Slimani Nadia, Bingham Sheila, Runswick Shirley, Ferrari Pietro, Day Nicholas E, Welch Ailsa A, Key Timothy J, Miller Antony B, Boeing Heiner, Sieri Sabina, Veglia Fabrizio, Palli Dominico, Panico Salvatore, Tumino Rosario, Bueno-De-Mesquita Bas, Ocké Marga C, Clavel-Chapelon Françoise, Trichopoulou Antonia, Van Staveren Wija A, Riboli Elio
Unit of Nutrition and Cancer, IARC-WHO, 69372 Lyon Cedex 08, France.
Cancer Epidemiol Biomarkers Prev. 2003 Aug;12(8):784-95.
A calibration approach was developed to correct for systematic between-cohort dietary measurement errors in the European Prospective Investigation into Cancer and Nutrition (EPIC), a large multicenter cohort study. To validate the 24-h diet recalls (24-HDRs) as reference measurements for between-cohort calibration, we estimated the agreement between center mean nitrogen (N) and total energy intakes and mean 24-h urinary N. Similar analyses using N and energy intake data from different dietary questionnaires (DQs) used at study baseline were conducted to estimate the effect of the calibration approach. This study was conducted between 1995 and 1999, and involved 1103 volunteers of both genders from 12 centers participating in European Prospective Investigation into Cancer and Nutrition. Pearson's correlation coefficients were weighted for study center sample size. When both genders were considered together (n = 22), the correlation coefficients between the center mean log-transformed urinary estimates and the center mean log-transformed dietary N estimates from the 24-HDRs were 0.86 and 0.94 after exclusion of outliers. The corresponding correlation with the DQs was 0.53. When center mean total energy intakes were regressed on center mean urinary N, the correlation remained slightly higher with 24-HDRs (0.91; 0.95 after exclusion of outliers) than DQs (0.86). When stratified by gender, these correlations were systematically higher in men than women with both dietary methods. The beta regression coefficients were not significantly different from 1 when mean N (or total energy intakes) from 24-HDR or DQ were regressed on urinary estimates, except with N from 24-HDRs in men and, in most cases, after adjustment for age, body mass index, and sex with both genders together. This suggests that overall the systematic bias across centers is of uniform magnitude. Although relatively high correlations were observed between urinary N and both dietary methods in men, the errors in DQs tend to vary in both directions (under- and over-reporting) in contrast with 24-HDRs in women. This observation may have implications on the dietary measurement error characteristics and support the potential benefit of between-cohort calibration.
在一项大型多中心队列研究——欧洲癌症与营养前瞻性调查（EPIC）中，开发了一种校准方法，以校正队列间饮食测量的系统误差。为了验证24小时饮食回顾（24-HDR）作为队列间校准的参考测量方法，我们估计了各中心平均氮（N）摄入量、总能量摄入量与平均24小时尿氮之间的一致性。使用研究基线时不同饮食问卷（DQ）中的氮和能量摄入数据进行了类似分析，以评估校准方法的效果。本研究于1995年至1999年进行，涉及来自12个参与欧洲癌症与营养前瞻性调查中心的1103名男女志愿者。皮尔逊相关系数根据研究中心样本量进行加权。当将男女一起考虑时（n = 22），排除异常值后，各中心平均对数转换后的尿氮估计值与24-HDR中各中心平均对数转换后的饮食氮估计值之间的相关系数分别为0.86和0.94。与DQ的相应相关系数为0.53。当各中心平均总能量摄入量对各中心平均尿氮进行回归时，与24-HDR的相关性（排除异常值后为0.91；0.95）仍略高于DQ（0.86）。按性别分层时，两种饮食方法在男性中的这些相关性均系统地高于女性。当24-HDR或DQ中的平均氮（或总能量摄入量）对尿氮估计值进行回归时，β回归系数与1无显著差异，但男性中24-HDR的氮以及大多数情况下男女一起校正年龄、体重指数和性别后除外。这表明总体而言，各中心的系统偏差幅度一致。尽管在男性中观察到尿氮与两种饮食方法之间的相关性相对较高，但与女性的24-HDR相比，DQ中的误差往往在两个方向（少报和多报）上有所不同。这一观察结果可能对饮食测量误差特征有影响，并支持队列间校准的潜在益处。
