Continuing Education Questionnaire, page 769
Meets Learning Need Codes 3000, 3010, 9000, and 9070
Development and Evaluation of a Short
Instrument to Estimate Usual Dietary
Intake of Percentage Energy from Fat
FRANCES E. THOMPSON, PhD, MPH; DOUGLAS MIDTHUNE, MS; AMY F. SUBAR, PhD, MPH, RD; VICTOR KIPNIS, PhD; LISA L. KAHLE;
ARTHUR SCHATZKIN, MD
Objective To describe the data-based development of a
short dietary assessment instrument, a 16-item screener;
and to evaluate the performance of the screener, compar-
ing its performance with a complete 120-item food fre-
quency questionnaire (FFQ) in assessing percentage en-
ergy from fat intake.
Design A subsample (n?404) of participants in the Na-
tional Institutes of Health-AARP Diet and Health Study,
who had completed an FFQ and two 24-hour dietary
recalls, also completed the fat screener. Percentage en-
ergy from fat from the screener and from the FFQ were
compared with estimated true usual intake using a mea-
surement error model.
Results For men, the mean percentage energy from fat
estimates for the different methods were: recalls, 30.1%,
screener, 29.9%; FFQ, 30.4%. For women, the results
were: recalls, 31.3%, screener, 28.4%, FFQ, 30.0%. Esti-
mated correlations between true intake and screener
were 0.64 and 0.58 for men and women, respectively, and
between true intake and FFQ were 0.67 for men and 0.72
for women. Estimated attenuation coefficients for the
screener were 1.29 (men) and 0.98 (women) and for the
FFQ were 0.56 (men) and 0.57 (women).
Conclusions The percentage energy from fat screener,
when used in conjunction with external reference data,
may be useful to compare mean intakes of fat for different
population subgroups, and to examine relationships be-
tween fat intake and other factors.
J Am Diet Assoc. 2007;107:760-767.
Year 2010 (1), the Dietary Guidelines for Americans 2005
(2), and the American Heart Association (3) all set goals
or guidelines that limit the percentage of energy from
total fat, limit saturated fats, and encourage that fat
intake derive primarily from polyunsaturated and mono-
unsaturated fats within one’s energy requirements.
These recommendations are all consistent with the 2002
Dietary Reference Intakes (4), which set an acceptable
distribution range of 20% to 35% of energy from total fat.
The rationale for this range is that exceeding the upper
end is associated with increased intake of both saturated
fat and energy intake, both of which are associated with
increased risk of chronic disease, whereas consuming less
than the lower end is associated with increased plasma
triacylglycerol concentrations and decreased high-density
lipoprotein cholesterol concentrations from high-carbohy-
drate diets (4).
All of these guidelines and recommendations recognize
that healthful diets must include intakes of a range of
food groups to meet nutrient requirements but must do so
while maintaining energy balance and, thus, cannot in-
clude excessive fats. Therefore, percentage of energy from
fat continues to serve as a useful indicator of this dimen-
The interviewer-administered 24-hour dietary recall
provides the most accurate and complete self-reported
information about an individual’s diet for a given day (5).
otal intakes of fat and types of fat continue to figure
prominently in national health objectives and di-
etary guidance. The 2010 Health Objectives for the
F. E. Thompson is an epidemiologist and A. F. Subar is
a nutritionist, Applied Research Program, Division of
Cancer Control and Population Sciences, D. Midthune
and V. Kipnis are mathematical statisticians, Biome-
try Research Group, and A. Schatzkin is chief, Nutri-
tional Epidemiology Branch, Division of Cancer Epide-
miology and Genetics, all with the National Cancer
Institute, Bethesda, MD. L. L. Kahle is senior systems
analyst, Information Management Services, Inc, Silver
Address correspondence to: Frances E. Thompson, PhD,
MPH, EPN 4016, 6130 Executive Blvd, MSC 7344, Be-
thesda, MD 20892-7344. E-mail: email@example.com
Copyright © 2007 by the American Dietetic
Journal of the AMERICAN DIETETIC ASSOCIATION
© 2007 by the American Dietetic Association
In the United States, the 24-hour dietary recall has been
used in national surveys to characterize the population’s
dietary intake and to monitor progress toward the
Healthy People 2010 dietary intake objectives (1). How-
ever, because the 24-hour dietary recall requires highly
trained interviewers and substantial coding and data pro-
cessing, it is prohibitively expensive for many applica-
tions. In some situations, self-administered food fre-
quency questionnaires (FFQs), often optically scannable
for inexpensive data entry, are used. Complete quantita-
tive FFQs themselves often require up to an hour to
complete and, thus, are not feasible in many studies.
