Key Points for Decision Makers
Because numerous long-term prophylactic treatments with varying benefit-risk profiles are available for hereditary angioedema, it is important to understand what treatment attributes are most meaningful to patients.Among the long-term prophylactic attributes evaluated in this discrete-choice experiment study, individuals with hereditary angioedema cared most about reducing the frequency of hereditary angioedema attacks, mode and frequency of administration, and reducing gastrointestinal side effects.
Introduction
Hereditary angioedema (HAE) is an unpredictable debilitating genetic disorder characterized by recurrent episodes of localized, cutaneous, or submucosal swelling that most commonly affects the face, skin, gastrointestinal (GI) tract, and extremities [1–3]. Abdominal pain is a hallmark indicator of HAE in addition to swelling, and patients often experience symptoms such as fatigue, nausea, and flu-like symptoms [3–7]. Hereditary angioedema is caused by mutation of the C1-inhibitor (C1INH) gene and has a prevalence rate of up to 1.6–1.9 per 100,000 people [8–10].
Some HAE episodes are unpredictable and are not initiated by specific triggers; however, local trauma, emotional stress, and infections may contribute to the onset of an attack [3, 4, 6]. Although episodes of HAE typically resolve within 1–5 days, swelling of the larynx or throat can be life threatening [3, 6] and if an HAE attack is left untreated, it can lead to increased morbidity and mortality [11–13]. Patients often experience anxiety and depression as a result of functional disability and facial disfigurement, both of which negatively impact health-related quality of life [11, 12, 14, 15].
International guidelines for the management of HAE recommend that on-demand treatments should be considered for alleviating all HAE attacks, and patients should have constant access to these treatments [16–18]. When a patient is at an increased risk of an attack because of known triggers (e.g., surgery), short-term prophylactic treatment is recommended and long-term prophylactic (LTP) treatments are recommended to adequately decrease the HAE attack rate and attack severity [16–18]. The overall goal of HAE treatment is to improve patients’ well-being by achieving total disease control (i.e., no HAE attacks), which is sometimes achievable with currently available LTPs, although some patients still experience breakthrough attacks [16–19].
As of July 2025, the US Food and Drug Administration had approved nine medications for the treatment of HAE, where in the last decade, several new treatments, notably LTP treatments, have been investigated and indicated as first-line treatments for HAE (i.e., berotralstat and lanadelumab) [20].
All approved on-demand and LTP medications are administered either orally, or via intravenous or subcutaneous routes, which can be associated with injection-site reactions (ISRs) [e.g., site swelling and site erythema], and typically require dosing regimens every 3–4 days, although newer LTPs have less frequent dosing regimens (every 2, 4, or 8 weeks) [16, 21–23]. In 2020, an oral LTP, berotralstat, was approved by the Food and Drug Administration [24], which offered an alternative mode of administration to intravenous or subcutaneous medications that mitigates ISRs. However, berotralstat must be administered once daily and is associated with adverse events, including abdominal pain, vomiting, diarrhea, back pain, and gastroesophageal reflux disease [24].
Specifically, the APeX-S trial phase II and III studies of berotralstat reported GI adverse events in 41–50% of patients [25, 26]. Some therapies, such as antisense prophylactic treatments and factor XIIa inhibitors, are associated with adverse events that are mild or moderate in severity [27]. Further, some individuals with HAE may be prescribed only short-term medications, suggesting an unmet demand and/or opportunity for new prophylactic treatments that meet the untreated patients’ needs [28].
Because numerous treatment options with varying benefit-risk profiles are available and are becoming available for HAE, it is important to better understand what treatment attributes are most meaningful to patients.
Discrete-choice experiments (DCEs) are widely used across a range of therapeutic areas to quantify patients’ and physicians’ treatment preferences, and the trade-offs they are willing to accept between the benefits and risks of those treatments [29–33]. Thus, the objectives of this study were to (1) quantify patient preferences for the attributes of LTP HAE treatments, (2) understand the trade-offs people with HAE are willing to make among treatment features, and (3) identify the factors that predict patients starting or switching to a new treatment.
