Medio‐dorsal thalamic dysconnectivity in chronic knee pain: A possible mechanism for negative affect and pain comorbidity

Abstract The reciprocal interaction between pain and negative affect is acknowledged but pain‐related alterations in brain circuits involved in this interaction, such as the mediodorsal thalamus (MDThal), still require a better understanding. We sought to investigate the relationship between MDThal circuitry, negative affect and pain severity in chronic musculoskeletal pain. For these analyses, participants with chronic knee pain (CKP, n = 74) and without (n = 36) completed magnetic resonance imaging scans and questionnaires. Seed‐based MDThal functional connectivity (FC) was compared between groups. Within CKP group, we assessed the interdependence of MDThal FC with negative affect. Finally, post hoc moderation analysis explored whether burden of pain influences affect‐related MDThal FC. The CKP group showed altered MDThal FC to hippocampus, ventromedial prefrontal cortex and subgenual anterior cingulate. Furthermore, in CKP group, MDThal connectivity correlated significantly with negative affect in several brain regions, most notably the medial prefrontal cortex, and this association was stronger with increasing pain burden and absent in pain‐free controls. In conclusion, we demonstrate mediodorsal thalamo‐cortical dysconnectivity in chronic pain with areas linked to mood disorders and associations of MDThal FC with negative affect. Moreover, burden of pain seems to enhance affect sensitivity of MDThal FC. These findings suggest mediodorsal thalamic network changes as possible drivers of the detrimental interplay between chronic pain and negative affect.


Supplementary Material Analyses overview and accompanying formulas
Imaging data were analysed with FSL.
For all fMRI analyses, subject-level analyses are required first. The corresponding formula for this manuscript can be depicted as: for each voxel i in each subject BOLDi = B_Thi*BOLDTh + B_Thi *CSF + B_Thi *WM + B0_0i Where Th : mediodorsal thalamus seed, CSF : cerebrospinal fluid, WM: white matter Mean CSF and WM time series per subject were used as confounding regressors.

The manuscript contains 3 main analyses:
1) A group comparison analysis on statistical significant differences in mdThal FC networks in pain participants and controls.
The analysis controlled for age, sex, and movement.
2) A regression analysis focusing on affect in the pain participants only: BOLDThgroup = B_Thi * affect+ B_Thi*age+ B_Thi *sex + B_Thi *FD + B0_2i The regression analysis is independent of the significant clusters for the group difference analysis

3) A moderation analysis for the interdependence of affect:
This analysis is based on clusters identified as significant from the regression analysis. Mean z values from significant clusters were entered as the focal independent variable (X), affect score as the outcome variable (Y), and ICOAP total score as moderating variable (W). This was performed with PROCESS v3.4 for SPSS (Hayes, 2017) and essentially requires 2 models: Affect = X + ICOAP + X*ICOAP + error (2) Figure S1: Mediodorsal thalamic nuclei (MDThal) seed region.
These maps are not suitable for drawing statistical inference but serve to enhance interpretation of the significant group differences by visual comparison to the statistically significant differences. Without such reference, a significant cluster in figure 1 (main manuscript) can mean either 1) one group has a higher mean FC value in this cluster than the other group, or 2) either group has a mean FC value in this cluster while the other group does not show any connection between the seed region and this cluster (i.e. the pattern only exists in one group), or 3) the significant cluster is based on negative connectivity and depicts the reverse of 1).
As example: The sgACC is one cluster that was found to be significantly different in the group comparison (figure 1 of main manuscript), is circled in yellow. This connection is negative (anticorrelation) in both groups while most clusters are based on positive connectivity. Mean z values of the ten largest positively correlated clusters plotted against affect score.

Medication effects
In consideration for potential medication effects, the partial correlation and moderation analysis was repeated excluding any patients who had either taken opioid medication or were prescribed antidepressant medication, resulting in a total of 51 patients. The partial correlation remained significant (r = .57, p < .001). To visualise whether there was any systematic effect of medication, we plotted using a colour-code of those on each type of medication ( Figure S5).

Figure S4. Regression analysis of the MD thalamus functional connectivity coloured according to medication type
Regression analysis of the MD thalamus functional connectivity related to affect score in chronic knee pain patients. Mean z values of the ten largest positively correlated clusters plotted against affect score and colour-coded according to medication type: blue: on antidepressant medication only; green: taken opioid medication only within the last 24 hours; orange: on antidepressant medication and taken opioid medication within the last 24 hours; red: all other patients.
The moderation analysis also remained significant with a significant moderating effect of ICOAP score on the strength of the relationship between the MD thalamus FC and affect score (interaction: b = .11, t(42) = 3.6, p < .001). This relationship was again significant in those with 1SD below the mean, at the mean, and 1SD above the mean for ICOAP score, with the effect being strongest in those with high ICOAP scores. The point at which this effect became significant and remained significant was at a Rasch-transformed ICOAP score of -2.78.