BACKGROUND Variceal bleeding is a frequent and serious complication of cirrhosis. Early detection of varices by videogastroscope (VGC) is recommended in all patients with cirrhosis to determine the need for prophylactic treatment. Have been described noninvasive markers of the presence of esophageal varices, which could prevent the realization of VGC for that purpose. OBJECTIVE To determine and compare noninvasive (longitudinal diameter of spleen, platelet count, platelet reason / spleen) as predictors of the presence of esophageal varices. METHODS We retrospectively studied 125 patients with cirrhosis from any cause. They had VGC, blood count and abdominal ultrasonography. The diagnostic accuracy for determining the presence of esophageal varices or large varices according to the different variables was studied using the area under the ROC curve (AUROC). RESULTS The prevalence of esophageal varices was 63.2% and 42.4% were diagnosed with large varices. The reason platelets/spleen and platelet count showed an AUROC of 0.74 for the detection of esophageal varices. The cut-off for the ratio platelets / spleen was 1.010 (sensitivity 72.15% and specificity 71.74%) for the presence of varices and 870 for the presence of clinically significant varices (sensitivity 62.26% and specificity 62.50%). The analysis according to these breakpoints showed that 23.6% of patients with scores higher than 1,010 had large varices and 45% of patients with values lower than 870 had not large varices. CONCLUSIONS Although the reason platelets/spleen showed an AUROC acceptable, its implementation would entail a risk of not diagnosing large varices in almost a quarter of the population studied.