Brain-to-brain mechanisms underlying pain empathy and social modulation of pain in the patient-clinician interaction

The study addresses how patient–clinician social interactions modulate pain experience and the underlying dynamic brain-to-brain mechanisms. Prior research indicates supportive presence can reduce pain, but mechanisms from both first-person (patient) and second-person (clinician) perspectives are often studied separately and in noninteractive designs. The authors hypothesize that the presence of a clinician reduces patients’ evoked pain intensity compared to isolation, that enhancing therapeutic alliance via a prior clinical intake improves clinicians’ empathy and understanding of patients’ pain, contributing to pain relief, and that these psychosocial effects are reflected in changes in patients’ pain-related brain processes and in increased brain-to-brain concordance between patients and clinicians.

The paper situates its work within several strands: meta-analyses and experimental studies on social support and pain show that supportive presence can reduce pain intensity or physiological arousal, though effects vary by relationship type and context. Clinical communication and empathy training have small-to-moderate beneficial effects on patient outcomes. Neuroimaging studies implicate networks including insula, prefrontal regions (dlPFC, vlPFC, vmPFC), and theory-of-mind circuits in social modulation of pain, placebo analgesia, and empathy for pain. Prior hyperscanning work by the authors observed increased patient–clinician concordance in social mirroring and theory-of-mind circuits when a clinical relationship was established. The literature also highlights the importance of sociocultural concordance (gender/ethnicity) in patient–clinician interactions and the role of synchronization in cooperative social behavior.

Design: fMRI hyperscanning of chronic pain patients (fibromyalgia) and licensed acupuncturists during live video-based dyadic interaction versus patient-alone control. Two factors: social presence (Dyadic vs Solo within visit) and clinical relationship (Clinical Interaction vs No Interaction across visits), with counterbalanced order and crossover to different partners to avoid carryover. Participants: 23 female fibromyalgia patients (final analyzable: 20 patients) and 22 acupuncturists (final analyzable: 20 clinicians), yielding 37 analyzable dyads (19 Clinical Interaction, 18 No Interaction). Prior clinical intake: In Clinical Interaction dyads, clinicians conducted a naturalistic intake (mean ~38 minutes) on a separate day to establish therapeutic alliance; No Interaction dyads had no intake and only brief introduction. Pain stimulation: Deep-tissue leg cuff pressure to the left lower leg (Hokanson Rapid Cuff Inflator). Individually calibrated Moderate pain (~40/100 VAS; mean ± SD pressure 157.92 ± 95.59 mmHg) and Nonpainful control (30 mmHg). Each run had six 15-s trials (3 Moderate, 3 Nonpainful) preceded by anticipation cues, with pseudorandomized sequence. Social manipulation: Within each MRI visit, patients underwent two runs: Solo (patient only, still image of clinician shown to control for visual input) and Dyadic (live two-way video stream; both patient and clinician scanned simultaneously). Instructions allowed nonverbal communication while minimizing motion. Behavioral measures: After each trial, patients rated pain intensity and clinician’s understanding (empathy); clinicians rated vicarious pain and empathy. Therapeutic alliance: CARE questionnaire completed by both parties at each session; dyad-wise alliance computed by averaging patient- and clinician-rated CARE. MRI acquisition: Simultaneous BOLD fMRI on Siemens 3T (Patient: Skyra; Clinician: Prisma). EPI parameters: TR 1,250 ms, TE 33 ms, flip 65°, 2 mm isotropic voxels, 75 slices, multiband 5, 624 volumes across two runs. Structural multi-echo MPRAGE for registration. Two MRI coils configured to allow facial video capture; synchronized two-way video (~20 Hz) over local network with <40 ms delay; custom software synchronized stimuli and acquisitions. Preprocessing: FSL v6.0.0—slice timing, motion correction (MCFLIRT), susceptibility correction (TOPUP), BET brain extraction, 4 mm FWHM smoothing, high-pass filter f = 0.011 Hz, grand-mean scaling; registration via bbregister (Freesurfer) to anatomy, FLIRT/FNIRT to MNI152. Motion exclusions applied for >2° rotation or >2 mm displacement. Statistical analyses: Self-report in R with α = 0.05. Repeated measures ANOVAs tested effects of Run (Solo vs Dyadic), Clinical context (Clinical Interaction vs No Interaction), Order, and Stimulus (Moderate vs No pain), plus correlations with CARE. fMRI: First-level GLMs (FILM with local autocorrelation) modeled Moderate pain vs No pain; rating periods and motion parameters as nuisance. Second-level fixed-effects contrasts derived Dyadic vs Solo per subject. Group-level (FSL FLAME 1+2) whole-brain analyses with cluster correction (z ≥ 3.1, α = 0.05) conducted for Clinical Interaction, No Interaction, and Clinical Interaction vs No Interaction. Brain-to-brain concordance: Identified patient pain-responsive ROI from independent Solo runs (Moderate > No pain), selecting contralateral S2 cluster (excluding patients with Moderate <70 mmHg; n = 16 for localizer). Extracted each patient’s S2 time series (Dyadic run) and used as regressor in clinician’s first-level GLM. Computed interaction between patient S2 time series (demeaned) and vicarious pain block regressor (centered; +1 pain, −1 nonpain). Resulting contrast maps reflected clinician voxelwise concordance with patient S2 during pain. Correlated magnitude (dlPFC cluster) with therapeutic alliance.

