To determine the diagnostic accuracy of computer tomography in the detection of venous collaterals surrounding the esophagus in patients with portal hypertension, preoperative computer tomography interpretations of these veins in 15 patients who were candidates for the Sugiura procedure for treatment of esophageal varices were correlated with those of the intraoperative assessment. Laparotomy revealed severe paraesophageal varices in five patients; four of them were found to have paraesophageal varices in computer tomography films. The sensitivity and specificity of computer tomography in diagnosing severe paraesophageal varices were 80% and 100%, respectively. A second assessment was performed in 59 additional patients with esophageal variceal hemorrhage to investigate the influence of paraesophageal varices on the efficacy of endoscopic sclerotherapy in the treatment of varices. The patients were divided into two groups: Group A included 17 patients with and group B 42 patients without paraesophageal varices on presclerotherapy computer tomography. All patients underwent elective sclerotherapy after being deemed hemodynamically stable. Patients in group A required more treatment sessions, more sclerosant and longer periods to obliterate varices completely than did group B patients. Eight patients in group A and six in group B (57% vs. 16%, p < 0.05) had variceal recurrence after obliteration during mean follow-ups of 20.8 and 19.9 mo, respectively. The mean time elapsed before variceal reappearance was shorter for group A than for group B (4.1 +/- 3.3 vs. 11.8 +/- 2.7 mo, p < 0.05). Among patients who developed new varices, five patients in group A and one in group B experienced repeat bleeding.(ABSTRACT TRUNCATED AT 250 WORDS)