Disability weights for castration-resistant prostate cancer: an empirical investigation

Structured qualitative interviews

We carried out structured, individual qualitative interviews with a purposive sample of Italian medical oncologists and urologists, selected to maximize representativeness of clinical practice across different Italian regions and on the basis of their recognized expertise in the management of nmCRPC (e.g., authors of peer-reviewed publication on nmCRPC). Although purposive sampling can be subject to selection bias (19), it is a technique widely used in qualitative research for the identification and selection of information-rich cases, that is, individuals who are especially knowledgeable about or experienced with the central phenomenon in the study (20).

The objectives of the interviews were to characterize different health states within the spectrum of CRPC disease (from nmCRPC to mCRPC), and to gather qualitative information on the diagnostic and treatment pathways of nmCRPC patients. Experts were invited via email to partake in the project, and an informed consent form was sent to those who agreed to participate. The interview questions (Supplementary file S1) were shared in advance. We carried out a pilot phase with two clinicians in order to validate the interview guide. Interviews were conducted in Italian either by phone or face-to-face, recorded and transcribed. Transcripts were analyzed qualitatively and synthesized using standard thematic analysis (21), and results were reported in an anonymized format. Quotes from the experts interviewed were included to illustrate key points, and a numerical code reported in parentheses was used to refer to different responders (i.e., experts were labelled from I to X). Descriptors were used in the text to inform readers on the degree of agreement of clinicians interviewed on a certain topic (see Supplementary Table A for interpretation of these descriptors).

Survey on DWs

An online survey addressed to urologists and medical oncologists was carried out in order to estimate DWs for CRPC.

The information collected during the interviews and the evidence from the literature (22) contributed to the drafting of the health states of CRPC. Four health states and their respective labels were identified, each assumed to be internally homogeneous in terms of disease severity, disability, treatment and prognosis. In line with several studies on DWs elicitation (23-25), also a description of the health states through EQ-5D-3L descriptive system was provided on five different dimensions: mobility, self-care, usual activities, pain/discomfort, anxiety/depression (26). Each dimension presents three levels of severity: no problems, moderate problems and severe problems. The EQ-5D descriptions for the health states were drafted using the results of qualitative interviews, where clinicians were asked to describe how a nmCRPC and mCRPC patient feels, and the published literature on mCRPC (27). For each health state, a standardized vignette was prepared, reporting the disease label, a brief clinical description and the generic description through EQ-5D dimensions (Figure 1). The vignettes were validated by one clinician and two health economists. We decided to use the 3-level version of the EQ-5D as, during the validation step of the vignettes, it resulted easier to implement than the 5-level version and was judged as being more easily comprehensible.

An online questionnaire was designed through Qualtrics XM and administered to clinicians, whose preferences on the health states were elicited using two valuation methods, the Person-Trade-Off (PTO) and the Visual Analogue Scale (VAS). With the PTO method, experts were asked to trade-off a number of individuals with perfect health and a number of individuals in the health state to be assessed (Supplementary File S2) (23,28). Due to the complexity of PTO exercise, respondents were given the possibility to check their answers and revise them if necessary. The PTO method attempts to measure social preference instead of individual preferences, as interviewees are asked to state preferences between people (28). Therefore, the estimates derived from PTO should reflect the perspective of a policy-maker, and PTO may represent the preferred method for estimating DWs for burden of disease studies. With the VAS method, instead, experts were asked to give a direct ranking of the health states in two steps: (1) by ordering the health states from the best (I) to the worst (IV); (2) by assigning a score to each health state through a graduated scale, which assumes values from 0 (worst score = death or worst possible health status) to 100 (best score = perfect health) (Supplementary File S2) (23). The VAS exercise is cognitively simpler than the PTO (29) as it does not entail a trade-off feature, and it was used to double-check the reliability of DWs obtained through the PTO method (23). The VAS scores provide information about the relative desirability of health states rather than giving evidence on the trade-offs that people are willing to make (30). All health states were valued independently according to both methods. The DWs were derived according to the following formulas (23):

DW (PTO) = 1 − (1,000/N)

DW (VAS) = 1 − (VAS score/100)

where N is the number of subjects in a certain health state for which 1 year of life extension is equivalent to the extension of 1 year of life for 1,000 healthy individuals. Data analyses were carried out using Stata^® (v16; StataCorp LLC, College Station, TX). A complete case analysis was carried out, that is, only complete questionnaires were considered (31). Individual DWs for both PTO and VAS were averaged across responders to obtain the DWs for the whole sample. Means and corresponding confidence intervals were computed using the command ci means in Stata. To establish whether the disease health states were valued in a consistent manner by individuals between the two valuation methods, the DWs obtained from the PTO and VAS methods were compared using Kendall rank correlation test, which is intended for use on small- and moderate-sized datasets (32). Kendall tau-a and tau-b statistics can assume values from −1 to 1, where −1 means perfectly negative correlation and 1 means perfectly positive correlation. A p-value <0.05 was considered significant.

