Brain inflammation and its predictive value for post-operative pain in total knee arthroplasty patients

Abstract

Recent evidence suggests that chronic pain patients exhibit elevated brain levels of the neuroinflammation marker 18 kDa translocator protein (TSPO). However, the clinical significance of brain TSPO elevations, and their responses to pain interventions, remain unknown. To explore these questions, we studied patients with knee osteoarthritis (KOA) undergoing total knee arthroplasty (TKA), a procedure which is curative for most, but carries a relatively high risk of persistent post-surgical pain. Pre-surgical KOA patients (n = 41) and healthy controls (n = 22) underwent brain positron emission tomography/magnetic resonance imaging, using the TSPO radioligand [¹¹C]PBR28. A subset of KOA patients (n = 27) returned for a second scan one-year post-TKA. When compared groups, pre-surgical KOA patients exhibited widespread [¹¹C]PBR28 PET signal elevations (Standardized Uptake Value Ratio), with pituitary uptake positively correlating with knee pain severity (rho = 0.51; p = 0.003). A voxel-wise paired t-test revealed that while most brain regions showed no change post-surgery, the [¹¹C]PBR28 PET signal significantly decreased in the thalamus and caudate, reaching control levels. Additionally, a Support Vector Machine model based on pre-surgical imaging, clinical, and demographic features, achieved a correlation of rho = 0.487 (p = 0.001) between the predicted and actual pain improvement. Top predictive features included [¹¹C]PBR28 uptake in the pituitary gland, cuneal cortex, amygdala and other regions. This study suggests that neuroinflammation 1) is widespread in KOA and, in some regions, 2) is linked to pain severity, 3) undergoes normalization following TKA, and 4) can predict post-surgical TKA outcomes. Understanding the neuroinflammatory mechanisms in KOA and post-surgical pain may guide targeted interventions and improve patient outcomes.

Keywords: Knee arthritis; Neuroinflammation; PET; Prediction; Total knee arthroplasty.

Figure 1:. Voxel-wise group differences in [¹¹C]PBR28 PET signal.

(A) Mean [¹¹C]PBR28 SUVR in KOA patients (top panel) and controls (bottom panel). (B) Maps displaying areas with significantly elevated [¹¹C]PBR28 SUVR in KOA compared to controls in a voxel-wise analysis, adjusted for BMI, genotype and age. (C) Average + SD SUVR extracted from several subclusters identified as statistically significant in the voxel-wise PET analysis from A (adjusted for BMI, genotype and age). The range of the y-axis is set depending on the distribution of individual data-points. PCC = posterior cingulate cortex, pMCC = posterior mid cingulate cortex, ACC = anterior cingulate cortex, WM = white matter.

Figure 2:. [¹¹C]PBR28 SUVR correlated with WOMAC pain scores.

The SUVR from pituitary gland, which showed elevated [¹¹C]PBR28 PET signal in figure 1, was extracted and plotted against WOMAC pain severity scores (adjusted for BMI, genotype and age).

Figure 3:

Post-TKA reductions in pain and PET signal. (A) WOMAC pain severity scores before and after TKA, demonstrating a significant reduction in pain severity. (B) Map displaying regions from the voxel-wise paired t-test results that exhibit a significant decrease in PET signal one year after TKA. (C) Extracted PET signal from (B), plotted for pre-TKA, post-TKA and controls, showing that post-TKA PET signal returned to control levels.

Figure 4:. ML-based prediction modeling of pain improvement one year post-TKA using [¹¹C]PBR28 regional uptake.

(A) True vs. predicted pain improvement, where each point represents the model's output for each test subject in the leave-one-out cross-validation loop. (B) Top-ranked features visualized on the MNI standard space, with the color scale representing their average weight across all iterations. (C) Mean absolute weights for each feature from all training loops, presented in descending order, with positive weights in dark gray and negative weights in light gray. Positive and negative weights indicate that higher values in that region are associated with greater and lower pain relief, respectively (assuming all other weights related to other regions remain constant). Unless right or left laterality is specified, labels are bilateral as per the Harvard-Oxford atlas.