Ther Clin Risk Manag. 2026 May 12;22:581371. doi: 10.2147/TCRM.S581371. eCollection 2026.
ABSTRACT
BACKGROUND: Caudal block in clinical practice relies on empirical drug administration without reliable prediction of anesthetic level, thereby limiting its application. This study aimed to develop a predictive model for sensory block level based on sacral canal anatomical factors and patient characteristics.
METHODS: A total of 132 patients undergoing elective anal surgery with caudal block were included. Ultrasound measured anatomical parameters (the width of the base of the sacrum (BSW), the anterior-posterior diameter of the sacral hiatus at its apex (SHAP), the anterior-posterior diameter of the sacral hiatus at the first segment inferior to the apex (SHFSIA-AP), the length of the sacral hiatus in the sagittal plane (SHLS) and demographic data (age, gender, height, weight) were collected. Sensory block level was categorized into five ordinal grades. Multivariable ordinal logistic regression identified independent predictors, with variance inflation factor (VIF) assessing multicollinearity and Brant test verifying proportional odds assumption. Model performance was evaluated via concordance index (C-index) and Lipsitz test, with internal validation performed using bootstrapping with 1000 resamples, and a nomogram was constructed.
RESULTS: Univariable analysis showed SHAP, SHLS, SHFSIA-AP, height, age, gender, and weight correlated with sensory block level (P<0.05). Final independent predictors were SHAP, SHLS, and height. No multicollinearity (VIF<5) and valid proportional odds assumption (Brant test P=0.36) were confirmed. The model had good predictive performance (C-index=0.881, 95% CI [0.809-0.953]) and fit (Lipsitz test P=0.49), with the nomogram visualizing probabilities of each block level category.
CONCLUSION: This study identified SHAP, SHLS, and height as predictors for caudal block sensory level. The nomogram enables individualized, precise drug administration, shifting the technique from experience-based to precision prediction.
PMID:42153217 | PMC:PMC13180391 | DOI:10.2147/TCRM.S581371
