Health state utilities associated with treatment options for acute myeloid leukemia (AML)

Abstract

Aims: Acute myeloid leukemia (AML) treatment typically involves remission induction chemotherapy followed by consolidation chemotherapy. New treatments for AML have recently been introduced, including a chemotherapy formulation called CPX-351, which is administered via less time-intensive IV infusion than the standard “7 + 3” continuous infusion regimen of cytarabine plus an anthracycline. The purpose of this study was to estimate utilities that could be used in economic modeling of AML treatment.

Materials and methods: In time trade-off interviews, participants from the UK general population valued 12 health states drafted based on literature and clinician interviews. To identify disutility associated with chemotherapy, two types of induction and four types of consolidation were added to an otherwise identical health state describing AML. The decrease in utility when adding these chemotherapy regimens represents the disutility of each regimen. Five additional health states were valued to estimate utilities associated with other AML treatments.

Results: Two hundred participants completed interviews. Mean (SD) utilities were 0.55 (0.31) for pre-treatment AML and 0.66 (0.29) for AML in temporary remission. Adding any chemotherapy significantly decreased utility (p < 0.0001). Induction had a mean disutility of –0.11 with CPX-351 and –0.15 with 7 + 3. Mean disutility for consolidation ranged from –0.03 with outpatient CPX-351 to –0.11 with inpatient 5 + 2. Utilities are also reported for other AML treatments (e.g. transplant, low-intensity chemotherapy).

Limitations: One limitation is that the differences in adverse event profiles between the treatment regimens were based on clinician opinion. Future use of CPX-351 in clinical trials or clinical settings will provide additional information on its adverse event profile.

Conclusions: While all chemotherapy regimens were associated with disutility, regimens with shorter hospitalization and less time-intensive infusion were generally perceived as preferable. These utilities may be useful in cost-utility models comparing the value of AML treatments.

Keywords:

• Acute myeloid leukemia
• chemotherapy
• consolidation
• induction
• time trade-off
• utility

JEL CLASSIFICATION CODES:

• H51
• I18

Introduction

Acute myeloid leukemia (AML) is a rare hematologic malignancy characterized by over-proliferation of myeloid blasts in the bone marrow, blood, and other tissues, leading to decreased health-related quality-of-life and increased risk of bleeding, infection, and mortality^1–4. Treatment involves eradicating leukemic cells from the blood and bone marrow to induce remission, as indicated by normalization of blood counts and a lack of visible leukemia cells in the blood or bone marrow⁵. Combination chemotherapy regimens are the standard of care for remission induction. One of the most commonly used induction regimens for AML is the “7 + 3” regimen, which consists of cytarabine administered as a 7-day continuous infusion plus an anthracycline (e.g. daunorubicin) administered once daily for 3 days. The regimen is typically administered in the inpatient setting^5,6. If complete remission is achieved with induction, it is followed by post-remission consolidation chemotherapy (i.e. typically administered in a 5-day continuous infusion [5 + 2 consolidation] or as intermittent 4-h infusions every 12 h on the first, third, and fifth days of each consolidation cycle) and/or hematopoietic stem cell transplantation.

Recently, new treatments for AML have been introduced, including a new formulation of intensive chemotherapy called CPX-351^7–12. CPX-351 is a nano-scale liposomal co-formulation of cytarabine and daunorubicin approved in the US and Europe for the treatment of adults with newly-diagnosed, therapy-related AML and AML with myelodysplasia-related changes^7,13–16. Phase 3 data have demonstrated significantly prolonged survival and an increased complete response rate for CPX-351 compared with the 7 + 3 regimen in older adults (aged 60–75 years) with newly-diagnosed high-risk/secondary AML¹⁷. The safety profile of CPX-351 in phase 3 was generally consistent with the known profile of the conventional 7 + 3 regimen. In addition, CPX-351 is administered via a less time-intensive infusion process than the standard of care: whereas the 7 + 3 and 5 + 2 regimens require continuous intravenous infusion 24 h per day for 7 and 5 days, respectively, CPX-351 is administered as a 90-min infusion on days 1, 3, and 5 for first induction and on days 1 and 3 for consolidation. Due to its administration schedule, CPX-351 can potentially be administered as an outpatient, rather than inpatient, treatment¹⁶.

As new treatments for AML are introduced, cost-utility models (CUAs) are often used to examine their value and inform decision-making on healthcare resource allocation^18–20. CUAs require utilities, which are values anchored to 0 (dead) and 1 (full health), representing the strength of preference for various health states²¹. However, published options for utilities associated with AML and its treatment are limited. For example, some models have used estimates from clinicians, rather than utilities based on patient preferences or general population values^22,23. Other studies have reported utilities derived from measures, such as the EQ-5D, completed by patients long after being treated for AML^24–26. Because of the lag between AML treatment and utility assessment, the values estimated in these studies do not adequately represent the utility impact of AML symptoms or treatment.

