Brain mediators of predictive cue effects on perceived pain

Abstract

Information about upcoming pain strongly influences pain experience in experimental and clinical settings, but little is known about the brain mechanisms that link expectation and experience. To identify the pathways by which informational cues influence perception, analyses must jointly consider both the effects of cues on brain responses and the relationship between brain responses and changes in reported experience. Our task and analysis strategy were designed to test these relationships. Auditory cues elicited expectations for barely painful or highly painful thermal stimulation, and we assessed how cues influenced human subjects' pain reports and brain responses to matched levels of noxious heat using functional magnetic resonance imaging. We used multilevel mediation analysis to identify brain regions that (1) are modulated by predictive cues, (2) predict trial-to-trial variations in pain reports, and (3) formally mediate the relationship between cues and reported pain. Cues influenced heat-evoked responses in most canonical pain-processing regions, including both medial and lateral pain pathways. Effects on several regions correlated with pretask expectations, suggesting that expectancy plays a prominent role. A subset of pain-processing regions, including anterior cingulate cortex, anterior insula, and thalamus, formally mediated cue effects on pain. Effects on these regions were in turn mediated by cue-evoked anticipatory activity in the medial orbitofrontal cortex (OFC) and ventral striatum, areas not previously directly implicated in nociception. These results suggest that activity in pain-processing regions reflects a combination of nociceptive input and top-down information related to expectations, and that anticipatory processes in OFC and striatum may play a key role in modulating pain processing.

Figures

Figure 1.

Experimental design and behavioral results. a, Trial structure. Each trial consisted of an auditory predictive cue followed by an anticipatory delay and 10 s of noxious thermal stimulation. fMRI analyses used single trial analysis to model brain responses evoked during anticipation and noxious thermal stimulation periods. After a fixed interstimulus interval, participants reported trial-by-trial perceived pain using a visual analog scale. b, Conditions of interest. During the first two runs of the task, low-pain cues always preceded low-pain stimulation (LL) and high-pain cues preceded high-pain stimulation (HH). In runs three through eight, trials were evenly divided between these conditions and trials in which each predictive cue was followed by a stimulus calibrated to elicit moderate pain [high cue plus medium pain (HM); low cue plus medium pain (LM)]. We included only HM and LM trials in our mediation analyses to examine cue-based expectancy effects during a single level of noxious thermal stimulation. c, Behavioral results. All participants reported greater pain on HM than LM trials. This difference comprises the “total effect” (path c) in our pain-period multilevel mediation analysis (analysis 1). ****p < 0.0001.

Figure 2.

Mediation hypothesis framework: analysis 1. Our primary analysis examines the dynamic relationships between pain predictive cue [top left; X; green, low cue plus medium pain (LM); red, high cue plus medium pain (HM)], voxelwise pain-evoked responses (bottom, M; single trial analysis estimates of AUC, the area under the curve), and trial-by-trial pain reports (top right; Y). The four components of multilevel mediation analysis address the current study's key questions. Top, Path c/c′: Do predictive cues affect perceived pain as measured by trial-by-trial pain reports? Path c reflects the total relationship between predictive cue and reported pain on medium trials, and path c′ reflects the direct behavioral relationship, controlling for activity in the mediator—in this case a brain voxel or region. Left, Path a (“cue effects”) provides inferences on whether brain voxels are modulated by predictive cue during a constant level of noxious thermal stimulation (Fig. 3, Fig. S2, available at www.jneurosci.org as supplemental material). Right, Path b (“report-related responses”) provides inferences on whether brain activity in each voxel predicts trial-by-trial pain reports, controlling for cue (stimulus temperature was held constant) (Fig. 4, Fig. S3, available at www.jneurosci.org as supplemental material). Middle, The a × b mediation effect provides inferences on whether brain voxels explain a significant amount of the covariance between predictive cues and perceived pain (Fig. 5, Fig. S4, available at www.jneurosci.org as supplemental material).

Figure 3.

