Special Poster Session 50th International Society for the Study of the Lumbar Spine Annual Meeting 2024

Introduction: Despite a growing rate of spine surgery in the elderly, and awareness of clinical prognostic factors, our ability to predict who will benefit and who will fail to respond to surgery remains poor. Approximately one third of patients have minimal change in their pain following lumbar spinal stenosis (LSS) surgery, creating a costly unmet need to identify patient biometrics that correlate with successful surgical outcomes. The purpose of this study was to identify preoperative serum epigenetic markers that are predictive of 1-year postoperative pain recovery following surgery for symptomatic LSS due to spine facet osteoarthritis (S-FOA).

Methods: We used a case-control approach to look for epigenetic signatures in bio-banked blood from 40 patients that either had a super response (>70% improvement of baseline pain scores at 12-months) or non-response to surgery (<30% change): 23 patients were categorized as super responders and 17 were non-responders. DNA methylation libraries and high throughput sequencing was performed. DNA methylation status was determined at 24,500 gene promoters in all patient samples. β-values, representing the ratio of methylated to total reads (β = methylated/(methylated + unmethylated)) averaged across all CpG’s (DNA methylation sites)  within a gene promoter region (defined as 2kb upstream, 1 kb downstream of transcript start site), were determined. Genes with significant differences in promoter methylation status between responders and non-responders were determined using Limma in R. Lasso logistic penalized regression was carried out using the R package ‘glmnet’ to identify candidate epigenetic polygenic risk scores (E-PRS) predictive of pain response. The 34 differentially methylated genes that had >1.5 fold differences in β-scores were used to identify biologically relevant pathways using the integrated pathway database pathDIP.

Results: We identified 210 differentially methylated candidate gene promoters (nominal uncorrected p-value <0.01) in responders relative to non-responders. Thirty-four of these candidate promoters had greater than 1.5-fold difference in methylation status in super- vs. non-responders, and, based on a literature search, twelve of these were associated with inflammatory or neurological processes or diseases. A very high degree of accuracy in differentiating pain response was achieved with candidate E-PRSs dependent on as few as 6 (AUC=0.957, accuracy 90.0%) or 11 (AUC=0.990, accuracy 92.5%) genes (Figure 1) . In comparison, the clinical factors model performed less well (77.5% accuracy). Focussing on the 34 genes that had >1.5-fold difference in methylation status, we performed a preliminary pathway enrichment analysis based on negatively and positively differentially methylated promoters. We identified 38 and 14 pathways respectively. Enriched pathways include inflammatory and neuronal signaling.

Discussion: We have identified biologically plausible candidate E-PRSs that may enable a personalized/precision medicine approach to accurately identify pain response following surgery for LSS due to facet OA. Currently, our ongoing efforts are focused on further validation of the E-PRS predictive ability in a larger cohort.