Reported neuroimaging findings in chronic pain are manifold but remarkably inconsistent. Brain network connectivity analysis arguably offers the most sensitive marker to detect dysfunctional brain plasticity underlying the maladaptive nature of chronic pain. Early functional connectivity (fc) studies reveal altered functional connectivity in chronic pain states, but to the best of our knowledge no studies have focussed upon the amygdala. The amygdala is a key region known to enhance pain sensitivity (e.g., fear, threat) with animal pain models recently showing the amygdala to be at the core of pain chronification mechanisms1,2. Support for an important role of this brain region also comes from emerging human chronic pain studies3,4. Altered amygdala FC has been found after successful pain intervention in complex paediatric pain syndrome5 and a mix of several musculoskeletal pain disorders6, but none of these studies were specific for the amygdala, its subnuclei or osteoarthritis. We aimed to investigate whether patients with painful chronic knee OA show altered amygdala connectivity compared to pain-free controls.

30 chronic knee osteoarthritis patients (mean age 67yrs, 15 males, mean pain rated before scans 27.5/100, mean pain duration 115 months) without other co-morbidities and 20 healthy volunteers (mean age 64yrs, 8 males) underwent phenotyping of negative affect and pain. As part of the multimodal MRI biomarker discovery study (3T, MR750, GE Healthcare) all patients underwent a resting state BOLD fMRI (GE-EPI: TE/TR=32/2000ms, 35 slices, voxel size=3.75x3.75mmx3.6mm,160 volumes) using a 32-channel head coil. Image analysis was performed using standard procedures in FSL 5.0.8 (FMRIB software library). Seed-based connectivity controlled for nuisance time-series regressors (CSF, white matter and 6 motion parameters) was calculated using 4 ROI’s (regions of interest) in each left and right amygdala (whole amygdala, superficial (SF), centromedial (CM) & basolateral (BL) nuclei) according to Roy et al7 and a control ROI in the primary visual cortex (V1) (3mm sphere diameter centred at x= -8 y= -96 z= 2 MNI).Between groups analyses were carried out using FLAME (a mixed-effects model within FSL; multiple-test corrected Z>2.3, cluster significance p<0.05) additionally controlled for between-subject motion.Relative to controls, OA patients displayed increased connectivity of the left CM to pre-and postcentral gyri and insula cortices (bilat.), increased connectivity of the right CM with right pre-and postcentral gyri and insula cortices, as well as increased connectivity of the left BL with the left middle frontal and inferior frontal gyri alongside frontal pole. No significant group differences in fc were found for bilateral whole amygdala, SF nuclei, right BL nuclei and the V1 control ROI. Preliminary covariate analysis further indicates that these alterations do no merely reflect low mood.This study identified increased functional connectivity of specific amygdala subnuclei in chronic OA pain patients compared to healthy subjects. These results indicate that pain-related changes would not have been uncovered if the amygdala had only been investigated as a single region as opposed to distinct nuclei. Given that findings focussed on the amygdala have to date been under reported in human pain research, this is a relevant point to consider in future analyses. These results also provide observational evidence on lateralization and suggest differential functions of amygdaloid nuclei in chronic OA pain. This is an important step in understanding developments in chronic pain better as literature on animal arthritis models hints at differential functions of the amygdaloid nuclei: anterior central amygdaloid neurons develop sensitization to pain disorder-related pain stimuli8, while the BL may have a more nonspecific action on pain as it has been shown “trigger pain behaviour in animals in the absence of nociceptive input”9 through stress hormones. The findings are promising to enable the discovery of a dysfunctional brain network signature of chronic pain. Additional graph analysis and effective connectivity analysis in relation to psychometry and pain phenotypes are underway to derive a network-based biomarker model of chronic pain.