Longitudinal statistical analysis of anti-CGRP migraine treatment outcomes, investigating the mismatch between biological improvement and patient-perceived benefit.
Anti-CGRP therapies are designed to reduce migraine frequency by targeting the Calcitonin Gene-Related Peptide (CGRP) pathway. Clinical efficacy is evaluated using Monthly Migraine Days (MMDs).
However, improvements in migraine frequency do not always correspond to a proportional improvement in patient-perceived disease burden.
This project investigates the potential objective-subjective mismatch in treatment response using a real-world longitudinal dataset of migraine patients.
The analysis explores two main questions:
Objective vs Subjective Response
Does a reduction in migraine frequency correspond to improvements in patient-reported outcomes such as HIT-6?
Determinants of Discordance
Which baseline characteristics may explain cases where clinical improvement does not translate into perceived benefit?
Longitudinal Trend in Monthly Migraine Days
Anti-CGRP therapy shows a clear reduction in migraine frequency over time across treatment cycles.
Objective vs Subjective Treatment Response
This scatter plot highlights the relationship between objective reduction in migraine days and change in patient-perceived disease buerden (HIT-6).
The analysis uses a longitudinal dataset of migraine patients treated with anti-CGRP therapies.
The data contains two main components.
Baseline dataset
Clinical and demographic information collected before treatment:
- age and gender
- migraine history
- comorbidities
- baseline disability and pain measures
Longitudinal dataset
Repeated measurements collected during treatment:
- Monthly Migraine Days (MMDs)
- HIT-6 (Headache Impact Test)
- MIDAS disability score
- HADS anxiety and depression scores
Patients were followed across three treatment cycles, with visits at: Month 1 - Month 3 - Month 6 - Month 9 - Month 12
Longitudinal Modeling Treatment trajectories were analyzed using linear mixed-effects models to account for repeated measurements within patients.
Model specification:
MMDs ~ CYCLE + MONTH + (1 | SUBJECT_ID)
HIT6 ~ CYCLE + MONTH + (1 | SUBJECT_ID)
These models estimate population-level treatment effects while capturing individual variability.
Response Definitions
Treatment response was evaluated at the end of Cycle 1.
Objective response
≥50 % reduction in Montlhy Migraine Days
Subjective response
≥5-points improvement in HIT-6 score
Patients were classified into response profile, with particular focus on the discordant subgroup showing biological improvement without perceived benefit.
-
Data integration
Merge baseline and longitudinal datasets. -
Missing data handling
Multiple Imputation (MICE) for baseline variables and linear interpolation for longitudinal outcomes. -
Longitudinal modeling
Linear mixed-effects models to evaluate treatment trajectories. -
Response classification
Identification of objective and subjective responders. -
Exploratory regression analysis
Logistic regression to explore baseline predictors of discordance. -
Clinical interpretation
Evaluation of the mismatch between biological response and patient-perceived benefit.
The analysis reveals important patterns.
Anti-CGRP therapy shows consistent reductions in migraine frequency across treatment cycles.
Improvements in HIT-6 scores are more heterogeneous, indicating that subjective benefit does not always follow objective improvement.
A small subgroup of patients exhibits objective improvement without subjective benefit, highlighting a clinically relevant mismatch.
Exploratory regression analysis suggests that this discordance is not explained by baseline clinical variables alone, indicating that additional psychosocial or dynamic factors may influence perceived treatment-benefit.
These findings suggest that treatment success in migraine prevention cannot be evaluated solely through biological markers such as migraine frequency.
Patient-reported outcomes provide complementary information about disease burden and perceived beneift, highlighting the importance of integrating subjective measures into treatment evaluation.
Language: R
Libraries
- tidyverse
- lme4
- mice
- ggplot2
- dplyr
- tidyr
Methods:
- longitudinal data analysis
- mixed-effects models
- multiple imputation (MICE)
- logistic regression
Additional materials for this project are available below.
Project report Full methodological details and results. Read the report
Project Presentation Slides used to present the project and main findings. View the presentation
Gallo Sabina
Marrali Irene
BSc in Artificial Intelligence @ Università degli Studi di Milano, Università degli Studi di Pavia, Università degli Studi di Milano-Bicocca