Shorter tools that measure a limited number of dietary
factors rather than the entire diet have been developed
(5). Among them are short dietary instruments that aim
to score and rank individuals based on their fat intakes
(6-11). In 1995 when this study began, there were no
short screening instruments that aimed to quantify fat
intake as a percentage of total energy. Our objective was
to develop such an instrument using national food con-
sumption data and to test its relative validity.
Development of Instrument
Identification of Foods. To identify which foods best pre-
dicted percentage energy from fat, we analyzed dietary
intake data (one 24-hour recall and two 1-day records)
collected from a nationally representative sample of the
United States in the US Department of Agriculture’s
1989-91 Continuing Survey of Food Intakes by Individu-
als (CSFII) (12), the most recent survey available at the
time. We categorized all foods reported on this survey
into 193 mutually exclusive food groups. For each survey
individual and food group, we calculated the number of
times that food group was reported over the 3 days. The
resulting food group variables represented possible fre-
quency of use questions that could be asked on a short
instrument to estimate percentage energy from fat.
We then calculated weighted least-squares regression
for all individuals age 50 to 69 years, using stepwise
selection procedures to select a small number of food
groups most predictive of percentage energy from fat. We
limited the analyses to this age range because we planned
to test the instrument in a study population composed
only of those ages. The response variable was percentage
energy from fat, and the independent variables were
mentions of each food group over the 3 days of report.
The strongest predictors of percentage energy from fat
were, in order: nonfat milk; eggs, fat added; bananas;
rice, no fat added; cheese, regular fat; margarine, regular
fat; salad dressing, regular fat; fried potatoes; ready-to-
eat cereals; hot dogs, regular; sausage, regular; butter;
citrus juice; mayonnaise, regular; and apples/applesauce.
Some of these foods were directly related to percentage
energy from fat (eg, eggs and cheese); some were in-
versely related to percentage energy from fat (eg, nonfat
milk, bananas, and rice). Taken together, these items
accounted for 34% of the variability in percentage energy
from fat in the CSFII 1989-91 adults age 50 to 69 years.
Although other food groups explained additional variabil-
ity, none explained more than 2% of total variability.
These results and information from cognitive testing to
enhance responses on food frequency type questions (13)
were used to guide the construction of the screener. Some
larger food groupings were used rather than individual
foods (eg, all fruits rather than bananas and applesauce),
because cognitive testing suggested that respondents
think of these interchangeably. The resulting percentage
energy from fat screener consists of frequency questions
for 15 food groups, and a separate question querying use
of reduced-fat margarine or butter (14). That question is
used in conjunction with three previous frequency ques-
tions about fat use to create a variable to estimate overall
frequency of non–reduced-fat margarine and butter.
