Objective: Aim of this study was to develop a random forest (RF)-based PD model to determine the postoperative motor outcome at 2 years follow-up after the implementation of LCIG by using clinico-demographic, motor and non- motor factors.
Background: Despite careful patient selection for Levodopa carbidopa intestinal gel (LCIG), some Parkinson’s disease (PD) patients show poor motor improvement. Innovative machine learning methods hold potential to develop tool for clinicians that reliably determine postoperative motor response, by using clinical variables before and after starting LCIG therapy.
Method: We selected 59 PD subjects (36male, 23female; age at baseline: 69.45± 8.54) from the ForHealth S.A. database. Data included demographics, disease duration, motor and non-motor clinical measures in years 0 and 2. Motor status was performed with the Unified Parkinson’s Disease Rating Scale part III (UPDRS- III). Hours of “Off” and dyskinesias time were assessed with the UPDRS-IV. Non-motor symptoms were evaluated by Montreal Cognitive Assessment (MoCA), Geriatric Depression Scale (GDS), the King’s Parkinson’s Disease Pain Questionnaire (KPPQ), Parkinson’s Disease Sleep Scale (PDSS). The severity of PD (Hoehn and Yahr) and quality of life (PD questionnaire- 39) were also evaluated. RF analysis, with 10000 trees, was used to combine both non-motor and motor variables to determine motor outcome (UPDRS-III year 2: 23.57 ± 14.22).
Results: We demonstrated that the proper combination of motor and non-motor measures significantly improved (p < 0.001)the prediction of outcome, reducing the RMSE (root-mean-square-error) of predicting UPDRS-III from 3.487279 to 3.066292 which signifies that the model was optimized quite well. Based on the “IncNodePurity”, the improvement factors of UPDRS-III (year 2) were, in descending order of magnitude, UPDRS-III (year 0), Disease Duration, MoCA (year 2), KPPQ (year0), Time “Off” (year2), Time Dyskinesia (year 0), PDQ39 (year 2) after the implementation Levodopa-carbidopa intestinal gel.
Conclusion: At 2-year follow-up, the decrease of motor outcome (UPDRS-III) was determined by UPDRS-III (year0), disease duration and non-motor symptoms of PD patients with LGIG therapy. These results are promising to improve patient counseling, expectation management, and patient satisfaction with LCIG therapy.
To cite this abstract in AMA style:
E. Efthymiopoulou, A. Antonoglou, B. Loupo, A. Bougea. Determination of the motor status after the implementation Levodopa-carbidopa intestinal gel in patients with advanced Parkinson’s disease using a Machine learning Algorithm [abstract]. Mov Disord. 2021; 36 (suppl 1). https://www.mdsabstracts.org/abstract/determination-of-the-motor-status-after-the-implementation-levodopa-carbidopa-intestinal-gel-in-patients-with-advanced-parkinsons-disease-using-a-machine-learning-algorithm/. Accessed September 8, 2024.« Back to MDS Virtual Congress 2021
MDS Abstracts - https://www.mdsabstracts.org/abstract/determination-of-the-motor-status-after-the-implementation-levodopa-carbidopa-intestinal-gel-in-patients-with-advanced-parkinsons-disease-using-a-machine-learning-algorithm/