How Fast Your Pain Map Shrinks After Knee Surgery May Predict Recovery

Total knee replacement is one of the most common and most successful operations in modern medicine, yet for a stubborn minority of patients the promise of pain relief never fully arrives. Roughly one in seven patients still reports moderate-to-severe knee pain a full year after surgery, a problem that has frustrated surgeons because it is difficult to predict from preoperative factors alone. Now a new study offers an unexpectedly simple clue: the speed at which a patient’s overall body map of pain shrinks in the first three months after surgery may reveal who is on the road to chronic pain and who is recovering well. The research, published in Frontiers in Pain Research, suggests that how pain changes over time carries information that no single snapshot can capture.

The study, led by Jun Liu, Wenheng Chen and Shan Gao of Zhumadian Central Hospital in China, followed 624 patients who underwent primary unilateral total knee arthroplasty, the technical term for a complete artificial knee replacement. Instead of relying only on pain intensity scores, the researchers asked patients to complete whole-body pain drawings at two weeks, six weeks, and three months after their operation. On these standardized body silhouettes, patients shaded every region where they had felt pain during the previous week. A semiautomated image-processing pipeline built on Python, OpenCV, SimpleITK and scikit-image then scanned, aligned and quantified these drawings, converting the shaded areas into a single number called the Pain Extent Index, or PEI: the percentage of body surface area marked as painful.

The technical machinery behind this measurement was rigorous. Paper drawings were scanned at high resolution and electronic drawings were exported as lossless image files, then corrected for orientation, stripped of stray annotations, and registered to a standardized anatomical template using affine and nonrigid registration. Every processed map passed through structured quality control, with flagged images independently reviewed and resolved by consensus. Reproducibility testing on a 20 percent sample of patients produced strikingly high agreement: the inter-rater intraclass correlation coefficient for the PEI was 0.987, and the median Dice coefficient, a measure of overlap between independently processed pain masks, was 0.949. In other words, two different analysts processing the same drawing arrived at nearly identical maps of where the patient hurt.

With the PEI quantified at each visit, the team calculated a patient-specific contraction velocity, essentially the rate at which the spatial footprint of pain shrank between two weeks and three months. This was done with linear mixed-effects modeling, a statistical technique that estimates both a population-average trajectory and an individual slope for each patient. Across the whole cohort, the median contraction rate was about 0.065 log units per week, meaning most patients’ pain maps contracted substantially during the subacute phase. But the spread was wide, and that spread turned out to matter enormously. Patients who went on to report moderate-to-severe knee pain at twelve months contracted at only about half the rate of those who recovered well, and roughly one in five of them actually showed spatial pain expansion rather than contraction.

The headline result is statistically robust. At twelve months, 92 patients, or 14.7 percent of the cohort, met the strict definition of persistent moderate-to-severe operated-knee pain, which required a pain score of at least four out of ten, pain on at least four days per week, and continuous or recurrent pain lasting at least three months. Each standard deviation of slower spatial contraction was associated with 42 percent higher odds of falling into this group, even after the researchers adjusted for age, sex, body mass index, comorbidities, preoperative pain intensity, preoperative pain extent, psychological factors such as catastrophizing and anxiety, neuropathic pain features, and crucially, the static pain measurements taken at both ends of the recovery window. That final adjustment is the study’s key innovation: knowing where a patient’s pain stood at two weeks and at three months still did not explain away the signal carried by the rate of change itself.

The pattern held up under an unusually thorough battery of sensitivity analyses. The association persisted when the exposure was redefined as absolute PEI change, relative contraction rate, time-weighted pain-area burden, or slopes restricted to the operated limb or the knee alone. A Bayesian joint model that explicitly accounted for uncertainty in the estimated slopes yielded a similar result, with a posterior odds ratio of 2.05. When the researchers excluded the twelve patients whose twelve-month pain had an adjudicated alternative explanation, such as infection or a mechanical prosthesis problem, the association actually strengthened. Inverse-probability weighting to address missing data, and separate analyses of paper versus electronic drawings, all produced consistent estimates, suggesting the finding is not an artifact of one analytical choice.

