Aziz Hatim Poonawalla1, Sushmita Datta1, Vaibhav Juneja1, Flavia Nelson2, Jerry Wolinsky2, Gary Cutter3, Ponnada Narayana1
1Diagnostic and Interventional Imaging, University of Texas Medical School at Houston, Houston, TX, USA; 2Neurology, University of Texas Medical School at Houston, Houston, TX, USA; 3Biostatistics, University of Alabama at Birmingham School of Public Health, Birmingham, AL, USA
In the relapsing-remitting multiple sclerosis (RRMS) literature, r-values for correlation with Expanded Disability Status Scale (EDSS) of various quantitative MRI metrics range from 0.310.61, following a strong (r=-0.7) inverse trend with population sample size. To better characterize this dependence, we conducted numerical simulations on ever-larger subsets of a large (n=139) RRMS patient cohort. The simulations confirmed that the correlation with EDSS was inflated for smaller population sizes, converging to the baseline correlation value as n approached maximum, with a metric-dependent threshold for convergence. The results suggest that reported correlations may be overstated for small studies and understated for large ones.