A weighted logistic regression model for estimation of recurrence of adenomas
A weighted logistic regression model for estimation of recurrence of adenomas
复制标题
DOI：
10.1002/sim.2648
复制
发表时间：
2007-03-30
影响因子：
2
通讯作者：
He, Yulei
中科院分区：
文献类型：
--
作者：
Hsu, Chiu-Hsieh;Green, Sylvan B.;He, Yulei
在结肠直肠息肉预防试验中，一些参与者可能会在预定时间之前（即在试验结束时）进行结肠镜检查。这导致受试者随访时间长短不一，试验结束时的复发状态数据可视为当前状态数据。在本文中，我们使用加权逻辑回归模型来估计试验结束时腺瘤数据的复发率。权重用于调整变量跟踪。我们发现逻辑回归倾向于低估复发率。在一项模拟研究中，我们表明，与逻辑回归相比，从当前状态数据的右端点导出的Kaplan-Meier估计量倾向于高估复发率，而加权逻辑回归方法与传统逻辑回归和Kaplan-Meier估计相比，即使在高不合规率下也能产生合理的复发率估计。这里用一个结肠癌研究的例子来说明这种方法。版权所有(c) 2006约翰威利父子有限公司
In a colorectal polyp prevention trial, some participants might have their follow-up colonoscopy conducted before the scheduled time (i.e. at the end of the trial). This results in variable follow-up lengths for participants and the data of recurrence status at the end of the trial can be considered as current status data. In this paper, we use a weighted logistic regression model to estimate recurrence rate of adenoma data at the end of the trial. The weights are used to adjust for variable follow-up. We show that logistic regression tends to underestimate recurrence rate. In a simulation study, we show that Kaplan-Meier estimator derived from the right endpoint of the current status data tends to overestimate recurrence rate in contrast to logistic regression and the weighted logistic regression method can produce reasonable estimates of recurrence rate even under a high non-compliance rate compared to conventional logistic regression and Kaplan-Meier estimator. The method described here is illustrated with an example from a colon cancer study. Copyright (c) 2006 John Wiley & Sons, Ltd.