Machine‐learning‐based approach for predicting response to anti‐calcitonin gene‐related peptide (CGRP) receptor or ligand antibody treatment in patients with migraine: A multicenter Spanish study

• Published in: European journal of neurology Vol. 29; no. 10; pp. 3102 - 3111
• Main Authors: Gonzalez‐Martinez, Alicia, Pagán, Josué, Sanz‐García, Ancor, García‐Azorín, David, Rodríguez‐Vico, Jaime Samuel, Jaimes, Alex, García, Andrea Gómez, Terán, Javier Díaz, González‐García, Nuria, Quintas, Sonia, Belascoaín, Rocío, Casas Limón, Javier, Latorre, Germán, Calle de Miguel, Carlos, Sierra, Álvaro, Guerrero‐Peral, Ángel Luis, Trevino‐Peinado, Cristina, Gago‐Veiga, Ana Beatriz
• Format: Journal Article
• Language: English
• Published: Oxford John Wiley & Sons, Inc 01-10-2022
• Subjects: Antibodies; Antibody response; Calcitonin; Calcitonin gene-related peptide; CGRP; Data collection; Demographic variables; erenumab; fremanezumab; galcanezumab; Headache; Headaches; Impact tests; Learning algorithms; Ligands; Machine learning; Migraine; Model accuracy; Monoclonal antibodies; Patients; Peptides; Prediction models; predictor; Receptors; response; Response rates
• ISSN: 1351-5101; 1468-1331
• DOI: 10.1111/ene.15458

Background and purpose
Several variables have been reported to be associated with anti‐calcitonin gene‐related peptide (CGRP) receptor or ligand antibody response, but with differing results. Our objective was to determine whether machine‐learning (ML)‐based models can predict 6‐, 9‐ and 12‐month responses to anti‐CGRP receptor or ligand therapies among migraine patients.
Methods
We performed a multicenter analysis of prospectively collected data from patients with migraine receiving anti‐CGRP therapies. Demographic and clinical variables were collected. Response rates in the 30% to 50% range, or at least 30%, in the 50% to 75% range, or at least 50%, and response rate of at least 75% regarding the reduction in the number of headache days per month at 6, 9 and 12 months were calculated. A sequential forward feature selector was used for variable selection and ML‐based predictive models for the response to anti‐CGRP therapies at 6, 9 and 12 months, with model accuracy not less than 70%, were generated.
Results
A total of 712 patients were included, 93% were women, and the mean (SD) age was 48 (11.6) years. Eighty‐four percent of patients had chronic migraine. ML‐based models using headache days/month, migraine days/month and the Headache Impact Test (HIT‐6) yielded predictions with an F1 score range of 0.70–0.97 and an area under the receiver‐operating curve score range of 0.87–0.98. SHAP (SHapley Additive exPlanations) summary plots and dependence plots were generated to evaluate the relevance of the factors associated with the prediction of the above‐mentioned response rates.
Conclusions
Our results show that ML models can predict anti‐CGRP response at 6, 9 and 12 months. This study provides a predictive tool that can be used in a real‐world setting.
In this study, the ability of machine‐learning (ML) algorithms to predict response to anti‐calcitonin gene‐related peptide (CGRP) receptor or ligand‐antibody therapies in a large cohort of patients with migraine was evaluated. The main finding was that ML can generate predictive models for response to anti‐CGRP receptor at 6, 9 and 12 months with high accuracy, using clinically relevant headache variables and scales. A calculator tool for the prediction of response rate was also generated. This study shows that ML may have an important role to play in improving prediction of anti‐CGRP treatment responses in a real‐world setting and could serve as a decision support system aid.