Journal of Geriatric Oncology 11 (2020) 24–30 Contents lists available at ScienceDirect Journal of Geriatric Oncology Treatment decision-making in chronic lymphocytic leukaemia: Key factors for healthcare professionals. PRELIC study Concepción Boqué a, María Reyes Abad b, María José Agustín b, Manuel García-Goñi c, Carolina Moreno d, Clara Gabás-Rivera e,⁎, Enrique Granados f, Antonio Castro-Gómez f, Carlos Pardo f, Luis Lizán e,g a Instituto Catalán de Oncología, Barcelona, Spain b Hospital Miguel Servet, Zaragoza, Spain c Universidad Complutense, Madrid, Spain d Hospital Santa Creu y Sant Pau, Barcelona, Spain e Outcomes'10, Castellón, Spain f Gilead Sciences, Madrid, Spain g Departamento de Medicina, Universitat Jaume I, Castellón, Spain a r t i c l e i n f o a b s t r a c t Article history: Objective: To explore the preferences of Spanish healthcare professionals (haematologists and hospital pharma- Received 20 November 2018 cists) for the treatment selection of active Chronic Lymphocytic Leukaemia (CLL) patients at first relapse, Received in revised form 13 March 2019 condition that mainly afflicts older adults. Accepted 16 March 2019 Methods: A discrete choice experiment (DCE) was conducted among haematologists and hospital pharmacists. A Available online 3 April 2019 literature review and a focus group informed the DCE design. CLL treatment settings were defined by seven at- tributes: four patient/disease-related attributes (age, functional status, comorbidities, and risk of the disease) Keywords: Chronic lymphocytic leukaemia and three treatment-related attributes (efficacy [hazard ratio of progression-free survival, HR-PFS], rate of Discrete choice experiment discontinuations due to adverse events and cost). A mixed-logit model was used to determine choice-based pref- Preferences erences. Relative importance (RI) of attributes was calculated and compared between stakeholders. Willingness- Healthcare professionals to-pay (WTP) was estimated through the DCE. Besides, nine ad-hoc questions were posed, to explore more in Physiological age depth CLL treatment decision making. Elderly Results: A total of 130 participants (72 haematologists and 58 hospital pharmacists) answered the DCE. All attri- Ageism butes were significant predictors of preferences (p b 0.05) in the multinomial model. Higher RI was obtained for Decision-making treatment-related attributes: the highest rated being ‘cost’ (23.8%) followed by ‘efficacy’ (20.9%). Regarding Drug cost patient-related attributes, the highest RI was obtained for ‘age’ (18.1%). No significant differences (p N 0.05) in RI between haematologists and pharmacists were found. WTP for the treatment was higher for younger CLL patients. Ad-hoc questions showed that patient age and functional status influence treatment decisions. Conclusions: For healthcare professionals, ‘cost’ and ‘efficacy’ (treatment-related attributes) and age (patient- related attribute) are the main factors that determine CLL treatment selection at first relapse. WTP decreases as patient's age increases. © 2019 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http:// creativecommons.org/licenses/by-nc-nd/4.0/). 1. Introduction population, the incidence and prevalence of CLL among older persons is expected to continue to increase. Chronic lymphocytic leukaemia (CLL) is the most common leukae- Treatment of older patients remains complicated due to multiple mia in the Western world [1]. It typically afflicts older adults [2], with factors, such as the variability in fitness [3], cognition and comorbidi- a median age at diagnosis of 72 [1]. With the rapid ageing of the ties [2]. Particularly, treatment of relapsed CLL is highly personalized due to the advanced age of the patient population, the risk of severe adverse events (AEs), and the often chronic nature of this disease Abbreviations: AEs, Adverse events; CLL, Chronic lymphocytic leukaemia; DCE, [4]. Traditional intensive chemotherapy treatments for CLL can be dif- Discrete-choice experiment; HR-PFS, Hazard ratio of progression-free survival; NHS, ficult for older patients to tolerate [2,5]. More recently, novel oral National health system. ⁎ Corresponding author at: Universitat Jaume I, Parc Cientific Tecnológic i Empresarial, agents are proving to be both well tolerated and effective in this pop- Edificio Espaitec 2, Avda. Sos Baynat s/n, 12071 Castellón de la Plana, Castellón, Spain. ulation, redefining the field of CLL therapy. Finally, new treatments E-mail address: cgabas@outcomes10.com (C. Gabás-Rivera). and combination therapies currently undergoing further study have https://doi.org/10.1016/j.jgo.2019.03.010 1879-4068/© 2019 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). C. Boqué et al. / Journal of Geriatric Oncology 11 (2020) 24–30 25 potential utility in older patients with CLL [2], and the capacity to in the field of haematology and hospital pharmacy from the SNS. modify the natural course of the disease [6]. Haematologists were eligible if they had at least two years of profes- Treatment choices for older patients must account for the interac- sional experience, their professional time dedicated to clinical tions of disease, age, and therapy, since ageing involves a progressive assistance was over 70%, and they had expertise in CLL patients. Phar- decline in functional reserve and greater comorbidity; all these factors macists were eligible if they had at least two years of professional expe- combine to reduce life expectancy and tolerance of stress [7,8]. These rience and were familiar with onco-haematological drugs. changes are universal, but occur at different rates in different individ- The minimum sample size necessary for the DCE was established ac- uals; hence age alone is a poor prognostic factor for life expectancy cording to the recommendation proposed by Orme [22]: n ≥ 500c ta , where and