Weekly MSK Literature Review
Week of September 21 to 16, 2026
Monday: AI-Based Patient Digital Twins in Rehabilitation Decision Support
The question: can digital twins move rehab from protocol-driven care to continuously updated, individualized decision support, and what stands between the concept and the clinic?
Rehab needs its own digital twin model. The authors separate general medical twins from rehabilitation twins, a distinction that must capture biomechanical, neurological, and functional dimensions.
The promised value is fewer therapy dead ends. Twins are positioned to simulate scenarios, predict outcomes, and reduce trial-and-error in therapy planning.
The data burden is heavy. A working twin draws on EHRs, motion capture, wearables, EMG, imaging, gait labs, environmental sensors, and population datasets.
Validation is thin. The authors note that few studies rigorously test twins in real-world clinical settings, and their gap table also flags multimodal integration, real-time personalization, scalability, and ethics and privacy.
Rehab is not acute care. Long timelines and heavy reliance on patient engagement mean that twins must sit within the therapist-patient relationship rather than act as standalone decision-makers.
Why it matters: Any vendor pitching predictive personalization for PT should be asked for prospective outcome data and a workflow integration plan, because this review shows the field has neither at scale. The opportunity is real, but the buyer’s leverage right now is in demanding validation before adoption.
Mikołajewska, E., Masiak, J., & Mikołajewski, D. (2024). Applications of artificial intelligence-based patient digital twins in decision support in rehabilitation and physical therapy. Electronics, 13(24), 4994. https://doi.org/10.3390/electronics13244994
Tuesday: Trajectories of Disability in Low Back Pain
The question: do patients with low back pain follow distinct disability paths over 12 months, and which baseline psychological factors sort them?
In a Danish cohort (N = 1048; analytic sample = 747), latent growth mixture modeling of RMDQ-23 scores at baseline, 6 months, and 12 months produced four trajectories.
High-stable (22%) carried severe disability with minimal improvement. Medium-stable (29.7%) held moderate disability with little change.
High-decreasing (20.4%) started at a severe level and improved by roughly 75%. Low-decreasing (27.9%) started mild and kept improving.
Baseline pain intensity, depressive symptoms, and pain catastrophizing predicted membership in all three symptomatic trajectories. Pain intensity combined with high depressive symptoms was especially important for the high-stable group.
Why it matters: Severe baseline disability does not predict a poor outcome, since one in five patients starts severe and improves substantially. Hence, the screening question is psychological load rather than starting score. Early depression and catastrophizing screening is what separates patients who need a different intervention from those who will improve with standard care.
Andersen, T. E., Karstoft, K.-I., Lauridsen, H. H., & Manniche, C. (2022). Trajectories of disability in low back pain. PAIN Reports, 7(6), e985. https://doi.org/10.1097/PR9.0000000000000985
Wednesday: Shared Pain and Theory of Mind Networks in Processing Others’ Suffering
The question: does the brain process another person’s physical pain and emotional suffering through the same machinery?
It does not. Using fMRI with 144 narrative scenarios across six conditions, along with pain and theory-of-mind localizers, the authors mapped two networks.
Physical pain stories activated the shared pain network, including the left insula, cingulate cortex, and secondary sensory regions. Verbal narratives were enough to do this, without images.
Emotional suffering stories engaged the theory-of-mind network: dmPFC, TPJ, and precuneus. Stories of distress caused by a false belief resembled emotional pain rather than physical pain.
The left insula responded selectively to physical pain and the dmPFC selectively to emotional pain. The aMCC and anterior thalamus responded to both.
Rating pain and actively empathizing produced comparable activation patterns, so the distinctions held across task instructions.
Why it matters: Recognizing a patient’s physical pain and understanding their emotional suffering appear to be separable cognitive tasks, which gives clinical communication training a neural rationale for explicitly teaching perspective-taking rather than assuming it follows from clinical experience.
Bruneau, E. G., Pluta, A., & Saxe, R. (2012). Distinct roles of the “Shared Pain” and “Theory of Mind” networks in processing others’ emotional suffering. Neuropsychologia, 50(2), 219–231.
Thursday: Reattribution to Mind-Brain Processes and Recovery From Chronic Back Pain
The question: does changing what patients believe causes their pain explain why pain reprocessing therapy works?
Get Rehab Think Tank’s stories in your inbox
Join Medium for free to get updates from this writer.
This secondary analysis of an RCT (151 adults with primary chronic back pain; PRT, open-label placebo injection, or usual care) coded patients’ free-text top three pain causes.
