Estimating health state utilities associated with a rare disease: familial chylomicronemia syndrome (FCS)

Abstract

Aims: Familial chylomicronemia syndrome (FCS) is a rare genetic disorder with no currently approved therapies. Treatments are in development, and cost-utility analyses will be needed to examine their value. These models will require health state utilities representing FCS. Therefore, the purpose of this study was to estimate utilities for FCS and an associated episode of acute pancreatitis (AP).

Methods: Because it is not feasible to gather a large enough sample of patients with this extremely rare condition to complete standardized preference-based measures, vignette-based methods were used to estimate utilities. In time trade-off interviews, general population participants in the UK and Canada valued health state vignettes drafted based on literature review, clinician input, and interviews with patients. Four health states described variations of FCS. A fifth health state, describing AP, was added to one of the other health states to evaluate its impact on utility.

Results: A total of 308 participants provided utility data (208 UK; 100 Canada). Mean utilities for FCS health states ranged from 0.46 to 0.83, with higher triglycerides, more severe symptoms, and a history of AP associated with lower utility values. The disutility (i.e. utility decrease) of AP ranged from –0.17 to –0.25, with variations depending on the health state to which it was added. Utility means were similar in the UK and Canada.

Conclusions: The vignette-based approach is useful for estimating utilities of a rare disease. The health state utilities derived in this study would be useful in models examining cost-effectiveness of treatments for FCS.

Keywords:

• Utility
• familial chylomicronemia syndrome
• FCS
• rare disease
• acute pancreatitis
• time trade-off

JEL CLASSIFICATION CODES:

• I10
• I19

Introduction

In recent years, there has been significant growth in development of medications for treatment of rare diseases^1–9. Economic modeling, including cost-utility analysis (CUA), is often needed to examine the value of these treatments¹⁰. These models require utilities, which are values representing the strength of preference for health states, to calculate quality-adjusted life years (QALYs). Health technology assessment (HTA) reviewers often prefer that utilities are derived from generic preference-based measures such as the EQ-5D completed by patients.^11–15 However, to derive utilities from patient-completed instruments, a sufficiently large sample of patients is needed to represent health states included in economic models. For rare diseases, it may not be feasible to recruit a large enough sample of patients living in the relevant health states. It can also be difficult to gather a sample of patients with rare conditions to validate generic instruments for use in the target population^16,17.

Despite these challenges, cost-effectiveness analyses are needed to inform resource allocation decisions for treatment of rare diseases, and these analyses often require health state utilities. Therefore, alternative methods for estimating utilities should be considered. Vignette-based methodology is one approach that can be used to estimate utilities for rare diseases when it is not feasible to collect preference-based data from a large enough sample of patients. Health state descriptions (often called vignettes or health states) can be drafted based on the best available information, possibly including input from patients, to ensure that the health states accurately represent the typical patient experience. Then, utilities can be estimated in a valuation study with general population respondents.

One rare disease that will soon require economic modeling is familial chylomicronemia syndrome (FCS). This genetic disorder is most commonly linked to mutations in the gene encoding lipoprotein lipase (LPL), an enzyme that breaks down chylomicron lipoprotein particles. It can also be linked to mutations in genes which encode other proteins that are necessary for proper LPL function. This leads to an increase in triglyceride levels^18,19. Patients with FCS often experience acute pancreatitis (AP) episodes, fatty deposits on the skin (eruptive xanthomas), abdominal pain, fatigue, impaired cognition, and enlargement of the liver or spleen. FCS is extremely rare, occurring in about one in one million people^20,21. Until recently, there were no approved therapies for FCS, and most patients attempt to manage the condition through an extremely restrictive diet^18,22. A novel treatment called volanesorsen has recently been authorized for use in the European Union²³, and it has been shown to reduce triglyceride levels in phase II and III trials^24,25. In the phase III trial, 77% of patients with FCS had triglycerides less than 750 mg/dL at three months of treatment²⁵. As treatments are introduced, utilities are needed for economic modeling to examine their value.

The purpose of this study was to estimate health state utilities representing FCS using vignette-based methodology. To ensure that the health state vignettes were a reasonably accurate representation of the patient experience, patients with FCS and physicians who treat FCS were interviewed to support the development of health state content. The health states were valued in time trade-off (TTO) interviews with general population respondents in the UK and Canada.