Some other utilities used to represent AML were derived in samples including patients with diseases other than AML, such as myelodysplastic syndrome, or patients receiving treatments that are no longer commonly used^27–30. Two recent presentations reported vignette-based studies yielding potentially useful utilities representing AML treatment^31,32, but provided little detail regarding the methods or vignette content. Importantly, no previous studies examined utility differences between the standard of care and newer chemotherapy, despite differences in infusion duration, length of hospitalization, and adverse event profiles that could have an impact on patient quality-of-life and health state preference.

The purpose of the current study was to estimate utilities that could be used in economic modeling of AML treatment, including utilities representing both standard and newer chemotherapy regimens. Although health technology assessment agencies often favor utilities derived from generic preference-based measures, such as the EQ-5D³³, this approach may not be appropriate in this situation for two reasons. First, there would be substantial challenges in having patients complete an instrument like the EQ-5D while undergoing chemotherapy, as it would be difficult to access patients during hospitalization and may not be ethical to add questionnaire completion tasks to the considerable patient burden during chemotherapy. Second, generic instruments are not designed to be sensitive to aspects of the chemotherapy treatment process that could be important to patients, such as duration of hospitalization and differences in infusion regimens.

Like most studies estimating utilities associated with the treatment process³⁴, this study used a vignette-based approach, which is well suited for isolating the utility impact of specific treatment attributes. Twelve health states (often called “vignettes”) were drafted to represent the typical patient experience in various phases of AML treatment. Health states representing either induction or consolidation treatment varied by length of hospital stay, side-effect profile, and infusion procedures so that utility differences among chemotherapy approaches could be estimated and used in CUAs. Additional health states represented newly-diagnosed AML, hematopoietic stem cell transplant, remission, and treatments for situations when chemotherapy is no longer indicated (e.g. best supportive care, hypomethylating agents [HMA]).

Methods

Overview of study design

This study was designed to estimate utilities associated with AML and various chemotherapy regimens used to treat AML. Health state descriptions (often called vignettes or health states) were developed and refined based on published literature, clinician interviews, and a pilot study. The health states each describe a 1-year period for a patient with AML. To calculate the disutility of the various types of treatment, induction and consolidation treatment regimens were added to otherwise identical 1-year health states in which chemotherapy was followed by temporary remission. This type of health state can be considered a “path state” because it describes a sequence of health-related events that change over time^35–37.

Utilities for these health states were then elicited in a time trade-off (TTO) task with a 1-year time horizon. The interviews were conducted with general population participants in May and June 2017 in Edinburgh and London, UK. Participants were required to provide written informed consent before completing study procedures, and all procedures and materials were approved by an independent Institutional Review Board (Ethical & Independent Review Services; Study Number 17036).

Health state development

Health states were drafted based on published literature, online data, and interviews with clinicians (described below). A targeted literature and online search was performed to support the health state content and inform the development of questions to be asked in the clinician interviews. The literature search focused on the impact of AML^38,39, symptoms^40,41, chemotherapy treatment regimens^{1,15,16,42–48}, treatment-related adverse events⁴⁹, best supportive care^45,50, and hematopoietic stem cell transplant^1,43,51.

Multiple rounds of telephone interviews were conducted with six clinicians. The clinicians included three physicians and three nurses with extensive experience in treatment of patients with hematologic malignancies: one oncologist specializing in AML; two physicians specializing in hematology, leukemia, and hematopoietic stem cell transplants; and three oncology nurses (two RNs and one nurse practitioner), one of whom specialized in hematopoietic stem cell transplants. Five of these healthcare professionals were from the US, and one was from Canada. Additionally, five of these six healthcare professionals were involved with the clinical program for CPX-351.

Health states were developed through an iterative process with the physicians and nurses, and each clinician participated in up to three discussions so they could respond to multiple drafts of the health states. Initial questions for the healthcare professionals focused on patients’ typical experience with AML prior to treatment, during chemotherapy (including both induction and consolidation phases), during the transplant process, and while in remission. Follow-up discussions focused on reviewing and editing health state drafts to ensure the descriptions of the treatment processes, symptoms, and adverse events were clear and accurate representations of the typical patient experience.

A total of 12 health states were drafted, each describing 1 year of a patient’s life. Health state A described a patient with active AML, typically a newly-diagnosed patient, prior to receiving treatment. Health states B1 and B2 both described remission periods. Health state B1 described the temporary remission that typically occurs following induction or consolidation^1,43, and B2 described the remission that follows a successful hematopoietic stem cell transplant⁴³. Two remission health states were developed because the physicians and nurses all explained that remission following transplant was different and more “durable” than the “temporary” remission following induction or consolidation. For the purposes of this vignette study, health states A, B1, and B2 were considered chronic health states, meaning that they remain unchanged over the 1-year period.