Mind or brain attributions rose from 10% of attributions at baseline to 51% after PRT, with effect sizes of g = 1.95 to 2.06.
Greater increases in mind-brain attribution were associated with lower pain intensity (B = -0.25; P = .04).
Reattribution partially mediated PRT’s effect on pain intensity at one year (B = -0.35).
Natural language processing recovered a principal dimension separating mind and brain explanations from biomechanical ones. Agreement with human coders was moderate before treatment and stronger after (Cohen kappa = 0.42 and 0.68).
Why it matters: This is a mechanistic signal, not just an efficacy signal, and it makes patient explanatory beliefs a measurable treatment target. Automated scoring of free-text attributions is a plausible EMR feature, though the pretreatment agreement of 0.42 says it is not ready to drive decisions alone.
Ashar, Y. K., Lumley, M. A., Perlis, R. H., Liston, C., Gunning, F. M., & Wager, T. D. (2023). Reattribution to mind-brain processes and recovery from chronic back pain: A secondary analysis of a randomized clinical trial. JAMA Network Open, 6(9), e2333846. https://doi.org/10.1001/jamanetworkopen.2023.33846
Friday: Pain Neuroscience Education Combined With Physical Therapy
The question: does adding pain neuroscience education to physical therapy improve pain and disability across chronic pain conditions?
This systematic review and meta-analysis pooled data from 19 RCTs (n = 693) across low back pain, neck pain, TMD, cervicogenic headache, migraine, knee OA, and carpal tunnel syndrome.
Mean pain scores fell from 5.89 to 3.03 after intervention, a reduction of roughly 49%.
Disability measures, including ODI, RMDQ, NDI, and DASH, improved across conditions.
PNE, when paired with therapeutic exercise, was reported in 52.6% of included studies and yielded the most consistent outcomes.
Heterogeneity was substantial. Intervention length ran from 2 to 24 weeks, PNE sessions from 1 to 12, and physical therapy modalities and sociocultural contexts varied. The authors call for cautious interpretation.
Why it matters: PNE belongs inside an exercise-based plan of care rather than as a standalone talk, and the field needs a standardized dose before payers or health systems can build it into a reimbursable pathway. The pre-post pain change is a descriptive figure across mixed trials, so it should not be quoted as a treatment effect.
Sánchez-Robalino, A., Sinchi-Sinchi, H., & Ramírez, A. (2025). Effectiveness of pain neuroscience education in physical therapy: A systematic review and meta-analysis. Brain Sciences, 15(6), 658. https://doi.org/10.3390/brainsci15060658
This Week at a Glance
Psychological load drives persistence: Andersen showed that depression, catastrophizing, and pain intensity sort patients into stable versus improving disability paths. Ashar approached the same territory from the treatment side, showing that shifts in beliefs about pain affect outcomes.
Belief change is a mechanism: Ashar found reattribution mediated one-year pain outcomes after PRT. Sánchez-Robalino’s pooled PNE data point in the same direction at the population level, though with far less mechanistic precision.
Education works best with movement: Sánchez-Robalino found the strongest and most consistent results when PNE was paired with therapeutic exercise. The implication is that reconceptualizing pain and testing that new belief through activity are complementary steps.
Empathy has distinct components: Bruneau separated the networks for sensing another’s physical pain and understanding their emotional suffering. Clinicians who treat both as one skill may be under-training the second.
Infrastructure lags the science: Mikołajewska describes twins that could personalize rehab in real time, but validation, workflow fit, and data integration remain unsolved. The psychological predictors in Andersen and the attribution scoring in Ashar are the kinds of inputs those systems would need.
What This Means in Practice
Clinicians: Screen for depression and catastrophizing at intake, and ask patients in their own words what they believe is causing their pain.
Clinical educators: Teach perspective-taking as a distinct skill alongside pain neuroscience content, and pair every educational session with graded activity.
EMR developers: Build structured fields for psychological screening and patient-reported pain attributions, and treat NLP scoring as decision support to be confirmed by a clinician.
Value-based care strategists: Use trajectory logic to target additional psychological resources to high-risk patients rather than applying them to everyone.
Digital health buyers require evidence of prospective validation and workflow integration before purchasing any predictive or twin-based platform.
Bottom line: The field has converged on patient beliefs and psychological risk as modifiable drivers of chronic pain outcomes, but it still lacks standardized dosing, validated tools, and the data infrastructure to deliver that care at scale. Closing that implementation gap is now a bigger opportunity than producing another efficacy trial.
