Assessing the predictive value of peak alpha frequency for the sensitivity to pain

Figure 1.. Peak alpha frequency (PAF). Peak alpha frequency is derived from the power spectrum of an individual's electroencephalography or magnetoencephalography (EEG/MEG) recording and is defined as the frequency with the highest power and a peak within the alpha range (8-13 Hz). As to the exact computation of PAF, different approaches exist.

Figure 2.. Outline of research questions, associated analyses, and results.

Figure 3.. Paradigm. Participants took part in 2 identical experimental sessions separated by approximately 1 month. Each session started with a 5-minute eyes-closed resting-state recording. This was followed by a pain paradigm consisting of 80 laser stimuli with 4-millisecond duration, and 4 fixed stimulus intensities (2.5, 3, 3.5, and 4 J) applied to the dorsum of the left hand in a pseudo-randomized order. Three seconds after each stimulus, an auditory cue prompted participants to verbally rate the perceived pain intensity on a scale from 0 (“no pain”) to 100 (“worst tolerable pain”). The interstimulus interval was 8 to 12 seconds. Brain activity was recorded using EEG. EEG, electroencephalography; ISI, interstimulus interval.

Figure 4.. Interindividual and intraindividual variations of pain perception and S1 PAF. (A) Single-participant mean pain ratings and mean S1 resting-state PAFs obtained during primary analyses for sessions 1 and 2. To obtain mean pain ratings, ratings were first averaged separately for trials of the 4 different stimulus intensities and then averaged across them. Mean S1 PAFs were obtained by averaging PAF across 5 bins based on sorted pain ratings (see below). (B) Single-participant SDs of pain ratings and S1 prestimulus PAFs obtained during primary analyses for sessions 1 and 2. SDs of pain ratings were calculated separately for trials of the 4 stimulus intensities and then averaged across them. SDs of S1 PAFs were calculated across bins. Raincloud plots show unmirrored violin plots displaying the probability density function of the data, boxplots, and individual data points. Boxplots depict the sample median, first (Q1), and third quartiles (Q3). Whiskers extend from Q1 to the smallest value within Q1 − 1.5* interquartile range (IQR) and from Q3 to the largest values within Q3 + 1.5* IQR. NRS, numerical rating scale; PAF, peak alpha frequency.

Figure 5.. The relationship between S1 resting-state PAF and interindividual variations of pain. (A) Primary analysis. Scatter plots show the results of Bayesian linear correlations between S1 resting-state PAF and pain intensity, corrected for age. (B) Multiverse analysis. The top panels show Pearson correlation coefficients obtained for all 144 specifications, sorted by effect size. The primary analysis is highlighted by a vertical bar. Middle panels mark the corresponding analytical choices of every specification (Table 1). Results are color-coded based on Bayes factors. The scale's upper and lower ends indicate moderate evidence for and against a relationship. The bottom panels display the inferential specification curve analysis showing the observed specification curve and the median and 1-sided 95% confidence intervals of 500 specification curves obtained from permutations of the data under the null hypothesis, all sorted by effect size. BF, Bayes factor; cog, center of gravity; PAF, peak alpha frequency; R, Pearson correlation coefficient; S1, primary somatosensory cortex.

Figure 6.. The relationship between S1 prestimulus PAF and intraindividual variations of pain. (A) Primary analysis. Raincloud plots show single-participant correlation coefficients of the relation between S1 prestimulus PAF and pain intensity. Bayes factors are obtained from Bayesian t-tests of correlation coefficients vs zero. (B) Multiverse analysis. The top panels show Pearson correlation coefficients obtained for all 216 specifications, sorted by effect size. The primary analysis is highlighted by a vertical bar. Middle panels mark the corresponding analytical choices of every specification (Table 1). Results are color-coded based on Bayes factors. The scale's upper and lower ends indicate moderate evidence for and against a relationship. The bottom panels display the inferential specification curve analysis showing the observed specification curve and the median and 1-sided 95% confidence intervals of 500 specification curves obtained from permutations of the data under the null hypothesis, all sorted by effect size. BF, Bayes factor; cog, center of gravity; PAF, peak alpha frequency; R, Pearson correlation coefficient; S1, primary somatosensory cortex.

Figure 7.. The relationship between global resting-state PAF and interindividual variations of pain. (A) Primary analysis. Scatter plots show the results of Bayesian linear correlations between global resting-state PAF and pain intensity, corrected for age. (B) Multiverse analysis. The top panels show Pearson correlation coefficients obtained for all 72 specifications, sorted by effect size. The primary analysis is highlighted by a vertical bar. Middle panels mark the corresponding analytical choices of every specification (Table 1). Results are color-coded based on Bayes factors. The scale's upper and lower ends indicate moderate evidence for and against a relationship. The bottom panels display the inferential specification curve analysis showing the observed specification curve and the median and 1-sided 95% confidence intervals of 500 specification curves obtained from permutations of the data under the null hypothesis, all sorted by effect size. BF, Bayes factor; cog, center of gravity; PAF, peak alpha frequency; R, Pearson correlation coefficient.

Figure 8.. The relationship between global prestimulus PAF and intraindividual variations of pain. (A) Primary analysis. Raincloud plots show single-participant correlation coefficients of the relation between global prestimulus PAF and pain intensity. Bayes factors are obtained from Bayesian t-tests of correlation coefficients vs zero. (B) Multiverse analysis. The top panels show Pearson correlation coefficients obtained for all 108 specifications, sorted by effect size. The primary analysis is highlighted by a vertical bar. Middle panels mark the corresponding analytical choices of every specification (Table 1). Results are color-coded based on Bayes factors. The scale's upper and lower ends indicate moderate evidence for and against a relationship. The bottom panels display the inferential specification curve analysis showing the observed specification curve and the median and 1-sided 95% confidence intervals of 500 specification curves obtained from permutations of the data under the null hypothesis, all sorted by effect size. BF, Bayes factor; cog, center of gravity; PAF, peak alpha frequency; R, Pearson correlation coefficient.