Development of Scoring Algorithms to Estimate Percentage Energy
from Fat. Identification of the foods composing the
screener used CSFII 1989-91 data, but the CSFII 1994-96
survey data (15) (released after development and admin-
istration of the screener in this study), were used to
develop scoring algorithms to translate an individual’s
responses into estimated dietary intake. The US Depart-
ment of Agriculture’s 1994-96 CSFII is also a represen-
tative sample of the US population, with data from two
nonconsecutive 24-hour recalls. We created food group
variables similar to those created for the earlier 1989-91
CSFII dataset and consistent with the questions on the
Initially, scoring algorithms were estimated only for
adults age 50 to 69 years. However, subsequent use of the
screener in other populations led us to re-estimate the
scoring algorithms for the general adult population (aged
18 and older). Because portion size varies with sex and
age (16), we incorporated sex–age (by 10-year age groups)
portion size estimates into the scoring procedure. For
each sex–age group, we computed the median amount
(grams) of each food group consumed per mention on the
1994-96 CSFII (14). We then estimated regression coeffi-
cients, using the CSFII 1994-96 and the following model:
Exp (pcalfat)??0??1NFG1P1??2NFG2P2?. . .??13NFG13P13
Exp (pcalfat) is the expected value of percentage energy
NFGkis the mean number of times per day an individ-
ual consumed food group k
Pkis the median portion size of group k for that indi-
vidual’s sex–age group
k indexes the 13 food groups.
We calculated weighted least-squares estimates of the
regression coefficients ?k, k?0, . . . , 13, on CSFII 1994-96
adults aged 18 and older, stratifying by sex and excluding
outliers (defined as values more than 2 interquartile
ranges below the first quartile or 2 interquartile ranges
above the third quartile of percentage energy from fat).
Because of the complex survey design in CSFII, the anal-
ysis was done with survey weights using the Survey Data
Analysis computer package (Research Triangle Institute,
Cary, NC) (17).
The food groups included in the screener account for
27% and 30% of the variability in intake of percentage
energy from fat for men and women, respectively, in the
CSFII 1994-96. (See reference 14 for the estimated re-
gression parameters and all scoring algorithms.)
May 2007 ● Journal of the AMERICAN DIETETIC ASSOCIATION
TESTING OF THE INSTRUMENT
To test the performance of the percentage energy from fat
screener, we offered it to a random sample of members of
the Calibration Study of the National Institutes of Health
(NIH)-AARP Diet and Health Study. The NIH-AARP
Diet and Health Study is a large prospective cohort of
approximately 567,000 individuals, age 50 to 69 years at
baseline, formed to examine the relationship of baseline
diet with later outcomes of cancer (18). The cohort study
included a calibration study to calibrate the relationship
of dietary intake estimates based on an FFQ with esti-
mates from two nonconsecutive 24-hour recalls, described
more fully by Thompson and colleagues (19). Approxi-
mately 2,000 individuals, with extremes of intake in at
least one of several dietary variables of interest (percent-
age energy from fat, fiber, fruits and vegetables, and red
meats) were chosen into the calibration study using a
stratified sampling design.
Two 24-hour recalls were administered from February
through August 1996. A second FFQ was mailed in Octo-
ber 1996. In June 1997, individuals in the calibration
study (N?1,975) were randomly divided into thirds, and
mailed one of three short dietary assessment instru-
ments: two different fruit and vegetable screeners, and
the percentage energy from fat screener. Results compar-
ing the fruit and vegetable screeners are published (20).
Of the 658 percentage energy from fat screeners mailed,
474 (72%) were returned. We excluded participants who
were subsequently excluded from the AARP cohort (n?3)
and 67 individuals with missing data. Percentage energy
from fat was estimated for 404 individuals (51% men,
35% age 50 to 59 years).
Data Collection and Processing and Variable Creation: 24-Hour
Dietary Recall. Two nonconsecutive 24-hour dietary recalls
were administered by trained interviewers via the tele-
phone. The second recall was obtained a median of 21
days after the first recall. A set of measuring guides was
sent to participants for use during the interview. Partic-
ipants were not told in advance when they would be
interviewed; interviews were scheduled to reflect the dis-
tribution of weekdays and weekend days. Participants
were asked to report all foods and beverages consumed on
the day before the interview, from midnight to midnight.
Interviewers used a food probe list containing standard-
ized probes. Data were coded using version 2.3 of the
Food Intake Analysis System (21). Percentage energy
from fat was computed for each individual and day of
recall as: 9 (kcal/gm fat)?fat (gm)/energy (kcal)?100(%).
