underreporting (rant)

[Wuzzy]

Right place to comment? Underreporting of kcal intake is an
odd phenomenon:

In particular,

1 or 2 convicing studies show higher kcal in lower bmi
persons(of same gender). Some have attributed this to
"correlated underreporting", slightly more bodyweight means
more risk of embarassment about foods. This reasoning isn't
entirely convincing to me since my thinking is that
"restrained" eaters (the people who eat healthy) are more
aware of their diet and more embarassed of admitting to eating
unhealthy foods daily - these are also the healthier and more
educated persons at supermarkets (perhaps my bias) The paradox
is also attributed to physical activity:
Ie. the sudden drop in kcal with increasing weight at certain
ages depends on sedentary work. The relationship of kcal
on weight is highly sensitive to age. This is also
convincing since most people around the world consume
similar kcal about 2000kcal while work activity varies
highly. This is unquestionable but I think that
"unhealthy" persons are eating "unhealthy" diets and that
this should be more easily demonstrable..

aside, my thinking is that it is easy to test whether it is
correlated underreporting or activity: match by age and weight
and thus there is no reason to think that person will
(un)consciously underreport compared to their match. In this
case if the paradox persists then it must be due to activity..

Another way is to use a variable defined as kcal/bmi and
exclude the extremes ie., those who eat too much or too little
for their bmi, since these are obviously cheating. Anyway it
is frustrating not to know whether to trust measures of
absolute kcal.. hence most researchers adjust for it, eg., %
fat is believed to be accurate whereas total fat may not be.
Also, types(quality) of foods may be a more strategic measure
of diet that overcomes this doubt.

okay, thats why I think there is discrepancy in this topic..

[Pbeyer]

wuzzy wrote:

> Right place to comment? Underreporting of kcal intake is an
> odd phenomenon:
>
> In particular,
>
> 1 or 2 convicing studies show higher kcal in lower bmi
> persons(of same gender). Some have attributed this to
> "correlated underreporting", slightly more bodyweight means
> more risk of embarassment about foods. This reasoning isn't
> entirely convincing to me since my thinking is that
> "restrained" eaters (the people who eat healthy) are more
> aware of their diet and more embarassed of admitting to
> eating unhealthy foods daily - these are also the healthier
> and more educated persons at supermarkets (perhaps my bias)
> The paradox is also attributed to physical activity:
> Ie. the sudden drop in kcal with increasing weight at
> certain ages depends on sedentary work. The relationship
> of kcal on weight is highly sensitive to age. This is
> also convincing since most people around the world
> consume similar kcal about 2000kcal while work activity
> varies highly. This is unquestionable but I think that
> "unhealthy" persons are eating "unhealthy" diets and
> that this should be more easily demonstrable..
>
> aside, my thinking is that it is easy to test whether it is
> correlated underreporting or activity: match by age and
> weight and thus there is no reason to think that person will
> (un)consciously underreport compared to their match. In this
> case if the paradox persists then it must be due to
> activity..
>
> Another way is to use a variable defined as kcal/bmi and
> exclude the extremes ie., those who eat too much or too
> little for their bmi, since these are obviously cheating.
> Anyway it is frustrating not to know whether to trust
> measures of absolute kcal.. hence most researchers adjust
> for it, eg., % fat is believed to be accurate whereas total
> fat may not be. Also, types(quality) of foods may be a more
> strategic measure of diet that overcomes this doubt.
>
> okay, thats why I think there is discrepancy in this topic..

Good ideas, Wuz-- seems like in my incomplete-but
sometimes-correct recall that somebody has studied subjective
responses from dietary histories/records against objective
evidence of intakes (tagged foods, cameras, weighed foods at
food checkouts, photos, etc)-- and then correlated that with
the characteristics of the individuals-- I'll check with one
of my friends who studies that kind of thing, but in the
meantime-- anyone else recall any study showing objective
intake vs reported intake by body size, etc? Pete

[Wuzzy]

Thanks Pete, to play on the "exercise-causes high kcal intake"
explanation of the paradox (BMI=-kcal) see a thread in Am J of
Epidemiology that I cite below - here the authors as well as
Willett, who also criticizes the author, agree that BMI=-kcal
is due to exercise, not underreporting. The authors show
convincingly that max VO2 correlates with kcal intake and that
this explains the paradox negative trend: bmi=-kcal.

