Phenotypic Classification of Low Back Pain in Rheumatic and Spinal Disorders: A Data-Driven Approach Using Latent Tree Analysis.
Source: PubMed, NCBI / U.S. National Library of Medicine
To examine the empirical correspondence between data-driven pain phenotypes and classical Traditional Medicine (TM) syndromes in a cohort with specific low back pain (LBP). We aimed to identify distinct phenotypes associated with musculoskeletal and spinal disorders using Latent Tree Models (LTM) and evaluate their alignment with traditional diagnostic constructs. This cross-sectional study was conducted at a tertiary specialty hospital in Ho Chi Minh City, Vietnam, recruiting 260 patients with specific LBP resulting from identifiable musculoskeletal and spinal disorders. The LTM was used to identify pain phenotypes from clinical symptom data. Risk factors for each phenotype were analyzed by multivariable logistic regression. The LTM identified three primary phenotypes. A "Stiffness/Cold/Heavy Pain" phenotype was strongly associated with radiating pain and a diagnosis of herniated disc. A "Dull, Localized Pain" phenotype was associated with bilateral pain. A "Sharp, Stabbing Pain" phenotype was strongly associated with a history of smoking. These data-driven clusters showed a clear alignment with the classical TM syndromes of Cold-Dampness, Kidney Deficiency, and Blood Stasis, respectively. LTM is an effective tool for identifying data-verifiable clusters that correspond to TM syndromes in patients with specific LBP. Linking these phenotypes to clinical risk factors and pathoanatomical diagnoses provides an evidence-based framework for stratifying heterogeneous LBP related to
Abstract
To examine the empirical correspondence between data-driven pain phenotypes and classical Traditional Medicine (TM) syndromes in a cohort with specific low back pain (LBP). We aimed to identify distinct phenotypes associated with musculoskeletal and spinal disorders using Latent Tree Models (LTM) and evaluate their alignment with traditional diagnostic constructs. This cross-sectional study was conducted at a tertiary specialty hospital in Ho Chi Minh City, Vietnam, recruiting 260 patients with specific LBP resulting from identifiable musculoskeletal and spinal disorders. The LTM was used to identify pain phenotypes from clinical symptom data. Risk factors for each phenotype were analyzed by multivariable logistic regression. The LTM identified three primary phenotypes. A "Stiffness/Cold/Heavy Pain" phenotype was strongly associated with radiating pain and a diagnosis of herniated disc. A "Dull, Localized Pain" phenotype was associated with bilateral pain. A "Sharp, Stabbing Pain" phenotype was strongly associated with a history of smoking. These data-driven clusters showed a clear alignment with the classical TM syndromes of Cold-Dampness, Kidney Deficiency, and Blood Stasis, respectively. LTM is an effective tool for identifying data-verifiable clusters that correspond to TM syndromes in patients with specific LBP. Linking these phenotypes to clinical risk factors and pathoanatomical diagnoses provides an evidence-based framework for stratifying heterogeneous LBP related to spinal disorders. This approach offers a potential tool to move toward mechanism-based therapeutic strategies, bridging traditional observation with modern data science.