Background
Although a type of rare disease, acute myeloid leukemia (AML) is the most common form of acute leukemia in adults, which accounts for approximately 25% of all leukemia in adults in the Western world [
1]. The median diagnosis age of AML is 68 years [
2]. As the population ages, the incidence of AML is expected to rise. AML progresses rapidly if left untreated; even with treatment, the survivorship remains poor, particularly among a few subtypes of AML. The 5-year survival rate is low at approximately 27% [
2]. In the United States (US), an estimated 21,380 new AML cases and 10,590 deaths related to AML were expected in 2017 [
2,
3].
The diagnosis and treatment of AML can take a toll on a patient’s life. Studies have shown a substantial drop of health-related quality of life (HRQL) shortly after the AML diagnosis and persistent through the course of therapy [
4,
5]. The current standard of care for AML consists of induction therapy to induce remission and post-remission consolidation chemotherapies. The treatment strategy for an individual is informed by a number of factors such as age, comorbidities, performance status, disease subtype [
6]. Although many patients achieve complete remission (CR) from the initial induction chemotherapy, most of them relapse, with less than one-fifth maintaining CR after 1 year [
2,
7,
8]. The intent of post-remission consolidation chemotherapies is to solidify CR and to delay relapse. However, adverse events are commonly associated with chemotherapies and can further compromise the HRQL of patients with AML. Eligible patients who achieve CR may receive allogeneic hematopoietic stem cell transplant (SCT), a potentially curative procedure, though relapse may still occur. Moreover, the process of SCT is logistically difficult (i.e., finding a donor match) and costly; it can substantially compromise patients’ HRQL due to the complexity of this invasive procedure, the associated intensive care for an extended hospitalization and the risks for adverse events (AEs) and complications [
9]. A large HRQL decrement during the short-term SCT recovery period is expected within the first few months following the SCT procedure, though the majority of patients who survive the procedure gradually return to pre-procedure levels of HRQL within 1–4 years [
9,
10].
New therapies are emerging to treat AML and to improve the HRQL of these patients. A proper assessment of utility values associated with health states of a disease and the treatment is important for the appraisal of new therapies during regulatory approval and for establishing their value for reimbursement [
11]. Societal preferences are recommended for health economic evaluation in order to assess the optimal allocation of health resources among the general population [
12,
13]. However, existing utility values are frequently lacking for rare diseases. AML disease specific utility values and those for toxicities associated with AML chemotherapies have yet to be established in the US [
14].
Discrete choice experiment (DCE) is a well-established method for eliciting preferences for medical products or services in the healthcare field [
15‐
19]. While standard gamble (SG) and time trade-off (TTO) are two conventional methods for estimating health state utility values, the DCE approach has been increasingly assessed for estimating health utilities [
20‐
23]. However, previous studies using the DCE approach for utility assessment focused on the methodology feasibility itself and were based on EQ-5D, which is a generic utility assessment measure. The choice-based method has become one of the recommended methods for eliciting preferences [
13], but the feasibility of using this approach to generate disease specific health state utilities is yet to be evaluated. This study applied the DCE method to estimate utilities from a US societal perspective for treatment-related health states, and dis-utilities of severe adverse events (SAEs) among patients newly diagnosed with AML who are fit for standard induction chemotherapy and for whom SCT is a potentially curative therapy. The health state utility values from this study can be used to generate the quality adjusted life years (QALYs) in cost-effectiveness/cost-utility analysis of a new AML intervention.
