Quarterly Sleep-Pain Coupling in Knee Pain: Pain-to-Sleep Dominance, Localization Dependence, and Reward-Circuit Moderation of Sleep-to-Pain Coupling

Sleep-pain coupling is bidirectional, with consistent day-to-day evidence of asymmetry favoring sleep-to-pain effects, yet within-person dynamics at longer timescales remain poorly understood, despite directionality shaping which symptom clinicians should target first. We examined sleep-pain coupling quarterly over three years in 229 adults. Numerical rating scales captured past-week self-reported sleep quality and eight self-reported pain-related items-the latter were factor-analyzed into general severity and a knee-versus-body contrast. A Bayesian bivariate autoregression with correlated innovations and person-specific slopes was fit to within-person deviations. Pain-to-sleep coupling, but not sleep-to-pain coupling, was credible (95% credible interval excluding zero)-stronger self-reported pain was associated with next-quarter worse sleep-reversing the typical day-to-day asymmetry. Quarters in which pain was more knee-dominant showed stronger pain-to-sleep coupling. We also tested, under a theory-driven approach, neuroimaging-based predisposition markers of these couplings. A prior study identified brain regions, including the nucleus accumbens (NAcc), where pain-evoked activation tracked sleep-deprivation-induced next-day pain amplification. Preclinical evidence also implicates the anterior cingulate cortex (ACC) and NAcc as parallel dopaminergic-gated nodes. We therefore tested, in a subset (N=174), whether baseline pain-evoked fMRI activation in these regions moderated sleep-to-pain coupling as a possible trait-level marker. The left NAcc and the ACCs credibly moderated sleep-to-pain coupling, which were stronger at greater deactivation and credible in nearly all deactivating individuals. In the pain-to-sleep direction, testing five preclinical pain-arousal relay nodes (fMRI and gray matter volume) as moderators yielded no credible effect. PERSPECTIVE: In knee osteoarthritis, worse pain predicts worse sleep months later, but sleep does not predict pain, reversing the usual day-to-day asymmetry. Treating pain may more directly improve sleep, while baseline pain-evoked deactivation in reward and cingulate regions may identify whose pain instead follows sleep. Future trials should test both.