Although measurements
of dietary intake based on a single 24–hour recall per individual
of the large sample size can provide a good probable
estimate of the mean for a study population, the
standard deviation will be greatly overestimated. Furthermore,
the random within-person variation may reduce the
strength of associations under the study, possibly to the
point of being undetectable. However, trained dietitians
employed standard protocols in helping the subjects in
reporting their daily diet to minimize plausible errors.
Third, recruitment from voluntary participation as well
as non-participation or non-responses might have introduced
a selection bias. But, general characteristics did not
differ significantly between included and excluded subjects
and high recruitment rate in main centers of the study
minimizes the chance that results were distorted by selection
bias