Analysis of Risk Factors and Development of a Predictive Model for Early- versus Late-Stage Residual Low Back Pain Following Single-Level Transforaminal Lumbar Interbody Fusion

Abstract Objective To identify risk factors for residual low back pain (LBP) at early (1-month) and late (6-month) stages following single-level transforaminal lumbar interbody fusion (TLIF) and construct predictive models. Methods Clinical data from 359 patients undergoing single-level TLIF at Zhongda Hospital, Southeast University (Jan–Dec 2024) were retrospectively analyzed. Groups were stratified based on Visual Analog Scale (VAS) scores: Early-stage Group A (VAS ≤ 4, n = 281; 127M/154F; age 60.00 [53.00–68.00] years) and Group B (VAS > 4, n = 78; 20M/58F; age 62.00 [54.25–70.00] years); Late-stage Group C (VAS ≤ 4, n = 317; 145M/172F; age 60.00 [52.00–68.00] years) and Group D (VAS > 4, n = 42; 2M/40F; age 66.50 [61.00–71.75] years). The included assessment indicators encompassed demographics, preoperative and postoperative VAS (Pre-VAS, Post1m-VAS, Post6m-VAS), Patient Health Questionnaire-9 (Pre-PHQ-9, Post1m-PHQ-9, Post6m-PHQ-9), and Barthel Index (Pre-BI, Post1m-BI, Post6m-BI), paraspinal muscle lean cross-sectional area/gross cross-sectional area (LCSA/GCSA), facet joint degeneration (FJD), disc height index (DHI), bone metabolism markers (TP1NP, OSTEOC, 25(OH)D, CTX), hematologic parameters (RBC, HGB, Creatinine), bone mineral density (BMD), surgical variables and other relevant metrics. Univariate and multivariate logistic regression analyses were used to identify factors associated with postoperative residual LBP in patients undergoing single-level TLIF. Predictive models were then constructed and evaluated using the area under the ROC curve (AUC), sensitivity, and specificity to assess predictive value. Results Univariate analysis revealed early residual LBP associated with Pre-VAS, Pre-PHQ-9, Post1m-PHQ-9, Height, FJD, LCSA/GCSA, BMD, TP1NP, OSTEOC, 25(OH)D, CTX, RBC, HGB, Creatinine, Sex, Radicular Pain and Diabetes (P < 0.05); late residual LBP correlated with Age, Pre-VAS, Post1m-VAS, Pre-PHQ-9, Post1m-PHQ-9, Post6m-PHQ-9, Pre-BI, Post6m-BI, Height, Weight, FT, LCSA/GCSA, BMD, TP1NP, OSTEOC, 25(OH)D, CTX, RBC, HGB, HCT, Creatinine, Sex, Radicular Pain, Diabetes,Smoking and Alcohol (P < 0.05). Multivariate analysis identified protective factors for early residual LBP: Pre-VAS (OR = 0.426, 95%CI [0.25–0.73]; P = 0.002), Post1m-PHQ-9 (OR = 0.098, 95%CI [0.05–0.19]; P < 0.001), FJD (OR = 0.507, 95%CI [0.30–0.84]; P = 0.009); risk factors: Pre-PHQ-9 (OR = 2.459, 95%CI [1.36–4.45]; P = 0.003), female sex (OR = 4.824, 95%CI [1.08–21.53]; P = 0.039). For late residual LBP, protective factors were Pre-PHQ-9 (OR = 0.33, 95%CI [0.16–0.69]; P = 0.003) and OSTEOC (OR = 0.91, 95%CI [0.83–0.99]; P = 0.035); risk factors were 25(OH)D (OR = 1.19, 95%CI [1.03–1.36]; P = 0.016) and Creatinine (OR = 1.01, 95%CI [1.00–1.02]; P = 0.038). Predictive models showed AUC = 0.962 (sensitivity 84.7%, specificity 96.2%) for early residual LBP and AUC = 0.974 (sensitivity 92.4%, specificity 95.2%) for late residual LBP, with concordance rates of 78.27% and 88.3% respectively. Conclusion Pre-VAS, Pre-PHQ-9, Post1m-PHQ-9, FJD, OSTEOC, 25(OH)D, Creatinine, and female sex significantly impact residual LBP post-TLIF. The models demonstrate high predictive accuracy for early-stage and late-stage residual LBP, supporting clinical risk assessment.