Pain intensity: Patients reported significantly lower pain during Dyadic versus Solo (Dyadic M = 29.43, SD = 17.69; Solo M = 34.27, SD = 19.81), F(1,181) = 9.26, P = 0.003, ηρ² = 0.05. No significant difference in pain intensity between Clinical Interaction (M = 34.07, SD = 19.11) and No Interaction (M = 29.04, SD = 18.33), F(1,181) = 2.44, P = 0.12, ηρ² = 0.01. Empathy ratings: Patients rated clinicians’ understanding higher in Clinical Interaction (M = 54.33, SD = 26.55) than No Interaction (M = 49.29, SD = 26.45), F(1,181) = 5.64, P = 0.019, ηρ² = 0.03; this correlated with therapeutic alliance (CARE), r = 0.42, P < 0.01. Agreement between patient pain and clinician vicarious pain: Greater agreement for Clinical Interaction (M = 86.98, SD = 10.09) vs No Interaction (M = 76.15, SD = 17.47); main effect of Clinical context F(1,34) = 5.254, P = 0.028, η² = 0.13; no Order effect; agreement correlated with alliance r = 0.41, P = 0.01. Patients’ brain responses: In Clinical Interaction dyads, Dyadic > Solo showed increased activation in left dpIns and m/aIns, bilateral S1 and S2, and bilateral dlPFC and right vlPFC; no significant Solo > Dyadic effects. No significant Dyadic vs Solo differences in No Interaction dyads. Direct contrast (Clinical Interaction vs No Interaction) of Dyadic–Solo revealed increased responses in vlPFC, S1, S2/dpIns, and dlPFC. dlPFC Z-stat increases (Dyadic–Solo) correlated positively with therapeutic alliance, r = 0.40, P = 0.01. Brain-to-brain concordance: Clinicians exhibited increased dlPFC concordance with patients’ contralateral S2 activity during pain for Clinical Interaction relative to No Interaction; the magnitude of S2–dlPFC concordance correlated with therapeutic alliance, r = 0.57, P < 0.001.

Findings demonstrate that interacting with a supportive clinician reduces patients’ evoked pain compared to isolation and that enhancing therapeutic alliance via prior clinical intake improves perceived empathy and clinician accuracy in estimating patient pain. Neural analyses indicate that social interaction engages patients’ nociceptive and modulatory circuits (dpIns, S1/S2) and prefrontal control regions (dlPFC, vlPFC) more strongly during Dyadic versus Solo, particularly when a prior clinical relationship exists. Critically, clinicians’ dlPFC activity dynamically couples with patients’ S2 during pain, and this concordance scales with therapeutic alliance, suggesting a two-brain mechanism where clinicians’ higher-order social-cognitive processing aligns with patients’ somatosensory pain processing. The involvement of dlPFC aligns with prior work on empathy for pain, theory of mind, placebo analgesia, and cooperative synchrony. While the presence of a supportive other can reduce pain, the specific quality of the relationship may influence neural and empathic processes even if direct effects on pain ratings between Clinical Interaction and No Interaction were not observed here. The results underscore the importance of the therapeutic relationship and suggest that boosting alliance can enhance brain-to-brain concordance during pain.

The study provides evidence for social modulation of pain within the patient–clinician interaction, demonstrating reduced pain with clinician presence, heightened patient activation in somatosensory and prefrontal circuits during dyadic interaction, and increased clinician–patient brain concordance (clinician dlPFC with patient S2) tied to therapeutic alliance. These results support a two-brain mechanism for pain empathy and supportive care and highlight neural targets (dlPFC, S2) and relational processes (alliance) that may be leveraged to improve pain outcomes. Future research should examine longitudinal development of alliance across repeated clinical visits, formally manipulate supportive behaviors (warmth, empathy, competence), match or examine effects of sociocultural concordance/discordance, incorporate modalities allowing assessment during verbal interaction (e.g., EEG, fNIRS), increase sample size and diversity (including more men with fibromyalgia), and explore how enhanced brain-to-brain concordance mediates clinical benefits.

Single intake session may not capture relationship development over time; sample had more socioculturally concordant dyads and could not be formally balanced, though proportions did not differ between Clinical Interaction and No Interaction groups; inability to assess brain activity during verbal communication due to fMRI motion constraints; limited sample size (n = 37 analyzable dyads) potentially reducing power and increasing type II error; predominantly female fibromyalgia patients, limiting generalizability and preventing assessment of sex/gender differences.