Percentage Energy from Fat Screener. Each frequency cate-
gory was assigned a value equal to the midpoint of that
category and converted to times per day. The scoring
algorithm described previously, using regression param-
eters developed from CSFII data, was applied to the
respondent’s percentage energy from fat screener an-
swers and the appropriate sex-age specific portion sizes to
estimate individual intake of percentage energy from
FFQ. The FFQ asked both frequency and portion size for
approximately 120 foods, and additional questions spe-
cific to fat intake. Survey data from the 1994-96 CSFII
were used to construct a database that included, for each
food on the FFQ, mean nutrient values for each of six
different strata, defined by sex and portion size (small,
medium, or large) (22). For each individual, total daily fat
and energy intake were computed by multiplying that
individual’s reported frequency by the nutrient amount
assigned for that individual’s stratum for each food on the
FFQ, summing across all foods, and standardizing to
reflect daily mean intake. Percentage energy from fat was
computed as described earlier for the screener.
Analytical Methods. The variable of interest, true usual in-
take of percentage energy from fat, is not observable in
free-living populations. However, we can estimate its dis-
tribution in the population, and its relation to reported
intake as measured by a test instrument (screener or
FFQ), by use of appropriate reference data and statistical
methods. Our reference instrument is two nonconsecutive
24-hour dietary recalls. We applied a measurement error
model (23) to estimate the relationship between true and
reported intake, assuming that the reference instrument
(24-hour recall) is unbiased at the individual level and
includes only within-person error. The measurement er-
ror model is:
where Fijis the jth repeat reference measurement (24-
hour recall) for the ith individual, Tiis the (unobservable)
true usual intake for the ith individual, uijis within-
person random error in the reference instrument, Qijis
reported intake from the jth repeat screener or question-
naire, ?0is the intercept and ?1is the slope in the linear
regression of Qijon Ti, and eijis within-person random
error in the test instrument.
The model assumes that random variables Ti, uij, and eij
are uncorrelated, that Tihas mean ?Tand variance ?T
and that uijand eijhave mean zero and variances ?u
Other important measures can be derived from the
parameters of the measurement error model, including
the correlation coefficient (R), or correlation between true
and reported intake, and the attenuation factor (?), which
is the slope in the regression of T on Q. The squared
correlation coefficient (R2) measures the proportion of
variation in true intake that is explained by reported
intake. Both the correlation coefficient and attenuation
factor are important for evaluating an instrument used in
studies of diet–disease relationships. The squared corre-
lation coefficient is inversely proportional to the sample
size needed to attain a specified power to detect a rela-
tionship, whereas the attenuation factor is the multipli-
cative bias in observed log relative risk that is due to
measurement error in the dietary instrument. In dietary
studies, ? is typically between 0 and 1, and therefore
attenuates (biases toward 0) the observed log relative
The scoring algorithm for the screener uses a regres-
sion model to calibrate the screener in an external data-
set (CSFII 1994-96), using the 24-hour recall as the ref-
erence instrument. As a result, under the measurement
error model the screener should have an attenuation
factor of approximately 1 (leading to unbiased estimates
May 2007 Volume 107 Number 5
of log relative risk in a diet-disease study), and should
have approximately the same mean but smaller variance
than T. Thompson and colleagues (24) describe a method
of adjusting the screener to have approximately the same
mean and variance as T. The variance-adjusted screener is
where (?T??Q), called the “variance adjustment factor,”
is the ratio of the standard deviation of true usual intake
to the standard deviation of screener-estimated intake,
and ?Qis the mean of Qij. Because the AARP calibration
study has both a screener and repeat 24-hour recalls, we
can estimate the variance adjustment factor using the
measurement error model described earlier. In AARP,
the estimated variance adjustment factor for the percent-
age energy from fat screener was 2.0 for men and 1.7 for
women. We used the variance-adjusted screener to esti-
mate the prevalence of attaining the recommended diet of
less than 30% energy from fat, and estimated the ad-
justed screener’s sensitivity and specificity at two differ-
ent levels of true intake.
