Cirrhosis Characteristics Impact the Platelet Count to Spleen Diameter Ratio Threshold to Predict Esophageal Varices: A Meta-Regression Analysis: 456

Introduction: To assess the basis of controversial results of platelet count to spleen diameter ratio (PSR) in predicting esophageal varices (EV) in cirrhosis. Methods: Studies using PSR were allocated to 2 groups: 12, which were designed to validate the PSR cutoff 909 (group A), and 16, which introduced individual cutoffs (group B). Meta-regression analysis was performed to assess the effect of cirrhosis etiology (the proportion of patients with viral hepatitis cirrhosis [PVHC]) and the proportion of patients with Child-Pugh class B and C (PCP) on the PSR diagnostic value. A hypothetical model was suggested to predict EV. Results: In group A, meta-analysis showed a pooled sensitivity of 84% (95% confidence interval [CI]: 77-89%), and meta-regression revealed that PSR sensitivity is correlated with PVHC (coefficient= 0.014; R2=0.12 p=0.089), and PCP (coefficient=0.027; R2=0.42; p=0.005) (Fig 1). Upon multiple regression, PCP was still correlated with sensitivity (coefficient=0.032, R2=0.27; p=0.016). In group B, studies with PCP of 80% or higher showed a higher variation in their introduced cutoffs (standard deviation 354 vs. 103; p<0.009).Figure 1: Sensitivity increasing as the proportion of patients with viral hepatitis cirrhosis and Child-Pugh class B & C increase; circles representing the studies; size correlates with study weight in analysis.Conclusion: When the PSR 909 applied, as the proportion of patients with CP class B and C or viral hepatitis cirrhosis in the cohort increased, this cutoff produced a better diagnostic value. In group B, individual cutoffs varied by increase in proportion of patients with CP class B and C. We concluded that no universal cutoff value derived from one nonhomogeneous cirrhosis cohort may be extrapolated to another; patients should be stratified based on CP class and etiology; each stratum has a unique cutoff. Alternatively, a hypothetical logistic model may predict EV probability (Figure 2).Figure 2: p: esophageal varices probability; e: base of natural logarithm; PSR: plate count/spleen diameter; etiology: viral, alcohol, etc.; CP: Child-Pugh class A vs. B&C; Bi: regression coefficient.