Noninvasive measurements of variceal pressure adequately reflect the hemodynamic effects of propranolol on portal hypertension. However, the prognostic value of variceal pressure responses during continued propranolol therapy has not been evaluated, and it is unclear whether this may substitute invasive measurements of portal pressure response. Fifty-five portal hypertensive patients with cirrhosis were studied before and at 4 months of continued propranolol therapy. Variceal pressure was measured using an endoscopic pressure gauge. Portal pressure was evaluated as the hepatic venous pressure gradient (HVPG). Over a 28 +/- 11 month follow-up, 16 patients experienced variceal bleeding. Baseline characteristics were similar in bleeders and nonbleeders. At 4 months, reduction in variceal pressure was less marked in bleeders than in nonbleeders (5% +/- 20% vs. -15% +/- 24%; P =.03). A fall in variceal pressure 20% or greater of baseline was an independent predictor of absence of variceal bleeding; which occurred in 5% of patients with a 20% or greater fall in variceal pressure versus 42% of patients with less than a 20% reduction (P =.004). The HVPG response had similar independent prognostic value (decrease > or =20%: 6% bleeding; decrease <20%: 45% bleeding; P =.004) but identified different patients. Achieving a 20% decrease in either variceal pressure or HVPG was highly sensitive (85%) and specific (93%) identifying patients not bleeding on follow-up. Endoscopic measurements of variceal pressure response to continued pharmacotherapy provide useful prognostic information on the risk of variceal bleeding. As with HVPG response, a fall in variceal pressure of 20% or greater is associated with a very low risk of variceal bleeding. The combination of both parameters allows almost optimal prognostication.