Just as striking were the secondary outcomes. Slower spatial contraction predicted higher knee pain scores at both six and twelve months, lower scores on the KOOS-JR joint-replacement quality-of-life questionnaire, and lower Forgotten Joint Scores, which measure how completely patients stop noticing their artificial joint in daily life. Each standard deviation of slower contraction was associated with a 4.6-point drop in KOOS-JR and a 6.3-point drop in the Forgotten Joint Score, and it raised the odds of failing to achieve a clinically meaningful improvement in joint function by 46 percent. The population-level heat maps tell the story visually: in patients who recovered well, pain density around the knee swelled briefly after surgery and then collapsed toward a low plateau, while in those with persistent pain the map remained stubbornly wide from three months onward.

Why should the spatial spread of pain carry prognostic power that intensity scores lack? The authors are careful not to overclaim a mechanism, but they outline several plausible biological pathways. A pain map that refuses to shrink could reflect persistent peripheral input from the healing joint, such as ongoing inflammation or synovitis. It could indicate guarding and altered movement patterns that spread mechanical load to the thigh, lower leg and contralateral side. It could also signal central sensitization, a state in which the nervous system itself amplifies and generalizes pain signals, a phenomenon long implicated in poor arthroplasty outcomes and connected to psychological distress. A whole-body slope cannot distinguish these mechanisms, but it may integrate them into a single, clinically readable signal of whether recovery is genuinely progressing.

The clinical implications are tantalizing but deliberately cautious. Serial pain drawings are cheap, quick and, as this study demonstrates, remarkably reproducible whether collected on paper or digitally, making them a plausible addition to routine follow-up visits or even remote monitoring programs. A patient whose pain map is contracting slowly at the six-week check could prompt clinicians to look harder for wound complications, prosthesis problems, stiffness, neuropathic features, sleep disturbance or rehabilitation barriers, rather than simply waiting for the twelve-month appointment to reveal a bad outcome. The authors stress that a slow contraction rate is a signal to investigate, not a diagnosis, and certainly not an indication for automatic imaging or revision surgery, since revision outcomes vary widely depending on the underlying cause of pain.

Important limitations temper the excitement. This was a single-center retrospective analysis of prospectively collected data, so residual confounding cannot be excluded, and the cohort may not generalize to other populations or healthcare systems. Summarizing recovery with a linear slope from three time points inevitably smooths over nonlinear dynamics, and pain drawings ultimately depend on patient comprehension and recall. The study also cannot establish causality: slow contraction may be a cause of chronic pain, a consequence of an underlying pathological process, or simply another face of the same phenomenon. What it does establish is that the direction and speed of spatial pain recovery carry real prognostic information beyond anything measured at a single visit. If future multicenter studies validate the approach and demonstrate that acting on an early contraction alert improves outcomes, the humble body diagram, a tool as old as clinical medicine itself, could become a powerful digital window into how recovery is truly unfolding.

Subject of Research: Association between early postoperative contraction of spatial pain extent and persistent moderate-to-severe knee pain 12 months after total knee arthroplasty

Article Title: Early spatial pain contraction and moderate-to-severe knee pain 12 months after total knee arthroplasty: a retrospective cohort study

Article References: Liu, J., Chen, W., & Gao, S. (2026). Early spatial pain contraction and moderate-to-severe knee pain 12 months after total knee arthroplasty: a retrospective cohort study. Frontiers in Pain Research, 7, Article 1948322. https://doi.org/10.3389/fpain.2026.1948322

Image Credits: AI Generated

DOI: 10.3389/fpain.2026.1948322

Keywords: total knee arthroplasty, chronic postsurgical pain, pain extent index, pain drawings, spatial pain contraction, subacute recovery, central sensitization, KOOS-JR, prognostic markers, knee osteoarthritis, pain trajectories, mixed-effects modeling