Hepatic hydrothorax (HH) is a rare but severe manifestation of cirrhotic ascites. Whether HH development relates to ascites severity is uncertain and simple clinical models to predict HH from all stages of ascites are missing. The recently published CIRrhotic Ascites Severity (CIRAS) model using only ascites-related variables may serve this purpose.

We investigated if the CIRAS model within one year predicts the development of HH requiring thoracentesis in patients with cirrhosis and ascites.

We used data from 1090 patients with cirrhosis and all severities of ascites enrolled in three randomized clinical trials with available CIRAS model scores and no history of HH. Fine and Gray regression was applied to estimate the CIRAS model's ability to predict HH.

Thirty-five patients developed HH requiring thoracentesis. The CIRAS model stratified patients at different risks for HH and increasing CIRAS score was associated with a higher risk for HH (sHR 1.49 [95% CI: 1.19–1.86]). The CIRAS model's discriminatory ability achieved an AUC of 0.67 (95% CI: 0.56–0.77); higher than of the cirrhosis severity scores Child-Pugh and MELD variants.

The CIRAS model predicts the development of HH in cirrhosis patients with any grade of ascites, suggesting a potential for improved pre-emptive HH management. This complements the general movement towards personalised treatments and care.