Leveraging questionnaire-based physical activity levels and the Goldberg method to evaluate the biological plausibility of self-reported energy intake: a doubly labeled water validation study

The Goldberg method is a statistical approach suggested for identifying implausible energy intakes (EI) in nutritional epidemiology. However, estimating physical activity levels (PAL), required for the method, is challenging. We quantified the accuracy of the Goldberg method combined with PAL derived from the Sedentary Time and Activity Reporting Questionnaire (Goldberg-PALSTAR-Q). Total energy expenditure derived from doubly labeled water (TEEDLW) was the comparator. Between 2009 and 2011, men (n=40) and women (n=59) completed a two-week DLW protocol, a food frequency questionnaire, and the STAR-Q. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy of the Goldberg-PALSTAR-Q were determined. Fifty-eight percent of men and women were classified as EI under-reporters by the Goldberg-PALSTAR-Q compared with 60% of men and 56% of women by TEEDLW. Among men, sensitivity, specificity, PPV, NPV and accuracy and 95% confidence intervals were 88% (68%-97%), 87% (61%-98%), 91% (72%-99%), 81% (56%-94%), and 87% (72%-95%), respectively; and among women 79% (62%-90%), 69% (50%-84%), 77% (60%-88%), 72% (52%-86%), and 75% (62%-84%), respectively. In studies where objective EI measures are lacking, the Goldberg method and individual-level questionnaire-derived PALs provide a viable option in sensitivity analyses exploring the influence of implausible EIs on diet and health-related associations.