Because of the stratified sampling design of the cali-
bration study, we used a weighted method of moments to
estimate parameters in the measurement error model.
Because percentage energy from fat was approximately
normally distributed for all instruments, no transforma-
tion to normality was necessary before fitting the model.
We excluded from the analysis subjects with outlier val-
ues, as previously defined, for screener, FFQ, or 24-hour
dietary recall, to avoid their undue influence. One man
and two women were excluded from the analysis because
Data collection for the NIH-AARP Diet and Health
Study was approved by the National Cancer Institute and
Westat Institutional Review Boards.
Estimates of Mean Intakes and Distributions
Table 1 shows the estimated mean percentage energy
from fat for the screener relative to true intake and prev-
alence of consuming the recommended diet of less than
30% energy from fat. Among men, percentage energy
from fat estimates from the screener and the recalls were
nearly identical. However, among women, the screener
significantly underestimated the true mean intake of per-
centage energy from fat. The same relationships were
found between the FFQ and estimated true intake: sim-
ilar estimates among men and underestimates from the
FFQ among women. These differences were mirrored in
prevalence estimates of meeting the recommended in-
takes: prevalences as estimated by the FFQ and the
screener were similar to that of true intake among men
and were somewhat different among women.
Estimates of Regression Parameters
Table 2 shows estimated parameters from regressing per-
cent energy from fat from the screener and the FFQ on
true intake, by sex. For the screener, the slope was well
less than 1.0, showing the “flattened slope syndrome”; ie,
that those who consumed higher percentage energy from
fat diets underreported on the screener and those who
consumed lower percentage energy from fat diets overre-
ported on the screener relative to the 24-hour dietary
recall. However, the correlation coefficients for both men
and women were 0.6. Attenuation coefficients for both
sexes were near 1.0, showing that the calibration proce-
dures incorporated in the scoring algorithms were effec-
For the FFQ, estimated slope coefficients were statis-
tically significantly higher than those for the screener.
Conversely, estimated attenuation coefficients were sta-
tistically significantly lower than those for the screener.
Correlation coefficients for the screener and the FFQ
Sensitivity and Specificity
Table 3 shows estimates of sensitivity and specificity by
sex for the screener and the FFQ, at two different cutoff
levels. Among men, at a cutoff of 30% energy from fat, the
screener and the FFQ are similar in sensitivity and spec-
ificity, approximately 70% for both. Among women at this
cutoff, the FFQ is somewhat more sensitive than the
screener because the mean values are better estimated by
the FFQ than the screener. At this level, which is close to
the mean value of the sample, variance adjustment of the
screener does not significantly improve its estimates.
However, at higher cutoff values, shown here as 35%,
further into the tails of the distributions, the FFQ is
substantially more sensitive than the screener, and the
variance adjustment for the screener dramatically im-
proves its sensitivity.
Two differing methods have been used to query about fat
intake. One method asks about particular behavior pat-
terns (eg, not eating the skin on chicken, eating fruit for
Table 1. Estimated mean percentage energy from fat and preva-
lence of intakes less than or equal to 30% energy from fat for true
intake,ascreener, and FFQ,bby sex: NIHc-AARP Diet and Health
Energy from Fat
Energy from Fat
4 ™™™™™ mean (SEd) ™™™™ 3
4™™™™™™ % ™™™™ 3
aDistribution of true intake was estimated from two nonconsecutive 24-hour recalls
using a measurement error model. All parameters were estimated by weighted method
of moments from the Freedman and colleagues (23) measurement error model.
bFFQ?food frequency questionnaire.
cNIH?National Institutes of Health.
eStatistically significantly (P?0.05) different from true intake.
