Exploring Natural Clusters of Chronic Migraine Phenotypes: A Cross-Sectional Clinical Study

Abstract

Heterogeneity in chronic migraine (CM) presents significant challenge for diagnosis, management, and clinical trials. To explore naturally occurring clusters of CM, we utilized data reduction methods on migraine-related clinical dataset. Hierarchical agglomerative clustering and principal component analyses (PCA) were conducted to identify natural clusters in 100 CM patients using 14 migraine-related clinical variables. Three major clusters were identified. Cluster I (29 patients) - the severely impacted patient featured highest levels of depression and migraine-related disability. Cluster II (28 patients) - the minimally impacted patient exhibited highest levels of self-efficacy and exercise. Cluster III (43 patients) - the moderately impacted patient showed features ranging between Cluster I and II. The first 5 principal components (PC) of the PCA explained 65% of variability. The first PC (eigenvalue 4.2) showed one major pattern of clinical features positively loaded by migraine-related disability, depression, poor sleep quality, somatic symptoms, post-traumatic stress disorder, being overweight and negatively loaded by pain self-efficacy and exercise levels. CM patients can be classified into three naturally-occurring clusters. Patients with high self-efficacy and exercise levels had lower migraine-related disability, depression, sleep quality, and somatic symptoms. These results may ultimately inform different management strategies.

Conflict of interest statement

The authors declare no competing interests.

Figures

Figure 1

Hierarchical agglomerative clustering results. (a) Radial dendrogram. Hierarchical agglomerative clustering resulted in determining 3 major clusters of chronic migraine patients based on their migraine-related phenotype and comorbidities. Cluster I (red; 29 patients) – the severely impacted patient featured higher levels of depression and migraine-related disability. Cluster II (green; 28 patients) – the minimally impacted patient exhibited higher levels of pain self-efficacy and exercise. Cluster III (blue; 43 patients) – the moderately impacted patient showed most features ranging midway between Cluster I and II. CM patients with MOH are displayed next to each color. Cluster I CM patients (the severely impacted) were 4 times higher odds of having MOH compared to Cluster II CM (the minimally impacted) (p = 0.02). (b) Inter-median comparison of clinical variables among the 3 clusters. This graph shows the comparison among the median centroids of each cluster across the 14 clinical phenotype variables. Significant differences in levels of depression, pain self-efficacy, migraine-related disability, and exercise were found between Cluster I (red; the severely impacted) compared to Cluster II (green; the minimally impacted). Statistically significant difference between median centroids of all 14 variables among the three clusters was examined using Kruskal-Wallis and Dunn’s post-hoc test. Bonferroni correction was used to adjust for multiple testing by dividing p value of 0.05 to 14, and using new p value of 0.0036 as significance threshold. Significant inter-median differences with p value < 0.0036 are displayed. Statistically significant results are shown with asterisks as ***p < 0.001, ****p < 0.0001. Abbreviations: BMI for body mass index; Freq for frequency of headache days; Sev for severity of headache; Dur for duration of chronic migraine in years; PHQ-9 instrument for assessing depression; GAD-7 for assessing anxiety; PCS instrument for assessing pain catastrophizing; PSQI for assessing sleep quality; PTSD for post-traumatic stress duration assessed by PC-PTSD; PHQ-15 for assessing level of somatic symptoms; PSEQ for pain self-efficacy questionnaire; MIDAS for migraine disability assessment; Ex for weekly exercise minutes. The scores were all rescaled from 0–1 as shown on “y” axis. (c) Agglomeration coefficients schedule. The first large increase between two consecutive agglomeration coefficients is indicated by blue star at stage 97, eliminating stages 98 and 99 with resultant 3 clusters as shown in. (d) Dendrogram with red line indicating optimal stopping point of clustering. The red line crosses 3 vertical  lines corresponding to 3 clusters. The last 2 horizontal  lines represent the last 2 agglomeration stages (stages 98 and 99). Agglomeration coefficients schedule is shown in (c).

Figure 2

Results from PCA. (a) Scree plot for PCA. The first five principal components (PCs) were found to have eigenvalues greater than 1 and explained 65% of the cumulative variability (red line) within the CM phenotype dataset. The steep part of the slope (blue line) was made by the first two PCs which were used to construct the biplot in (b). PCs are displayed on the ‘x’ axis. The right ‘y’ axis shows the percentage of cumulative variability explained by each variable (red line). The left ‘y’ axis shows the eigenvalues for each PC. (b) Biplot of PCA clinical variables and chronic migraine patients. Biplot of clinical variables and participants showed differential aggregation of the 3 clusters identified on hierarchical agglomerative clustering. Ellipses show 95% confidence interval of clusters aggregation. Cluster I CM patients (red ellipse; the severely impacted) aggregated around higher migraine disability while Cluster II (green ellipse; the minimally impacted) assembled around higher self-efficacy and exercise levels. Cluster III (blue ellipse; the moderately impacted) patients were scattered between Clusters I and II.