Parametric trial-by-trial prediction of pain by easily available physiological measures

Pain is commonly assessed by subjective reports on rating scales. However, in many experimental and clinical settings, an additional, objective indicator of pain is desirable. In order to identify an objective, parametric signature of pain intensity that is predictive at the individual stimulus level across subjects, we recorded skin conductance and pupil diameter responses to heat pain stimuli of different durations and temperatures in 34 healthy subjects. The temporal profiles of trial-wise physiological responses were characterized by component scores obtained from principal component analysis. These component scores were then used as predictors in a linear regression analysis, resulting in accurate pain predictions for individual trials. Using the temporal information encoded in the principal component scores explained the data better than prediction by a single summary statistic (i.e., maximum amplitude). These results indicate that perceived pain is best reflected by the temporal dynamics of autonomic responses. Application of the regression model to an independent data set of 20 subjects resulted in a very good prediction of the pain ratings demonstrating the generalizability of the identified temporal pattern. Utilizing the readily available temporal information from skin conductance and pupil diameter responses thus allows parametric prediction of pain in human subjects.