May 2007 ● Journal of the AMERICAN DIETETIC ASSOCIATION
dessert), and seeks to rank individuals by their fat intake,
not estimate the actual amount of fat. A second method
uses a food frequency format. For the FFQ type of instru-
ments, the dominant approach used to identify which
foods to ask about has been the “nutrient consumption
approach,” which selects foods based on their nutrient
contribution in the population. The nutrient consumption
approach has been used by Block and colleagues to de-
velop short fat assessment questionnaires in the United
States (6,25,26). The alternative “variance-based ap-
proach” selects foods based on their ability to predict
differences in the variability of nutrient consumption.
The variance-based approach has been used to develop a
shortened food questionnaire for assessment of fats in
Germany (27,28). We adopted this latter approach to
identify the foods that are most predictive of variability in
percentage energy from fat in the United States. Not all
foods that are major sources of fat are major predictors of
percentage energy from fat. For example, in the United
States, beef, which is a major contributor to fat, is not a
major factor in explaining the variability in percentage
energy from fat because of its widespread consumption.
The short screener tested in this sample seems to do
nearly as well as a longer, more quantitative FFQ. Among
men, both the screener and the FFQ estimated mean
population intake within 0.3% energy from fat, and
among women, both the screener and the FFQ underes-
timated mean population intake. Both instruments were
similar in their ability to rank individuals, with correla-
tions of 0.6 to 0.7. For epidemiological purposes of relat-
ing diet to disease, the screener seems to have a theoret-
ical advantage over the FFQ. The scoring algorithms used
for the screener include calibration to multiple 24-hour
dietary recalls of the US adult population, assuring an
attenuation coefficient close to 1.0 under the assumption
that the 24-hour dietary recall is an unbiased reference
instrument. In this NIH-AARP study sample, the esti-
mated attenuation coefficients for the screener were 1.3
(men) and 0.9 (women), and, as expected, compare favor-
ably to those for the uncalibrated FFQ (0.6 for both men
and women). Thus, in this sample risk estimates based on
the uncalibrated FFQ would be attenuated substantially,
whereas those based on the screener would be relatively
Calibration of the FFQ to multiple 24-hour dietary
recalls or food records through calibration/validation sub-
studies has been proposed for use in epidemiologic stud-
ies (29). Such substudies have the virtue of calibrating
the FFQ within the population being studied, but they
are expensive and logistically complex, and are impracti-
cal for the kinds of studies that would use the screener. In
contrast, because the screener is precalibrated, albeit
within an external population (CSFII), a calibration sub-
study is not necessary.
Good distributional information is necessary to make
judgments about prevalence, for example, the percentage
of the population attaining a certain dietary goal. Be-
cause the variance of the screener is much smaller than
the variance of true usual intake, the screener is not
appropriate for estimating prevalence except near the
mean. We have attempted to overcome this limitation by
estimating variance adjustment factors, in this case from
Table 2. Estimates of slope in the regression of reported on true intake,acorrelation between reported and true intakes, and attenuation
coefficient (and standard errors) resulting from measurement error in reported intake for percentage energy from fat for screener and FFQb, by
sex: NIHc-AARP Diet and Health Study
Sex ScreenerFFQ Screener
Correlation CoefficientsAttenuation Coefficients
4 ™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™ estimate (SEd) ™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™ 3
0.31 (0.05)0.80 (0.13)0.64 (0.08)
0.34 (0.06)0.90 (0.14)0.58 (0.07)
aDistribution of true intake was estimated from two nonconsecutive 24-hour recalls using a measurement error model.
bFFQ?food frequency questionnaire.
cNIH?National Institutes of Health.
Table 3. Sensitivity and specificity of the screener and the FFQarelative to true intake, at 30% and 35% energy from fat cutoff values, by sex:
NIHb-AARP Diet and Health Study
>30% Energy from Fat
>35% Energy from Fat
>30% Energy from Fat
>35% Energy from Fat
Screener FFQ FFQFFQ
aFFQ?food frequency questionnaire.
bNIH?National Institutes of Health.
cEstimates without variance adjustment.