Methods

Health state development

Five health state descriptions were drafted and refined based on published literature, interviews with FCS patients, and interviews with clinicians specializing in FCS. The literature search was performed to inform development of the patient and clinician interview guides and identify characteristics of FCS that should be included in the health states. This literature search focused on hypertriglyceridemia,^26–28 FCS case studies providing descriptions of patients,^29–32 impact of FCS^18,33, the FCS diet^22,34, and pancreatitis.^35–38

Multiple rounds of telephone interviews were conducted with three clinicians specializing in lipidology (endocrinologist, cardiologist, nurse practitioner). As FCS is a rare disease, the clinicians saw relatively few patients with the condition, but all reported that they had sufficient experience to describe the typical patient experience of the disease, its treatment, its impact, pancreatitis, and the restrictive diet. The endocrinologist, cardiologist, and nurse practitioner reported seeing approximately 10, five, and four patients with FCS per year, respectively. The three clinicians were located in Texas, Arizona, and New York.

The first interview with each clinician included a series of open-ended questions designed to elicit descriptions of patients’ typical experience with FCS, its impact, the FCS diet, treatment, and AP. Descriptions of these patient experiences were generally consistent across the three clinicians, and no contradictory information emerged from the interviews. Clinician’s responses were used to inform the development of the first draft of the health states. During subsequent follow-up discussions, each clinician reviewed the health states and suggested edits as necessary. These discussions continued until all three clinicians agreed that the health states provided clear and accurate representations of the typical patient experience. In addition, the three clinicians agreed that, taken together, these five health states provided reasonable coverage of the range of patient experiences with FCS. Given their professional experience and the state of the research on this rare disease, they did not believe that additional health states with finer distinctions among triglyceride levels and associated symptoms could be supported by available data or clinician input.

Interviews with three patients (two female, one male; ages 47, 56, and 57; all living in the US) were conducted to confirm that the health state descriptions were consistent with their personal experience of FCS. All three patients reported that they could comment on health states with high or low triglycerides because their triglyceride levels and associated symptoms had varied during their lifetime with FCS. In addition, all had experienced AP, and they were therefore able to comment on the AP health state. One of the patients had experienced only one episode of AP, while the others had experienced multiple episodes. These interviews began with discussion of their personal experience with FCS, including symptoms, diet, impact, and pancreatitis. Then, the patients reviewed draft health states and provided feedback on clarity, accuracy, and relevance of the health state descriptions. Edits were made based on patient input. All three patients agreed that the health states were consistent with their personal experience of FCS.

Five health state vignettes were developed for valuation in this study. Four of these health states were chronic (i.e. unchanging over time, valued with a 10-year time horizon), describing FCS varying by triglyceride levels and history of AP: low triglycerides and no AP history (Health State A); high triglycerides and no AP history (B); low triglycerides and history of AP that had resolved (C); and high triglycerides and history of AP that had resolved (D). Key differences between the four chronic health states include symptom severity, concerns about recurring pancreatitis, and the amount of missed work due to symptoms. The fifth health state described an AP attack (E) as a temporary event including symptoms, duration, hospitalization, and treatment. The FCS diet was described in an introductory information page that applied to all health states.

In the health states and the introductory information page, FCS was described in simple language so that the information would be easily comprehensible to general population participants. When medical terminology was used, it was defined in simple terms. For example, triglycerides were defined as “fats.” In addition, all health states included descriptions of symptom severity and impact so that respondents could form impressions and preferences based on the description of patient experience, rather than the triglyceride level.

The health states were formatted as a series of bullet points on individual cards. The bullet points were organized into categories with headings intended to facilitate comprehension. For example, the headings for health state A were disease and diet; physical symptoms; cognitive symptoms; and impact on usual activities and quality of life. For health states C and D, the section describing lingering symptoms from previous pancreatitis was titled “previous experience.” See Supplementary Appendix A for the introductory information page and Supplementary Appendix B for the full text of the final health state vignettes.

