Department of Population Science, American Cancer Society, Atlanta, GA, USA.
Department of Epidemiology, Rollins School of Public Health Emory University, Atlanta, GA, USA.
J Nutr. 2021 Mar 11;151(3):636-648. doi: 10.1093/jn/nxaa358.
Valid assessment of dietary intake in diverse populations is important for studies of chronic disease risk in the United States.
We evaluated the reproducibility and validity of a food frequency questionnaire (FFQ) modified for the American Cancer Society's Cancer Prevention Study-3 (CPS-3) prospective cohort, among a racially/ethnically diverse subgroup.
The Diet Assessment Substudy included 677 CPS-3 participants (64% women; 61% non-Hispanic white, 24% non-Hispanic black, 15% Hispanic), aged 31-70 y, who completed 2 FFQs 1 y apart (FFQ1, FFQ2), 4-6 telephone-administered 24-h dietary recalls (24HRs), and 2 fasting blood samples and 24-h urine collections ∼6 mo apart in the interim. Spearman rank correlation coefficients (ρ) were used to evaluate FFQ reproducibility and validity compared with 24HRs for 67 nutrient exposures. For 18 of these nutrients, we used the method of triads to calculate validity coefficients (VCs, ρ) from pairwise correlations of FFQ2, 24HRs, and biomarkers. Analyses were stratified by sex, race/ethnicity, education, and BMI.
Mean (range) FFQ reproducibility correlations were ρ = 0.65 (0.50-0.91) for men and ρ = 0.63 (0.37-0.89) for women; mean (range) energy-adjusted, deattenuated correlations of FFQ2 with 24HRs were ρ = 0.60 (0.33-0.84) for men and ρ = 0.55 (0.21-0.79) for women. FFQ2 VCs (ρ) among men ranged from 0.42 for β-cryptoxanthin to 0.91 for omega-3 (n-3) fatty acids and, among women, from 0.41 for sodium to 0.79 for total vitamin D. Mean FFQ reproducibility and validity were highest among whites (ρ = 0.68, ρ = 0.58, respectively) and slightly lower among blacks (ρ = 0.57, ρ = 0.49, respectively) and Hispanics (ρ = 0.59, 0.55, respectively). FFQ reproducibility and validity were slightly lower among those with less than a 4-y college degree, and those with a BMI ≥30 kg/m2.
Reproducibility and validity of the CPS-3 FFQ were comparable with similar studies for most nutrients, among all subgroups. These findings support future dietary analyses in the contemporary CPS-3 cohort and other similar cohorts.
在美国，为了研究慢性病的风险，对不同人群进行饮食摄入的有效评估非常重要。
我们评估了美国癌症协会癌症预防研究-3（CPS-3）前瞻性队列中经过修改的食物频率问卷（FFQ）在一个种族/民族多样化亚组中的可重复性和有效性。
饮食评估子研究包括 677 名 CPS-3 参与者（64%为女性；61%为非西班牙裔白人，24%为非西班牙裔黑人，15%为西班牙裔），年龄在 31-70 岁之间，他们在 1 年内完成了 2 次 FFQ（FFQ1、FFQ2），4-6 次电话进行的 24 小时膳食回忆（24HR），以及在间隔期间进行了 2 次空腹血样和 24 小时尿液采集。Spearman 等级相关系数（ρ）用于评估 FFQ 与 67 种营养素的 24HR 之间的可重复性和有效性。对于其中的 18 种营养素，我们使用三联体法从 FFQ2、24HR 和生物标志物的两两相关关系中计算有效性系数（VC，ρ）。分析按性别、种族/民族、教育程度和 BMI 进行分层。
男性的平均（范围）FFQ 可重复性相关系数为 ρ=0.65（0.50-0.91），女性为 ρ=0.63（0.37-0.89）；男性能量调整后，FFQ2 与 24HR 之间的去衰减相关性平均（范围）为 ρ=0.60（0.33-0.84），女性为 ρ=0.55（0.21-0.79）。男性的 FFQ2 VC（ρ）范围从β-隐黄质的 0.42 到 n-3 脂肪酸的 0.91，女性的 VC 范围从钠的 0.41 到总维生素 D 的 0.79。白人的平均 FFQ 可重复性和有效性最高（ρ=0.68，ρ=0.58），黑人的稍低（ρ=0.57，ρ=0.49），西班牙裔的更低（ρ=0.59，ρ=0.55）。具有不到 4 年大学学历的参与者以及 BMI≥30kg/m2 的参与者的 FFQ 可重复性和有效性略低。
在所有亚组中，CPS-3 FFQ 的可重复性和有效性与大多数营养素的类似研究相当。这些发现支持在当代 CPS-3 队列和其他类似队列中进行未来的饮食分析。
