Department of Biostatistics, Boston University School of Public Health, Boston, MA, USA.
Stat Med. 2012 Jun 30;31(14):1517-30. doi: 10.1002/sim.4500. Epub 2012 Feb 17.
We present a nonparametric test to validate surrogate endpoints based on measure of divergence and random permutation. This test is a proposal to directly verify the Prentice statistical definition of surrogacy. The test does not impose distributional assumptions on the endpoints, and it is robust to model misspecification. Our simulation study shows that the proposed nonparametric test outperforms the practical test of the Prentice criterion in terms of both robustness of size and power. We also evaluate the performance of three leading methods that attempt to quantify the effect of surrogate endpoints. The proposed method is applied to validate magnetic resonance imaging lesions as the surrogate endpoint for clinical relapses in a multiple sclerosis trial.
我们提出了一种基于分歧度量和随机排列的非参数检验方法，用于验证替代终点。该检验方法是对 Prentice 替代终点统计学定义的直接验证，不对终点进行分布假设，并且对模型误设具有稳健性。我们的模拟研究表明，该非参数检验在大小和功效的稳健性方面均优于 Prentice 准则的实用检验。我们还评估了三种尝试量化替代终点效果的领先方法的性能。该方法应用于验证多发性硬化症试验中磁共振成像病变作为临床复发的替代终点。