Discussion
AML is a rapidly progressive disease, and as a result, patients usually receive aggressive and often complex, treatments such as chemotherapies and SCT. These treatments may be life-saving, but can also negatively affect patient HRQL profoundly over both the short term and long term [
4], particularly among older patients [
5]. A 2004 literature review by Radaelli et al. assessed HRQL among patients with AML and found that HRQL score dropped shortly after diagnosis and persisted low throughout the course of therapy [
4]. Furthermore, while long-term survivors regained HRQL nearly completely, in some cases, substantial negative impacts may persist well after treatment has ended. As such, clear definitions of health states for AML and proper utility valuations of these health states are necessary to reflect the benefits and risks associated with new therapies [
29]. The availability of such values enables the incorporation of QALYs for cost-effectiveness/cost-utility analysis of these therapies. This study estimated the utility values for pre-defined, based on literature and consultation of clinical experts, treatment-related AML health states, and disutilities of treatment-related SAEs from a US societal perspective. A combination of the attributes, which considered patient-reported symptoms, function, mental health, and the care process, performed well in the descriptive system used in this study for AML health states categorization. Health states associated with a greater degree of independence and ability to function or work normally were highly preferred.
SG and iterative TTO are two frequently used methods to eliciting preference utility values. However, these methods usually involve in-person interviews and thus, most of the time data would be collected from a convenient sample from the general population when eliciting preferences. Moreover, the concept of SG or TTO are not always easily grasped by the respondents. Frequently, data from respondents who fail to understand those tasks are excluded from final analyses. The DCE method has attracted researchers in the healthcare field in recent years with its relatively simpler design and easier implementation than the traditional methods. Data from all respondents can be included for analysis and having a sample that is national representative is practical, as demonstrated in this study. In a comparative study, the health utility measurement scale, EQ-5D, was used to facilitate the utility estimation using both DCE and iterative TTO; the results of that study supported DCE to be a valid tool for this purpose [
20,
21].
The current study is the first estimation of societal preferences using the DCE approach for disease-specific utilities, namely the AML treatment-related health states. Using the DCE approach, this study applied the statistical approach to constrain the upper and lower limit for utility estimation to be 0 and 1 to represent death and full health. The pre-defined health states utility estimation considered the intercept from the model to account for the residual factors not captured by the included attributes. This is because even when an AML patient is in remission, this patient is not likely in full health due to those residual factors. This approach is different from the one typically applied with EQ-5D, where a person who has no problem with any of the five dimensions of interest is assigned to have a value of 1 (full health). Admittedly, this definition of full health has been criticized for not considering a person’s general well-being [
30‐
32]. On the other hand, this study arbitrarily constrained the lower bound of the utility value to be 0, meaning no health states worse than dead. Our study focused on estimating the utility values for the five pre-defined health states, which are believed to be preferred states to death. Therefore, this constraint is reasonable. Within the current attributes considered for the health states, a negative utility value cannot be derived. Cognizant a health state being worse than dead possible (e.g., in the EQ-5D value set), in extreme situations, an AML patient could have a negative utility value when considering the dis-utilities due to SAEs should be experienced.
Results from this study estimated the utility values for five health states associated with newly diagnosed patients with AML fit for standard chemotherapy. The utility values ranged from 0.875 for CR to 0.158 for SCT event. Several of the utility values estimated in this study were within the range of values reported previously for other hematologic neoplasms, regardless of methodology for assessment. For example, the utility value of CR for acute lymphoblastic leukemia (ALL) using TTO was 0.86 [
33] and for chronic lymphoblastic leukemia (CLL) using SG was 0.91 [
34] in the UK general population; the utility value of CR for chronic myelogenous leukemia (CML) was 0.84 using the TTO approach taking responses from convenience samples in Canada, the US, the UK, and Australia [
35]. The study for CML included a similar health state of “not responding to treatment” with a utility value of 0.21 [
35], similar to our “no CR” health state with a utility value of 0.262. In addition to similarities in utility values associated with CR and no CR, we also found similar values from these studies when evaluating safety. For example, the study for CLL utility assessment also included grade 3/4 pneumonia and found a disutility of − 0.20 [
34]. This value is also similar to the finding of − 0.218 for grade 3/4 serious infection in our study.
