Categories
Nevin Manimala Statistics

Splenic stiffness does not predict esophageal varices in children with portal hypertension

J Pediatr Gastroenterol Nutr. 2025 Oct 27. doi: 10.1002/jpn3.70247. Online ahead of print.

ABSTRACT

OBJECTIVE: To investigate ultrasound parameters, particularly splenic stiffness, as predictors of pediatric esophageal varices.

METHODS: We included all children aged 0-19 years who underwent esophagogastroduodenoscopy, for variceal screening or surveillance, and abdominopelvic ultrasound with splenic elastography measurement. We also recorded biological parameters (platelets count, albumin) to determine a clinical prediction rule (CPR). Derivation and validation cohorts were defined according to measurement date. Receiver-operating characteristic (ROC) statistics and sensitivity, specificity, positive predictive value, and negative predictive value for the optimal threshold value were calculated, and used to assess the performance of each parameter.

RESULTS: Eighty derivation cohort children and 58 validation cohort children formed the study sample. Cohort characteristics did not differ for age, sex, distribution of varices but differ for some etiologies, abdominal surgery, spleen size, splenic stiffness measurement (SSM), ascites, and the presence of spontaneous portosystemic shunts. In the derivation cohort, splenic stiffness was the best independent predictor of esophageal varices, with an area under the ROC curve (AUC) of 0.83, a sensitivity of 0.90, and specificity of 0.73 at a threshold of 22 kPa. In the validation cohort, SSM was no longer significantly associated with esophageal varices at endoscopy, had a lower sensitivity of 0.26, and no other threshold could be found. Liver stiffness measurement (LSM) and CPR had a correct predictive value (AUC 0.70 for LSM; 0.78 for CPR in the derivation cohort and 0.64 for LSM; 0.71 for CPR in the validation cohort) for esophageal varices.

CONCLUSIONS: SSM cannot be used as a single parameter to predict esophageal varices. LSM and CPR despite their lower AUC appear to much more robust measures with consistent results across cohorts.

PMID:41144851 | DOI:10.1002/jpn3.70247

By Nevin Manimala

Portfolio Website for Nevin Manimala