Six path states (C1, C2, D1, D2, D3, and D4) represented a year beginning with chemotherapy followed by remission. Health states C1 and C2 described a year beginning with induction therapy followed by remission (B1). Health state C1 described the 7 + 3 induction regimen^1,44,48, while C2 described induction with CPX-351¹⁵. Health states D1 to D4 described a year beginning with post-remission consolidation therapy, with subsequent continued remission (B1). Health state D1 described the 5 + 2 consolidation regimen^1,48, and D2 described high-dose cytarabine (HiDAC)^1,43,47. When used as consolidation therapy, CPX-351 can be administered in either an inpatient or outpatient setting^16,46, and this difference in setting was hypothesized to have an influence on health state preferences. Therefore, two health states were drafted to represent these two types of CPX-351 consolidation therapy: inpatient (D3) and outpatient (D4).

Key differences among the chemotherapy regimens included length of hospital stay, duration of the treatment infusion, and the common treatment side-effects. The hospital stay was 4 weeks for 7 + 3 induction; 5 weeks for CPX-351 induction; and 1 week for 5 + 2, HiDAC, and CPX-351 inpatient consolidations. There was no hospital stay associated with CPX-351 outpatient consolidation. The difference in hospital stay was supported by treatment patterns in clinical trials^46,52 and by interviews with the clinicians. Infusion durations were based on current treatment guidelines and prescribing information^15,47,48 and were corroborated by the physicians and nurses. The infusion durations were 24 h per day for 7 days for 7 + 3 induction; 90 min on days 1, 3, and 5 for CPX-351 induction; 24 h per day for 5 days for 5 + 2 consolidation; 4 h twice each day on days 1, 3, and 5 for HiDAC consolidation; and 90 min on days 1 and 3 for CPX-351 consolidation.

The differences among adverse events associated with chemotherapy were based on communications with clinicians (three physicians and three nurses described above) who had personal experience treating patients with the more established regimens (e.g. 7 + 3, 5 + 2, HiDAC) and CPX-351. These clinicians reported anecdotal information about hair loss and fatigue experienced by their patients on these treatments.

Health state F described the hematopoietic stem cell transplant process followed by durable remission (B2). This health state presented a successful hematopoietic stem cell transplant without all the potential complications that can result from this procedure^1,43,51. In addition to these health states, two more chronic health states were developed to capture additional types of treatments commonly used in patients with AML who are not considered good candidates for chemotherapy and transplants: non-intensive therapy with hypomethylating agents (E1)^42,44,45,48 and best supportive care (E2)^45,50.

The health states were presented to respondents on individual cards, each with a series of bullet point descriptions. The bullet points were organized into categories with headings intended to help the respondents understand the health state content. For example, the headings for the remission health states were disease status, symptoms, impact, and treatment follow-up. For health states C1 to D4, the sections describing chemotherapy regimens had the following headings: hospital, treatment process, side-effects, and recovery from treatment. To help respondents understand the sequence of events described in each path state, a timeline was depicted at the bottom of each health state card. See Supplementary Appendix for full text of the final health states used in the valuation study.

Participants

Participants in the pilot and main studies were recruited from the general population and were required to be at least 18 years old; able to understand the utility assessment procedures; able and willing to provide written informed consent; and a UK resident. These inclusion criteria did not include clinical characteristics, because this study aimed to estimate utilities for cost-utility analyses in submissions to health technology assessment agencies, which often prefer that utilities represent general population values^33,53. Participants were recruited via newspaper advertisements (including five newspapers in Edinburgh and two newspapers in London) and online advertisements on Gumtree.com. Potentially interested participants called or emailed to provide their contact information, and then they were called by phone to be screened using a standardized screening script. Some demographic details were requested during these screening calls to ensure that no particular demographic group was over-represented relative to the UK general population (e.g. gender, age, ethnic/racial background, employment status).

Pilot study

To assess the clarity of the health states and utility assessment methodology, a pilot study was conducted in London with 20 general population participants (50.0% male; mean age = 40.2 years; age range = 22–80 years). After an introductory ranking task, the participants valued the health states using TTO methods with a 1-year time horizon and 1-month trading increments. The first two participants in the pilot study reviewed and ranked all health states at once. However, it was difficult for the participants to maintain logical rankings among the large number of detailed health states. Therefore, it was decided to present the health states in two groups for the remaining participants. The chemotherapy health states (e.g. B1 plus the induction and consolidation states) were included in the first group (group 1), as they all included B1 representing temporary remission. The remaining five health states (A, B2, E1, E2, and F) comprised the second group (group 2).

Subsequent participants completed the ranking and TTO separately for the two groups of health states. With the health states separated into two groups, the ranking and TTO tasks were feasible for the great majority of respondents, and they generally reported that the health state language and content was clear and comprehensible. Some participants suggested minor revisions in formatting and word choice, and the health states were edited accordingly.

