Department of Medical Oncology, University Hospital La Paz, Paseo de la Castellana, Madrid 261-28046, Spain.
J Natl Cancer Inst. 2011 Nov 2;103(21):1613-20. doi: 10.1093/jnci/djr388. Epub 2011 Oct 4.
Determining life expectancy in terminally ill cancer patients is a difficult task. We aimed to develop and validate a nomogram to predict the length of survival in patients with terminal disease.
From February 1, 2003, to December 31, 2005, 406 consecutive terminally ill patients were entered into the study. We analyzed 38 features prognostic of life expectancy among terminally ill patients by multivariable Cox regression and identified the most accurate and parsimonious model by backward variable elimination according to the Akaike information criterion. Five clinical and laboratory variables were built into a nomogram to estimate the probability of patient survival at 15, 30, and 60 days. We validated and calibrated the nomogram with an external validation cohort of 474 patients who were treated from June 1, 2006, through December 31, 2007.
The median overall survival was 29.1 days for the training set and 18.3 days for the validation set. Eastern Cooperative Oncology Group performance status, lactate dehydrogenase levels, lymphocyte levels, albumin levels, and time from initial diagnosis to diagnosis of terminal disease were retained in the multivariable Cox proportional hazards model as independent prognostic factors of survival and formed the basis of the nomogram. The nomogram had high predictive performance, with a bootstrapped corrected concordance index of 0.70, and it showed good calibration. External independent validation revealed 68% predictive accuracy.
We developed a highly accurate tool that uses basic clinical and analytical information to predict the probability of survival at 15, 30, and 60 days in terminally ill cancer patients. This tool can help physicians making decisions on clinical care at the end of life.
确定终末期癌症患者的预期寿命是一项艰巨的任务。我们旨在开发和验证一种列线图，以预测终末期疾病患者的生存时间。
从 2003 年 2 月 1 日至 2005 年 12 月 31 日，连续纳入 406 例终末期患者进行研究。我们通过多变量 Cox 回归分析了 38 个与终末期患者预期寿命相关的特征，并根据赤池信息量准则通过向后变量消除法确定了最准确和简约的模型。将 5 个临床和实验室变量构建到一个列线图中，以估计患者在 15、30 和 60 天时的生存概率。我们使用 2006 年 6 月 1 日至 2007 年 12 月 31 日期间治疗的 474 例患者的外部验证队列验证和校准了该列线图。
训练组的中位总生存期为 29.1 天，验证组为 18.3 天。东部合作肿瘤学组表现状态、乳酸脱氢酶水平、淋巴细胞水平、白蛋白水平以及从初始诊断到终末期疾病的诊断时间是生存的独立预后因素，保留在多变量 Cox 比例风险模型中，并构成了该列线图的基础。该列线图具有较高的预测性能，经 bootstrap 校正的一致性指数为 0.70，并且具有良好的校准度。外部独立验证显示预测准确率为 68%。
我们开发了一种高度准确的工具，该工具使用基本的临床和分析信息来预测终末期癌症患者在 15、30 和 60 天时的生存概率。该工具可以帮助医生在生命末期做出临床决策。
