Machine learning models using just one month of early treatment data can predict 3-, 6-, and 12-month response to anti-CGRP migraine antibodies.
1-month data predicts 12-month responseML models using early treatment data can forecast long-term CGRP antibody response, potentially saving months of ineffective treatment
What the researchers found
Machine learning models using early treatment data (up to one month) can predict 3-, 6-, and 12-month response to anti-CGRP antibodies in migraine patients.
Why it matters
CGRP antibodies are expensive and not every patient responds. Being able to predict who will benefit early could save time, money, and help patients get to effective treatment faster.
The numbers in context
Models predicted responses at 3, 6, and 12 months using predictors available within the first month of treatment. Data included monthly headache days and acute medication use.
How the study worked
Prospective cohort study collecting demographic and monthly clinical data from migraine patients receiving anti-CGRP mAbs for 12 months, with machine learning model development.
Who was studied
Migraine patients receiving anti-CGRP monoclonal antibody treatment for 12 months
What this study cannot tell us
The model needs external validation in other patient populations. Machine learning predictions are probabilistic, not certain, for individual patients.
How to read the evidence
Rated moderate: prospective cohort study with ML modeling. Needs external validation in other patient populations and clinical settings.
When this study was published
Published in 2024. Part of the growing application of machine learning to personalized medicine in neurology.
The bigger picture
CGRP antibodies are expensive and not all patients respond. Predicting response within one month could save months of ineffective treatment and thousands of dollars per patient, while getting non-responders to alternative therapies faster.
Questions still open
- What accuracy do the models achieve for each prediction timepoint?
- Would the model work across different CGRP antibodies?
Common questions
Can doctors predict if CGRP drugs will work for you?
How does the prediction tool work?
Read the original research
An evolving machine-learning-based algorithm to early predict response to anti-CGRP monoclonal antibodies in patients with migraine.
Cephalalgia : an international journal of headache, 44(12), 3331024241262751
Citation
Romozzi, Marina; Lokhandwala, Ammar; Vollono, Catello; Vigani, Giulia; Burgalassi, Andrea; García-Azorín, David; Calabresi, Paolo; Chiarugi, Alberto; Geppetti, Pierangelo; Iannone, Luigi Francesco. (2024). An evolving machine-learning-based algorithm to early predict response to anti-CGRP monoclonal antibodies in patients with migraine.. Cephalalgia : an international journal of headache, 44(12), 3331024241262751. https://doi.org/10.1177/03331024241262751