The 1-year TTO time horizon

TTO valuation studies differ with regard to the duration of time in the health state being rated, which is often called the time horizon. Previous research has indicated that the time horizon could have an impact on the resulting utility values, as well as differences among health state utilities⁵⁴. Therefore, the time horizon for each TTO study must be selected carefully. For the current study, a 1-year time horizon was selected for several reasons. First, and most importantly, the goal was to identify the utility impact of relatively brief events (e.g. a 1-month course of chemotherapy), and these relatively brief events would be unlikely to affect valuations of a longer time horizon, such as 10 years. With the 1-year time horizon, the impact of the relatively brief event was clearly important to the respondents. Second, with the 1-year time horizon, the differences among health states can be applied in subsequent cost-utility models as quality-adjusted life year (QALY) decrements. Third, prognosis with AML is known to be poor, with ∼ 70% of elderly patients dying from the disease within 1 year of diagnosis^5,55. Therefore, the relatively short 1-year time horizon provides a realistic representation of disease course.

Utility interview procedures and scoring

After finalizing the health states and methods based on the pilot study, health state utilities were elicited in a TTO valuation study in May and June 2017. To control for order effects, participants were randomized to review either the group 1 health states (B1, C1, C2, D1, D2, D3, and D4) or group 2 health states (A, B2, E1, E2, and F) first, followed by the other group. Within each group, health states were presented in random order. Both the ranking and TTO for the first health state group were completed before proceeding to ranking and TTO for the second health state group. If participants appeared fatigued after their first health state group, they were given the option to discontinue the interview before beginning the second group. Along with the health states, participants were shown a background information page briefly describing AML and its symptoms.

After ranking the health states within each group, participants valued the health states in a TTO task with a 1-year time horizon and 1-month trading increments. Participants were offered a choice between living 1 year in the health state being rated or a shorter duration in full health. For each health state, choices were presented in an order that alternated between longer and shorter amounts of time in full health, specified in months: 12, 0 (i.e. dead), 11, 1, 10, 2, 9, 3, 8, 4, 7, 5, and 6 months. For health states that the respondent perceived as better than dead, utility scores (u) were calculated based on the point of indecision as the number of months in full health (x) divided by the number of months in the health state being rated (u = x/12 months), yielding a utility score on a scale with the anchors of dead (0) and full health (1).

When participants indicated that a health state was worse than dead, the task and scoring procedures were altered as described in previous literature^56,57. Participants were offered a choice between dead (choice 1) and a 1-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 1-year life span. The resulting negative utility scores were calculated with a bounded scoring approach commonly used to avoid highly skewed distributions for negative utility scores (u = –x/12, where x is the number of months in full health, and 12 is the number of months in the total life span of choice 2).

Statistical analysis procedures

Statistical analyses were completed using SAS. Continuous variables including utilities and differences between health state utilities are summarized in terms of means and standard deviations, and categorical variables such as gender and race are summarized as frequencies and percentages. Disutilities for induction and consolidation regimens were calculated by subtracting the utility of the remission health state from the utility of the induction/consolidation health state. For example, health state B1 described temporary remission, and health state C1 described induction with 7 + 3 followed by temporary remission. Therefore, the utility difference between health states B1 and C1 can be attributed entirely to the addition of the 7 + 3 induction therapy regimen, and the disutility of the 7 + 3 induction regimen can be computed by subtracting the utility of B1 from the utility of C1. Because the health states each lasted 1 year, this disutility can be considered a QALY decrement.

Demographic sub-groups (age, gender, geographic location) were compared with Chi-square analyses (for categorical variables) and t-tests (for continuous variables). Pairwise t-tests were conducted to test whether there were significant differences between the remission health states and the other health states, between the two induction health states, or between each of the consolidation health states.

Results

Sample description

A total of 232 potential participants were scheduled for interviews. Of these, 205 attended the scheduled interview, but five had difficulty understanding the health states and/or TTO procedures and were therefore unable to provide valid TTO data. Thus, 200 valid TTO interviews were conducted (100 in Edinburgh and 100 in London). The sample was 49% female, with a mean age of 45.5 years (Table 1). There were no statistically significant differences between the Edinburgh and London sub-groups in age, gender, marital status, employment status, or level of education. The only statistically significant demographic difference was that a greater percentage of participants in Edinburgh reported ethnic/racial background as white (90% vs 54%; p < 0.0001). The most commonly reported health conditions were depression (19.0%), anxiety (18.5%), arthritis (8.5%), hypertension (7.0%), and diabetes (6.5%). No participants reported having AML, although one participant (0.5%) reported having another type of blood cancer (non-Hodgkin lymphoma) and nine participants (4.5%) reported having other types of cancer.

Table 1. Demographic characteristics.