Methods
Study Design
This study employed a DCE survey instrument conducted among individuals with HAE, who evaluated a series of choices between pairs of hypothetical LTP HAE treatment alternatives. Each hypothetical treatment alternative was defined by differing levels of treatment attributes, which included efficacy (i.e., time to maximum efficacy, reduction in HAE attack frequency), safety attributes (i.e., risk of ISRs, risk of GI side effects), and process attributes (i.e., frequency of and mode of administration). A top-down approach was used to select the attributes as suggested by the International Academy of Health Preference Research (Table 1) [34–40].
Selection of key attributes that differentiated the comparators of interest (berotralstat, lanadelumab, garadacimab-gxii, donidalorsen, and C1 esterase inhibitor subcutaneous [human]) was informed by the product labels for approved products at the time this study was conducted [41–43] and clinical data, including clinical trials [26, 44–48] as well as a literature review of studies looking at preferences for HAE treatment and unmet needs [14, 28, 49, 50]. For example, the efficacy attribute was explained as a reduction in attack frequency to reflect the primary endpoint of most HAE trials, which aim to demonstrate a significant reduction in the rate of angioedema attacks in the treatment arm (e.g., NCT05139810 [51]; NCT06007677 [52]). The text in the survey carefully described the concept of time to maximum efficacy and how it related to the attribute describing reduction in attack frequency.Table 1Attributes and levels for the discrete-choice experimentTechnical attribute labelPatient-facing attribute labelAttribute levelsTime to maximum efficacyTime until the prophylactic treatment provides full benefit1 week2 weeks5 weeks10 weeksReduction in attack frequencyReduction in attacks over 6 months once the treatment is providing full benefits45% (average of approximately 6 or 7 attacks in the 6 months once the treatment provides full benefit)85% (average of approximately 2 attacks in the 6 months once the treatment is providing full benefits)90% (average of approximately 1 attack in the 6 months once the treatment is providing full benefits)Mode of administrationHow you take the treatmentTablet a,bVial and syringePrefilled syringeAutoinjectorDosing frequencyHow often you take the treatmentOnce a day aTwice a week aOnce every 4 weeksOnce every 8 weeksRisk of ISRRisk of a mild-to-moderate ISR0% b5%20%55%Risk of a GI side effect (grade 1 or 2)Risk of a mild-to-moderate GI side effect0%7%15%25%GI gastrointestinal, ISR injection-site reactionaThe experimental design incorporated a restriction to ensure that the dosing frequency was always once a day or two times a week when the treatment was administered as a tabletbThe experimental design incorporated a restriction to ensure that the risk of ISR was always 0% when the treatment was administered as a tablet
The treatment attributes and levels were used to create a set of DCE questions (Fig. 1). A full factorial experimental design, involving all possible combinations of attributes and levels, has resulted in an infeasible number of choice sets (12,288 potential profiles and 150,994,944 possible choice sets).
Therefore, a fraction of this full factorial design was created using D-optimality criteria (i.e., searching algorithms to maximize D-efficiency that minimizes the variance-covariance matrix of the estimated parameters), aligning with good research practice guidelines [53]. In addition to main effects, the experimental design accommodated restrictions between the risk of ISR and oral administration and between oral administration and frequency to prevent unrealistic combinations from occurring (i.e., ISR with a tablet) [54, 55].The experimental design constructed a fraction of the full factorial design (all possible combinations of attribute levels), including 72 questions that were split into six blocks of 12 DCE questions. Respondents were randomly assigned to one of the six blocks of 12 DCE questions, generating cross-sectional data for estimation of a multivariate preference model.Fig.
1Example of a discrete-choice question
In addition to the DCE, the preference survey included the Angioedema Control Test (AECT), a patient-reported outcome instrument, as a measure of respondents’ angioedema control. Demographic and clinical characteristics also were collected, as were self-reported data about previous treatments for HAE, including willingness to hypothetically switch LTP medications and the importance of prescribing frequency in their treatment decisions. All data collected from the online survey were deidentified.
The final survey instrument is presented in Exhibit S1 in the Supplementary Material (ESM).
The entire survey, including the wording of the attributes and levels, was pretested [37] with a convenience sample of 17 eligible participants with self-reported HAE using 1-hour 1:1 semi-structured cognitive debriefing interviews (called “think-aloud” interviews), conducted via a teleconference. Survey comprehension questions were used to confirm respondent comprehension and cognitive burden. If comprehension questions were misunderstood, clarification was provided and key information was repeated.
