Kipnis V, Freedman L S, Brown C C, Hartman A M, Schatzkin A, Wacholder S
Division of Cancer Prevention and Control, National Cancer Institute, Bethesda, MD 20892-7354, USA.
Am J Epidemiol. 1997 Nov 15;146(10):842-55. doi: 10.1093/oxfordjournals.aje.a009202.
The use and interpretation of energy-adjustment regression models in nutritional epidemiology has been vigorously debated recently. There has been little discussion, however, regarding the effect of dietary measurement error on the performance of such models. Contrary to conventional assumptions invoked in the standard treatment of the effect of measurement error in regression analysis, reporting errors in dietary studies are usually biased, correlated with true nutrient intakes and with each other, heteroscedastic, and nonnormally distributed. Methods developed in this paper allow for this more complex error structure and are therefore more appropriate for dietary data. For practical illustration, these methods are applied to data from the Women's Health Trial Vanguard Study. The results demonstrate considerable shrinkage in the magnitude of the estimated main exposure effect in energy-adjustment models due to attenuation of the true effect and contamination from the effect of an adjusting covariate. In most cases, this shrinkage causes a sharply reduced statistical power of the corresponding significance test in comparison with measurement without error. These results emphasize the need to understand the measurement error properties of dietary instruments through validation/calibration studies and, where possible, to correct for the impact of measurement error when applying energy-adjustment models.
能量调整回归模型在营养流行病学中的应用和解读最近引发了激烈争论。然而，关于膳食测量误差对这类模型性能的影响，却很少有讨论。与回归分析中测量误差效应的标准处理所采用的传统假设相反，膳食研究中的报告误差通常存在偏差，与真实营养素摄入量相关且相互关联，具有异方差性，并且呈非正态分布。本文所开发的方法考虑了这种更为复杂的误差结构，因此更适用于膳食数据。为了进行实际说明，这些方法被应用于妇女健康试验先锋研究的数据。结果表明，由于真实效应的衰减以及调整协变量效应的干扰，能量调整模型中估计的主要暴露效应大小出现了显著收缩。在大多数情况下，与无误差测量相比，这种收缩导致相应显著性检验的统计功效大幅降低。这些结果强调了通过验证/校准研究了解膳食工具测量误差特性的必要性，并在可能的情况下，在应用能量调整模型时校正测量误差的影响。