CLL treatment tolerance [3,7]. Therefore age should not be the n is the number of respondents, c is the maximum number of levels per sole determinant in decision-making and choice of therapy [9]. Ageism attribute, t is the number of tasks, and a is the number of alternatives is discriminating against individuals or groups on the basis of their per task. age [10] and is a topic that has been studied at length in the field of on- Professionals were invited to participate through the Spanish Society cology [11], since it is an attitude frequently observed in health profes- of Haematology and by a direct invitation to the pharmacy service sionals [12]. To rely totally on chronological age in making treatment of hospitals and clinics. Therefore, the total number of professionals decisions is ethically inappropriate [13,14]. Nowadays, the assessment invited was not known. They were requested via email to complete a of physiological age, its influence on treatment tolerance, and personal- questionnaire (Supplementary Table S1) administered on a dedicated ized treatment of CLL are emerging as key issues in the management of study website. The study questionnaire was available from March to CLL in older patients [7,8]. May 2017. Cost considerations have also been described to be a factor in treating A multidisciplinary scientific committee constituted by experts in older CLL patients [15]. In National Healthcare Systems (NHS) with public haematology (n = 2), hospital pharmacy (n = 2) and health economy funding, like the Spanish NHS, the Sistema Nacional de Salud (SNS), pre- (n = 1), each with N15 years' professional experience, provided scien- scribing physicians can be considered as double agents acting as patients' tific advice during the development of the study. advocates but also as society's gatekeepers of resource use. On the one hand, public healthcare organizations put pressure on physicians to con- 2.2. Discrete-Choice Experiment (DCE) trol pharmaceutical spending and, on the other hand, doctors are pa- tients' agents and as such prescribe treatments that maximize the Stated-preference surveys are one of the most reliable and valid utility, effectiveness and quality of care in the face of economic restric- techniques available for quantifying participants preferences. The tions [16]. In the hospital setting, the decision of incorporating drugs in most commonly used stated-preference format in health care is the formulary is the responsibility of the Pharmacy and Therapeutics Com- DCE [23]. The DCE allows the identification and evaluation of partici- mittee [17]. In this regard, some cost awareness studies have suggested pants' preferences for treatment and their relative importance in deci- that physicians only have an approximate knowledge of drug prices sion making [24]. Respondents are presented with several treatment [16] and tend to neglect price differences between products of identical choice sets, each one comparing two hypothetical treatment options composition [18]; while other findings show that doctors feel pharma- (scenarios) and are asked to choose between them. Each choice set con- ceutical costs are important [16,19]. sists of multiple hypothetical profiles consisting of a fixed set of criteria Healthcare professionals' preference over alternative treatments can (attributes), with varying values (levels) between the profiles [25]. The be addressed through the discrete choice experiment (DCE), a common DCE allows for the estimation of overall preferences for any combina- technique to elicit individual preferences for health care services or tion of attributes and is shown to be one of the most sensitive methods technologies [20]. The DCE allows for aproximation of the decision- to elicit preferences, so the DCE is considered to better reflect actual de- making process performed by patients and professionals in real life. It cision making than rating scale exercises [26]. The DCE was developed consists of presenting the participants with a free choice of two hypo- according to good research practices recommended by International thetical treatment options (scenarios), determined by different charac- Society for Pharmacoeconomics and Outcomes Research (ISPOR) [27]. teristics. The answers are analyzed to estimate the relative importance (RI) of the investigated attributes in the choice of treatments [21]. A few studies have explored patients' or professionals' preferences 2.2.1. Identification of Attributes and Levels for the treatment of onco-haematological malignancies, mostly related Attributes and their associated levels were initially identified from to treatment efficacy or AEs. However, to our knowledge, neither the literature review. After, a focus group with the scientific committee to importance of drug cost nor patient-or disease-related attributes have assess the relevance of the potential attributes identified in the litera- been previously considered in preference elicitation studies in CLL, ture, to identify attributes not retrieved in the literature and finally to despite they constitute key factors in treatment decision-making. assess the comprehensibility of the attributes and levels proposed. The The aim of PRELIC study is to explore the preferences of Spanish result of literature review and the focus group produced seven attri- healthcare professionals (haematologists and hospital pharmacists) butes with a maximum of three levels each (Table 1). Various reasons for the selection of treatment for CLL patients at first relapse through a lay behind the final selection of these attributes. The “age” attribute DCE. Although health professionals are not decision makers in the was limited to elderly patients (70, 75 and 85 years). Two attributes (ge- price and reimbursements processes within the SNS, we investigate netic alterations and relapse) were redefined into one (“risk of the dis- how important different attributes of the therapeutic scenario of CLL ease”), according to the group criteria for patients at relapse proposed in are for them, and what factors determine their preferences, in regards ESMO guidelines and in the Spanish national guidelines for the manage- to both the patient and therapeutic characteristics. In particular, we ment of patients with CLL [1,28]. The range of drug costs was chosen focus on patient's advanced age as a possible decision bias, and on the taking into account the annualized list prices of the main drugs used cost of the drug. for relapsed CLL in 2017 in Spain. Hospital medicines or hospital- dispensed medicines are fully reimbursed in Spain; therefore, no out- 2. Materials and Methods of-pocket expenses are expected from the patients. 