Résumé : BACKGROUND & AIMS: The MELDNa score was developed to improve the prognostic value of the MELD score in cirrhosis and was built for serum sodium concentrations numerically capped between 125 and 140mmol/L. This model is not validated in a well-defined population of patients with cirrhosis and refractory ascites in whom severe hyponatremia (⩽125mmol/L) is frequent. This study assessed the prognostic value of severe hyponatremia and the MELDNa score in these patients. METHODS: A consecutive, single-centre, observational, prospective study was performed in patients with cirrhosis and refractory ascites defined according to the International Ascites Club criteria. The prevalence of low serum sodium was assessed in this population. Predictive factors of mortality were analyzed and compared. RESULTS: One hundred seventy-four patients were included. Sixty-six (37.9%) had low serum sodium (<130mmol/L). Sixty-one (35.1%) had diuretic-intractable ascites due to severe hyponatremia (⩽125mmol/L). The median MELDNa score was 23 (10-33). The 1-year cumulative incidence of death was 55% (95% CI: 55-56%). The best predictive factors of mortality were the following: severe hyponatremia (⩽125mmol/L) as an underlying cause of refractory ascites, a higher Child-Pugh score, beta-blocker therapy, and a high frequency of large-volume paracentesis. The Child-Pugh score had a higher area under receiver operating curve to predict mortality than MELDNa. CONCLUSIONS: In patients with cirrhosis and refractory ascites, severe hyponatremia and Child-Pugh score are better predictors of mortality than MELDNa.