Participants

Participants were required to be at least 18 years old; reside in the UK or Canada; and be able to give informed consent and complete protocol requirements. Inclusion criteria did not specify clinical criteria because interviews were intended to yield utilities that may be used in cost-utility analyses for submission to HTA agencies, which often prefer that utilities represent general population values^11,14,15. Participants in this general population sample were recruited via newspaper and online advertisements. Potentially interested participants who responded to the advertisements were screened by telephone to assess eligibility prior to attending their study interview, and eligibility was confirmed at the beginning of the interview. During screening, an effort was made to recruit a sample that was similar to the general populations of the UK and Canada with regard to age, gender, and racial/background. In addition, employment status was monitored to ensure that the rate of unemployed individuals was roughly comparable to that of the general population of each country.

Pilot study

Methods were tested in a pilot study with 20 participants in London (10 women; mean age = 41.75 years; age range 19–52 years). Health states were valued in a TTO task, and then revised based on feedback from pilot participants. Revisions affected formatting and organization, rather than content. For example, information common to all the health states (e.g. FCS diet) was removed from each health state and presented in a background information page that applied to all health states. Participants had no difficulty using the 10-year time horizon for the four chronic states, followed by the one-year time horizon for AP.

Utility interview procedures and scoring

After the pilot study, the TTO valuation study was performed in the UK (Edinburgh and London) in April 2018 and Canada (Montreal and Quebec City) in August 2018. The strengths and limitations of TTO methods, as well as comparisons to other approaches for valuing health states, have been discussed extensively in previous literature³⁹. All participants provided written informed consent, and the study was approved by an independent institutional review board (Ethical & Independent Review Services, Study 18016). In Canada, participants were given the choice of English or French for the interview and study materials.

First, participants reviewed the introductory information page. Then, they ranked the four chronic health states (A, B, C, D) from most preferable to least preferable to familiarize themselves with the health state content. After the ranking task was completed, these four health states were valued in a TTO utility elicitation task with a 10-year time horizon, following commonly used methodology³⁹. For each health state, participants were offered a series of choices between spending 10 years in the health state or spending varying amounts of time in full health. Full health was described as “You are healthy. You do not have any health problems. You can perform your usual activities without difficulty (getting around the community, work, school, social, family, and physical activities).” Choices were presented in six-month (5%) increments, with time in full health alternating between longer and shorter durations (i.e. 10 years, 0 [dead], 9.5, 0.5, 9.0, 1.0, 8.5, 1.5…). The utility value was calculated based on the point of indifference between y years in the health state being valued and x years in full health (utility = x/y).

When respondents perceived a health state to be worse than dead, the task and scoring procedures were altered as described in previous literature⁴⁰. Participants were offered a choice between dead (choice 1) and a 10-year life span (choice 2) beginning with varying amounts of time in the health state being rated, followed by full health for the remainder of the 10-year life span. The negative utility scores were calculated with a scoring approach commonly used to avoid highly skewed distributions (utility = –x/10, where x is the number of years in full health, and 10 is the number of years in the total life span of choice 2).

To estimate disutility associated with an attack of AP, respondents rated a sequence of two health states in a TTO task with a one-year time horizon with one-month trading increments. Each participant first rated either C or D, followed by the identical health state with the addition of an attack of AP (health state E) occurring in the middle of the one-year time period. The one-year time horizon was used to assess the impact of this temporary event because the addition of a brief event is unlikely to have a measurable impact on preference for a 10-year time horizon. Because the time horizon was exactly one year, the utility decrease associated with the addition of health state E to either C or D represents the QALY decrement associated with AP. Participants were randomized as to whether health state C or D was used as the context for AP. The one-year TTO follows the same procedures and scoring as those described above for the 10-year TTO.

Interviewers were trained to identify illogical responses in order to maintain data quality. For example, ranking a health state with an attack of AP as preferable to the same health state without an attack of AP would be illogical. When respondents provided unexpected preferences in the ranking or TTO task, the interviewer would query the response to ensure that the respondent understood the task and the health state and was stating their preferences accurately. Interviewers were trained to avoid biasing any responses or suggesting that any response was correct or incorrect.

