Functional network architecture predicts psychologically mediated analgesia related to treatment in chronic knee pain patients

Abstract

Placebo analgesia is an indicator of how efficiently the brain translates psychological signals conveyed by a treatment procedure into pain relief. It has been demonstrated that functional connectivity between distributed brain regions predicts placebo analgesia in chronic back pain patients. Greater network efficiency in baseline brain networks may allow better information transfer and facilitate adaptive physiological responses to psychological aspects of treatment. Here, we theorized that topological network alignments in resting state scans predict psychologically conditioned analgesic responses to acupuncture treatment in chronic knee osteoarthritis pain patients (n = 45). Analgesia was induced by building positive expectations toward acupuncture treatment with verbal suggestion and heat pain conditioning on a test site of the arm. This procedure induced significantly more analgesia after sham or real acupuncture on the test site than in a control site. The psychologically conditioned analgesia was invariant to sham versus real treatment. Efficiency of information transfer within local networks calculated with graph-theoretic measures (local efficiency and clustering coefficients) significantly predicted conditioned analgesia. Clustering coefficients in regions associated with memory, motivation, and pain modulation were closely involved in predicting analgesia. Moreover, women showed higher clustering coefficients and marginally greater pain reduction than men. Overall, analgesic response to placebo cues can be predicted from a priori resting state data by observing local network topology. Such low-cost synchronizations may represent preparatory resources that facilitate subsequent performance of brain circuits in responding to adaptive environmental cues. This suggests a potential utility of network measures in predicting placebo response for clinical use.

Keywords: brain network; chronic pain; placebo; predictive analysis; resting state; synchronization.

Figures

Figure 1.

Psychological conditioning significantly reduced evoked pain in OA patients. A, The patients were psychologically conditioned to expect analgesia in the test side of the arm. Patients reported greater analgesia (percentage decrease postconditioning to preconditioning) in test side relative to the control side in both sham and acupuncture-treated groups. B, The psychologically conditioned analgesia was invariant to the type of treatment (sham vs real); both groups showed pain decrease specific to the psychological conditioning. C, Histogram shows variability in response to psychological conditioning between individuals. Error bars indicate ± SEM, **p < 0.01.

Figure 2.

Parcellation scheme used for measuring network properties in the brain of resting OA patients. A, The network was constructed from 125 regions based on the Harvard Oxford Atlas (see Table 2 for list of coordinates). B, Circle diagram showing local connectivity near the perimeter and long distance connections between nodes shown in randomly colored edges (mean of n = 42 displayed at T = 0.6). The right (R) and left (L) sides of the circle represent the two brain hemispheres and the ordering of the nodes from top to bottom is organized based on their respective positions on the brain from anterior (A) to posterior (P). S, superior; I, inferior. For abbreviations in B, see Table 2.

Figure 3.

Network clustering in resting brain of OA patients predicts subsequent psychologically induced analgesia. A, Mean clustering coefficients (CCoef) and local efficiency (LocEff) significantly correlates with percentage pain decrease at different network thresholds (FDR corrected for multiple comparisons q < 0.05). Mean characteristic path length (Plength) and mean global efficiency (GloEff) did not show a significant correlation after correction. B, Correlation matrix (n × n) for the 125 brain regions (binarized at T = 0.25) visually shows greater clustering in brain networks of responders (RSP) than in nonresponders (NRSP). The two patient groups were derived based on a median split of analgesic response to psychological conditioning (percentage decrease in pain). The ordering of the 125 regions is the same as in Table 2. C, Regression plots with prediction intervals show that the relationship between mean clustering coefficients (measured at T = 0.25) and analgesia was not affected by the type of treatment (sham vs real). D, Relation between mean clustering coefficients and percentage pain decrease was significant when networks were thresholded using a constant link density of 0.25 with no significant interaction for sham versus real acupuncture. *p < 0.05.

Figure 4.

Sex difference in brain network properties and analgesic response to psychological conditioning. A, Psychologically induced analgesia was significantly greater in women than in men. B, Women suffering with OA showed significantly greater clustering coefficients than men at most network thresholds. C, Circle diagrams showing network connectivity in women and men at T = 0.6, *p < 0.05. Error bars indicate ± SEM. For abbreviations, see Table 2.

Figure 5.

Brain regions with topological properties associated with analgesic response to psychological conditioning. A, Brain networks that showed a significant relationship between clustering coefficient and percentage pain decrease in regions represented with randomly colored circles (FDR corrected T = 0.25). Regions involved in cognitive modulation of pain and emotion (SFG, superior frontal gyrus; DMPFC, dorsal medial prefrontal cortex; ACC, anterior cingulate–caudal and subgenual), motivational brain circuitry (MPFC, medial prefrontal cortex; NAc, nucleus accumbens, putamen, globus pallidus), memory (MT, temporal regions; Cingp, posterior cingulate; Hip, hippocampus), and visual circuitry showed higher clustering coefficients in patients that later showed greater psychologically conditioned analgesia. B, The centrality (degree) in regions shown in randomly colored circles was positively correlated with greater analgesic response to conditioning (FDR correction T = 0.25). These regions included bilateral MPFC regions (dorsal, ventral, and middle), the hippocampus (Hipp L), the left frontal pole (FPL), and the OFP. C, The right ventral MPFC (vMPFC, shown in red) showed greater functional connectivity in relation with greater analgesic response in regions shown in yellow including bilateral ACC caudal regions (ACCcaud), right middle frontal gyrus (MFG), right thalamus (Thal R), left hippocampus (Hipp L), left parahippocampal gyrus anterior division (paraHippaL), and left middle temporal gyrus anterior (MTGaL). For other abbreviations, see Table 2.