Contributions
Abstract: 195
Type: Oral
Abstract Category: Clinical aspects of MS - 8 Clinical assessment tools
Most of people with Multiple Sclerosis (PwMS) in the relapsing remitting (RR) phase of the disease develops a secondary progressive (SP) course within 15-20 years if untreated, or if the adopted pharmacological and rehabilitative protocols are not continuously adjusted according to the evolution of the disease. Therefore, the prediction of the transition from RR to SP is one of the most important methodological gaps to be addressed. The availability of a statistical model able to predict disease worsening is one of the major unmet needs that could significantly improve timeliness, personalization and, consequently, the efficacy of the treatments.
Here, we propose a machine learning (ML) approach that, leveraging on Patient Reported (PRO) and Clinician Assessed Outcomes (ClinAO), aims at forecasting the transition of PwMS from RR to SP. The data have been collected within the Italian MS Foundation (FISM) initiative PROMPRO-MS dataset. The set of PRO and ClinAO consisted of 165 items from: EDSS; FIM; Edinburgh Handedness Inventory; Abilhand; MoCA; PASAT 3; SDMT; HADS; LSI; OAB-q; MFIS. Up to 8 time points (data acquisition every 4 months) for a total of 2699 samples from about 900 PwMS followed by Italian MS Society Rehabilitation Centre (Genoa, Padua, Vicenza) were the base for ML analysis.
The ML algorithms developed for our aims consisted of three consecutive steps: the first, testing the capacity to distinguish, only using PRO and ClinAO, RR from SP PwMS; the second, predicting the data pattern at the next time point; finally, based on the previous two steps, predicting the disease course at the next time point.
Main results showed that the adopted models achieved the following performance scores in predicting the disease course based on about all the PRO and ClinAO items: accuracy 0.841, precision 0.900.
To the best of our knowledge, the proposed model, which is soon going to be validated in clinical practice, is the first attempt to solve this delicate task leveraging on patient-friendly measures and machine learning. This will help clinicians to take important decisions concerning treatment and therapies that can substantially improve the quality of life of their patients.
Disclosure:
Andrea Tacchino: nothing to disclose
Samuele Fiorini: nothing to disclose
Michela Ponzio: nothing to disclose
Annalisa Barla: nothing to disclose
Alessandro Verri: nothing to disclose
Mario Alberto Battaglia: nothing to disclose
Giampaolo Brichetto: nothing to disclose
Abstract: 195
Type: Oral
Abstract Category: Clinical aspects of MS - 8 Clinical assessment tools
Most of people with Multiple Sclerosis (PwMS) in the relapsing remitting (RR) phase of the disease develops a secondary progressive (SP) course within 15-20 years if untreated, or if the adopted pharmacological and rehabilitative protocols are not continuously adjusted according to the evolution of the disease. Therefore, the prediction of the transition from RR to SP is one of the most important methodological gaps to be addressed. The availability of a statistical model able to predict disease worsening is one of the major unmet needs that could significantly improve timeliness, personalization and, consequently, the efficacy of the treatments.
Here, we propose a machine learning (ML) approach that, leveraging on Patient Reported (PRO) and Clinician Assessed Outcomes (ClinAO), aims at forecasting the transition of PwMS from RR to SP. The data have been collected within the Italian MS Foundation (FISM) initiative PROMPRO-MS dataset. The set of PRO and ClinAO consisted of 165 items from: EDSS; FIM; Edinburgh Handedness Inventory; Abilhand; MoCA; PASAT 3; SDMT; HADS; LSI; OAB-q; MFIS. Up to 8 time points (data acquisition every 4 months) for a total of 2699 samples from about 900 PwMS followed by Italian MS Society Rehabilitation Centre (Genoa, Padua, Vicenza) were the base for ML analysis.
The ML algorithms developed for our aims consisted of three consecutive steps: the first, testing the capacity to distinguish, only using PRO and ClinAO, RR from SP PwMS; the second, predicting the data pattern at the next time point; finally, based on the previous two steps, predicting the disease course at the next time point.
Main results showed that the adopted models achieved the following performance scores in predicting the disease course based on about all the PRO and ClinAO items: accuracy 0.841, precision 0.900.
To the best of our knowledge, the proposed model, which is soon going to be validated in clinical practice, is the first attempt to solve this delicate task leveraging on patient-friendly measures and machine learning. This will help clinicians to take important decisions concerning treatment and therapies that can substantially improve the quality of life of their patients.
Disclosure:
Andrea Tacchino: nothing to disclose
Samuele Fiorini: nothing to disclose
Michela Ponzio: nothing to disclose
Annalisa Barla: nothing to disclose
Alessandro Verri: nothing to disclose
Mario Alberto Battaglia: nothing to disclose
Giampaolo Brichetto: nothing to disclose