2.1. Study Design and Participants 2.2.2. Experimental Design Experimental design techniques were then used to construct a series PRELIC is an observational study developed in the Spanish of choice tasks from combinations of the attribute levels (called scenario healthcare setting. The participants of the study were professionals alternatives). The design was orthogonal (each pair of levels appears 26 C. Boqué et al. / Journal of Geriatric Oncology 11 (2020) 24–30 Table 1 regarding the cost in a scale from a reference cost of €20,000 to ≥ Attributes and levels included in the scenarios. €100,000, using intervals of €10,000. Attributes Levels Patient/Disease Age (years) 70 2.4. Survey Instrument 75 85 Performance Good (ECOG 0-1) The questionnaire consisted of four parts: (i) screening questions status Poor (ECOG ≥2) (selection criteria), (ii) sociodemographic data, (iii) preferences (DCE) Comorbidities No comorbidities and (iv) ad-hoc questions. To avoid response bias due to the framing Renal failure (with or without other effect of question sequence, the order of ad-hoc questions about youn- comorbidities) Other comorbidities (with functional ger and older patients were randomized in the questionnaires. repercussion and normal renal function) Risk of the Low [without 17p-deletion/p53 mutation and disease late relapse (N3 years)] 2.5. Statistical Analysis High [with 17p-deletion/p53 mutation and/or early relapse (b3 years)] 2.5.1. DCE Treatment Efficacy HR = 0.30 (70% reduction in the risk of (HR-PFS) progression or death) Based on respondents' answers to the series of choice questions, the HR = 0.45 (55% reduction in the risk of mixed logit model was applied to calculate the preference weights for progression or death) all attribute levels, using Stata software [29]. The model relates the HR = 0.65 (35% reduction in the risk of probability of choosing one treatment over another to observable differ- progression or death) ences between treatment options. The model accounts for preference Discontinuation 5% due to AEs 15% heterogeneity among respondents. It yields both a mean effect and a 25% standard deviation (SD) of effects across the sample [30]. The depen- Cost €10,000/year dent variable was choice, whereas independent variables were the €35,000/year attribute levels. €70,000/year The mixed logit model was firstly considered with all the variables as ECOG: Eastern Cooperative Oncology Group; HR: Hazard Ratio; HR-PFS: Hazard Ratio of categorical, except the cost, which was considered linear to estimate Progression-Free Survival; AEs: Adverse Events. WTP. In a second model, the attributes age, treatment efficacy and dis- continuation for adverse effects were considered as linear variables. equally often across all pairs of attributes within the design) and bal- The two models were compared using the Akaike Information Criterion anced (each level of an attribute occurs the same number of times) and the Bayesian Information Criterion [31]. In both cases, values [24]. The fractional factorial analysis (main effects orthogonal matrix) obtained for the linear model were lower than those obtained for the generated 36 scenarios, with a mix-and-match algorithm being used categorical model, indicating that the linear transformations provided to generate the choice sets. Dominated scenarios were identified and a better adjustment. minimized. To avoid participant fatigue, the 36 scenarios were distrib- The regression estimates from the mixed logit model are interpreted uted among three versions of questionnaires of 12 multiple-choice as the utilities associated with each respective attribute level and sets. Each participant received a randomly assigned version of the ques- referred to as part-worth utilities, which represent the relative contri- tionnaire. Every task required to choose between two scenario alterna- bution of changes in treatment attributes to the choice. A higher part- tives (Supplementary Table S2). worth utility for a specific attribute level indicates a greater likelihood that a treatment scenario including that attribute level will be chosen, 2.3. Ad-hoc Questions all else being equal. Part-worth utilities were subsequently used to esti- mate RI of each studied attribute. RI was calculated as the percentage Nine ad-hoc questions were included in the questionnaire to com- ratio of the utility difference between the highest and lowest levels plement the information of the DCE. The questions raised were based within each attribute to the sum of differences between the highest on the contributions of the scientific committee about the most relevant and lowest levels across all attributes in total. RI ranges from 0 to attributes for decision-making in CLL. This included three questions 100% per attribute (sum to 100% across all attributes) [32]. The inclusion about the importance given to age, four about functional status, and of a cost attribute in the DCE allows for a monetary estimate of utility, two to estimate the willingness-to-pay (WTP) in a situation of thera- the WTP [33], by applying the methodology of the marginal rate of sub- peutic choice in CLL (Supplementary Table S1). stitution [34]. This method consists in dividing the part-worth utility of In question 1, professionals were asked about the “Importance of the attribute levels (in this case, age [per year], efficacy [HR-PFS] [1%] patient's age when making decisions or recommendations about phar- and % discontinuation due to AEs [1%]) by the part-worth utility of the macological treatment in CLL” using a Likert scale from 1 (not impor- additional cost per month; and expressed in absolute value. tant) to 5 (very important). The range “important” was defined as the values 4 (important) and 5 (very important) of the scale, and “not important” as the values 1 (not important) and 2 (slightly important). 