Framework for personalized prediction of treatment response in relapsing remitting multiple sclerosis

Abstract

Background: Personalized healthcare promises to successfully advance the treatment of heterogeneous neurological disorders such as relapsing remitting multiple sclerosis by addressing the caveats of traditional healthcare. This study presents a framework for personalized prediction of treatment response based on real-world data from the NeuroTransData network. Methods: A framework for personalized prediction of response to various treatments currently available for relapsing remitting multiple sclerosis patients was proposed. Two indicators of therapy effectiveness were used: number of relapses, and confirmed disability progression. The following steps were performed: (1) Data preprocessing and selection of predictors according to quality and inclusion criteria; (2) Implementation of hierarchical Bayesian generalized linear models for estimating treatment response; (3) Validation of the resulting predictive models based on several performance measures and routines, together with additional analyses that focus on evaluating the usability in clinical practice, such as comparing predicted treatment response with the empirically observed course of multiple sclerosis for different adherence profiles. Results: The results revealed that the predictive models provide robust and accurate predictions and generalize to new patients and clinical sites. Three different out-of-sample validation schemes (10-fold cross-validation, leave-one-site-out cross-validation, and excluding a test set) were employed to assess generalizability based on three different statistical performance measures (mean squared error, Harrell's concordance statistic, and negative log-likelihood). Sensitivity to different choices of the priors, to the characteristics of the underlying patient population, and to the sample size, was assessed. Finally, it was shown that model predictions are clinically meaningful. Conclusions: Applying personalized predictive models in relapsing remitting multiple sclerosis patients is still new territory that is rapidly evolving and has many challenges. The proposed framework addresses the following challenges: robustness and accuracy of the predictions, generalizability to new patients and clinical sites and comparability of the predicted effectiveness of different therapies. The methodological and clinical soundness of the results builds the basis for a future support of patients and doctors when the current treatment is not generating the desired effect and they are considering a therapy switch.

Digital Twins for Multiple Sclerosis

“…For example, Stühler et al and Kalincik et al have investigated the individual response of pwMS to disease-modifying therapies using generalized linear models. However, in both studies, data density and quality were insufficient because, among other reasons, the cohorts were too small or there were data gaps in MRI data or data could not be comprehensively included (189)(190)(191). With the DTMS, all historically and currently available data should be continuously included in the analysis, if possible, to increase predictive power.…”

The Development and Validation of a Disability and Outcome Prediction Algorithm in Multiple Sclerosis Patients

“…However, the statistical measures used for validation (mean squared error, negative log-likelihood and C-Index) were different than the ones used in the present study and hence, a relevant comparison cannot be performed from this point of view, even though both studies were based on a probabilistic approach in terms of prediction. The authors therefore conclude that the main advantage of the present work over the one conducted by Stühler was the EDSS threshold probability prediction, which might lead to a more accurate quantification of MS patients evolution as opposed to the sole estimation of disability progression [17]. Therefore, despite the limitations and the disadvantages which were highlighted in detail, the developed algorithms present a consistent degree of clinical relevance, being able of retrieving patient-tailored disability and outcome profiles.…”

“…Therefore, it should be mentioned that that the final model which is used in the computations is a statistical model embedded in a disease severity scale, and not a machine learning model [20]. By contrast, a study conducted by Stühler aimed at developing a proper machine learning algorithm (based on Bayesian generalized linear models) which estimated the treatment response in MS patients [17]. The validation results were considered clinically relevant for both the relapse and the confirmed disability progression prediction.…”

“…Regarding multiple sclerosis diagnosis, monitoring and management, such artificial intelligence techniques are mainly based on multiple sclerosis disease course classification based on central nervous system lesions [7,21]. In addition, other studies used machine learning or statistical approaches for predicting the treatment response or the EDSS (Expended Disability Status Scale) score evolution [17,22], while others computed a more sensitive and complex disability scale, namely the Combinatorial Weight-Adjusted Disability Score, in order to perform relevant simulations of MS disease progression [20]. Therefore, considering the high prevalence and severity of this chronic disease, as well as the fact that until now only few machine learning methods for MS patients monitoring were developed, the aim of this study was to develop a disability and outcome prediction algorithm in MS patients.…”

Impact of drug diversity on treatment effectiveness in relapsing-remitting multiple sclerosis (RRMS) in Germany between 2010 and 2018: real-world data from the German NeuroTransData multiple sclerosis registry

“…Promising techniques emerge based on biomarker like neurofilament light chain 20 or B-cell activity response 21 or RWD-based statistical predictive algorithms. 22 As treatment decisions are driven currently by European label definitions, national cost control regulations and perceptions of physicians and patients, personaliseddata-based decision support is required to further improve individual care.…”