"correlated underreporting" might not explain the "neatness"
of the trend, bmi=-kcal since it is highly linear (R^2=high)
and not confined to the extreme BMIs.

A debate to keep an eye on, at least, particularly since low
kcal intake has benefits other than on BMI such as on
cardiovascular risk and on lowering insulin requirements which
can be shown experimentally: This benefit of low kcal intake
might not show in large-scale studies unless well conducted.

REFERENCES:
1)Dietary measures of physical activity. Am J. Epi. 120 (6)
1984. -A good study, there are two follow ups to this in
later AJE issues: Willett argues that the authors'
conclusions are correct but their methods are wrong: their
regression equation was bodyfat=-kcal/weight which may only
reflect the inverse mathematical relation here. The authors
subsequently corrected their logic by following Willett's
energy residual method and came to same conclusions.

2)An Assessment of Caloric Intake as an Indicator of Physical
Acitvity. Prev Med. 1989 (18) 444-451. -Same author as (1),
"David R Jacobs", but this time uses a large population with
single 24-hh.

[Jay Tanzma]

pbeyer wrote:

[snip]

> Good ideas, Wuz-- seems like in my incomplete-but
> sometimes-correct recall that somebody has studied
> subjective responses from dietary histories/records against
> objective evidence of intakes (tagged foods, cameras,
> weighed foods at food checkouts, photos, etc)-- and then
> correlated that with the characteristics of the
> individuals...

The accuracy of dietary recalls has been studied using a much
more rigorous "gold standard": total energy expenditures
measured by the doubly labeled water technique. In one study
(see below), the general population underreported intake by
about 18%. Obese and formerly obese women were worse,
underreporting their intakes by 27% and 36%, respectively.

-Jay

J Am Diet Assoc 1993 May;93(5):572-9

Measurements of total energy expenditure provide insights into
the validity of dietary measurements of energy intake.

Black AE, Prentice AM, Goldberg GR, Jebb SA, Bingham SA,
Livingstone MB, Coward WA.

Dunn Clinical Nutrition Centre, Cambridge, United Kingdom.

The quantification of errors inherent in methods of measuring
dietary intake has been handicapped by the absence of
independent markers for testing their validity. The doubly
labeled water technique permits a precise measure of energy
expenditure in free-living persons. Because energy expenditure
must equal energy intake in populations in energy balance,
this technique may be used to validate the assessment of
energy intake. A series of studies demonstrated good agreement
between mean energy intake and mean energy expenditure when
food intake was recorded by observers or when it was
self-reported by normal-weight, self-selected, highly
motivated volunteer subjects using weighed records. However,
in randomly recruited men and women, energy intake by weighed
records was 82% and 81%, of energy expenditure, respectively,
indicating underestimation of habitual intake. Men and women
in the lowest third of reported intake recorded energy
expenditure of only 69% and 61%, respectively. Reported intake
of obese and previously obese women was only 73% and 64% of
expenditure, whether measured by weighed record or by diet
history, confirming suspicions that these subjects
misrepresented their intake. Acceptable weighed records were
obtained from 7- and 9-year-olds whereas 15- and 18-year-olds
underestimated intake. Diet histories taken from the same
children tended to overestimate intake. These studies suggest
that, ideally, all dietary studies should include independent
measures of validity.