2.5.2. Subgroup DCE Analysis In questions 2–3, professionals assessed the “Frequency in which Professional specialty (haematology and pharmacy) and time of ex- patient's advanced age modifies the therapeutic regimen indicated/rec- perience were studied as possible explanatory variables for the prefer- ommended for CLL” and the “Frequency in which patient's advanced ences, estimated independently for each participant. Traditional age impedes/hinders active treatment”, respectively, using a Likert parametric independent t-tests or the non-parametric Mann-Whitney scale from 1 (never) to 5 (always). In question 2, the range “modifies” U test, according to data distribution, were used to assess whether RI was defined as the values 3 (sometimes), 4 (frequently) and 5 (always) of attributes differed between groups. Regarding time of experience, of the scale, and “does not modify” as the values 1 (never) and 2 two subgroups were established from the cut-off point of mean years (rarely). In question 3, the range “prevents/hinders” was defined as of professional experience of the sample: professionals with lower or the values 3 (sometimes), 4 (frequently) and 5 (always), and “does higher time of experience than the mean of participants. Preferences not prevent/hinders” as the values 1 (never) and 2 (rarely). based on time of professional experience were then analyzed and com- In questions 4–7, participants indicated the specific value for age in pared from the total of participants and by specialty subgroups each situation. In questions 8–9, participants indicated their preferences (haematologists and pharmacists). C. Boqué et al. / Journal of Geriatric Oncology 11 (2020) 24–30 27 2.5.3. Ad-hoc Questions Table 3 The importance given to age in treatment decision-making was de- Professionals' preferences. scribed by relative and absolute frequencies of the responses. The age Attribute Level Coefficients SE p-value limit for treatment recommendation according to performance status Age (years) Per year −0.123 0.017 0.000’ was estimated as mean (SD). WTP from hypothetical scenarios was de- 70 −8.608 – – scribed by relative and absolute frequencies of response and by 75 −9.223 – – weighted mean values. 85 −10.453 – – Performance Good (ECOG 0–1) 0 – – status Poor (ECOG ≥2) −1.195 0.173 0.000 3. Results Comorbidities No comorbidities 0 – – Renal failure (with or without −0.328 0.148 0.027 3.1. Demographic Characteristics other comorbidities) Other comorbidities (with −0.532 0.176 0.003 functional repercussion and A total of 130 professionals from the different Spanish regions com- normal renal function) pleted the electronic questionnaire and were included in the final data Risk of the Low 0 – – analysis. Of these, 72 (55.4%) were haematologists and 58 (44.6%) disease High 0.753 0.197 0.000 were pharmacists. Participants' characteristics are presented in Efficacy 1% reduction in the risk of 0.061 0.007 0.000’ (HR-PFS) progression/death Table 2. This sample size accomplishes the minimum number of partic- HR = 0.30 (70% reduction in the 4.263 – – ipants according to the sample estimation, n ≥ 500c/ta, where c is the risk of progression/death) maximum number of levels per attribute (three), t is the number of HR = 0.45 (55% reduction in the 3.349 – – tasks (36 tasks), and a is the number of alternatives per task (two). risk of progression/death) HR = 0.65 (35% reduction in the 2.131 – – risk of progression/death) 3.2. Discrete Choice Experiment Discontinuation 1% −0.066 0.009 0.000’ due to AEs 5% −0.329 – – 3.2.1. Preferences 15% −0.988 – – 25% −1.647 – – Table 3 and Fig. 1 present the results of the analysis of professionals' Cost €1/month −0.000^ 0.000 0.000 preferences for the treatment scenario in CLL. All attributes were statis- €833.33/month (€10,000/year) 0.404 – – tically significant determinants of professionals' preferences (p b 0.05). €2916.67/month (€35,000/year) −1.413 – – Among all participants, higher RI was obtained for treatment-related €5833.33/month (€70,000/year) −2.826 – – attributes, the highest rated being ‘cost’ (23.8%) followed by ‘efficacy’ (haz- ‘Age, efficacy and discontinuation due to adverse effects were considered as linear vari- ard ratio of progression-free survival, HR-PFS) (20.9%). Regarding patient- ables, as linear transformations provided a better adjustment than the categorical related attributes, the highest RI was obtained for ‘age’ (18.1%) (Table 4). mixed-logit model. ^0.00048; AEs: % of discontinuation due to adverse events; HR-PFS: Hazard Ratio of Progression-Free Survival; SE: Standard error. 3.2.2. Willingness-to-Pay Professionals' WTP calculated from the DCE for the attributes age, ef- with higher time of experience (p b 0.026). The analysis of prefer- ficacy (HR-PFS) and % discontinuation due to AEs is shown in Table 5. ences in pharmacists with an experience b17 years (n = 31) and WTP for CLL treatment increases for any of the following: lower patient ≥17 years (n = 27) showed no significant differences (data not age, higher PFS or lower discontinuation rates due to AEs. shown). 