Int J Med Sci 2021; 18(9):1966-1974. doi:10.7150/ijms.53743 This issue

Research Paper

Development and validation a simple model for identify malignant ascites

Ying-Yun Guo1, Xiu-Lan Peng2, Na Zhan3, Shan Tian1, Jiao Li1, Wei-Guo Dong

1. Department of Gastroenterology, Renmin Hospital of Wuhan University, Wuhan, Hubei, 430060, China.
2. Department of Oncology, The Fifth Hospital of Wuhan, Wuhan, Hubei, 430050, China.
3. Department of Pathology, Renmin Hospital of Wuhan University, Wuhan, Hubei, 430060, China.

This is an open access article distributed under the terms of the Creative Commons Attribution License ( See for full terms and conditions.
Guo YY, Peng XL, Zhan N, Tian S, Li J, Dong WG. Development and validation a simple model for identify malignant ascites. Int J Med Sci 2021; 18(9):1966-1974. doi:10.7150/ijms.53743. Available from

File import instruction


Graphic abstract

The differential diagnosis of benign ascites and malignant ascites is incredibly challenging for clinicians. This research aimed to develop a user-friendly predictive model to discriminate malignant ascites from non-malignant ascites through easy-to-obtain clinical parameters. All patients with new-onset ascites fluid were recruited from January 2014 to December 2018. The medical records of 317 patients with ascites for various reasons in Renmin Hospital of Wuhan University were collected and reviewed retrospectively. Thirty-six parameters were included and selected using univariate logistic regression, multivariate logistic regression, and receiver operating characteristic (ROC) curve analyses to establish a mathematical model for differential diagnosis, and its diagnostic performance was validated in the other groups. Age, cholesterol, hypersensitivity C-reactive protein (hs-CRP) in serum, ascitic fluid adenosine deaminase (AF ADA), ascitic fluid lactate dehydrogenase (AF LDH) involvement in a 5-marker model. With a cut-off level of 0.83, the sensitivity, specificity, accuracy, and area under the ROC of the model for identifying malignant ascites in the development dataset were 84.7%, 88.8%, 87.6%, and 0.874 (95% confidence interval [CI], 0.822-0.926), respectively, and 80.9%, 82.6%, 81.5%, and 0.863 (95% CI,0.817-0.913) in the validation dataset, respectively. The diagnostic model has a similar high diagnostic performance in both the development and validation datasets. The mathematical diagnostic model based on the five markers is a user-friendly method to differentiate malignant ascites from benign ascites with high efficiency.

Keywords: diagnosis, differential, ascites, carcinoma, logistic models