ations, unless one postulates that drinkers of coffee (with or
without caffeine) have a better memory for unhealthy behavior
than non-drinkers.
The diet data reported here were not directly validated;
we know of no way to confirm dietary fat and saturated/
unsaturated fat ratios short of direct observation or a fat
biopsy.28 Indirect validation of both food frequency and
24-hour diet recall was obtained in this cohort by demonstrating
expected diet-disease associations. We have reported
the expected cross-sectional associations of sodium,
calcium, and potassium with blood pressure,29.30 and inverse
associations of fiber with heart attack;31 potassium with
stroke32 and calcium with hip fracture in prospective studies.33
Other attributes were also indirectly validated: smoking
was inversely associated with HDL and directly with waisthip-ratio;34
alcohol was associated directly with waist-hipratio,
aspartate aminotransferase and HDL;35 exercise was
positively associated with HDL and inversely associated
with obesity, pulse rate, and plasma insulin.36 Any misclassification
of self-reported behaviors would be expected to
reduce these associations, and not have caused them.