between patient and public preferences suggest that it is difficult for those without RP to understand the burden of the disease. Not only do patients with RP place more importance on maintaining visual function, they also prioritize different elements of the impact of visual function deterioration. Incorporating patient preferences into treatment valuation will bring important information on disease burden to the valuation process. Further, the existence of heterogeneity that cannot be attributed to specific patient characteristics suggests that it may be important to engage individual patients in their assessment of treatment options.

Some limitations should be noted in this study. First, the study population was not designed to be fully representative of the patient with RP population; rather, sample stratification was employed to allow estimation of the impact of characteristics on preferences. Soft quotas were applied to ensure generally comparable samples sizes within age categories (which were defined based on literature regarding progression of RP). Second, given the challenges of recruiting a large enough sample of patients with XLRP, a broader sample of patients with RP was enrolled and a strategy assessing for differences in patients with and without XLRP was adopted. Although differences between the two were analyzed and determined to be not significant, small differences may exist that we could not detect given the study power. Third, it was necessary to make assumptions about the pace of deterioration of visual function following variation in the first step in deterioration that comprised the experimental design. Specifically, the design assumed linear progression as it was uncertain how changes in time to progression would impact the progression of future health states. Fourth, significant preference parameters for the life expectancy attribute for patients were not identified due to much lower preference being given to survival than anticipated during the design stage. Further work should include a wider range of survival attribute levels. Lastly, novel methods were adopted to assess preferences for productivity, which may have affected the findings.

Study limitations were managed by following preference research good practice [29]. These practices include tenets such as ensuring all participants had a good understanding of the choice tasks through provision of detailed training materials; pilot testing the study design; assessing the internal validity of responses using consistency and rationality tests (dominance and stability/repeated task tests); response pattern analysis (serial nonparticipation and dominated decision-making) and response time metrics; conducting health literacy and numeracy comprehension assessments; and testing the sensitivity of preferences to variation in participant internal validity. Implementing a best/worst DCE approach ensured that preference data were captured effectively; simplifying participants’ decision-making process also helped mitigate limitations [30, 31].

Conclusions

This analysis using an online DCE demonstrated that patients with RP had different preferences than the public, with very large value placed on delaying visual deterioration. This reinforces the substantial impact of RP on patients’ lives and demonstrates the importance of considering patient preferences in the assessment of treatments for RP.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

This study was sponsored by Johnson & Johnson.

Declarations

Author contributions

K. Marsh and M. Trapali contributed to methodology, software validation, and formal analysis. M. Ingham contributed to conceptualization, methodology, writing, visualization, supervision, project administration, and funding acquisition. S. Martin contributed to conceptualization, writing, supervision, and funding acquisition. All authors fully reviewed and commented on all previous versions of the manuscript and approved the final manuscript.

Data availability

The data sharing policy of Janssen Pharmaceutical Companies of Johnson & Johnson is available at https://www.janssen.com/clinical-trials/transparency. As noted on this site, requests for access to the study data can be submitted through Yale Open Data Access (YODA) Project site at http://yoda.yale.edu.

Disclosures

KM and MT are employees of Evidera. MI was an employee of Johnson & Johnson at the time the study was conducted. SM is an employee of Johnson & Johnson.

Ethics approval

This study was conducted in multiple stages following good research practices and received ethical approval from Ethical & Independent Review Services (#21142), in accordance with the Declaration of Helsinki and the Common Rule.

Consent to participate

All study participants provided consent via a web survey by opting into the project following determination of eligibility but prior to the commencement of any research-related survey administration or project involvement.

Consent for publication (from patients/participants)

All study participants provided written informed consent, which included notification that the study may lead to published results.

Code availability

The code used in this study is available from the corresponding author upon reasonable request.