Attribute descriptions and survey graphics and text were revised according to feedback from patients. After pretesting, the final survey instrument was administered online. The study team used a holistic approach to review all data to ensure that each returned survey corresponded to a unique respondent (e.g., by investigating repeat IP addresses).
Study Population
Individuals with HAE were recruited by a US-based HAE patient advocacy group and provided informed consent. Individuals eligible for study participation included adults aged ≥18 years residing in the USA who were proficient in English and had a self-reported HAE diagnosis. The RTI International Institutional Review Board reviewed the study protocol (protocol ID STUDY00022609) and deemed the study exempt from full review on 13 September, 2023.
Sample size calculations represent a challenge in DCE studies because the models used to estimate preference weights simultaneously estimate multiple coefficients. The sample size for this study was chosen based on usual practice in preference studies [56] and the feasibility of recruiting patients with this rare disease.
Statistical Analyses
Analyses of the DCE data followed good research practices [34, 57] and were performed in Stata 17. The analyses used a random-parameters logit (RPL) model that relates the choices respondents made in the treatment-choice questions to the differences in attribute levels across the alternatives in each choice question [57]. This approach produced preference weights for all attribute levels included in the survey.
All levels in each of the attributes were effects coded, and the delta method was used to calculate standard errors for the preference weight for each omitted attribute level [58]. To determine the conditional relative importance of each attribute, the difference between the attribute level with the highest preference weight and the level with the lowest preference weight was calculated. This difference represents the relative importance of each attribute relative to all other attributes included in the study, conditional on the range of levels of the attribute.
Minimum acceptable benefits (MABs) and maximum acceptable risks (MARs) were also calculated using the RPL regression parameters to measure patients’ tolerance for desirable and undesirable treatment features. Maximum acceptable risks and MABs were calculated using mean coefficient estimates for attribute levels of effects-coded variables and assumed linearity of preference between each pair of levels included in the survey, but did not assume linearity across the full range of levels. These calculations determined the average risk or the average benefit that yielded a level of disutility equal to the utility generated by changing from one level to another of a risk or benefit attribute.
A latent class (LC) model was used to explore heterogeneity in the DCE results by probabilistically assigning responses to a set of classes based on the observed pattern of choices. The optimal number of classes was determined using the Bayesian information criterion [59], the Akaike information criterion [60], and the Akaike information criterion 3 [61], followed by the calculation of preference weights and conditional relative importance for each class. Logistic regression models with selected covariates (chosen through correlation analyses to minimize multi-collinearity) were used to determine the association between respondent characteristics and the likelihood of a respondent expressing an interest in starting a new prophylactic treatment when asked.
Only completed surveys with at least one response to a DCE question were included in the final analysis. Analyses were performed on observed data only; no imputation for missing values was performed. The number of missing responses was captured for each question.
Results
Respondent Characteristics
Of the 1221 participants invited to the survey, 278 (22.8%) completed the screening questions and were determined to be eligible. Twenty-five (9.0%) of the eligible participants were excluded because of the recruitment goals having been achieved, and three (1.1%) were removed because of showing no variability in their answers to DCE questions.
The final sample included 250 respondents: 204 (81.6%) identified as female and 213 (85.2%) identified as White (Table 2). The mean age was 39 years (range, 18–79 years). On average, respondents scored 9.7 out of 16 (standard deviation, 4.2) on the AECT, for which a score of 10 or better is considered evidence of well-controlled disease [62].
Respondents had been diagnosed with HAE an average of 21 years before starting the survey (maximum, 54 years); 217 respondents (86.8%) reported experiencing a hospitalization or emergency department visit because of their HAE during that time frame, and had most recently been treated in the emergency department or hospital because of an HAE attack 7 years (average) before the survey. Of those who had more than one HAE attack per year before starting treatment (n = 247 [98.8%]), 81 (32.8%) reported five or more attacks in the past 6 months. Overall, 201 respondents (80.4%) were currently taking LTP medication, 97 (48.3%) of whom were extremely satisfied with their current treatment.
