Grading portal gastropathy: validation of a gastropathy scoring system

The specific aim of this study was to independently evaluate the reproducibility and validity of the Baveno portal hypertensive gastropathy (PHG) grading system.

This is a prospective study of 100 consecutive patients with cirrhosis and portal gastropathy. An endoscopy was performed at entry on all subjects and on 70 patients during follow-up (median 15 months). Interobserver reproducibility was assessed by the kappa statistic. Patients with PHG-related bleeding were compared with those without bleeding. The relationship of PHG-related bleeding to the PHG score was assessed by a receiver operating characteristic curve and by multiple logistic regression analysis.

Interobserver agreement about the presence of the mosaic pattern and red marks was high (kappa >0.75). However, the agreement about the extent of the lesion was poorer (kappa = 0.4-0.45). There was a stepwise increase in PHG-related bleeding risk with increasing PHG scores. Multiple logistic regression confirmed that the PHG score independently correlated with PHG-associated bleeding (OR = 2.5, 90% CI = 1.4-4.6, p < 0.009). During follow-up, red marks developed de novo in one of 70 patients. The severity of red marks worsened in five of 25 patients, whereas they disappeared in eight of 62 individuals. Nodular lesions in the antrum were found in five subjects. The risks of recurrent bleeding during follow-up were related to severe PHG scores (>4), presence of gastric antral vascular ectasia, and nodular lesions in the antrum.

The PHG scoring system is reproducible and accurately reflects the risks of PHG-related bleeding in patients with cirrhosis.