May 2007 Volume 107 Number 5
multiple 24-hour dietary recalls on the same sample. This
procedure effectively adjusts the variance of the screener
to mimic true intake, as estimated from multiple recalls.
However, generally one would not have recalls or records
in a study using screeners. External variance adjustment
factors, those estimated from other studies, could be used
to better approximate the true distributions, under the
assumption that the factor is constant over populations.
At this time, this study provides the only variance adjust-
ment factors available for the percentage energy from fat
The measurement error model used in these analyses
relies on 24-hour dietary recall data. Recalls are subject
to error in measuring usual intake from day-to-day vari-
ability and from reporting error. The model does not
require that the 2 days of intake perfectly estimate the
individual’s usual intake, but it does require that the
24-hour dietary recall be unbiased at the individual level,
with only within-person random error. However, studies
using doubly labeled water and/or urinary nitrogen have
shown that energy and protein intakes are often under-
reported with the 24-hour dietary recall (30-34). If the
amount of underreporting were constant across all indi-
viduals, the mean of true intake would be underesti-
mated, but the variance of true intake and the regression
parameters (the slope, correlation coefficient, and atten-
uation coefficient) would be unchanged. However, there is
evidence that individuals may systematically differ in
their reporting accuracy (35-39). Kipnis and colleagues
have proposed the potential of intake-related bias and
person-specific bias in the 24-hour dietary recall (40);
analysis of data from the Observing Protein and Energy
Nutrition study (41) provided empirical evidence that
biases exist for reported energy, protein, and percentage
energy from protein, and that use of the 24-hour dietary
recall as a reference instrument leads to overestimation
of the correlation and attenuation coefficients for these
nutrients. Although we do not know the error-structure
for percentage energy from fat, it is quite possible that
correlation and attenuation coefficients are overesti-
mated in this study. We therefore consider the compari-
son of screener and FFQ in this study to be a comparison
of relative, rather than absolute, validity.
As with other short instruments, the percentage energy
from fat screener’s ability to accurately assess the indi-
vidual’s diet is limited. In addition, its use in intervention
studies with self-selected and potentially biased partici-
pants has not been evaluated, nor has it been evaluated
in populations with unusual diets or with limited ability
to self-report diet. The performance of the percentage
energy from fat screener in four different intervention
sites of the NIH Behavior Change Consortium (42) is
currently being evaluated and should help address these
issues of potential biases in report and generalizability.
The screener requires the use of more accurate refer-
ence data to enable appropriate and timely estimates of
portion size and calibration coefficients. In this regard,
the 2003-2004 National Health and Nutrition Examina-
tion Survey dietary data will be used to update the in-
strument, including its scoring algorithms. Finally, dis-
factors should be considered approximate rather than
In the absence of more accurate dietary methods, the
percentage energy from fat screener may be useful to
characterize population intakes of percentage of energy
from fat, allowing comparisons across subpopulations
and across time for the same population. In addition, the
brevity of the instrument may allow incorporation into
larger study instruments, in which interrelationships be-
tween percentage energy from fat and other factors could
be usefully examined. An abbreviated version of the per-
centage energy from fat screener, using similar develop-
ment and validation methods (43), has been effectively
used in this manner in the 2000 National Health Inter-
view Survey (24). Finally, with incorporation of appropri-
ate variance adjustment, the screener may allow approx-
imate estimates of prevalence and serve as an effective
The authors thank the participants in the NIH-AARP
Diet and Health Study for their outstanding cooperation.
In addition, the authors acknowledge the work of
Westat, Inc, in conducting the NIH-AARP Diet and
Health Study. In particular, the authors thank Westat’s
NIH-AARP Diet and Health Study Project Director, Paul
Hurwitz, and the study nutritionist, Susan McNutt, for
their work on the Calibration Substudy.
1. US Department of Health and Human Services.
Tracking Healthy People 2010. Washington, DC: US
Government Printing Office; 2000.