One hundred and ninety-one respondents (76.4%) had at some point stopped taking an LTP medication, with the top reason for stopping (n = 111 [58.1%]) being that the medication was not controlling HAE symptoms well. On average, respondents considered reducing their HAE treatment frequency to be important (mean score of 4.0 on a scale from 1 [“Not at all important”] to 5 [“Extremely important”]) and placed high importance on discussing with their healthcare professional the option of taking their treatment once every 4 weeks or once every 8 weeks (mean score of 3.9 on the same scale).Table 2Respondent characteristicsQuestionRespondents (N = 250) aWhat is your age (in years)? (Eligible if ≥18 years) Mean (SD)39 (11) Min, max18, 79What gender identity best describes you?
(Select only 1 answer.) Female204 (81.6) Male44 (17.6) Gender fluid1 (0.4)Nonbinary1 (0.4) A gender identity not listed here0 (0.0)What race(s) or ethnicity(ies) do you consider yourself to be? (Please select all that apply.) African American or Black15 (6.0) Alaska Native, American Indian, or Native American5 (2.0) Asian5 (2.0) Hispanic, Latin American, or Latinx21 (8.4) Middle Eastern or North African2 (0.8) Native Hawaiian or Pacific Islander0 (0.0) White213 (85.2) A race or ethnicity not listed4 (1.6)Years since diagnosis Mean (SD)21 (13) Min, max0, 54An HAE attack is when you experience severe swelling (angioedema) in part of the body. Which of the following body parts have you ever experienced an HAE attack in?
(Please select all that apply.) Arms, legs, hands, or feet235 (94.0) Groin or genitals183 (73.2) Torso, back, stomach, or abdomen233 (93.2) Face, head, or ears192 (76.8) Throat, tongue, lips, or mouth205 (82.0) Other23 (9.2)Sometimes, people with HAE can have a serious attack that requires medical attention in a hospital. Have you ever received care or treatment in the emergency room or hospital because of an HAE attack? Yes217 (86.8) No33 (13.2)Among respondents ever hospitalized because of an HAE attackN217When were you last treated in the emergency room or hospital because of an HAE attack (years since treated last)?
Mean (SD)7 (7) Min, max0, 37All respondents Even if your current treatment or management routine is working, thinking back to when your HAE was at its worst, how often would you have an HAE attack? Every few days or more frequently112 (44.8) Every few weeks or so100 (40.0) Every few months34 (13.6) Once or twice a year1 (0.4) Less than once a year2 (0.8) Don’t know or not sure1 (0.4)Among respondents who experienced attacks more than once a year n247 How many HAE attacks have you had in the last 6 months? One or fewer79 (32.0) 2–5 attacks87 (35.2) 5–10 attacks38 (15.4) 11–15 attacks20 (8.1) 16 or more23 (9.3) Don’t know or not sure0 (0.0)Respondents currently taking an LTP medicine n201 Using the scale below, how satisfied are you with your current long-term prophylactic HAE treatment(s)?
1 [Not at all satisfied]11 (5.5) 29 (4.5) 325 (12.4) 459 (29.4) 5 [Extremely satisfied]97 (48.3) Mean (SD)4.1 (1.1)Among respondents who have stopped taking an LTP medicine n191 Which of the following statements best describe why you stopped taking a HAE treatment and started a new long-term prophylactic treatment? (Please select all that apply.) It was required by my insurer19 (9.9) Concerns about long-term affordability or insurance coverage25 (13.1) It wasn’t controlling my HAE symptoms well111 (58.1) There were concerning side effects49 (25.7) The administration burden was too much63 (33.0) It was not convenient or didn’t fit in with my lifestyle71 (37.2) My doctor recommended I try something new82 (42.9) Concerns about supply issues or being able to fulfill a prescription29 (15.2) Other25 (13.1)HAE hereditary angioedema, LTP long-term prophylactic, Max maximum, Min minimum, SD standard deviation“Select all that apply” questions may yield a total number of responses greater than the total number of respondents who answered the questionaValues are n (%) unless otherwise stated
Survey Comprehension Questions
Respondents were asked three questions to assess and reinforce their comprehension of attribute descriptions, risk grids used to present attribute levels, and choice questions. Most respondents (n = 200 [80.0%]) answered each comprehension question correctly; 38 respondents (15.2%) answered one comprehension question incorrectly, eight (3.2%) answered two comprehension questions incorrectly, and only four (1.6%) answered all three comprehension questions incorrectly. If responses to any comprehension question indicated that the respondent did not understand the information, graphic, or question, the survey provided feedback to the respondent and repeated the information that the respondent may not have understood.