Statistical analysis procedures

Statistical analyses were completed using SAS version 9.2 (SAS Institute, Cary, NC). Continuous variables were summarized in terms of means and SDs, and categorical variables were summarized as frequencies and percentages. Subgroups were compared with independent t-tests, and health state utilities were compared with paired t-tests.

Results

Sample characteristics

A total of 318 participants attended interviews. Ten participants were unable to complete the interview procedures, including three in Canada and seven in the UK (i.e. difficulty understanding the TTO task or health states). Therefore, the final sample consists of 308 participants (n = 208 in the UK and n = 100 in Canada; demographic results presented in Table 1). The most common health conditions were anxiety (n = 57, 18.5%), depression (n = 45, 15.9%), and hypertension (n = 33, 10.7%). No participants reported a diagnosis of FCS. One participant reported knowing somebody with FCS.

Table 1. Sample characteristics.

Characteristics	UK (N = 208)	Canada (N = 100)
Age (Mean, SD)	44.8 (15.23)	47.4 (17.35)
Gender (n, %)
Male	104 (50.0%)	50 (50.0%)
Female	104 (50.0%)	50 (50.0%)
Racial/Ethnic Background (n, %)
White	148 (71.2%)	88 (88.0%)
Mixed	13 (6.3%)	3 (3.0%)
Asian	18 (8.7%)	0 (0.0%)
Black	28 (13.5%)	3 (3.0%)
Other^a	1 (0.5%)	6 (6.0%)
Marital status (n, %)
Single	137 (65.9%)	57 (57.0%)
Married/cohabitating/living with partner	69 (33.2%)	38 (38.0%)
Other^b	2 (1.0%)	5 (5.0%)
Employment status (n, %)
Full-time work	83 (39.9%)	52 (52.0%)
Part-time work	57 (27.4%)	10 (10.0%)
Other^c	68 (32.7%)	38 (38.0%)
Education level (n, %)
University degree	95 (45.7%)	35 (35.0%)
No university degree	113 (54.3%)	65 (65.0%)

^aOther self-reported racial/ethnic background in the UK was Persian (n = 1). Other in Canada included Arab (n = 1), Haitian (n = 1), and unspecified (n = 4).

^bOther self-reported marital status in the UK was engaged (n = 2). Other in Canada included civil union (n = 1), couple (n = 1), and free union (n = 3).

^cOther self-reported employment status in the UK included apprentice (n = 11), business owner (n = 1), care for disabled child (n = 1), disabled (n = 4), ESA (n = 1), retired (n = 22), self-employed (n = 5), student (n = 18), unemployed (n = 14), and 0 h contract (n = 1). Other in Canada included autonomous (n = 1), business owner (n = 2), disabled (n = 1), homemaker (n = 2), retired (n = 25), student (n = 5), and unemployed (n = 2).

Health state utilities

In the introductory ranking task with health states A through D, there was little variation in rank order. In the UK, 203 of the 208 participants (97.6%) ranked the health states from most preferable to least preferable in the order A, C, B, D. The other five UK participants ranked the health states as A, B, C, D. In Canada, all 100 participants ranked the health states as A, C, B, D.

Mean utilities of the four chronic states (Table 2) followed logical patterns with more severe FCS associated with lower utilities. Values ranged from 0.46 to 0.80 in the UK and 0.47 to 0.83 in Canada. There were no statistically significant differences between the UK and Canada on any of these utilities. All pairwise comparisons between mean health state utilities (e.g. comparing A to B) were statistically significant in both the UK and Canada (all p < .0001).

Table 2. Health state utilities^a.

	Mean (SD) utilities			Disutility of acute pancreatitis
FCS health states	UK (N = 208)	Canada (N = 100)	Difference between UK and Canada	UK	Canada
10-year health states
A: Low TG, no AP	0.80 (0.21)	0.83 (0.18)	–0.03 (0.20)
B: High TG, no AP	0.60 (0.33)	0.61 (0.33)	–0.02 (0.33)
C: Low TG, previous AP	0.74 (0.23)	0.78 (0.19)	–0.04 (0.22)
D: High TG, previous AP	0.46 (0.42)	0.47 (0.36)	–0.01 (0.40)
1-year health states
Subgroup rating E relative to C	N = 107	N = 48
C without E	0.71 (0.30)	0.80 (0.17)	–0.08 (0.27)
C with E (adding an AP event to health state C)	0.46 (0.40)	0.60 (0.28)	–0.14 (0.37)	–0.25 (0.28)	–0.20 (0.20)
Subgroup rating E relative to D	N = 101	N = 52
D without E	0.53 (0.38)	0.57 (0.35)	–0.04 (0.37)
D with E (adding an AP event to health state D)	0.33 (0.43)	0.40 (0.34)	–0.07 (0.40)	–0.20 (0.23)	–0.17 (0.16)