3.2.3. Factors Defining Preferences: Subgroup Analysis In the analysis of preferences by specialty subgroup, no significant 3.3. Ad-hoc Questions differences in attributes RI were found between haematologists and pharmacists (Table 4). Questions 1–3: 90.7% of participants considered age to be important Two groups were established according to mean time of professional when making decisions about CLL treatment, while only 4.6% consid- experience (16.6 years): professionals with b17 years of experience ered that age is irrelevant for therapeutic decisions. 93% of participants (n = 64) and professionals with ≥17 years of experience (n = 66). Sub- considered that age modifies the recommended therapeutic regimen group analysis of mean attribute importance revealed that professionals for CLL. Additionally, 77.7% of participants expressed a view that the with larger experience gave stronger preference to “age” (p b 0.042) patient's age hinders access to active treatment. (15.6% those with b17 years of experience, and 18.7% those with Questions 4–7: In patients with good performance status, profes- ≥17 years of experience) (Supplementary Table S3). sionals would recommend treatment with chemo-immunotherapies The analysis of preferences between haematologists with an ex- up to 80.9 (SD: 9.2) years. On the other hand, treatment with targeted perience b17 years (n = 33) and ≥ 17 years (n = 39) revealed signif- therapies would be recommended until later ages (86.5 [SD: 8.5] icant differences, with higher RI given to “age” (p b 0.040) and lower years). When patient performance status is restricted, the age of recom- RI given to “% discontinuation due to AEs” by those professionals mendation of active treatment for CLL decreases, recommending Table 2 Participants sociodemographic characteristics. Total (n = 130) Haematologists (n = 72) Pharmacists (n = 58) Gender, male (%) 45.4 38.9 53.4 Age, years (mean, SD) 45.6 (8.4) 46.2 (7.5) 44.7 (9.4) Professional experience, years (mean, SD) 16.6 (8.4) 16.8 (7.7) 16.3 (9.3) Practice setting Public (%) 93.1 90.3 96.6 Public-private (%) 6.9 5.4 3.4 Position (%) Head of department (%) 31.5 20.8 44.8 Associate (%) 68.5 79.2 55.2 CLL patients per year (mean, SD) – 72.1 (71.1) – Results for the whole sample of participants are highlighted in bold. 28 C. Boqué et al. / Journal of Geriatric Oncology 11 (2020) 24–30 Fig. 1. Part-worth utilities within attributes. (utilities associated with each attribute level). AEs: % of discontinuation due to AEs. HR-PFS: hazard ratio of progression-free survival. chemo-immunotherapies up to 75.2 [SD: 10.7] years and targeted ther- treatment would be that with lower cost, higher efficacy, and for a apies up to 82.1 [SD: 9.6]. younger patient. Questions 8–9: When considering a treatment offering a 1 year- In many countries, pricing and public financing decisions are driven PFS gain with respect to standard treatment; professionals or supported by economic evaluation studies. However, it is arguable expressed that the Spanish SNS should pay 12.3% less for the treat- whether decisions should be guided by cost-related preferences in deci- ment in patients aged ≥80, compared to patients aged 70 (€36,769 sion makers or even in health professionals. The American College of Phy- vs. €41,923 per year, respectively). In patients aged 70, 15.3% of pro- sicians outlined that health resource allocation decisions must focus on fessionals considered that it would be acceptable to pay ≥€60,000 for medical efficacy, effectiveness, and need, with consideration of cost the treatment, whereas only 7.6% consider this range acceptable in based on the best available medical evidence [45]. In line with this, patients ≥80 years. Detailed results for ad-hoc questions are shown cost-consciousness has been widely studied in scientific literature. in Supplementary Table S4. Some studies have stated that doctors have a limited or inconsistent un- derstanding of therapeutic costs [36–38]. On the other hand, it has also been found that doctors' stated cost-consciousness appears to be gener- 4. Discussion ally high, and many decisions regarding medical treatments are influ- enced by factors other than the expected benefit to the patient, CLL is a disease of old age [35], and its incidence and prevalence are including concerns about cost and income [39]. Knowing what influences expected to increase with the ageing of the population. In this context, physicians' attitudes toward health care costs is an important matter, be- there is a need to identify the factors that drive treatment decision- cause most health care expenditures are the results of doctors' decisions making for this population group, in order to provide them with the [37]. This study confirms that cost is important for healthcare profes- best treatment strategies. sionals when making decisions about CLL treatment. High costs of new In this study, a DCE was carried out to assess the importance con- CLL therapies might have influenced these preferences. ferred by healthcare professionals (haematologists and pharmacists) From the DCE, we also found that WTP for the treatment depends on to the characteristics of the therapeutic scenario at first relapse of CLL. patients' age, being lower for older patients. In this study we do not ex- A total of 130 participants (72 haematologists and 58 pharmacists) an- plore the reasons why this happens. Considering the literature, there swered the study questionnaire. Cost (RI: 23.8%), efficacy (RI: 20.9%) could be at least two explanations. First, health professionals might per- and age (RI: 18.1%) were found to be the most important factors for pro- ceive that the value of some health states, and consequently the value of fessionals when making decisions about CLL treatment. The preferred the drugs aiming to correct them, are lower in older patients than