[Wuzzy]

mypcos@hotmail.com (wuzzy) wrote in message
> aside, my thinking is that it is easy to test whether it is
> correlated underreporting or activity: match by age and
> weight and thus there is no reason to think that person will
> (un)consciously underreport compared to their match. In this
> case if the paradox persists then it must be due to
> activity..

just what i learned is that matching is a tricky subject
since, for example when matching for gender there may be an
interaction with gender that'll confound your results: say
heavier males eat more but heavier females eat less, when you
match for gender the effects will cancel (average) out.. Hence
matching is sometimes inferior technique.(as you quickly
discover matching is extremely limited)

In the above case you might have an interaction of bmi, age
and appetite and matching may not be a good way to deal
with this..

[Wuzzy]

mypcos@hotmail.com (wuzzy) wrote in message
> The authors show convincingly that max VO2 correlates with
> kcal intake and that this explains the paradox negative
> trend: bmi=-kcal.

ie., assume most people are at constant kcal balance:

this can be seen using followup data, for instance follow up
at 5 years looks like:

Paired Samples Correlations Correlation Pair 1 KCAL & KCAL2
0.48533982 Pair 2 INSULIN & INSULIN2 0.854905549 Pair 3 Waist
circumference (cm) (2+ yea & WAISTCR2 0.975730549 Pair 4
Weight(kg) & WEIGHTK2 0.975953917 Pair 5 Height(cm) & HEIGHT2
0.98934447

Note the correlation between waist-circumference can be 98%
predicted within 5years.. ie., most people are in net
caloric balance in what is visible in a few years - over 15
years this may be change though too gradually to detect in
small intervals? note 24HH kcal correlates 0.50 over 5 year
period, which is still pretty good.. insulin is modest at
85% suggesting there is a bit of unexplained variance over 5
year period..

[Pbeyer]

Jay Tanzman wrote:

> pbeyer wrote:
>
> [snip]
>
> > Good ideas, Wuz-- seems like in my incomplete-but
> > sometimes-correct recall that somebody has studied
> > subjective responses from dietary histories/records
> > against objective evidence of intakes (tagged foods,
> > cameras, weighed foods at food checkouts, photos, etc)--
> > and then correlated that with the characteristics of the
> > individuals...
>
> The accuracy of dietary recalls has been studied using a
> much more rigorous "gold standard": total energy
> expenditures measured by the doubly labeled water technique.
> In one study (see below), the general population
> underreported intake by about 18%. Obese and formerly obese
> women were worse, underreporting their intakes by 27% and
> 36%, respectively.
>
> -Jay
>
> J Am Diet Assoc 1993 May;93(5):572-9
>
> Measurements of total energy expenditure provide insights
> into the validity of dietary measurements of energy intake.
>
> Black AE, Prentice AM, Goldberg GR, Jebb SA, Bingham SA,
> Livingstone MB, Coward WA.
>
> Dunn Clinical Nutrition Centre, Cambridge, United Kingdom.
>
> The quantification of errors inherent in methods of
> measuring dietary intake has been handicapped by the absence
> of independent markers for testing their validity. The
> doubly labeled water technique permits a precise measure of
> energy expenditure in free-living persons. Because energy
> expenditure must equal energy intake in populations in
> energy balance, this technique may be used to validate the
> assessment of energy intake. A series of studies
> demonstrated good agreement between mean energy intake and
> mean energy expenditure when food intake was recorded by
> observers or when it was self-reported by normal-weight,
> self-selected, highly motivated volunteer subjects using
> weighed records. However, in randomly recruited men and
> women, energy intake by weighed records was 82% and 81%, of
> energy expenditure, respectively, indicating underestimation
> of habitual intake. Men and women in the lowest third of
> reported intake recorded energy expenditure of only 69% and
> 61%, respectively. Reported intake of obese and previously
> obese women was only 73% and 64% of expenditure, whether
> measured by weighed record or by diet history, confirming
> suspicions that these subjects misrepresented their intake.
> Acceptable weighed records were obtained from 7- and
> 9-year-olds whereas 15- and 18-year-olds underestimated
> intake. Diet histories taken from the same children tended
> to overestimate intake. These studies suggest that, ideally,
> all dietary studies should include independent measures of
> validity.

Jay-- nice "dig", by the way. I was "off line for a while and
couldn't respond. Pete