Differing statistical approaches affect the relation between egg consumption, adiposity, and cardiovascular risk factors in adults.

Background: Associations between food patterns and adiposity are poorly understood.

Objective: Two statistical approaches were used to examine the potential association between egg consumption and adiposity.

Methods: Participants (n = 18,987) aged ≥19 y were from the 2001-2008 NHANES who provided 24-h diet recall data, body mass index (BMI) and waist circumference (WC)-determined adiposity measures, and blood pressure, circulating insulin, glucose, and lipid concentrations were considered cardiovascular risk factors (CVRFs). Covariate-adjusted least-squares means ± SEs were generated.

Results: The first statistical approach categorized participants into egg consumers or nonconsumers. Consumers had higher mean BMI (in kg/m(2); 28.7 ± 0.19; P = 0.006) and WC (98.2 ± 0.43 cm; P = 0.002) than did nonconsumers (28.2 ± 0.10 and 96.9 ± 0.23 cm, respectively). Second, cluster analysis identified 8 distinct egg consumption patterns (explaining 39.5% of the variance in percentage of energy within the food categories). Only 2 egg patterns [egg/meat, poultry, fish (MPF)/grains/vegetables and egg/MPF/grains], consumed by ≤2% of the population, drove the association (compared with the no-egg pattern) between egg consumption and BMI and WC. Another analysis controlled for the standard covariates and the other food groups consumed with eggs in those 2 egg patterns. Only the egg/MPF/other-grains pattern remained associated with BMI and WC (both P ≤ 0.0063). The pattern analyses identified associations between an egg pattern (egg/MPF/other grains/potatoes/other beverages) and diastolic blood pressure (DBP) and serum LDL cholesterol (both P ≤ 0.0063). A final analysis was conducted by adding percentage of energy from fast foods and medication use for diabetes to the covariates. The association between the egg/MPF/grains pattern and BMI and the egg/MPF/potatoes/other beverages and DBP and LDL cholesterol disappeared.

Conclusions: Care needs to be taken with data interpretation of diet and health risk factors and the choice of statistical analyses and covariates used in the analyses because these studies are typically used to generate hypotheses. Additional studies are needed to better understand these relations.