in Table 4 Relative importance of attributes. Attribute Total (n = 130) RI haematologists (n = 72) RI pharmacists (n = 58) p-value (haematologists vs. pharmacists) Cost 23.8% 20.4% 21.7% 0.529a Efficacy (HR-PFS) 20.9% 20.7% 17.7% 0.063b Age 18.1% 16.6% 17.9% 0.366a AEs 12.9% 12.9% 12.5% 0.164b Performance status 11.7% 10.7% 11.7% 0.239a Risk of the disease 7.4% 11.4% 12.0% 0.540b Comorbidities 5.2% 7.3% 6.5% 0.104b Results for the whole sample of participants are highlighted in bold. a t-test. b Mann-Whitney U test; AEs: % of discontinuation due to AEs; HR-PFS: Hazard Ratio of Progression-Free Survival. C. Boqué et al. / Journal of Geriatric Oncology 11 (2020) 24–30 29 Table 5 active treatment and determining the type of therapy. In this sense, Willingness-to-pay (WTP). despite a low RI was assigned to performance status when consid- Attribute Unit WTP/montha ered jointly in the DCE, this attribute is a significant determinant of Age ↓1 year €253.9 preferences, and therefore, ad-hoc questions show that it influences Efficacy (HR-PFS) ↑1% €125.7 CLL decision-making. For patients with good functional status, % discontinuation due to AEs ↓1% €136.0 chemo-immunotherapy was recommended up to 80.9 (SD: 9.2) AEs: % of discontinuation due to adverse events; HR-PFS: Hazard Ratio of Progression-Free years, while in those with limited status only up to 75.2 (SD: 10.7) Survival. years. With regard to new targeted therapies, they were recom- a WTP for each attribute was estimated dividing the part-worth utility of the attribute mended up to 86.5 (SD: 8.5) years for patients with good condition levels (in this case, age [per year], efficacy [HR-PFS] [1%] and % discontinuation due to AEs and up to 82.1 (SD: 9.6) years for patients with functional limitation. [1%]) by the part-worth utility of the additional cost per month; and expressed in absolute value. The additional questions further confirmed that patient age has an impact on health professionals' monetary utility of CLL treatment, in line with the results obtained from the DCE. Hence, professionals younger counterparts [40,41]. A second explanation might be that, even considered that monetary utility of the treatment for the SNS in considering that the value of the drug was the same regardless of age, young patients is greater than in older patients. That would suggest health professionals felt compelled by a utilitarian ethic to maximize so- health professionals might consider acceptable a higher price paid cial benefit by allocating the most expensive resources to younger pa- by the SNS for younger patients than for older patients. Analysis re- tients [42]. Of note, the greater RI assigned to chronological age than garding ad hoc questions looking at the WTP is limited by the refer- to performance status is a fundamental issue in geriatric oncology. A pri- ence cost used and the frequency of the interval of prices give to mary determination when considering the appropriate therapy for an health professionals. However, the reference cost (€20,000) was older patient with cancer is patient's physiologic age (which assesses, chosen based on real annual cost of CLL treatment in Spain and the among others, comorbidity and functional status) rather than chrono- intervals of €10,000 provide an ample range of values to health logic age [43]. professionals. This study emphasizes the importance of taking into account patient-centered aspects in decision-making about CLL treatment, 5. Conclusion including both patient/disease and treatment attributes in the therapeutic scenario of CLL. It is widely described in the literature The study shows that for the evaluated Spanish healthcare pro- that the assessment of patient-related factors, such as age, comor- fessionals, when making treatment decisions for CLL patients at bidities and functional status, is of crucial importance for an opti- first relapse, treatment cost is the most important attribute, mal treatment choice in CLL [3,7,8,44]. At the same time, a few followed by PFS and age. Age is a crucial factor, WTP for the treat- DCE studies have analyzed the role of drug cost in determining ment being lower in older patients. It remains to be seen if these treatment selection preferences in oncology [45–47]. However, to results, and results from similar studies, will impact in future our knowledge, this is the first DCE study including both patient- guidelines regarding therapeutic decision-making in CLL, and related characteristics and drug cost as attributes for CLL treatment whether those guidelines will take into account physiological age selection. Therefore, the importance given to the combination (instead of chronological age), cost/effectiveness and treatment of these factors in health professional preferences had not been allocation. Extensive further research is warranted to ascertain if previously quantified. the findings are confirmed in other oncological diseases of the An important limitation of a DCE study is that other attributes not elderly. That would provide a wider vision on whether health pro- included may also be relevant and different levels may provide other fessionals' preferences point to cost before age and that affects part-worth utilitity values and relative importance. However, we their decision-making process. Audits of drug utilization should tried to prevent this potential bias by selecting the most important address whether older patients are being treated according to attributes from the literature, using the input of an expert panel. guidelines to the same extent as younger ones. Finally, where eco- Moreover, it is not known whether these unconscious preferences nomic evaluation is used for pricing and financing, if studies translate to prescriptions [48] or into different patterns of choices confirmed different values of treatment by physiological age, in real