Department of Respiratory Medicine, The Second Affiliated Hospital, Dalian Medical University, Dalian, China; Department of Respiratory Medicine, The North Area of Suzhou Municipal Hospital, Suzhou, China.
Department of Clinical Laboratory, Peking University First Hospital, Beijing, China.
Cancer Lett. 2018 Apr 28;420:236-241. doi: 10.1016/j.canlet.2018.01.079. Epub 2018 Feb 21.
This study aimed to build a valid diagnostic nomogram for assessing the cancer risk of the pulmonary lesions identified by chest CT.
A total of 691 patients with pulmonary lesions were recruited from three centers in China. The cut-off value for each tumor marker was confirmed by minimum P value method with 1000 bootstrap replications. The nomogram was based on the predictive factors identified by univariate and multivariate analysis. The predictive performance of the nomogram was measured by concordance index and calibrated with 1000 bootstrap samples to decrease the overfit bias. We also evaluated the net benefit of the nomogram via decision curve analysis. Finally, the nomogram was validated externally using a separate cohort of 305 patients enrolled from two additional institutions.
The cut-off for CEA, SCC, CYFRA21-1, pro-GRP, and HE4 was 4.8 ng/mL, 1.66 ng/mL, 1.83 ng/mL, 56.55 pg/mL, and 63.24Lpmol/L, respectively. Multivariate logistic regression model (LRM) identified tumor size, CEA, SCC, CYFRA21-1, pro-GRP, and HE4 as independent risk factors for lung cancer. The nomogram based on LRM coefficients showed concordance index of 0.901 (95% CI: 0.842-0.960; P < 0.001) for lung cancer in the training set and 0.713 (95% CI: 0.599-0.827; P < 0.001) in the validation set. Decision curve analysis reported a net benefit of 87.6% at 80% threshold probability superior to the baseline model.
Our diagnostic nomogram provides a useful tool for assessing the cancer risk of pulmonary lesions identified in CT screening test.
本研究旨在建立一种有效的诊断列线图，用于评估胸部 CT 发现的肺部病变的癌症风险。
本研究共纳入中国 3 家中心的 691 例肺部病变患者。采用最小 P 值法和 1000 次 bootstrap 重复确定每个肿瘤标志物的截断值。列线图基于单变量和多变量分析确定的预测因素。采用一致性指数评价列线图的预测性能，并通过 1000 次 bootstrap 样本进行校准，以减少过拟合偏倚。我们还通过决策曲线分析评估了列线图的净获益。最后，使用另外 2 家机构的 305 例患者的独立队列对列线图进行外部验证。
CEA、SCC、CYFRA21-1、pro-GRP 和 HE4 的截断值分别为 4.8ng/ml、1.66ng/ml、1.83ng/ml、56.55pg/ml 和 63.24Lpmol/L。多变量逻辑回归模型（LRM）确定肿瘤大小、CEA、SCC、CYFRA21-1、pro-GRP 和 HE4 为肺癌的独立危险因素。基于 LRM 系数的列线图在训练集中的一致性指数为 0.901（95%CI：0.842-0.960；P<0.001），在验证集中为 0.713（95%CI：0.599-0.827；P<0.001）。决策曲线分析报告，80%阈值概率下的净获益为 87.6%，优于基线模型。
本研究构建的诊断列线图为 CT 筛查试验中发现的肺部病变的癌症风险评估提供了一种有用的工具。
