Pain Characterization,Identification,and Clustering Of Patients With Chronic Low Back Pain Using Principal Component Factor And Cluster Analysis

Objective:Chronic low back pain(CLBP)significantly impacts human quality of life and its clinical outcomes are often unsatisfactory due to its heterogeneity and various reasons.Thus,hierarchical management of CLBP is crucial for early etiological diagnosis and subsequent treatment selection.Neuropathic low back pain(NLBP),as a special subset,is a common clinical challenge in accurate identification.While research on NLBP evaluation has yielded fruitful results,there remains a lack of effective and simple methods for rapid and precise identification.This study combines the subjectively reported short-form of McGill Pain Questionnaire-2(SF-MPQ-2)and objectively measured infrared thermal imaging to explore CLBP subgroups through K-Means clustering,with the goal of providing a new approach for fast identification and hierarchical management of neuropathic pain of CLBP.Methods:From 2021 to 2022,CLBP patients who sought medical treatment at Zhujiang Hospital of Southern Medical University were included as study subjects.We collected sociodemographic data and evaluated the patients using a variety of clinical assessments,including the Brief Pain Inventory,DN4 scale,Oswestry Disability Index,Patient Health questionnaire-9,Generalized Anxiety Disorder-7,Pittsburgh Sleep Quality Index,SF-MPQ-2,and infrared thermal imaging.We used 18 pain descriptors from SF-MPQ-2 and the absolute value of the average temperature difference between the bilateral limbs of 8 body parts measured by infrared thermal imaging as characteristic variables.We reduced the dimension of data through principal component factor analysis to explore CLBP subgroups through KMeans clustering,and carried out subgroup validation through various assessment scales.Results:A total of 108 CLBP patients were included in the study,and we selected 26 characteristic variables for principal component analysis,obtaining nine principal components with characteristic values>1.The cumulative total variance explained was 65.86%,which was finally transformed into seven common factors by factor rotation.We then performed K-means clustering based on the values of the common factors,dividing the patients into two categories.The adaptive subgroup(n=70,64.8%)had low average body surface temperature differences and high continuous pain scores.The maladaptive subgroup(n=38,35.2%)had high average body surface temperature differences and high multidimensional pain scores.During subgroup validation we found significant differences between the subgroups(P<0.05)in 18 of the 26 characteristic variables.Compared to the adaptive subgroup,the maladaptive subgroup had a higher incidence of neuropathic pain(65.8%to 71.1%,P<0.05),higher temperature differences in the posterior thigh,posterior calf,and anterior calf(P<0.05),and higher multidimensional pain scores on the SF-MPQ-2,especially in the neuropathic pain dimension such as numbness,tingling,cold-freezing pain and hot-burning pain(P<0.001).In addition,the maladaptive subgroup had higher scores for functional impairment,anxiety,depression,and sleep disturbance(P<0.05),and higher scores for perceived pain and emotional interference(P<0.05).Conclusions:Using the SF-MPQ-2 and infrared thermography,we effectively classified CLBP patients by principal component factor analysis and unsupervised clustering.The integration of pain descriptors and body surface temperature as a novel pain characterization shows promise in identifying different CLBP.Our findings contribute to hierarchical management and personalized treatment with a novel approach.