Predictors of variceal bleeding: an analysis of clinical, endoscopic, and haemodynamic variables, with special reference to intravariceal pressure

Gut. 1989 Dec;30(12):1757-64. doi: 10.1136/gut.30.12.1757.

Abstract

To assist in the prediction of variceal bleeding we examined 12 clinical, endoscopic and haemodynamic variables in 126 patients with portal hypertension (72 bleeders and 54 non-bleeders). Multiple regression and discriminant analysis were utilised for 101 patients. Stepwise regression among the 12 variables explained only 54% of the variability existing between bleeders and non-bleeders. Six of the 12 variables accounted for 51% of the total bleeder/non-bleeder variability. A predicting equation was evolved [Yp = -0.60 + 0.19 X1 (colour of varices) + 0.16 X2 (cherry red spots) + 0.03 X3 (red wale, markings) + 0.01 X4 (haematocystic spots) + 0.15 X5 (Conn's grading of variceal size) + 0.03 X6 (intravariceal pressure)]. Of the six variables used, variceal size indicated 35%, and intravariceal pressure accounted for 12%, of the explained variability between bleeders and non-bleeders. The presence of the three red colour signs and the colour of the varices accounted for only 3% and 1% respectively of the total explained variation. Using the discriminant function analysis, we were able to correctly identify 85% of the bleeders and 81% of the non-bleeders. Additional studies are required to detect the variables responsible for the unexplained variability (46%) between bleeders and non-bleeders.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Adolescent
  • Adult
  • Aged
  • Child
  • Child, Preschool
  • Discriminant Analysis
  • Esophageal and Gastric Varices / complications*
  • Esophageal and Gastric Varices / pathology
  • Esophageal and Gastric Varices / physiopathology
  • Esophagoscopy
  • Female
  • Gastrointestinal Hemorrhage / etiology*
  • Hemodynamics
  • Humans
  • Hypertension, Portal / complications
  • Infant
  • Liver Diseases / pathology
  • Male
  • Middle Aged
  • Pressure
  • Regression Analysis