In the absence of a gold standard to evaluate the validity of DWs (6), we first compared the results obtained with the rankings of matching diseases drawn the GBD 2013 study (33), a web-based survey of participants aged 18-65 years in four European countries (Hungary, Italy, the Netherlands and Sweden). More specifically, we considered the DWs for the following cancer health states: (i) diagnosis and primary therapy; (ii) metastatic cancer; (iii) terminal phase with medication; (iv) terminal phase without medication. Then, we compared the DWs for objectively less severe and more severe health states.

A pilot of the survey was conducted by recruiting a sample of clinicians during two national scientific conferences of urologists in Italy. Clinicians were randomly identified and interviewed by one researcher at the congress venues. Following the pilot phase, a distribution list of urologists was created ad hoc by mapping all the urology units in Italy. A search on the hospitals’ website and on Google was performed to retrieve the email address of urologists operating in each unit. A snowballing procedure led to further enrich this list with Italian clinical experts in CRPC. Similarly to what discussed for the interviews, also in this case the purposive sampling approach adopted may have led to selection bias (19). However, due to the lack of a publicly available and comprehensive database on Italian clinicians dealing with nmCRPC patients, a purposive sampling represented the only doable and most cost-effective approach to recruit responders.

The survey was sent at the end of September 2020. Three reminders were sent through Qualtrics XM, one every 2 weeks, before closing the survey at the beginning of November 2020.

Health states description and HRQoL for CRPC patients

With respect to HRQoL, clinicians agreed that nmCRPC patients usually do not experience pain due to the disease (i.e., are asymptomatic) nor difficulties in performing daily activities. The main impact of the disease is on the emotional and psychological sphere, as «(…) it is not easy to live in a state in which the [Prostate-Specific Antigen] PSA continues to rise and you know that the disease will become metastatic sooner or later. The patient knows that he has a sword of Damocles above his head (…)» (III). mCRPC patients can be asymptomatic, symptomatic or heavily symptomatic. Generally, the presence and severity of pain is due to metastases and «(…) depends also from the site of metastases. Pain may become extremely relevant if the patient has diffused bone metastases and problems in lymph nodes» (I). For symptomatic or heavily symptomatic patients, the impact of the disease on the ability to perform daily activities and on the social sphere may be substantial, and can vary according to the patients’ age (e.g., older patients usually have a less aggressive disease and more leniently accept their conditions). The anxiety may be more pronounced when the patient receives the first diagnosis of metastatic disease, as he knows that his health state has changed, but usually «(..) the passage from first to second treatment line is not lived as a drama» (X). In some cases, the mCRPC patient is less worried and anxious than the nmCRPC one because he is more “prepared” («Sometimes the certainty, although negative, is better than the doubt» (VII)).

On the basis of the interviews and published literature (22), four health states for CRPC patients were identified: nmCRPC, asymptomatic mCRPC, symptomatic mCRPC, mCRPC in progression during or after chemotherapy. For these health states, DWs were estimated through a survey addressed to clinical experts (see the “Disability weights” section).

Diagnostic and treatment pathway of CRPC patients

nmCRPC patients usually face a long history of disease, and, according to the majority of clinicians interviewed, their clinical pathway is currently «not well defined» (I) and «needs to be ameliorated and codified» (X).

The diagnosis is made with a combination of laboratory tests (PSA) and imaging. Computed tomography (CT) scan and bone scan «are the gold standards for diagnosis» (VI) and, with very few exceptions, are always used for the diagnosis of nmCRPC, as indicated by clinical guidelines (34). Moreover, their use is compulsory to administer next-generation androgen receptors (e.g., apalutamide, enzalutamide, darolutamide), also for compassionate use. Despite the presence of a standard approach, however, nmCRPC patients «are diagnosed in a very heterogeneous manner» (IX) and «very often the choice of imaging is discretional» (X). Currently, prostate-specific membrane antigen (PSMA) positron emission tomography (PET) or choline PET is available, characterized by higher sensitivity and specificity. Sometimes diagnosis is performed through PET only, in contrast to clinical guidelines. According to some clinicians, the use of PET varies widely across different hospitals, «on the basis of the possibility of the center to perform these exams» (IV) and may depend on the preferences of clinician who is in charge of patient care. Despite the use of new-generation imaging, some responders recognize that this approach might not be appropriate in this setting «because in the phase III trials of the new molecules (…) the patient was diagnosed on the basis of a negative CT scan and bone scan» (X) and «in theory the staging in real life should follow the one performed in the trials, i.e., traditional imaging» (IX), to replicate similar results.