Survey responses were analyzed regardless of whether respondents answered comprehension questions correctly or incorrectly.
Discrete-Choice Experiment
Preference Weights
Most attributes had levels that were naturally ordered, and the results indicated that better levels were generally preferred to worse levels: respondents preferred treatments with lower dosing frequency, lower risks of GI side effects, and a higher reduction in attack frequency (Fig. 2; Table S1 of the ESM). Exceptions included time until maximum efficacy and risk of ISR.
Time until maximum efficacy had a numerically larger preference weight for 2 weeks until maximum efficacy (0.1) than 1 week (0.0); however, the difference between the two preference weights was not statistically different. Similarly, the risk of ISR had a numerically larger, but not statistically different, preference weight for a 5% risk of ISR (0.3) than a 0% risk (0.1). The differences between the preference weights for many of the levels of the mode of administration attribute were not statistically significant, although the results suggest that respondents preferred oral tablets over injections.
Although the preference weight for the autoinjector (− 0.2) was numerically larger than that for a prefilled syringe (− 0.3) or a vial and syringe (− 0.3), the difference in preference weights between different injections was not statistically different.Fig. 2Preference weights (N = 250). The parameter estimates are the preference weights corresponding to the effects-coded attribute levels.
With effects coding, 0 indicates the mean effect across all attribute levels rather than the omitted level. The effects-coded variables are categorical variables that range from − 1 to 1. The preference weights corresponding to the effects-coded variables are log odds, which are distributed symmetrically around 0.
The vertical bars surrounding each mean preference weight denote the 95% confidence interval of the point estimate (computed by the delta method for the omitted effects-coded level in estimation for model specification). GI gastrointestinal, ISR injection-site reaction
Conditional Relative Attribute Importance
Over the ranges presented in the survey, a reduction in attack frequency had the greatest conditional relative attribute importance (CRAI), with the difference between a 45% reduction in frequency and a 90% reduction accounting for 31.7% of the total possible change in utility (Fig. 3). Reducing the time to maximum efficacy from 10 to 2 weeks had the lowest CRAI, accounting for 4.4% of the total possible change in utility.
For changing the mode of administration from a vial and syringe to an oral tablet, reducing the risk of a GI side effect from 25 to 0%, and reducing the dosing frequency from once every day to once every 8 weeks, the CRAI values were not statistically significantly different from one another.Fig. 3Conditional relative attribute importance (N = 250). The conditional relative importance of an attribute is calculated as the ratio of the difference of the preference weight between the most desirable level and the least desirable level over the sum of the differences across all attributes that is scaled to 100.
The conditional importance of each attribute is expressed as a percentage of this total. The vertical bars surrounding each relative importance weight estimate denote the 95% confidence interval. Attributes are presented in the order in which they appeared in the discrete-choice experiment (DCE) questions.