practice. Therefore, the study participants might make other more research would be needed on equity and financial aspects choices in real life. It has to be also noted that no information explor- of establishing differential pricing and reimbursement schemes ing the ethical issues about balancing the dual role of participants as reflecting those differences. patient advocates and gatekeepers of system resources was col- lected. Sample size requirements in DCE are also a controversial Funding issue. No general recommendations exist, so sample sizes differ substantially between studies and the method used is not usually re- The study was funded by Gilead Sciences. ported [49]. This study followed the well-established recommenda- tions proposed by Orme [22]. According to the survey method Ethical Statement employed, the number of professionals invited was not known and therefore response rate cannot be estimated. Finally, the findings of The study was approved by the Ethics Committee (CEIC) of Bellvitge the study significantly reflect preferences of healthcare professionals University Hospital. in Spain and should be interpreted within their context. In any case, data provide new insight into treatment decision-making in CLL. Conflicts of Interest Ad-hoc questions allowed us to explore in more detail the charac- teristics of CLL treatment considered most relevant by experts, in- Gabás-Rivera C and Lizán L work for an independent research entity and cluding age, functional status and WTP. The vast majority of have received fees for their contribution to the development of the pro- participants (90.7%) considered that age is important when making ject and the writing of the manuscript. Boqué C, Abad MR, Agustín MJ, decisions about CLL treatment. For 93%, age modifies the recom- García-Goñi M and Moreno C have participated in advisory meetings mended therapeutic regimen; while 77.7% find older age impedes for the study. Granados E and Pardo C work in Gilead Sciences. Castro- access to active treatment. Functional status is a limiting factor for Gómez A has worked in Gilead sciences. All authors have approved accessing treatment approaches in CLL, restricting the provision of the final article. 30 C. Boqué et al. / Journal of Geriatric Oncology 11 (2020) 24–30 Author Contributions [22] Orme B. Getting started with conjoint analysis: strategies for product design and pricing research. . 2nd ed.Madison: Wis. Research Publishers LLC; 2010. [23] Hincapie AL, Penm J, Burns CF. Factors associated with patient preferences for dis- Study concept and design: Boqué C, Abad MR, Agustín MJ, García-Goñi ease-modifying therapies in multiple sclerosis. J Manag Care Spec Pharm 2017;23 M, Moreno C, Gabás-Rivera C, Granados E, Castro-Gómez A, Pardo C, (8):822–30. [24] Johnson FR, Lancsar E, Marshall D, Kilambi V, Mühlbacher A, Regier DA, et al. Con- Lizán L. structing experimental designs for discrete-choice experiments: report of the Data acquisition: Boqué C, Abad MR, Agustín MJ, García-Goñi M, Moreno ISPOR conjoint analysis experimental design good research practices task force. C, Granados E, Castro-Gómez A. Value Health 2013;16(1):3–13. [25] Erdem S, Thompson C. Prioritising health service innovation investments using Quality control of data and algorithms, data analysis and interpretation, public preferences: a discrete choice experiment. BMC Health Serv Res 2014;14 statistical analysis and Manuscript preparation: Gabás-Rivera C, Lizán L. (1):360. Manuscript editing: Gabás-Rivera C, Granados E, Castro-Gómez A, Pardo [26] Wijnen BFM, Van Der Putten IM, Groothuis S, De RJA, Noben CYG, Paulus ATG, et al. Discrete-choice experiments versus rating scale exercises to evaluate the C, Lizán L. importance of attributes. Expert Rev Pharmacoecon Outcomes Res 2015;15(4): Manuscript review and final approval: Boqué C, Abad MR, Agustín MJ, 721–8. García-Goñi M, Moreno C, Gabás-Rivera C, Granados E, Castro-Gómez [27] Bridges JFP, Hauber AB, Marshall D, Lloyd A, Prosser LA, Regier DA, et al. Conjoint A, Pardo C, Lizán L. analysis applications in health—a checklist: A report of the ISPOR good research practices for conjoint analysis task force. Value Health 2011;14(4):403–13. [28] García-Marco JA, Delgado J, Hernández-Rivas JA, Ramírez Payer Á, Loscertales Pueyo Appendix A. Supplementary Data J, Jarque I, et al. Update of the Grupo Español de Leucemia Linfocítica Crónica clinical guidelines of the management of chronic lymphocytic leukemia. Med Clin (Barc) 2017;148(8) (381.e1–381.e9). Supplementary data to this article can be found online at https://doi. [29] StataCorp. Stata statistical software: release 14. College Station, TX: StataCorp LP; org/10.1016/j.jgo.2019.03.010. 2015. [30] Hauber AB, González JM, Groothuis-oudshoorn CGM, Prior T, Marshall DA, Cunningham C, et al. ISPOR task force report statistical methods for the analysis of References discrete choice experiments: a report of the ISPOR conjoint analysis good research practices task force. Value Health 2016;19(4):300–15. [1] Eichhorst B, Robak T, Montserrat E, Ghia P, Hillmen P, Hallek M, et al. Chronic lym- [31] Lewis F, Butler A, Gilbert L. A unified approach to model selection using the likeli- phocytic leukaemia: ESMO clinical practice guidelines for diagnosis, treatment and hood ratio test 2011;155–62. follow-up. Ann Oncol 2015;26(Suppl. 