Path a: cue-based modulation of pain period activity. a, Cue effects on PPN regions. Our analysis of cue effects (“path a”) consists of a direct contrast between activity on HM and LM trials. Nearly all PPN regions were modulated by predictive cues. Extracted time courses (bottom) demonstrate that responses peak late in the pain period or after termination of the stimulus for most regions of interest, due at least in part to hemodynamic response lag. Pink bars on the x-axis indicate cue presentation, red bars indicate noxious stimulation, and gray bars indicate rating scale presentation. b, Cue effects on limbic and frontal regions. We also observed positive cue effects (HM > LM, yellow/orange) in left DLPFC and left amygdala, while medial OFC and right VLPFC showed greater activity with low-pain cues (LM > HM, blue).

Figure 4.

Path b: activity related to reported pain. a, Report-related PPN regions. Our pain report analysis (“path b”) identifies voxels whose pain period activity predicts trial-by-trial pain ratings, controlling for cue. This analysis revealed that a subset of pain processing regions, including right anterior insula, rdACC, pre-SMA, and bilateral cerebellum, correlated with variations in subjective intensity under a constant level of noxious thermal stimulation. For each PPN region, bottom graphs show the variability in path b coefficients across individual subjects (blue lines) and the group regression slope (dark black line) and 95% confidence interval (gray shading). b, Report-related frontal and limbic regions. In addition to PPN regions, right dmPFC, inferior frontal gyrus (IFG), and lateral PFC were also positively correlated with trial-by-trial pain reports, controlling for cue (yellow/orange). Medial OFC was inversely related to perceived pain (blue).

Figure 5.

Pain processing network mediators. a, Pain processing network mediators: surface map. Three key PPN regions were found to mediate expectancy effects on perceived pain. R. Thalamus, Right thalamus; L. Ant. Insula, left anterior insula. b, Pain processing network mediators. Slice views of PPNMs, clockwise from top left: left anterior insula (−36, 6, −4); right thalamus (18, −18, 14); left rdACC (−8, 18, 32). c, Mediation path diagram for the left anterior insula. Left anterior insula shows a positive path a effect, indicating more activity with high-pain cues. The mean standardized path coefficient is shown with standard error (in parentheses). Lower left, Individual subjects' regression lines and group average, as in Fig. 4. Lower right, Histogram of the bootstrap estimates of the path a distribution. The light shading shows the 95% bootstrap confidence intervals, and vertical line marks the null hypothesis value of zero. The right panels show the path b effect, which shows a link with trial-by-trial reported pain. Standardized path coefficients and standard error for the mediation effect are shown in white text on the brain surface map. The dashed arrow represents the direct effect of predictive cues on perceived pain (path c′). *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 6.

Functional response profiles in key mediator regions. Bar graphs depict functional patterns underlying observed cue effects in key mediator regions. The left panel illustrates theoretical patterns across levels of noxious stimulation. Regions depicted in red, including left anterior insula (L. Insula), right thalamus (R. Thalamus), left dorsolateral prefrontal cortex (L. DLPFC), and pons display a pain assimilation profile (LL LM > HM > HH). Dorsal anterior cingulate cortex, depicted in orange, shows a profile based on expected intensity, such that responses to the high-pain cue conditions (HH and HM) are greater than responses to the low-pain cue conditions (LL and LM). Rostrodorsal anterior cingulate, depicted in yellow, shows an expectancy violation signal, such that responses to medium trials are greater than responses to either of the validly cued conditions.

Figure 7.

Anticipatory mediators of cue effects on PPNM responses (analysis 2). a, Framework for analysis 2. We conducted a secondary mediation analysis to examine the relationship between cue-evoked anticipatory activity and pain processing network mediators depicted in Fig. 5 (left anterior insula, left rdACC, and right thalamus). We tested whether anticipatory responses mediate cue effects on PPNM pain-evoked responses on medium temperature trials. Activity was averaged across PPNMs to localize mediator voxels, and we then tested the relationship between mediators and each individual PPNM region. b, Analysis 2 results. Cue-evoked anticipatory responses in medial OFC (top) and right ventral striatum (bottom) mediated cue effects on all three PPNMs. All path coefficients were positive. Mean path coefficients and standard error for the individual pathways are presented in Table 1.