After diagnosis, nmCRPC patients treated with ADT «are always followed in outpatient setting» (III), «continue to do visits and exams with increasing frequency and continue to receive the therapy that was already ongoing» (II). The clinician in charge of nmCRPC patients’ care is mainly represented by the urologist, the oncologist or the radiotherapist. The prominence of one figure over the other may vary across different hospitals, as the management of nmCRPC is extremely heterogeneous in Italy, and may depend on the type of primary treatment, for example, urologist for patients who underwent surgical treatment or radiotherapist for those treated with radiotherapy. Despite the presence of a main responsible figure in the patient pathway, all clinicians interviewed declare that in their hospitals there is a multidisciplinary team for the management of PC patients, thus treatment decisions for nmCRPC patients are taken collectively and responsibility is shared. In the majority of hospitals, however, only some selected nmCRPC cases are discussed by the multidisciplinary team, usually identified by the main responsible clinician and limited to the most complex patients. The majority of responders report that there are no codified criteria that drive the choice of which cases should be discussed. The specialists usually involved in the multidisciplinary teams are urologists, medical oncologists, radiotherapists and pathologists. In some cases, also radiologists and nuclear medicine clinicians are involved to interpret the imaging results, as well as palliative care clinicians, endocrinologists, physiatrists and psycho-oncologists. Overall, clinicians believe that the discussion of cases in multidisciplinary teams should follow a generalized approach as «it ensures quality and gives a holistic vision of the patient. (…) This multidisciplinary approach is a guarantee of clinical pathway’s speed, from diagnosis to restaging to therapy choice» (I). The formal recognition of multidisciplinary teams through integrated care pathway, with either hospital or regional accreditation, is heterogeneous across Italian hospitals and in place in 50% of the organizations represented by interviewees.

At the time of the interviews, the reimbursement of next-generation androgen inhibitors (e.g., apalutamide, enzalutamide, darolutamide) in Italy was still under negotiation, the standard of care for nmCRPC patients was therefore ADT and active follow-up, although the majority of clinicians declared to administer the next-generation androgen inhibitors for compassionate use. Clinicians cited several factors, besides effectiveness, that may influence the choice of one treatment over the others, among which were the availability of alternative therapies, the patient’s profile, the impact of treatment on HRQoL, its cost, clinician’s preferences and also an explicit request coming from the patient. Clinicians agree that only one line (maximum two lines) of hormonal manipulation should be performed in this patient population, and that the use of successive manipulations, very common in the past, should be abandoned as they do not provide any added value for a nmCRPC patient (34). Clinicians expect that, with the diffusion of next-generation androgen inhibitors in clinical practice, «(…) the overall management of these patients will change both in terms of referral and in terms of more homogeneity in patients’ selection, follow-up and treatment» (X).

Disability weights

The online survey on DWs for CRPC health states was sent to 1,323 clinicians (of whom 1,227 were urologists). Among the experts invited, 54 started the online survey and 21 completed it (response rate = 1.6%; completion rate = 39%). Respondents reported an average experience in managing PC patients of approximately 16 years (standard deviation [SD] = 10.1), and the majority were urologists (66.7%) (Tab. I). They operated in 10 different Italian regions, and the majority in Northern Italy (66.7%).

According to the ranking of the health states provided in the VAS exercise, experts considered nmCRPC as the best state (with one exception), while responses were more heterogeneous as regards symptomatic and progressive metastatic CRPC (Tab. II).

The estimates (Tab. III, Fig. 2) clearly show that the DWs for metastatic disease states are higher than the non-metastatic one, and increase with disease progression. The absolute values of the weights for PTO ranged from 0.415 (95% confidence interval [CI], 0.208-0.623) in nmCRPC to 0.740 (95% CI, 0.560-0.920) in mCRPC, in progression during or after chemotherapy. As for the VAS method, weights ranged between 0.246 (95% CI 0.131-0.361) in nmCRPC and 0.689 (95% CI, 0.583-0.795) in mCRPC, in progression during or after chemotherapy. The strength and direction of association that exists between PTO and VAS values for each health state, as assessed by Kendall rank correlation test, was weak and not statistically significant (Tab. III). The lack of statistical significance might be partly explained by the limited sample size analyzed. However, at the aggregate level, the ranking of PTO and VAS is consistent (i.e., monotonically increasing), suggesting some internal coherence of data. Exploratory analyses indicate that in the PTO exercise higher experience corresponds to lower values assigned to the DW (negative correlation, although statistically significant only for the nmCRPC state).