5):v78–84. [32] Landfeldt E, Eriksson J, Ireland S, Musingarimi P, Jackson C, Tweats E, et al. Patient, [2] Rowswell-Turner RB, Barr PM. Treatment of chronic lymphocytic leukemia in older physician, and general population preferences for treatment characteristics in re- adults. J Geriatr Oncol 2017;8(5):315–9. lapsed or refractory chronic lymphocytic leukemia: a conjoint analysis. Leuk Res [3] Mauro FR, Salaroli A, Caputo MD, Colafigli G, Petrucci L, Campanelli M, et al. Manage- 2016;40:17–23. ment of elderly and unfit patients with chronic lymphocytic leukemia. Expert Rev [33] Ratcliffe J. The use of conjoint analysis to elicit willingness-to-pay values. Int J Hematol 2016;9(12):1165–75. Technol Assess Health Care 2000;16(1):270–5. [4] Nastoupil LJ, Flowers CR. Management of relapsed chronic lymphocytic leukemia: [34] Mandeville KL, Lagarde M, Hanson K. The use of discrete choice experiments to in- applying guidelines to practice. Community Oncol 2012;9(12):S85–92. form health workforce policy: a systematic review. BMC Health Serv Res 2014:1–14. [5] Pinilla-Ibarz J, Emole J. Chronic lymphocytic leukemia in the elderly, which investi- [35] Shanafelt T. Treatment of older patients with chronic lymphocytic leukemia: key gations are necessary: a map for the practicing oncologist. Cancer Control 2015;22 questions and current answers. Hematology 2013;2013:158–67. (4):7–16. [36] Allan GM, Lexchin J. Physician awareness of diagnostic and nondrug therapeutic [6] Barrientos C. Management of chronic lymphocytic leukemia in the elderly. Cancer costs: a systematic review. Int J Technol Assess Health Care 2008;24:158–65. Control 2015;22(4):17. [37] Bellian DP, King KA, Wahl J, Price JH. Psychiatrists' knowledge and attitudes about [7] Balducci L, Dolan D. Chronic lymphocytic leukemia in the elderly: epidemiology and costs of commonly prescribed treatments in psychiatry. J Community Health 2001; proposed patient-related approach. Cancer Control 2015;22(4):3–6. 26:11–22. [8] Balducci L. Systemic treatment of gastric and esophageal adenocarcinoma in elderly [38] Tsugawa Y, Jha A, Newhouse J, Zaslavsky A, Jena A. Variation in physician spending patients. J Gastrointest Oncol 2015;6(1):75–8. and association with patient outcomes. JAMA Intern Med 2017;177(5):675–82. [9] El-Jurdi NH, Saif MW. Should ageism be a stratification factor in patients with pan- [39] Bovier PA, Martin DP, Perneger TV. Cost-consciousness among Swiss doctors: a creatic cancer? JOP 2014;15(2):147–50. cross-sectional survey. BMC Health Serv Res 2005;5(1):72. [10] WHO. Ageism. [Internet]. Available from: http://www.who.int/ageing/features/faq- [40] Cubi-Molla P, Shah K, Garside J, Devlin N, Henderman M, Devlin N. Age and utilities: ageism/en/ issues for HTA. London: Office of Health Economics; 2017 Research Paper 17/03. [11] Shumway DA, Hamstra DA. Ageism in the undertreatment of high-risk prostate can- (April). cer: how long will clinical practice patterns resist the weight of evidence? J Clin [41] Tsuchiya A. Age-related preferences and age weighting health benefits. Soc Sci Med Oncol 2015;33(7):676–8. 1999;48(2):267–76. [12] Schrijvers D, European Society of Medical Oncology, editors. ESMO Handbook of [42] Mandal J, Ponnambath D, Parija S. Utilitarian and deontological ethics in medicine. cancer in the senior patient. Informa Healthcare; 2010. Tropenmed Parasitol 2016;6(1):5–7. [13] Appelbaum FR. Preparative regimens and ageism. Biol Blood Marrow Transplant [43] Korc-grodzicki B, Holmes HM, Shahrokni A. Geriatric assessment for oncologists. 2011;17(10):1419–20. 2015;12(4):261–74. [14] Kowdley GC, Merchant N, Richardson JP, Somerville J, Gorospe M, Cunningham SC. [44] Extermann M, Wedding U. Comorbidity and geriatric assessment for older patients Cancer surgery in the elderly. Sci World J 2012;2012:1–9. with hematologic malignancies: a review of the evidence. J Geriatr Oncol 2012;3 [15] Chen Q, Jain N, Ayer T, Wierda WG, Flowers CR, O'Brien SM, et al. Economic burden (1):49–57. of chronic lymphocytic leukemia in the era of oral targeted therapies in the United [45] Arellano J, Hauber AB, Mohamed AF, Gonzalez JM, Collins H, Hechmati G, et al. Phy- States. J Clin Oncol 2017;35(2):166–74. sicians' preferences for bone metastases drug therapy in the United States. Value [16] González López-Valcárcel B, Librero J, Sanfélix-Gimeno G, Peiró S. Are prescribing Health 2015;18(1):78–83. doctors sensitive to the price that their patients have to pay in the Spanish national [46] Regier DA, Diorio C, Ethier M-C, Alli A, Alexander S, Boydell KM, et al. Discrete choice health system? BMC Health Serv Res 2011;11(1):333. experiment to evaluate factors that influence preferences for antibiotic prophylaxis [17] Puigventós F, Santos-Ramos B, Ortega A, Durán-García E. Structure and procedures in pediatric oncology. Moormann AM, editor. PLoS One 2012;7(10). https://doi.org/ of the pharmacy and therapeutic committees in Spanish hospitals. Pharm World 10.1371/journal.pone.0047470 Internet. e47470. Available from:. Sci 2010;32(6):767–75. [47] Shafey M, Lupichuk SM, Do T, Owen C, Stewart DA. Preferences of patients and phy- [18] Alastrué Loscos JI, Meneu de Guillerna R, Peiró Moreno S. Analysis of opinions, atti- sicians concerning treatment options for relapsed follicular lymphoma: a discrete tudes and knowledge of physicians of health centers of Valencia concerning efficacy choice experiment. Bone Marrow Transplant 2011;46(7):962–9. and efficiency of drug prescriptions [in Spanish]. Aten Primaria 1998;21(3):165–71. [48] Krucien N, Gafni A, Pelletier-Fleury N. Empirical testing of the external validity of a [19] Allan G, Lexchin J, Wiebe N. Physician awareness of drug cost: a systematic review. discrete choice experiment to determine preferred treatment option: the case of PLoS Med 2007;4(9):e283. sleep apnea. Health Econ 2015;24(8):951–65. [20] Kleij K-S, Tangermann U, Amelung VE, Krauth C. Patients' preferences for primary [49] de Bekker Grob EW, Ryan M, Gerard K. Discrete choice experiments in health eco- health care – a systematic literature review of discrete choice experiments. BMC nomics: a review of the literature. Health Econ 2012;21:145–72. Health Serv Res 2017;17(1):476. [21] Brett Hauber A, Fairchild AO, Reed Johnson F. Quantifying benefit-risk preferences for medical interventions: an overview of a growing empirical literature. Appl Health Econ Health Policy 2013;11(4):319–29.