性别-年龄-肺功能分期与复合生理指数对特发性肺纤维化患者死亡风险的预测价值比较研究

背景　性别-年龄-肺功能（GAP）分期和复合生理指数（CPI）均作为预测特发性肺纤维化（IPF）患者死亡风险的工具，常被国内外学者使用，而哪一种模型更具有优势或临床应用价值，目前鲜见报道。目的　探讨GAP分期与CPI对IPF患者死亡风险的预测价值。方法　选取2012-2019年在同济大学附属上海市肺科医院确诊的有完整临床数据及回访记录的200例IPF患者为研究对象，其中男181例（90.5%），女19例（9.5%）；年龄41~85岁，平均年龄（66.0±7.6）岁。分别使用GAP分期与CPI评分对患者进行分组，GAP Ⅰ期109例（54.5%），GAPⅡ期75例（37.5%），GAPⅢ期16例（8.0%）；CPI评分≤41分99例（49.5%），CPI评分>41分101例（50.5%）。对不同GAP分期及CPI分组下IPF患者的临床、影像学、肺功能等特征进行比较，发现其符合GAP分期与CPI的模型分布特征。使用Cox比例风险回归模型对IPF患者的预后因素进行分析，并绘制两模型预测IPF患者死亡率的受试者工作特征曲线（ROC曲线）计算两模型ROC曲线下面积（AUC），利用Medcalc 软件分析两模型在总体以及不同时间点上对IPF死亡风险的预测能力。结果　Cox比例风险回归模型结果显示，GAP评分高〔HR=1.038，95%CI（1.023，1.054）〕、CPI评分高〔HR=1.509，95%CI（1.286，1.771）〕、GAP Ⅱ期〔HR=2.622，95%CI（1.536，4.475）〕、GAP Ⅲ期〔HR=4.002，95%CI（1.947，8.226）〕是IPF患者预后的独立危险因素（P<0.05）。GAP分期预测IPF患者1年内死亡率、2年内死亡率、3年内死亡率、总体死亡率的AUC分别为0.685〔95%CI（0.616，0.749）〕、0.675〔95%CI（0.606，0.740）〕、0.642〔95%CI（0.571，0.708）〕、0.668〔95%CI（0.598，0.733）〕，CPI预测IPF患者1年内死亡率、2年内死亡率、3年内死亡率、总体死亡率的AUC分别为0.750〔95%CI（0.684，0.809）〕、0.745〔95%CI（0.679，0.804）〕、0.735〔95%CI（0.669，0.795）〕、0.745〔95%CI（0.679，0.804）〕。GAP分期与CPI预测IPF患者2年内死亡率、3年内死亡率、总体死亡率的AUC比较，差异均有统计学意义（Z=2.193，P=0.028 3；Z=2.918，P=0.003 5；Z=2.529，P=0.011 4）；GAP分期与CPI预测IPF患者1年内死亡率的AUC比较，差异无统计学意义（Z=1.799，P=0.072 1）。结论　CPI与GAP分期对于IPF预后均是非常有效的预测指标，且CPI在总体死亡率以及2年、3年内死亡率的预测上明显优于GAP分期，但AUC仅为0.64~0.75，其准确预测能力有限，仍需探索建立更为实用有效的风险预测模型。

Comparison of the Predictive Value of GAP Staging and CPI for Risk of Death in Patients with Idiopathic Pulmonary Fibrosis

Background　Gender，age，and physiologic variables（GAP）staging and composite physiologic index（CPI），are often used alone for predicting the death risk of idiopathic pulmonary fibrosis（IPF），but which is more superior or has higher clinical value is still unclear. Objective　To explore the predictive value of GAP staging and CPI for risk of death in patients with IPF. Methods　A retrospective analysis was done in 200 cases of confirmed IPF with complete clinical data and follow-up records who hospitalized in Shanghai Pulmonary Hospital，Tongji University，from 2012 to 2019〔including 181 men（90.5%），and 19 women（9.5%），with an average age of（66.0±7.6）years（ranging from 41 to 85 years）〕. By using GAP staging，109（54.5%），75（37.5%），and 16（8.0%）cases were classified as at GAP stage Ⅰ，stage Ⅱ，and stage Ⅲ，respectively. And by using CPI score，99（49.5%）and 101（50.5%）cases were evaluated with CPI ≤ 41 points，and >41 points，respectively. Clinical，imaging，and physiological characteristics were compared across GAP stages and CPI groups to find the intergroup differences. Cox regression was used to perform univariate analysis of the prognostic factors of IPF. The area under the ROC curve（AUC）of each of the two models to predict the mortality of IPF patients was calculated to measure the prognostic accuracy. Medcalc was used to quantify the predictive accuracy of the two models for overall IPF mortality，and one-year，two-year，and three-year IPF mortality. Results　Cox regression analysis demonstrated that high GAP score〔HR=1.038，95%CI（1.023，1.054）〕，high CPI score〔HR=1.509，95%CI（1.286，1.771）〕，GAP stage Ⅱ〔HR= 2.622，95%CI（1.536，4.475）〕，GAP stage Ⅲ〔HR=4.002，95%CI（1.947，8.226）〕were independently associated with increased risk of poor prognosis of IPF patients（P<0.05）. For predicting one-year，two-year，and three-year mortality as well as overall mortality of IPF patients，the AUC of GAP staging was 0.685〔95%CI（0.616，0.749）〕，0.675〔95%CI（0.606，0.740）〕，0.642〔95%CI（0.571，0.708）〕，and 0.668〔95%CI（0.598，0.733）〕，respectively，and the AUC of CPI score was 0.750〔95%CI（0.684，0.809）〕，0.745〔95%CI（0.679，0.804）〕，0.735〔95%CI（0.669，0.795）〕，0.745〔95%CI（0.679，0.804）〕，respectively. GAP staging and CPI score had statistically significant differences in the AUC for predicting two-year，and three-year mortality as well as overall mortality of IPF（Z=2.193，P=0.028 3；Z=2.918，P=0.003 5；Z=2.529，P=0.011 4）， but had no statistically difference in one-year mortality（Z=1.799，P=0.072 1）. Conclusion　Both CPI and GAP staging may be efficient prognostic factors for IPF. CPI was significantly better than GAP staging in predicting the overall，two-year，and three-year mortality，but the AUC ranged 0.64 to 0.75，indicating that its prognostic performance may be limited. More practical and feasible death risk prediction models are still needed to be established.