Abbreviations. AP, Acute pancreatitis; FCS, Familial chylomicronemia syndrome; SD, Standard deviation; TG, Triglycerides.

^aTTO utilities are on a scale anchored to 0 representing dead and 1 representing full health.

The purpose of the one-year TTO was to estimate the disutility of an episode of AP, described in health state E (Table 2). An AP episode resulted in disutilities of –0.25 and –0.20 when added to health states C and D, respectively. There were no significant differences between the UK and Canada on the one-year valuations of health state D with or without E. The one-year utilities of C were significantly greater in Canada than the UK, both without the addition of E (t = –2.17, p = .03) and with E (t = –2.52, p = .01). However, there were no significant differences between the UK and Canada in disutilities associated with AP, regardless of whether C or D was used as the baseline state. All disutilities were statistically significant (all p < .0001). For example, in the UK, the one-year utility of 0.71 for health state C was significantly greater than the value of C with the addition of an AP episode (0.46).

Most participants were willing to trade time to avoid living in the four chronic health states. Only three participants (1.4%) in the UK and two participants (2.0%) in Canada were unwilling to trade time to avoid any of the chronic health states.

In the UK, negative utility scores were rare for health states A (1.0%), B (3.8%), and C (1.4%). Rates of negative scores were somewhat higher for health states D (9.6%) and the valuations of C and D with the addition of AP (9.3% and 14.9%, respectively). Rates of negative utilities were lower in Canada: A (0%), B (3.0%), C (0%), D (5.0%), C with E (2.1%), and D with E (7.7%).

Discussion

Results followed expected patterns, with lower utilities for health states representing more severe FCS as indicated by higher triglycerides, more severe symptoms, and history of AP. Utilities and differences between utility scores were similar in the UK and Canada. There are no previously published utilities for FCS that can be used for comparison, but the utility values appear to be in a reasonable range based on comparison to EQ-5D utilities reported for type 2 diabetes, another endocrine-related condition⁴¹. The higher utilities for FCS with low triglycerides are in a similar range to previously published utilities for type 2 diabetes without complications, while the lower utilities for FCS with high triglycerides are similar to those of type 2 diabetes with more severe complications such as peripheral neuropathic pain or retinopathy.

When using the resulting utilities in a CUA, modelers should be aware of the difference between chronic and temporary health states. Because the four chronic states did not change over time, the utilities can be applied for any duration in a model (consistent with constant proportional trade-off, a key assumption of the QALY model). In a CUA, these utilities would have to be applied to the proportion of patients who can be categorized as having either high or low triglycerides, with or without a history of AP. These proportions will likely vary across models, depending on the clinical trial or real-world data used as the source of information.

Unlike the utilities of the four chronic health states, the disutility of AP is tied to the time horizon of the task because the AP attack was described as temporary event that changed over time. These utility decreases associated with AP represent the impact of AP on a one-year period, and the disutility should be applied in a model as a QALY decrement. The AP episode was applied to two health states (C and D). Anecdotal evidence from previous studies has suggested that temporary medical events have more impact on utility when added to milder health states, rather than more severe states. Current results support this hypothesis. In both countries, the AP disutility was larger when added to the milder health state (C) than to the more severe state (D).

Limitations of vignette-based methodology should be acknowledged. Like most medical conditions, FCS symptoms range from mild to severe, and utilities derived from vignettes cannot represent every level of severity. Instead, the utilities represent preferences for discrete health states as described in the vignettes, which were drafted to represent typical patient profiles. The extent to which these utilities might differ from values derived from patient-completed measures such as the EQ-5D or Health Utilities Index^13,42 is unknown. However, methodology was selected to maximize comparability to standardized utility valuation methods. For example, chronic health states were valued by general population participants in a TTO task with a 10-year time horizon, similar to methods used to derive the original EQ-5D scoring tariffs⁴³.