	Edinburgh (n = 100)	London (n = 100)	Participants in the analysis sample (n = 200)	p-value^a
Age (Mean, SD)	45.0 (15.8)	46.1 (15.4)	45.5 (15.6)	0.64
Gender, n (%)
Male	50 (50.0%)	52 (52.0%)	102 (51.0%)	0.78
Female	50 (50.0%)	48 (48.0%)	98 (49.0%)
Ethnic/Racial background, n (%)
White	90 (90.0%)	54 (54.0%)	144 (72.0%)	<0.0001
Mixed^b	4 (4.0%)	6 (6.0%)	10 (5.0%)
Asian	5 (5.0%)	20 (20.0%)	25 (12.5%)
Black	1 (1.0%)	16 (16.0%)	17 (8.5%)
Other^c	0 (0.0%)	4 (4.0%)	4 (2.0%)
Marital status, n (%)
Not married^d	55 (55.0%)	66 (66.0%)	121 (60.5%)	0.11
Married/Cohabitating/Living with partner	45 (45.0%)	34 (34.0%)	79 (39.5%)
Other	0 (0.0%)	0 (0.0%)	0 (0.0%)
Employment status, n (%)
Full-time work	40 (40.0%)	39 (39.0%)	79 (39.5%)	0.88
Part-time work	24 (24.0%)	27 (27.0%)	51 (25.5%)
Other^e	36 (36.0%)	34 (34.0%)	70 (35.0%)
Education level, n (%)
University degree	44 (44.0%)	40 (40.0%)	84 (42.0%)	0.57
No University degree	56 (56.0%)	60 (60.0%)	116 (58.0%)

^a p-values for comparisons between Edinburgh and London are based on t-tests for continuous variables and Chi-square analyses for categorical variables.

^b Mixed includes White and Black Caribbean; White and Black African; White and Asian; and other mixed/multiple ethnic groups as reported by respondents.

^c Other Ethnic/Racial background includes Arab and other, as reported by respondents (verbatim text responses included: Iranian, half British/Iraqi, and British Turkish Cypriot).

^d Not married includes single, divorced, separated, and widowed, as reported by the respondents.

^e Other employment includes homemaker, student, unemployed, retired, disabled, and unspecified, as reported by the respondents.

Abbreviations. SD, standard deviation.

Health state utilities

Seven participants who started with the group 1 health states did not continue with group 2. Likewise, seven participants who began with group 2 did not complete group 1. Therefore, each health state was ranked and valued by 193 respondents.

In the introductory ranking task, the rankings ranged from 1 (most preferable health state) to 7 (least preferable health state) for the group 1 health states, and from 1 (most preferable health state) to 5 (least preferable health state) for the group 2 health states. In group 1, temporary remission (health state B1) was ranked first by all but two respondents (yielding a mean ranking of 1.03), followed by the health states describing a chemotherapy regimen and then B1 for the remainder of the 1-year period: D4 (2.21), D3 (3.06), D2 (4.20), C2 (5.28), D1 (5.44), and C1 (6.69). In group 2, the durable remission health state (B2) was always ranked first, with a mean ranking of 1.00, followed by F (2.50), A (3.53), E1 (3.76), and E2 (4.20).

Utilities of the group 1 and group 2 health states are presented in Tables 2 and 3, respectively. Health state B2 representing durable remission had the highest mean utility score (0.86), followed by B1 representing temporary remission (0.66). The addition of induction (C1, C2) or consolidation (D1, D2, D3, D4) chemotherapy resulted in lower health state utilities. Among the chemotherapy health states, longer hospital stays, more serious adverse event potential, and longer treatment infusion periods were generally associated with lower utilities (Table 2).

Table 2. Utilities of Group 1 health states (n = 193).

	Differences among the chemotherapy health states
Health states	Hospital stay (weeks)	Infusion duration	Potential side-effects	Mean utility^a	SD	95% CI
B1: Remission, between treatment				0.66	0.29	0.61–0.70
C1: Induction, 7 + 3	4	24 h per day for 7 days	Complete hair loss, risk of infection, feeling very fatigued	0.50	0.32	0.46–0.55
C2: Induction, CPX-351	5	90 min on days 1, 3, and 5	Thinning hair, risk of infection, feeling a little fatigued	0.55	0.32	0.50–0.59
D1: Consolidation, 5 + 2	1	24 h per day for 5 days	Complete hair loss, risk of infection, feeling very fatigued	0.54	0.31	0.50–0.59
D2: Consolidation, HiDAC	1	4 h twice each day on days 1, 3, and 5	Thinning hair, risk of infection, feeling very fatigued	0.58	0.30	0.54–0.62
D3: Consolidation, CPX-351,inpatient	1	90 min on days 1 and 3	Thinning hair, risk of infection, feeling a little fatigued	0.61	0.29	0.57–0.65
D4: Consolidation, CPX-351, outpatient	0	90 min on days 1 and 3	Thinning hair, risk of infection, feeling a little fatigued	0.62	0.29	0.58–0.66

^a TTO scores are on a scale anchored with 0 representing dead and 1 representing full health.

