Journal:Brain, Behavior, and Immunity
Year, Volume, Issue, Page(s):16, 53, , 183-193
Study used principal component analysis (PCA) and cluster analysis to identify the sets of markers that contribute independently to variability in cerebrospinal (CSF) inflammatory profiles after traumatic brain injury (TBI). PCA was applied to CSF inflammatory marker data collected from 114 adults with severe TBI. Using the results, groups (or clusters) of individuals with similar patterns of acute CSF inflammation were defined that were then evaluated in the context of outcome and other relevant CSF and serum biomarkers collected days 0-3 and 4-5 post-injury. Four significant principal components (PC1-PC4) were identified for days 0-3, and PC1 accounted for 31 percent of the variance. Cluster analysis then defined two distinct clusters, such that 100 percent of individuals in cluster 1 had a positive score for PC1, and a majority of individuals (61 percent) in cluster 2 had negative values for PC1. Multinomial logistic regression analyses showed that individuals in cluster 1 had a 10.9-times increased likelihood of severe disability at 6 months compared to cluster 2, after controlling for covariates. Cluster group did not discriminate between mortality compared to good recovery after controlling for age and other covariates. Cluster groupings also did not discriminate mortality or 12 month outcomes in multivariate models. In this study, PCA and cluster analysis established a subset of CSF inflammatory markers measured in days 0-3 post-TBI that distinguished individuals with poor 6-month outcome. Overall, this data-driven approach provides a novel assessment of the potential of patterns among TBI-associated inflammatory biomarkers to predict long-term outcomes after TBI.