Improving fMRI-based prediction of individual pain ratings by accounting for interindividual variance

Abstract

We challenge the pessimistic conclusion of a recently published paper by Hoeppli et al that fMRI-BOLD measures are useless in predicting interindividual differences in pain perception. By conducting a re-analysis of publicly available data of their study, we propose an alternative analysis approach that addresses the issue of interindividual variance differences in BOLD readouts, aiming to enhance the predictive power of fMRI-based measures. Instead of using absolute values of both, pain ratings and BOLD measures, we make use of robust intraindividual differences between the two experimental conditions in their study. Our findings demonstrate a statistically significant positive linear relationship between the neurologic pain signature (NPS) score, a multivariate pain classifier based on BOLD fMRI, and individual differences in perceived pain ratings. This relationship is driven by individuals that report pain sensitivity to both experimental conditions and can clearly distinguish between the two. Our results provide evidence for the potential of fMRI-BOLD measures in predicting interindividual differences in pain perception and allow for a more optimistic conclusion regarding the ongoing debate whether fMRI can be used as an objective measure for pain perception.

A replicable and generalizable neuroimaging-based indicator of pain sensitivity across individuals

“…We examined two likely explanations. One is that the physical intensity of painful stimuli may be too small to evoke clear pain sensation (Goltermann & Büchel, 2023). The power to detect correlations will be dampened if no clear pain sensations are elicited or BOLD signals are just too noisy due to insufficient physical intensity.…”

“…An early seminal study linked larger pain-evoked blood oxygen level dependent (BOLD) responses in these areas with higher pain sensitivity (Coghill et al, 2003). Yet, a recent study from the same lab with a larger sample size (N = 101) demonstrated that no single brain area significantly correlated with interindividual pain sensitivity (Hoeppli et al, 2022), which then ignited a new round of debate (Gim et al, 2023; Goltermann & Büchel, 2023). These discrepant findings in fMRI studies may represent a fresh example of the replication problem and power failure in empirical sciences (Baker, 2016; Button et al, 2013; Open Science Collaboration, 2015; Poldrack et al, 2017).…”