Abbreviations. CI, confidence interval; HiDAC, high-dose cytarabine; SD, standard deviation.

Table 3. Utilities of the Group 2 health states (n = 193).

Health state	Mean utility^a	SD	95% CI
A: AML	0.55	0.31	0.51–0.59
B2: Durable remission	0.86	0.18	0.83–0.88
E1: BSC	0.52	0.32	0.47–0.56
E2: HMA	0.48	0.33	0.43–0.53
F: Transplant	0.65	0.28	0.61–0.69

^a TTO scores are on a scale anchored with 0 representing dead and 1 representing full health.

Abbreviations. AML, acute myeloid leukemia; BSC, best supportive care; CI, confidence interval; HMA, hypomethylating agents; SD, standard deviation.

The majority of participants perceived each of the 12 health states to be better than dead, resulting in positive utility scores. B2 (durable remission) was never rated worse than dead; F (transplant) was rated worse than dead by four participants (2.1%); and A (AML), B1 (temporary remission), and both CPX-351 consolidation health states (D3, D4) were rated worse than dead by five participants (2.6%). The remaining health states were rated worse than dead slightly more often: health state D2 (consolidation, HiDAC; n = 6, 3.1%), C2 (induction, CPX-351; n = 7, 3.6%), E1 (best supportive care; n = 8, 4.1%), C1 (induction, 7 + 3; n = 9, 4.7%), and E2 (HMA; n = 11, 5.7%). No health states were rated as equal to dead (i.e. utility = 0).

Disutilities and comparisons among health states

The health states that included chemotherapy (i.e. the C and D health states) were identical to health state B1, except for the addition of the chemotherapy regimen. Therefore, any utility difference between B1 and the chemotherapy health states can be attributed to the chemotherapy regimen and interpreted as the disutility of the chemotherapy regimen. Induction chemotherapy regimens had the greatest disutilities: –0.11 for CPX-351 and –0.15 for the 7 + 3 regimen. Disutilities for consolidation treatment regimens ranged from –0.03 (CPX-351 outpatient treatment) to –0.11 (the 5 + 2 treatment regimen). Disutilities relative to B1 are presented in Table 4, along with utility differences between the induction health states and between the consolidation health states.

Table 4. Differences among health state utilities.^a

Health state utility differences	Mean	SD	95% CI
Disutilities of chemotherapy regimens
C1 – B1: Disutility of induction, 7 + 3	−0.15	0.18	−0.18 to −0.13
C2 – B1: Disutility of induction, CPX-351	−0.11	0.15	−0.13 to −0.09
D1 – B1: Disutility of consolidation, 5 + 2	−0.11	0.15	−0.13 to −0.09
D2 – B1: Disutility of consolidation, HiDAC	−0.08	0.12	−0.09 to −0.06
D3 – B1: Disutility of consolidation, CPX-351, inpatient	−0.05	0.09	−0.06 to −0.03
D4 – B1: Disutility of consolidation, CPX-351, outpatient	−0.03	0.07	−0.04 to −0.02
F – B2: Disutility of transplant	−0.21	0.26	−0.25 to −0.17
Utility difference between induction health states
C1 – C2: 7 + 3 vs CPX-351	−0.04	0.12	−0.06 to −0.03
Utility differences between consolidation health states
D1 – D2: 5 + 2 vs HiDAC	−0.03	0.08	−0.05 to −0.02
D1 – D3: 5 + 2 vs inpatient CPX-351	−0.07	0.12	−0.08 to −0.05
D1 – D4: 5 + 2 vs outpatient CPX-351	−0.08	0.13	−0.10 to −0.06
D2 – D3: HiDAC vs inpatient CPX-351	−0.03	0.08	−0.04 to −0.02
D2 – D4: HiDAC vs outpatient CPX-351	−0.04	0.10	−0.06 to −0.03
D3 – D4: Inpatient CPX-351 vs outpatient CPX-351	−0.01	0.05	−0.02 to −0.00

^a Of the 200 participants, seven who began with the group 1 health states did not continue with group 2. Likewise, seven participants who began with group 2 did not complete group 1. Therefore, each health state was valued by 193 respondents.

Abbreviations. CI, confidence interval; HiDAC, high-dose cytarabine; SD, standard deviation.

T-tests were conducted to examine pairwise differences between health state utilities, and all comparisons revealed statistically significant differences between utilities. The utility of health state B1 (temporary remission) was significantly (p < 0.0001) greater than the utilities of all chemotherapy health states (i.e. the C and D health states), suggesting every chemotherapy regimen was associated with a significant disutility. The difference between the CPX-351 and 7 + 3 induction health states was also statistically significant (difference = 0.045; p < 0.0001). In addition, all differences among the consolidation health states were found to be statistically significant: D1 vs D2 (difference = 0.035; p < 0.0001), D1 vs D3 (difference = 0.066; p < 0.0001), D1 vs D4 (difference = 0.078; p < 0.0001), D2 vs D3 (difference = 0.031; p < 0.0001), D2 vs D4 (difference = 0.042; p < 0.0001), and D3 vs D4 (difference = 0.011; p = 0.001).