GI gastrointestinal, ISR injection-site reaction
Minimum Acceptable Benefit
The MABs of reduction in attack frequency were calculated from both 45% and 85%, where mean MAB values ranged from 5.9 to 28.6% and from 3.2% to more than 5%, respectively (Table 3). Where the MAB of an attribute was reported as greater than 5%, the MAB of reduction in attack frequency was reported as greater than a total of 90% reduction in attack frequency (i.e., the highest level of reduction in attack frequency included in the survey).Table 3MAB of reduction in attack frequencyAttributeAttribute reductionMean MAB over a 45% reduction in attack frequency (95% CI)Mean MAB over an 85% reduction in attack frequency (95% CI)From this levelTo this levelMode of administrationTabletAutoinjector24.5% (18.0–31.0%)> 5% aDosing frequencyOnce every 8 weeksOnce every 4 weeks5.9% (1.5–10.2%)3.2% (0.5–5.9%)Risk of ISR 15 percentage-point increase5%20%7.9% (3.3–12.5%)4.3% (1.3–7.3%) 35 percentage-point increase20%55%10.2% (5.7–14.8%)> 5% a 50 percentage-point increase5%55%18.1% (13.2–23.1%)> 5% aRisk of a GI side effect 7 percentage-point increase0%7%8.3% (4.0–12.5%)4.5% (1.5–7.5%) 8 percentage-point increase7%15%12.3% (8.1–16.6%)> 5% a 15 percentage-point increase0%15%20.6% (16.2–25.0%)> 5% a 10 percentage-point increase15%25%8.0% (3.9–12.1%)4.4% (1.5–7.2%) 18 percentage-point increase7%25%20.4% (15.7–25.0%)> 5% a 25 percentage-point increase0%25%28.6% (23.7–33.5%)> 5% aCI confidence interval, GI gastrointestinal, ISR injection-site reaction, MAB minimum acceptable benefitaEstimates of MAB were not extrapolated outside the benefit range presented in the survey. Means reported as > 5% represent MAB estimates greater than a total of 90% reduction in attack frequency, which was the maximum level of reduction in attack frequency presented in the study
Maximum Acceptable Risk
The MAR for ISR ranged from 11.1% (for an improvement in dosing frequency from once every 4 weeks to once every 8 weeks) to more than 50% (for improvements in the reduction in attack frequency from 45 to 85% and 90% and for a change in mode of administration from autoinjector to tablet) [Table 4]. The MAR for GI side effects ranged from 5% (for an improvement in dosing frequency from once every 4 weeks to once every 8 weeks) to more than 25% (for improvements in reduction in attack frequency from 45 to 85% and 90%), the highest level of risk of a GI side effect included in the survey. Overall, respondents tended to be more tolerant of the risks of ISRs than GI side effects in exchange for the same improvements in other attributes.Table 4MAR of an ISR and a GI side effectAttributeAttribute improvementMean MAR of ISR (95% CI)Mean MAR of a GI side effect (95% CI)From this levelTo this levelReduction in attack frequency 40 percentage-point improvement45%85%> 50% a> 25% b 5 percentage-point improvement85%90%19.3% (0.0–38.6%)7.6% (4.0–11.2%) 45 percentage-point improvement45%90%> 50% a> 25% bMode of administrationAutoinjectorTablet> 50% a17.5% (12.7–22.3%)Dosing frequencyOnce every 4 weeksOnce every 8 weeks11.1% (0.4–21.8%)5.0% (0.7–9.2%)CI confidence interval, GI gastrointestinal, ISR injection-site reaction, MAR maximum acceptable riskEstimates of MAR were not extrapolated outside the risk range presented in the surveyaThe MAR of ISR is interpreted as the maximum acceptable increase in risk of ISR for a corresponding attribute improvement.
The preference weight for 0% risk of ISR was disordered to 5% risk of ISR and not statistically different from 20% risk of ISR. Therefore, MARs of ISR were calculated using a 5% risk of ISR as the baseline risk instead of 0% (as for risk of a GI side effect). This allows the assumption that, given an attribute improvement (i.e., increased reduction in attack frequency, a preferred change in mode of administration, or decreased dosing frequency), respondents will accept a corresponding increase in risk of ISR.
Estimates of MAR were not extrapolated outside the risk range presented in the survey. Hence, a mean MAR of ISR reported to be > 50% represents a total risk estimate > 55% (55% risk of ISR − 5% risk of ISR = 50% risk of ISR), which was the maximum level of risk of ISR presented in the studybMeans that are reported to be > 25% represent MAR estimates > 25% risk of a GI side effect, which was the maximum level of the risk of GI side effect presented in the study
Heterogeneity Analysis
Latent Class Analysis
After comparing various LC models, the optimal number of classes was determined to be 2. The two LCs of respondents were identified with a predicted probability of class membership of 51.9% of the sample in class 1 and 48.1% of the sample in class 2. Figures S1 and S2 of the ESM present the estimated mean preference weights for each attribute level and CRAI estimates for all attributes in each of the classes, respectively.
Class 1 (51.9%: “burden minimizers”) placed similar weight on changes in most attributes but prioritized the risks of side effects as well as mode of administration. Respondents associated with this class placed the most importance on avoiding the risks of ISR and GI side effects and were less sensitive to improvements in reduction in attack frequency than those in class 2.
