Background Biological disease-modifying anti-rheumatic drugs (bDMARDs) can be tapered in some rheumatoid arthritis (RA) patients in sustained remission. The purpose of this study was to assess the feasibility of building a model to estimate the individual flare probability in RA patients tapering bDMARDs using machine learning methods. Methods Longitudinal clinical data of RA patients on bDMARDs from a randomized controlled trial of treatment withdrawal (RETRO) were used to build a predictive model to estimate the probability of a flare. Four basic machine learning models were trained, and their predictions were additionally combined to train an ensemble learning method, a stacking meta-classifier model to predict the individual flare probability within 14 weeks after each visit. Prediction performance was estimated using nested cross-validation as the area under the receiver operating curve (AUROC). Predictor importance was estimated using the permutation importance approach. Results Data of 135 visits from 41 patients were included. A model selection approach based on nested cross-validation was implemented to find the most suitable modeling formalism for the flare prediction task as well as the optimal model hyper-parameters. Moreover, an approach based on stacking different classifiers was successfully applied to create a powerful and flexible prediction model with the final measured AUROC of 0.81 (95%CI 0.73–0.89). The percent dose change of bDMARDs, clinical disease activity (DAS-28 ESR), disease duration, and inflammatory markers were the most important predictors of a flare. Conclusion Machine learning methods were deemed feasible to predict flares after tapering bDMARDs in RA patients in sustained remission.

IntroductionAmidst growing consensus that stakeholder decision-making during drug development should be informed by an understanding of patient preferences, the Innovative Medicines Initiative project ‘Patient Preferences in Benefit-Risk Assessments during the Drug Life Cycle’ (PREFER) is developing evidence-based recommendations about how and when patient preferences should be integrated into the drug life cycle. This protocol describes a PREFER clinical case study which compares two preference elicitation methodologies across several populations and provides information about benefit–risk trade-offs by those at risk of rheumatoid arthritis (RA) for preventive interventions.Methods and analysisThis mixed methods study will be conducted in three countries (UK, Germany, Romania) to assess preferences of (1) first-degree relatives (FDRs) of patients with RA and (2) members of the public. Focus groups using nominal group techniques (UK) and ranking surveys (Germany and Romania) will identify and rank key treatment attributes. Focus group transcripts will be analysed thematically using the framework method and average rank orders calculated. These results will inform the treatment attributes to be assessed in a survey including a discrete choice experiment (DCE) and a probabilistic threshold technique (PTT). The survey will also include measures of sociodemographic variables, health literacy, numeracy, illness perceptions and beliefs about medicines. The survey will be administered to (1) 400 FDRs of patients with RA (UK); (2) 100 FDRs of patients with RA (Germany); and (3) 1000 members of the public in each of UK, Germany and Romania. Logit-based approaches will be used to analyse the DCE and imputation and interval regression for the PTT.Ethics and disseminationThis study has been approved by the London-Hampstead Research Ethics Committee (19/LO/0407) and the Ethics Committee of the Friedrich-Alexander-Universität Erlangen-Nürnberg (92_17 B). The protocol has been approved by the PREFER expert review board. The results will be disseminated widely and will inform the PREFER recommendations.

No abstract

Objective To quantify preferences for preventive therapies for rheumatoid arthritis (RA) across three countries. Methods A web-based survey including a discrete choice experiment was administered to adults recruited via survey panels in the UK, Germany and Romania. Participants were asked to assume they were experiencing arthralgia and had a 60% chance of developing RA in the next 2 years and completed 15 choices between no treatment and two hypothetical preventive treatments. Treatments were defined by six attributes (effectiveness, risks and frequency/route of administration) with varying levels. Participants also completed a choice task with fixed profiles reflecting subjective estimates of candidate preventive treatments. Latent class models (LCMs) were conducted and the relative importance of attributes, benefit–risk trade-offs and predicted treatment uptake was subsequently calculated. Results Completed surveys from 2959 participants were included in the analysis. Most participants preferred treatment over no treatment and valued treatment effectiveness to reduce risk more than other attributes. A five-class LCM best fitted the data. Country, perceived risk of RA, health literacy and numeracy predicted class membership probability. Overall, the maximum acceptable risk for a 40% reduction in the chance of getting RA (60% to 20%) was 21.7%, 19.1% and 2.2% for mild side effects, serious infection and serious side effects, respectively. Predicted uptake of profiles reflecting candidate prevention therapies differed across classes. Conclusion Effective preventive pharmacological treatments for RA were acceptable to most participants. The relative importance of treatment attributes and likely uptake of fixed treatment profiles were predicted by participant characteristics.

No abstract