A multiverse approach to heat-evoked skin conductance analysis: evaluating the influence of analytic pipeline on associations between skin conductance and pain

Indices of physiological arousal and sympathetic nervous system activity provide important information about pain that can supplement verbal reports. Numerous studies focus on electrodermal activity, which is easy to measure and can provide insights on relationships between pain and arousal through the quantification of skin-conductance response (SCR). Despite SCR's value, there is no standard way to quantify SCR in response to noxious stimulation, and conclusions may vary depending on analytic approaches. We used a multiverse analysis to evaluate associations between SCR and pain across 8 studies of thermal pain in healthy volunteers (total n = 567). For each data set, SCRs were quantified using 5 common analysis approaches (preprocessing only; manual scoring; automated scoring; model-based analysis; physiological pain classifier), leading to 18 distinct SCR measures. Associations with pain and temperature were strongest when automated scoring was combined with manual artifact detection, regardless of model comparison approach, study type, and whether pain was rated continuously or categorically. More specifically, Bayesian and model-based comparisons revealed the strongest associations with pain when the software Ledalab quantified the sum of amplitudes on each trial and we excluded trials with artifacts. We apply these findings across data sets both with and without psychological interventions to show that subjective pain statistically mediates associations between temperature and heat-evoked SCR, and that SCR is modulated by pain-predictive cues regardless of verbal instruction. This work advances a simple analysis pipeline that can be adopted in future pain studies to improve efficiency and reproducibility and aid the development of objective pain biomarkers.