Sub-group comparisons

No statistically significant differences in utility or disutility scores by geographic location (i.e. London vs Edinburgh) were found. There were no statistically significant differences in utility scores by gender. When dividing the sample into older and younger age groups (by median split), the older group was found to have slightly, but significantly, higher utility scores for all health states except B2 (t = –1.202, p = 0.23), C2 (t = –1.808, p = 0.07), and F (t = –0.239, p = 0.81). However, there were no statistically significant age differences in disutilities, indicating that the relationships between health state utilities did not differ between the older and younger age groups.

Discussion

All health states representing chemotherapy were associated with significant decreases in utility. As expected, respondents tended to prefer chemotherapy regimens with shorter hospital stays (unless the regimen with the shorter stay had other disadvantages), shorter infusion times, and less severe adverse events. Therefore, health states describing the CPX-351 treatment process had higher utilities than corresponding health states describing standard chemotherapy regimens (e.g. 7 + 3 induction and 5 + 2 consolidation). The chemotherapy health state associated with the smallest disutility was CPX-351 consolidation administered as outpatient treatment, suggesting respondents perceived value in avoiding hospitalization.

The health states in which a chemotherapy regimen was followed by remission (i.e. the C and D health states) can be considered “path states”, which describe a sequence of health-related experiences rather than a chronic unchanging state of health^35–37. Because each path represented exactly 1 year, the resulting disutilities calculated as differences among these health states can be used in economic modeling as QALY decrements. When used in a CUA, the difference scores reported in Table 3 should be used to adjust utility scores for a 1-year period. For example, health state C1 is identical to health state B1, except for the addition of the 7 + 3 induction chemotherapy regimen including hospitalization, treatment process, adverse events, and recovery. Therefore, the utility difference between these two health states can be attributed entirely to the addition of the chemotherapy regimen, and the difference score of –0.15 can be considered the disutility of the 7 + 3 induction regimen. When using this value in a model, a QALY decrement of –0.15 should be applied to each patient who experiences this chemotherapy regimen.

The other health states also followed the expected patterns. Durable remission (health state B2) was associated with the highest utility, whereas treatments for patients not considered to be appropriate for chemotherapy (E1 and E2) had relatively low utilities. These health states were rated as chronic states rather than path states, which means the utility values can be applied in a CUA for any period of time. The transplant procedure (F) was associated with a substantial disutility relative to B2, and this value of –0.21 can be used as a QALY decrement similar to the chemotherapy disutilities discussed above.

Although this study yielded utilities that are likely to be useful in modeling, the vignette-based approach has inherent limitations that should be considered when using these values. With any vignette study, the resulting utility scores represent the health states being rated, rather than actual patient experience. In the current study, health states were drafted based on detailed literature review and an extensive series of iterative clinician interviews. Every clinician provided input at multiple points during health state development to ensure the health states represented the typical patient experience as accurately as possible. Nevertheless, the extent to which these utilities derived from the preferences of a general population sample may differ from values reported by actual patients is not known.

The 1-year TTO time horizon and path state approach should also be considered when using the resulting utility values. Longer TTO time horizons, such as the 10-year time horizon in the original EQ-5D valuation study^58,59, are more common. However, a wide range of long and short time horizons are used⁶⁰, and shorter time horizons may be most appropriate when estimating utilities associated with temporary events or diseases with relatively short life expectancies. While some research has reported similar utility scores with longer and shorter time horizons⁵⁶, other studies suggest the time horizon can have an impact on the magnitude of utility scores or differences between health state utilities^54,61,62. In the current study, it was necessary to use the shorter time horizon so that the impact of chemotherapy regimens lasting ∼ 1 month would not be obscured in the context of a much longer time period, such as 10 years. In addition, the 1-year approach allows for the results to be used conveniently as QALY decrements in subsequent modeling. It is possible that preferences among chemotherapy regimens could differ if they were valued in the context of longer health states.

An additional limitation relates to the differences in adverse events represented in the health states. Because of the limited data on differences in adverse event profiles between CPX-351 and the other regimens at the time this study was conducted, the development of the health state vignettes relied on clinician opinion to identify key adverse events. Clinicians who had personal experience treating patients with the more established regimens (i.e. 7 + 3, 5 + 2, HiDAC) and CPX-351 reported that they had observed differences in the extent of hair loss and fatigue, which are two adverse events likely to have an impact on patients’ quality-of-life. Therefore, these two differences were represented in the health states. However, as additional clinical trial results are published for CPX-351, our understanding of its adverse events profile may continue to evolve. For example, a recently published phase 3 study reported rates of hair loss and fatigue with both treatments^17,63. In this trial, the reported alopecia rates were lower for CPX-351 than for 7 + 3, which supports the difference described in the health states. However, the trial did not report a notable between-group difference in fatigue, which differs from the personal experience reported by the clinicians interviewed for the current study.

