This was an observational, descriptive, exploratory and cross-sectional study based on a DCE. The study was performed within the Spanish healthcare public system from September 2017 to February 2018. The DCE was conducted in accordance with International Society for Pharmacoeconomics and Outcomes Research (ISPOR) good practice recommendations for conjoint analysis in healthcare .
A steering committee constituted by three Spanish experts in RA (CDT, JIC, AUA) led the project.
The study population included RA patients and rheumatologists with experience in the management of RA. Study participants were invited to participate by emails sent by the Spanish Patient Advocacy Group Coordinadora Nacional de Artritis (ConArtritis) (RA patients) and the Spanish Society for Rheumatology (SER) (rheumatologists).
Patients 18 years or older, on DMARD treatment for at least 12 months and who gave their consent to participate were included in the study. The participating rheumatologists had to have at least 3 years of experience in RA management and work in the Spanish Health System.
As per the approach proposed by Orme , the minimum sample size necessary for the DCE was based on an estimate of proportion. The criterion of maximum variability was applied, with a 95% confidence interval and 10% margin of error. Patient sample size was estimated on the basis of the adult population in Spain in 2016 (37,408,739)  and RA prevalence (0.5%) . The sample size for the rheumatologists was determined using the estimated number of rheumatologists practicing in the public Spanish Health System (629) . A minimum sample of 96 RA patients and 83 rheumatologists was required.
Discrete Choice Experiment
Selection of Attributes and Levels
In DCEs, patients choose between two hypothetical treatment alternatives described by attributes (characteristics) and their corresponding levels (different possible values of the attributes) . To select the attributes and levels for the DCE, three consecutive steps were conducted: (1) literature review; (2) RA patient focus group discussion; (3) rheumatologist focus group discussion.
Key terms related to the disease, treatment and stated-preferences studies were used to search the international Pubmed/Medline database. Publications referring to patient and physician preferences in relation to RA treatment as well as those that referred to their perspectives on the management of the disease were consulted. Articles published in Spanish or English up to 9 March 2016 were reviewed.
The results of the literature review  were used to provide inputs for discussion in both focus groups.
Focus Groups (Patients and Rheumatologists)
Following the literature review, two focus groups, one with RA patients and one with rheumatologists, were used to validate and assess the relevance of the attributes and levels identified in the literature as well as to identify attributes not previously described but that were relevant to the Spanish RA population.
A total of five RA patients, invited by the patient advocacy group “ConArtritis,” participated in the patient focus group. After the completion of a brief questionnaire on sociodemographic and clinical characteristics, the list of attributes and levels derived from the literature review was presented to the patients to discuss them. During the discussion, patients were encouraged to add new attributes and levels not previously identified in the literature review, but relevant for them. When all attributes and levels were identified, a ranking exercise was then performed to determine the relevance of the attributes and levels proposed. The interpretation of the qualitative analysis and the analysis of the ranking exercises allowed identifying the most important attributes.
After the patient focus group, one with four experienced rheumatologists (including the members of steering committee) was conducted. The objectives of this focus group were to discuss the relevance of the attributes identified in the literature review and proposed by patients from the focus group and define the attributes and levels to be included in the DCE.
As a result of the literature review and the two focus groups, seven attributes composed of two or four levels each were selected (Table 1).
The combinations of attributes and levels that defined each treatment pair were determined by an experimental design developed according to ISPOR recommendations. The DCE design encompassed two properties: orthogonality and balance . The orthogonal design guarantees that all attribute levels vary independently, and the balance design ensures that each attribute level occurs the same number of times. The pairs of choice (Fig. 1) were generated by the mix and match algorithm . To avoid dominance between alternatives, the resulting scenarios were evaluated for dominated alternatives.
A total of eight scenarios were created, which formed a single block. Additionally, an initial control scenario, in which one treatment was clearly superior to the other (dominant option), was included. Participants who answered this question incorrectly were excluded from analysis as this indicates that they did not comprehend what is required from them in this study .
Two online surveys were generated, one for patients and one for rheumatologists. Both contained the same DCE choice scenarios and included an information form and an electronic informed consent form that had to be read and accepted before completing the questionnaire. In addition, both questionnaires initially included a series of questions to verify that participants met the selection criteria.
The rheumatologist questionnaire included a set of sociodemographic and professional variables to characterize them. The patient questionnaire included sociodemographic and clinical variables, and a Health Assessment questionnaire (HAQ) to assess the patient’s functional status . A set of ad hoc questions was also included to collect the patient’s perception of: (1) the current degree of involvement in treatment decision-making and their expectations about their involvement ; (2) the satisfaction with the information received about the disease, current treatment and therapeutic alternatives (Likert scale: 1 = not at all satisfied; 5 = very satisfied).
Stata version 14 and R version 3.4.1 were used for the statistical analysis. A value of p < 0.05 was considered significant for all statistical tests.
For the descriptive analysis of the qualitative variables, the relative and absolute frequencies were calculated, and for the quantitative variables central tendency and dispersion measures were used for each group of participants.
To assess the utility and the relative importance (RI) value given to the attributes of RA treatments by patients and rheumatologists, a conditional logit model  was used. Respondents who did not select the dominant option in the control scenario were excluded. Substantial improvement of RA symptoms, time with optimal QoL, severe and mild adverse events and additional cost per month attributes were linearly transformed. Coefficients obtained in the conditional logit model represented the partial utilities, i.e., the preference for each level within each attribute. A statistically significant coefficient indicates that the attribute level influences the respondents’ treatment decisions. The RI of each attribute, defined as the relative preference weight for the attribute over all attributes, was calculated as the quotient between the range of the partial utility values of the attribute and the sum of the partial utility values ranges of the whole set of attributes. The greater the RI among the seven attributes, the more significant the attribute was for decision-making.
To evaluate which characteristics of the participants influenced decision-making, a hierarchical cluster analysis was applied to each group of participants based on DCE response, i.e., the scenario selected in each pairs of choice . Since scenario choices were dichotomous, a binary distance was applied. Stats and Nbclust packages were used to determine the optimal number of clusters [33, 34]. To assess differential characteristics between the clusters obtained, sociodemographic and clinical characteristics of each cluster were compared.
Maximum acceptable risk (MAR) was estimated as the quotient between the utility associated with a clinical benefit attribute (substantial improvement of RA symptoms) and the utility associated with risk (severe adverse events) .
To establish differences between the patient’s current role in the decision-making process and their expectations about their involvement in it, the answers given in the ad hoc questionnaire  were compared using McNemar Bowker’s test .
Statement of Ethics Compliance
This study was conducted in accordance with the principles of the Declaration of Helsinki. It was developed to ensure that Good Clinical Practices were observed, in keeping with ICH Harmonized Tripartite Guideline principles. The study protocol was submitted to the Spanish Agency of Medicines and Medical Devices and to the Clinical Research Ethics Committee of Puerta de Hierro-Majadahonda Hospital for approval. An electronic informed consent form was read and accepted by all participants before completing the questionnaire.