BACKGROUND & AIMS: Spontaneous portosystemic shunts (SPSS) have been associated with hepatic encephalopathy (HE). Little is known about their prevalence among patients with cirrhosis or clinical effects. We investigated the prevalence and characteristics of SPSS in patients with cirrhosis and their outcomes. METHODS: We performed a retrospective study of 1729 patients with cirrhosis who underwent abdominal computed tomography or magnetic resonance imaging analysis from 2010 through 2015 at 14 centers in Canada and Europe. We collected data on demographic features, etiology of liver disease, comorbidities, complications, treatments, laboratory and clinical parameters, Model for End-Stage Liver Disease (MELD) score, and endoscopy findings. Abdominal images were reviewed by a radiologist (or a hepatologist trained by a radiologist) and searched for the presence of SPSS, defined as spontaneous communications between the portal venous system or splanchnic veins and the systemic venous system, excluding gastroesophageal varices. Patients were assigned to groups with large SPSS (L-SPSS, ≥8 mm), small SPSS (S-SPSS, <8 mm), or without SPSS (W-SPSS). The main outcomes were the incidence of complications of cirrhosis and mortality according to the presence of SPSS. Secondary measurements were the prevalence of SPSS in patients with cirrhosis and their radiologic features. RESULTS: L-SPSS were identified in 488 (28%) patients, S-SPSS in 548 (32%) patients, and no shunt (W-SPSS) in 693 (40%) patients. The most common L-SPSS was splenorenal (46% of L-SPSS). The presence and size of SPSS increased with liver dysfunction: among patients with MELD scores of 6-9, 14% had L-SPSS and 28% had S-SPSS; among patients with MELD scores of 10-13, 30% had L-SPSS and 34% had S-SPSS; among patients with MELD scores of 14 or higher, 40% had L-SPSS and 32% had S-SPSS (P < .001 for multiple comparison among MELD groups). HE was reported in 48% of patients with L-SPSS, 34% of patients with S-SPSS, and 20% of patients W-SPSS (P < .001 for multiple comparison among SPSS groups). Recurrent or persistent HE was reported in 52% of patients with L-SPSS, 44% of patients with S-SPSS, and 37% of patients W-SPSS (P = .007 for multiple comparison among SPSS groups). Patients with SPSS also had a larger number of portal hypertension-related complications (bleeding or ascites) than those W-SPSS. Quality of life and transplantation-free survival were lower in patients with SPSS vs without. SPSS were an independent factor associated with death or liver transplantation (hazard ratio, 1.26; 95% confidence interval, 1.06-1.49) (P = .008) in multivariate analysis. When patients were stratified by MELD score, SPSS were associated with HE independently of liver function: among patients with MELD scores of 6-9, HE was reported in 23% with L-SPSS, 12% with S-SPSS, and 5% with W-SPSS (P < .001 for multiple comparison among SPSS groups); among those with MELD scores of 10-13, HE was reported in 48% with L-SPSS, 33% with S-SPSS, and 23% with W-SPSS (P < .001 for multiple comparison among SPSS groups); among patients with MELD scores of 14 or more, HE was reported in 59% with L-SPSS, 57% with S-SPSS, and 48% with W-SPSS (P = .043 for multiple comparison among SPSS groups). Patients with SPSS and MELD scores of 6-9 were at higher risk for ascites (40.5% vs 23%; P < .001) and bleeding (15% vs 9%; P = .038) than patients W-SPSS and had lower odds of transplant-free survival (hazard ratio 1.71; 95% confidence interval, 1.16-2.51) (P = .006). CONCLUSIONS: In a retrospective analysis of almost 2000 patients, we found 60% to have SPSS; prevalence increases with deterioration of liver function. SPSS increase risk for HE and with a chronic course. In patients with preserved liver function, SPSS increase risk for complications and death. ClinicalTrials.gov ID NCT02692430.