One potential risk with some health state vignettes is that it may be necessary to provide more information than would be included in health states derived from generic preference-based instruments. For example, whereas a health state based on the EQ-5D would have five bullet points, the length of health states in the current study ranged from five bullet points to 13 bullet points. Although efforts were made to keep the health states as brief as possible, more information was necessary to capture the ongoing impact of AP in health states C (10 bullet points) and D (13 bullet points). It is possible that these longer health states could have introduced reading comprehension and memory difficulties for some participants. To mitigate these challenges, interviewers carefully reviewed each bullet point with the respondents to ensure that they understood and attended to all details. In addition, illogical rankings and TTO valuations were queried to ensure that they were based on good comprehension and actual preference, rather than a misunderstanding of health state content. In general, it appears that these efforts were effective because the resulting utilities followed logical patterns. Still, it is possible that some participants may not have considered or remembered all parts of the two longer health states when performing the TTO task.

Another methodological limitation is that the health state vignettes were drafted and refined based on input from a relatively small number of clinicians and patients. Because FCS is extremely rare, it is difficult to recruit a large sample of patients and clinicians with relevant experience. Furthermore, the patients and clinicians were based in the US, rather than the UK and Canada, which were the locations of the subsequent utility valuation study. Because of differences between the healthcare systems, it is possible that treatment approaches for FCS may vary across these three countries. In sum, although health states were drafted based on the best available information, a larger and more geographically diverse sample of patients and clinicians could have led to slightly different health states and utility values.

Generalizability of the sample may also be a limitation. Although efforts were made to avoid over-representing any group with regard to age, gender, ethnic/racial background, or employment status, the sample was not recruited to be nationally representative in either the UK or Canada. Future research with a larger sample could be conducted to elicit utilities for FCS that are truly nationally representative.

Despite limitations, this study demonstrates that the vignette-based approach can be useful for estimating utilities of rare diseases. This methodology may be considered when it is not feasible to recruit a large enough patient sample to derive utilities necessary for economic modeling. The utilities estimated in this study can be used in models examining the cost-effectiveness of new treatments for FCS.

Transparency

Declaration of funding

This study was funded by Akcea Therapeutics.

Declaration of financial/other relationships

Louis Matza and Timothy Howell are employees of Evidera, a company that received funding from Akcea for time spent conducting this research. Glenn Phillips was an employee of Akcea Therapeutics at the time this study was conducted, including the time when this manuscript was drafted. Nicole Ciffone and Zahid Ahmad received funding from Akcea for time spent on this research.

A peer reviewer on this manuscript has disclosed that they have been involved in clinical trials of volanesorsen sponsored by Akcea and that their institution has received payments for their attendance at advisory board meetings. The peer reviewers on this manuscript have no other relevant financial relationships or otherwise to disclose.

Author contributions

LM and GP co-directed the study including study design, health state development, data analysis, and data interpretation. TH directed protocol development and the data collection, while assisting with health state development. NC and ZA provided clinical input to help shape the study design, health states, methods, protocol, and interpretation. LM and TH drafted the manuscript, and all other authors provided input and approval.

Ethics compliance

Ethical Approval/Informed Consent: All procedures performed in this study were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Informed consent was obtained from all individual participants included in the study.

Acknowledgements

The authors would like to thank Hayley Karn and Gordon Parola for assistance with the pilot study; Haylee Andrews, Ella Brookes, Christopher Langelotti, and Natalie Taylor for assistance with participant recruitment; Kristen Deger, Meredith Hoog, Christopher Langelotti, Haylee Andrews, Melissa Garcia, and Ella Brookes for assistance with the UK data collection; Beenish Nafees, Carole Charland, Karen Hofman, and Matthew Sparks for assistance with the Canadian data collection; Benjamin Arnold and the FACITtrans team for performing the translations; Ray Hsieh and Christine Thompson for statistical programming; and Amara Tiebout for editorial support.