Furthermore, in health state vignettes, it is not possible to represent every adverse events reported in clinical trials, and some common adverse events that did not appear to distinguish between treatment regimens were omitted (e.g. febrile neutropenia, pneumonia, hypoxia). It is possible that future research will reveal that some of the omitted adverse events differentiate between the two treatments. If so, the current utilities would fail to capture preferences associated with these differences. Future use of CPX-351 in clinical trials and clinical settings will provide additional information on its adverse event profile. Given these limitations associated with adverse events descriptions in the health states, researchers using the current utilities in economic models should interpret and use results with appropriate caution. In addition, modelers should review up-to-date published literature to determine whether new information on the CPX-351 profile is available.

Another limitation is that health state vignettes cannot represent the wide range of treatment approaches. Therefore, health states must be drafted to describe the most typical patient experiences. In this case, health states were designed to represent the most common chemotherapy regimens for AML (e.g. 7 + 3 induction, HiDAC consolidation, 5 + 2 consolidation) and one newer regimen (CPX-351). As new treatments for AML are introduced, it is possible that the current utilities could be useful beyond the specific regimens on which these health states are based. However, if future regimens are substantially different, it is possible that these health states may not be applicable.

It should also be noted that treatment for AML is highly variable, and the exact combination of chemotherapy regimens is likely to be determined based on the individual characteristics of each patient. For example, many patients may have multiple courses of consolidation rather than a single course, and, to represent these patients, researchers can apply the disutility of consolidation multiple times. In addition, some of the chemotherapy regimens described in the health states may not be appropriate for elderly or frail patients. Researchers should consider this segment of the patient population when using these utilities in a model, with the understanding that not all patients will be treated with the stronger chemotherapy regimens. To represent treatment of these patients in a model, researchers may want to use utility values for health states E1 (best supportive care) or E2 (hypomethylating agents), which describe more conservative treatment approaches. In sum, when using the current utility values, it will be important to confirm that the health state content is consistent with the treatment regimens and patient populations being modeled.

It should also be noted that the health states in the current study were longer and more complex than health states in most other vignette-based utility valuations. Based on the pilot study, several steps were taken to help simplify and clarify the health states. Health state language was simplified, formatting was edited (e.g. headers, separate sections for different concepts), and simple timeline graphs were added to the bottom of each health state card. In addition, interviewers tried to ensure respondents understood and considered all the health state content, and participants whose understanding was questionable were discontinued. Still, with long health states, it is possible some participants focused on specific details that were particularly salient to them, rather than being able to consider all health state content. It should also be noted that generalization to countries other than the UK is unknown, and the study was not powered to adjust for multiple comparisons across health states or sub-groups, and, therefore, findings of statistical significance may not be reproducible.

Conclusions

Despite limitations, findings suggest the vignette-based TTO method was feasible for quantifying health state utilities associated with AML treatment regimens. Furthermore, this is the first study to estimate differences among new and commonly used chemotherapy regimens for treatment of AML. A range of utilities are provided, which may be useful in models examining and comparing the value of treatments for patients with AML.

Transparency

Declaration of funding

This study was funded by Jazz Pharmaceuticals, Inc.

Declaration of financial/other interests

Louis S. Matza, Kristen A. Deger, and Timothy Howell are employees of Evidera, a company that received funding from Jazz Pharmaceuticals, Inc. for time spent conducting this research. Vicky Fisher and Arthur C. Louie are employees of and hold stock in Jazz Pharmaceuticals, Inc. Karen C. Chung is a former employee of Jazz Pharmaceuticals, Inc. and holds stock in Amgen, Baxalta, Baxter, Bayer, Gilead, Jazz Pharmaceuticals, Inc., and Shire. Kimberly Koetter, Andrew M. Yeager, and Donna Hogge have no relevant conflicts to disclose.

Previous presentations

Some of these results were previously presented at the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) 20th Annual European Congress in November 2017 in Glasgow, Scotland. The title of the presentation was “Health State Utilities Associated With Treatment Options for Acute Myeloid Leukaemia (AML) (PCN200)”.

Data availability statement

Data are available from the authors upon request.

Acknowledgements

The authors would like to thank Chantelle Browne, Regina Buachie, Owen Cooper, Ashley Duenas, Meredith Hoog, Julia Ingram, Christopher Langelotti, Laura Swett, Hayley Syrad, Amara Tiebout, and Erica Zaiser for assistance with data collection; Christine Thompson for statistical programming; and Amara Tiebout for editorial assistance. Editorial assistance was also provided by Kimberly Brooks, PhD, CMPP™, of SciFluent Communications and was financially supported by Jazz Pharmaceuticals, Inc.