A Clinical Prediction Model for Identifying a Nociplastic Pain Component: Development, Internal Validation, and Evaluation in a Tertiary Chronic Pain Cohort

Abstract

This project will develop and internally validate a clinical prediction model that estimates the probability of a nociplastic pain component in adults with chronic pain. Nociplastic pain describes pain arising from altered nociception without clear evidence of tissue damage or a lesion or disease of the somatosensory system sufficient to explain it. Recently published clinical criteria (Kosek et al., 2021) require a multidomain assessment, yet few studies have tested how well such features distinguish nociplastic pain from nociceptive and neuropathic pain in large, mechanistically mixed populations, and none has provided an individual patient’s probability together with an evaluation of calibration and clinical utility. The study will use existing data from consecutive adults assessed at a tertiary chronic pain clinic in Toronto, Canada. The outcome will be a physician-assigned nociplastic component (predominant or mixed with nociceptive or neuropathic pain) versus none. Twenty-two candidate scores representing the Kosek clinical domains (pain distribution, nociceptive and neuropathic features, history of pain hypersensitivity, associated symptoms, and mechanical sensitivity measured by pressure algometry) will be reduced in three prespecified phases using penalized (ridge) logistic regression, with every clinical domain required to remain represented. Model development will take place in a 75% development sample. Performance will be evaluated in a 25% held-out test sample, by nested cross-validation of the full selection procedure, and by bootstrap validation of the entire modelling procedure. Outcomes will include discrimination, calibration, classification accuracy, and net benefit from decision curve analysis, together with comparisons against alternative algorithms, alternative instruments, and each candidate input alone. Reporting will follow TRIPOD+AI. The expected outputs are a fully specified, parsimonious model with its equation and classification cutoff, internal estimates of its performance and selection stability, and an assessment of how far a comprehensive assessment can be shortened while preserving accuracy. The model is intended to help clinicians identify patients who warrant fuller mechanism-based assessment. External validation will be required before clinical use.