Limited predictive utility of spinal cord stimulation trial pain metrics for post-implant outcomes.

Abstract

Background: This study evaluated the predictive value of pain outcomes reported during spinal cord stimulation (SCS) trialing for short-term, intermediate-term, and long-term post-implantation pain relief, and whether baseline patient characteristics modify these relationships.

Methods: A retrospective review identified 124 patients who underwent SCS trial and implantation between 2015 and 2022 with a minimum of 6-months follow-up. Pain outcomes included numeric rating scale pain score (NRS), patient-reported percent pain relief (PR-PPR), and calculated percent pain relief (C-PPR) post-trial and at short-term (ST) 1-6 months (>1 to ≤6 months), intermediate term (IT) 6-15 months (>6 to ≤15 months), long-term (LT) 15-24 months (>15 to ≤24 months), and extended long-term (ELT) 24-36 months (>24 to ≤36 months) post-implantation. Univariable linear and logistic regression assessed predictive value, while Bland-Altman analysis and Lin's concordance correlation coefficient (CCC) evaluated agreement between PR-PPR and C-PPR. Interaction modeling assessed effect modification of baseline covariates.

Results: Patients experienced significant pain reduction following trial (∆NRS 4.7; 95% CI 4.3 to 5.2), with sustained, but attenuating, benefit through 36 months. ∆NRS demonstrated the strongest predictive value (peak R²=0.32 at 15-24 months), though overall explanatory power was modest. In adjusted models, post-trial NRS was the strongest predictor of achieving ≥50% long-term pain relief (OR 0.70, p=0.002). Bland-Altman analysis showed a mean difference of 10 percentage points between PR-PPR and C-PPR (95% limits of agreement -36% to +56%) with poor overall agreement (CCC=0.36). Greater than 50% PR-PPR was maintained in 81.5% at 1-6 months and 70.2% at 24-36 months, with no significant differences between high and moderate trial responders. Covariate interaction modeling did not identify any consistent patient-specific modifiers.

Conclusions: Trial-phase pain metrics demonstrated limited utility in explaining short, intermediate, and long-term pain relief. NRS-based measures showed the most consistent predictive value, while disagreement between PR-PPR and C-PPR raises concerns regarding the reproducibility of percentage-based metrics. These findings emphasize the need for refinement of reliable SCS trial predictors for post-implant pain relief outcomes.