目的 分析原发性肝癌（PHC）干预前并发肝性胸腔积液（HH）的影响因素，构建列线图风险预测模型并进行评价。 方法 回顾性收集2012年10月—2021年10月首次就诊于昆明市第三人民医院并诊断为PHC的353例住院患者的临床资料，根据是否并发HH分为胸腔积液组（n=153）和非胸腔积液组（n=200）。收集PHC患者的一般资料和入院首次临床检测资料。符合正态分布的计量资料2组间比较采用成组t检验；非正态分布的计量资料2组间比较采用Mann-Whitney U检验。计数资料组间比较采用χ²检验或Fisher确切概率法。单因素分析有统计学意义的变量经多重共线性检验后，进行多因素Logistic回归分析确定独立影响因素。使用“rms”程序包建立列线图风险预测模型，使用Hosmer-Lemeshow检验评估模型的拟合度，绘制受试者工作特征曲线。使用“Calibration Curves”程序包绘制校准曲线，使用“rmda”程序包绘制临床决策曲线及临床影响曲线对风险预测模型进行评价。 结果 353例PHC患者中存在胸腔积液有153例，患病率为43.34%。Child-Pugh分级B级（OR=2.652，95%CI：1.050～6.698，P=0.039）、Child-Pugh分级C级（OR=7.963，95%CI： 1.046～60.632，P=0.045）、总蛋白（OR=0.947，95%CI：0.914～0.981，P=0.003）、超敏C反应蛋白（OR=1.007，95%CI：1.001～1.014，P=0.025）、白细胞介素2（OR=0.801，95%CI：0.653～0.981，P=0.032）是PHC干预前并发HH的独立影响因素，据此建立列线图风险预测模型。Hosmer-Lemeshow检验显示该模型具有较好的拟合度（χ²=5.006，P=0.757），该模型的受试者工作特征曲线下面积为0.752（95%CI：0.701～0.803），灵敏度为78.40%，特异度为63.50%。校准曲线显示该模型预测PHC干预前并发HH具有较好的一致性。临床决策曲线和临床影响曲线显示在一定阈值范围内该模型具有较好的临床实用性。 结论 Child-Pugh分级、总蛋白、白细胞介素2、超敏C反应蛋白是PHC干预前发生HH的独立影响因素，据此建立的列线图模型可有效预测发生HH的风险。

Objective To investigate the influencing factors for hepatic hydrothorax （HH） before intervention for primary hepatic carcinoma （PHC）， and to construct and assess the nomogram risk prediction model. Methods A retrospective analysis was performed for the clinical data of 353 hospitalized patients who attended the Third People’s Hospital of Kunming for the first time from October 2012 to October 2021 and there diagnosed with PHC， and according to the presence or absence of HH， they were divided into HH group with 153 patients and non-HH group with 200 patients. General data and the data of initial clinical testing after admission were collected from all PHC patients. The independent-samples t test was used for comparison of normally distributed continuous data between two groups， and the Mann-Whitney U test was used for comparison of non-normally distributed continuous data between two groups； the chi-square test or the Fisher’s exact test was used for comparison of categorical data between groups. After the multicollinearity test was performed for the variables with statistical significance determined by the univariate analysis， the multivariate Logistic regression analysis was used to identify independent influencing factors. The “rms” software package was used to construct a nomogram risk prediction model， and the Hosmer-Lemeshow test and the receiver operating characteristic （ROC） curve were used to assess the risk prediction model； the “Calibration Curves” software package was used to plot the calibration curve， and the “rmda” software package was used to plot the clinical decision curve and the clinical impact curve. Results Among the 353 patients with PHC， there were 153 patients with HH， with a prevalence rate of 43.34%. Child-Pugh class B （odds ratio ［OR］=2.652， 95% confidence interval ［CI］： 1.050 — 6.698， P=0.039）， Child-Pugh class C （OR=7.963， 95%CI： 1.046‍ ‍—‍ ‍60.632， P=0.045）， total protein （OR=0.947， 95%CI： 0.914‍ ‍—‍ ‍0.981， P=0.003）， high-sensitivity C-reactive protein （OR=1.007， 95%CI： 1.001‍ ‍—‍ ‍1.014， P=0.025）， and interleukin-2 （OR=0.801， 95%CI： 0.653‍ ‍—‍ ‍0.981， P=0.032） were independent influencing factors for HH before PHC intervention， and a nomogram risk prediction model was established based on these factors. The Hosmer-Lemeshow test showed that the model had a good degree of fitting （χ²=5.006， P=0.757）， with an area under the ROC curve of 0.752 （95%CI： 0.701‍ ‍—‍ ‍0.803）， a sensitivity of 78.40%， and a specificity of 63.50%. The calibration curve showed that the model had good consistency in predicting HH before PHC intervention， and the clinical decision curve and the clinical impact curve showed that the model had good clinical practicability within a certain threshold range. Conclusion Child-Pugh class， total protein， interleukin-2， and high-sensitivity C-reactive protein are independent influencing factors for developing HH before PHC intervention， and the nomogram model established based on these factors can effectively predict the risk of developing HH.