Department of Thoracic Surgery, The Icahn School of Medicine at Mount Sinai, New York, New York.
Cancer Research and Biostatistics, Seattle, Washington.
J Thorac Oncol. 2023 Dec;18(12):1689-1702. doi: 10.1016/j.jtho.2023.08.005. Epub 2023 Aug 9.
The International Association for the Study of Lung Cancer developed an international pleural mesothelioma database to improve staging. Data entered from 1995 to 2009 (training data set) were analyzed previously to evaluate supplemental prognostic factors. We evaluated these factors with new clinical data to determine whether the previous models could be improved.
Patients entered into the database from 2009 to 2019 (validation cohort) were assessed for the association between previous prognosticators and overall survival using Cox proportional hazards regression with bidirectional stepwise selection. Additional variables were analyzed and models were compared using Harrell's C-index.
The training data set included 3101 patients and the validation cohort, 1733 patients. For the multivariable pathologic staging model applied to the training cohort, C-index was 0.68 (95% confidence interval [CI]: 0.656-0.705). For the validation data set (n = 497), C-index was 0.650 (95% CI: 0.614-0.685), and pathologic stage, histologic diagnosis, sex, adjuvant therapy, and platelet count were independently associated with survival. Adding anemia to the model increased the C-index to 0.652 (95% CI: 0.618-0.686). A basic presentation model including all parameters before staging yielded a C-index of 0.668 (95% CI: 0.641-0.695). In comparison, the European Organization for Research and Treatment of Cancer model yielded C-indices of 0.550 (95% CI: 0.511-0.589) and 0.577 (95% CI: 0.550-0.604) for pathologic staging and presentation models, respectively.
Although significant predictors differed slightly, the International Association for the Study of Lung Cancer training model performed well in the validation set and better than the model of the European Organization for Research and Treatment of Cancer. International collaboration is critical to improve outcomes in this rare disease.
国际肺癌研究协会(IASLC)开发了一个国际间胸膜间皮瘤数据库,以改善分期。此前已对 1995 年至 2009 年(训练数据集)输入的数据进行了分析,以评估补充预后因素。我们使用新的临床数据评估了这些因素,以确定以前的模型是否可以得到改进。
从 2009 年至 2019 年(验证队列)输入数据库的患者,使用双向逐步选择的 Cox 比例风险回归,评估先前的预后因素与总生存之间的相关性。分析了其他变量,并使用 Harrell 的 C 指数比较了模型。
训练数据集包括 3101 例患者,验证队列包括 1733 例患者。对于应用于训练队列的多变量病理分期模型,C 指数为 0.68(95%置信区间[CI]:0.656-0.705)。对于验证数据集(n=497),C 指数为 0.650(95%CI:0.614-0.685),病理分期、组织学诊断、性别、辅助治疗和血小板计数与生存独立相关。将贫血纳入模型后,C 指数增加到 0.652(95%CI:0.618-0.686)。一个基本的呈现模型,包含所有分期前的参数,其 C 指数为 0.668(95%CI:0.641-0.695)。相比之下,欧洲癌症研究与治疗组织(EORTC)的模型分别产生了 0.550(95%CI:0.511-0.589)和 0.577(95%CI:0.550-0.604)的病理分期和呈现模型 C 指数。
尽管显著的预测因素略有不同,但国际肺癌研究协会(IASLC)的训练模型在验证集中表现良好,优于欧洲癌症研究与治疗组织(EORTC)的模型。国际合作对于改善这种罕见疾病的预后至关重要。