Class 2 (48.1%: “efficacy maximizers”) preferred maximizing the reduction in attack frequency. Respondents associated with this class placed the greatest importance on improving the reduction in attack frequency offered by treatments and were largely indifferent between mode of administration, dosing frequency, and risk of GI side effects. However, in contrast to class 1, respondents in class 2 placed the least importance on the risk of ISR.
As shown in Table 5, membership in class 1 was statistically significantly associated with the characteristics “ever experienced an HAE attack in the groin or genitals” and “has poorly controlled HAE as indicated by an AECT score < 10.” Membership in class 2 was statistically significantly associated with “selected only White as a race or ethnicity category” and “ever experienced mild-to-moderate GI symptoms because of HAE.”Table 5Two-class logit analysis: class membership model parametersVariableModel parameters: class 1Selected only White as a race or ethnicity category− 1.0**Full-time, part-time, or self-employed− 0.4Ever experienced an HAE attack in the groin or genitals0.7**Has poorly controlled HAE as indicated by an AECT score < 100.7**Selected more comorbidities than the median respondent0.5Has received injections with an autoinjector for HAE or another health condition− 0.3Ever had a mild-to-moderate ISR− 0.5Ever experienced mild-to-moderate GI side effects because of HAE− 0.7**Log-likelihood− 1503.3AIC3092.7BIC3354.4AECT Angioedema Control Test, AIC Akaike information criterion, BIC Bayesian information criterion, GI gastrointestinal, HAE hereditary angioedema, ISR injection-site reaction, ** statistically significant at the 5% level (p < 0.05)Note: class 2 is used as the reference class. Positive model parameters indicate an association with membership in class 1. Negative model parameters indicate an association with membership in class 2
Treatment-Switching Logit Analysis
When asked hypothetically whether they would be willing to switch from their current treatment to a new prophylactic treatment with their preferred mode of administration, 197 respondents (78.8%) stated they would be open to starting a new medication, whereas 53 respondents (21.2%) were not. Respondent characteristics associated with an increased likelihood of preferring to start a new treatment (i.e., an odds ratio that is statistically significant and greater than 1), included poorly controlled HAE (AECT score < 10), having an autoinjector/EpiPen for HAE/another health condition but has not used it, currently taking Berinert as LTP medicine, and identifying as African American or Black (Table S2 of the ESM). Conversely, characteristics that were associated with a decreased likelihood of preferring to start a new treatment (i.e., an odds ratio of less than 1) included demographic variables (age, race/ethnicity), use of Medicare, receipt of co-pay assistance, diagnosis of GI conditions, receipt of treatment through an IV infusion for HAE/another health condition, and a history of switching treatments in the past.
Discussion
This study, which is the first to assess patient preferences for the attributes of LTP HAE treatment options, quantified the conditional relative importance of HAE treatment attributes and explored the trade-offs that people with HAE are willing to accept among treatment features. Our findings revealed that, in general, patients cared most about reducing the frequency of HAE attacks, followed by mode and frequency of administration and a reduction in side effects, and that patients with HAE exhibited heterogeneous preferences.
Generally, preferences were naturally ordered and preferences aligned with a priori expectations informed by observations made in formative pre-test interviews. Respondents tended to be more tolerant of the risks of ISRs than GI side effects in exchange for the same improvements in other attributes, as reflected in the CRAI of the risks of ISR and GI side effects. This lack of tolerance for GI side effects may be related to the fact that HAE attacks often occur in the GI tract and this side effect may be of concern to patients.
It was found that, on average, respondents were willing to trade increases in the reduction in attack frequency in exchange for worse levels of other attributes and, on average, willing to accept risks of ISR and GI side effects in exchange for improvements in other attribute levels, especially for efficacy.
As found in this study, preference for oral tablets has previously been reported by Geba et al. [63], who found that 98% of prophylactic HAE medication users would prefer oral medication over their current treatment; however, patients’ willingness to trade off other treatment features for their preferred administration was not investigated in that study. Our study found that respondents preferred oral tablets to injections, but this preference was contingent on safety and efficacy. Respondents were willing to accept a 17.5% risk of a GI side effect for a change in mode of administration from an autoinjector to a tablet, and on average, respondents were willing to accept an additional 24.5% reduction in attack frequency from a base of 45% to switch from an oral tablet to an autoinjector.