Because of a relatively small number of patients affected, therapy development for rare diseases has unique challenges, particularly when a large sample size is needed for assessing relevant outcomes. AML in particular has posed its own challenge due to the presence of many subtypes with various gene mutations [
14]. For example, only 30% of adults with newly diagnosed AML have mutations in the fms-related tyrosine kinase 3 gene (FLT3) which is associated with a high relapse rate and poor prognosis. The findings from this study support the use of DCE methodology as a valid approach for obtaining societal preferences for disease-specific treatment-related health states. This can be a particularly valuable tool for generating utility values for rare diseases such as AML. The findings of this study may be generalizable for AML, given that the utility values were estimated from the societal perspective, with a national representative sample in the US. The preference values estimated in this study can be applied in economic evaluations of AML interventions.
The findings from this study may also have implications in addressing the unmet needs when developing new therapies for AML. The results showed that the most preferred attributes were those involving the maintenance of independent functioning and avoidance of hospitalization. CR had the highest utility value (0.875), followed by post-SCT short-term recovery (0.398), relapse (0.355), no CR (0.262), and SCT (0.158). These values reflect the extent to which the society place on treatment-related AML health states though it is important to keep in mind that SCT remains the only potentially curative treatment for AML. The relatively low utility value of SCT may mainly be a reflection of the restriction associated with the procedure that people strongly preferred to avoid (e.g., hospitalized with frequent blood transfusion). Admittedly, the preferred attributes resulting from this study were from the perspective of the general population. For future AML therapy development, the preferences coming from patents themselves should also be taken into account.
Available AML therapies with curative intent have remained constant for several decades, although emerging therapies may change the treatment landscape [
36,
37]. Utility values and societal preferences, such as those produced by this study, can aid in economic assessments of new therapies for AML as well as identify unmet needs in the patient population. Future research on approaches to reduce the negative impact of AML treatment, with the goals of maintaining independence and minimizing life disruption as well as eliciting disease remission, would be valuable for the future.
Limitations
The online panel-based survey was answered by adults from the general population in the US, and the utility values were generated from a societal perspective. Participant eligibility for the survey was determined based on self-reported information. Overall, the participants reflected the demographics of the US population, although by study design, they had a higher educational attainment (47.7% with a bachelor’s degree or higher) than the general population (33%) [
38]. However, the preferences of patients with AML, and according utilities, may differ from those of the general US population and those of other countries. For example, in this study, we observed that the likelihood of participants choosing the scenario of being able to work normally was over 3 times that of choosing a scenario involving not being able to take care of oneself. When considering that the median age of AML diagnosis is 68, the study participants’ preference for ability to work vs. the ability to take care of themselves could be different from that of the general population.
Among all participants, 85% provided consistent responses on the choice cards for test and retest validity, suggesting that the choice cards were well understood and that participants made reasonable preference decisions [
39]. AML is a complex disease, which has heterogeneous course and therapeutic options. As such, AML patients could have more health states based on the given treatments and stage of disease. Our study included five health states for patients newly diagnosed with AML who are fit for standard induction chemotherapy and for whom SCT is a potentially curative therapy. However, the DCE design provides flexibility in deriving utility values for additional health states. For example, the CR health state was currently defined for
CR after all treatments are completed. By applying different levels of the attributes,
CR during consolidation may be defined with fever (occasionally), lack of energy (occasionally), problems with daily function (not able to work but able to perform daily activities), anxious/depressed (occasionally), blood transfusion (occasionally), and hospitalized (occasionally), and thus have an estimated utility value of 0.421.
It should be noted that the treatment-related AML health states and the attributes/levels used to define these health states were based on literature review and experts’ input. No separate patient focus groups were conducted to confirm the content validity and attributes’ relevance to health states. Ideally, focus groups should be conducted to include patients across the proposed health states and with some patients who have experienced most of the health states. However, due to the rarity of the disease, the age of the patients, and the frailty of the patients in certain health states, it was not practical to carry out focus groups among this particular patient population. We believe the current two approaches together used to determine health states, particularly the rounds of discussions with the experts, should have reflected the common health states that clinicians observe in their daily clinical practice. Of course, future studies applying the methodology in this research should consider including patient focus groups when possible.