As LTP treatment is the only way to gain complete disease control [16, 18], it is important for healthcare professionals to understand patient perspectives on treatment attributes of LTP medications, especially with several LTP medications with different administration routes currently in development [64].
Two classes were determined to describe clusters of preferences; burden minimizers in class 1 (51.9% of sample) placed the most importance on avoiding the risks of ISR and GI side effects, and they were relatively less sensitive to improvements in reduction in attack frequency than those who were efficacy maximizers in class 2. In contrast, efficacy maximizers in class 2 (48.1% of sample) preferred maximizing the reduction in attack frequency and were largely indifferent among mode of administration, dosing frequency, and the risk of GI side effects. Characteristics associated with membership in class 1 included ever experiencing an HAE attack in the groin or genitals and having poorly controlled HAE (AECT score < 10), whereas characteristics associated with membership in class 2 included selecting only White as a race or ethnicity category and having ever experienced mild-to-moderate GI side effects because of HAE.
In class 1, those with poorly controlled HAE may view any level of reduction in attack frequency described in the survey as a desirable outcome, thereby diminishing the overall importance of the attribute relative to other attributes included in the study.
Logit analysis of respondent willingness to start a new prophylactic treatment also revealed heterogeneity, indicating that many factors affect respondents’ preferences for long-term HAE treatments. Factors associated with a higher likelihood of starting a new prophylactic treatment included poorly controlled HAE, time since last hospitalization because of HAE, and having an autoinjector but never using it; characteristics associated with a lower likelihood of starting a new prophylactic treatment included increased age, race/ethnicity, and being diagnosed with GI conditions. Collectively, these findings may provide key insights to guide clinicians when engaging with patients to make individualized treatment decisions that align with patients’ priorities.
It is important to note that this study is subject to limitations. First, the survey was restricted to US-based respondents and was written only in English, limiting it to respondents who could speak or read English. Because of the absence of an interviewer assistant during AECT data collection, data may be subject to noise and measurement error.
Notably, voluntary surveys pose potential volunteer and selection biases, and how the characteristics of those who chose to respond compared with those who were invited but did not participate is unknown. Such biases may result in non-representativeness, which may lead to an underestimate or overestimate of respondent satisfaction and preferences. Moreover, respondents included in this survey were mainly female and were also not racially and ethnically diverse.
Therefore, this survey sample was not representative of the broader HAE population. With regard to the statistical modeling conducted, RPL provides mean relative preference estimates for each attribute level over a population and cannot be used to estimate preferences for an individual. Finally, since this study was conducted, the landscape of approved HAE treatments has evolved, notably with the recent approval of garadacimab-gxii, administered monthly by self-injection [65]; however, attributes of garadacimab-gxii were considered in the study design and the range of levels reflected the data available on garadacimab-gxii at the time (e.g., injection every 4 weeks).
There were several strengths of this study as well. This survey was pretested using in-depth interviews with patients with a self-reported diagnosis of HAE, and it used an experimental design developed using good research practices [53]. A large sample size (particularly for a rare disease) was used and recruited from a database exclusive to a US-based HAE patient advocacy group, resulting in strong participation by those active in the HAE patient community.
Treatment-choice data were analyzed using advanced RPL methods in alignment with good research practices [34, 57] and are consistent with work conducted by the Food and Drug Administration [66] that avoids estimation bias from unobserved variation in preferences across the sample and from within-sample correlation in the choice sequence for each respondent. There was a high pass rate for the comprehension questions (>90% for each question) compared with other preference studies previously reported [67], which is indicative of a methodologically robust and comprehensible survey instrument.
Conclusions
Individuals with HAE have distinct and heterogenous preferences, caring most about reducing the frequency of HAE attacks, mode and frequency of administration, and reducing GI side effects. Time to full clinical benefit of a treatment was lowest on the attribute list for patients, and patients with poorly controlled HAE are likely to be open to starting a new treatment. This study highlights the need for patient-physician communication during decision making about initiating a new long-term treatment for HAE.
Supplementary Information
Below is the link to the electronic supplementary material.Supplementary file1 (DOCX 215 KB)Supplementary file1 (PDF 1209 KB)