2. US Department of Health and Human Services, US
Department of Agriculture. Dietary Guidelines for
Americans, 2005. 6th ed. Washington, DC: US Gov-
ernment Printing Office, January 2005.
3. American Heart Association. An Eating Plan for
americanheart.org. Accessed May 25, 2006.
4. Institute of Medicine. Dietary Reference Intakes for
Energy, Carbohydrate, Fiber, Fat, Fatty Acids, Cho-
lesterol, Protein, and Amino Acids. Washington, DC:
National Academies Press; 2002/2005.
5. Thompson FE, Subar AF. Dietary assessment meth-
odology. In: Coulston AM, Rock CL, Monsen ER, eds.
Nutrition in the Prevention and Treatment of Disease.
San Diego, CA: Academic Press; 2001:3-30.
6. Block G, Clifford C, Naughton MD, Henderson M,
McAdams M. A brief dietary screen for high fat in-
take. J Nutr Educ. 1989;21:199-207.
7. Kristal AR, Shattuck AL, Henry HJ. Patterns of di-
etary behavior associated with selecting diets low in
fat: Reliability and validity of a behavioral approach
to dietary assessment. J Am Diet Assoc. 1990;90:214-
8. Kemppainen T, Rosendahl A, Nuutinen O, Ebeling T,
Pietinen P, Uusitupa M. Validation of a short dietary
questionnaire and a qualitative fat index for the as-
sessment of fat intake. Eur J Clin Nutr. 1993;47:765-
9. Dobson AJ, Blijlevens R, Alexander HM, Croce N,
Heller RF, Higginbotham N, Pike G, Plotnikoff R,
Russell A, Walker R. Short fat questionnaire: A self-
May 2007 ● Journal of the AMERICAN DIETETIC ASSOCIATION
administered measure of fat-intake behaviour. Aust J
Public Health. 1993;17:144-149.
10. Coates RJ, Serdula MK, Byers T, Mokdad A, Jewell
S, Leonard SB, Ritenbaugh C, Newcomb P, Mares-
Perlman J, Chavez N, Block G. A brief, telephone-
administered food frequency questionnaire can be
useful for surveillance of dietary fat intakes. J Nutr.
11. Retzlaff BM, Dowdy AA, Walden CE, Bovbjerg VE,
Knopp RH. The Northwest Lipid Research Clinic Fat
Intake Scale: Validation and utility. Am J Public
12. US Department of Agriculture. 1989-91 Continuing
Survey of Food Intakes by Individuals and 1989-91
Diet and Health Knowledge Survey [on CD-ROM].
Beltsville, MD: Agricultural Research Service, Belts-
ville Human Nutrition Research Center; 1996. NTIS
accession number: PB96-501747.
13. Subar AF, Thompson FE, Smith AF, Jobe JB, Ziegler
RG, Potischman N, Schatzkin A, Hartman A, Swan-
son C, Kruse L, Hayes RB, Lewis DR, Harlan LC.
Improving food frequency questionnaires: A qualita-
tive approach using cognitive interviewing. J Am
Diet Assoc. 1995;95:781-788.
14. Applied Research Program, National Cancer Insti-
tute, National Institutes of Health. Percent Energy
from Fat Screener. Available at: http://www.riskfactor.
cancer.gov/diet/screeners/fat. Accessed May 25, 2006.
15. US Department of Agriculture. What We Eat in
America 1994-96, Continuing Survey of Food Intakes
by Individuals (CSFII) 1994-1996 [public use CD-
ROM]. Washington, DC: Agricultural Research Ser-
vice, USDA; 1998. NTIS Accession No. PB98-500457.
16. Krebs-Smith SM, Guenther PM, Cook A, Thompson
FE, Cucinelli J, Udler J. Foods Commonly Eaten in
the United States: Quantities Consumed per Eating
Occasion and in a Day, 1989-91. NFS Report No.
91-3. Washington, DC: US Department of Agricul-
17. Research Triangle Institute. SUDAAN User’s Man-
ual, Release 8.0. Research Triangle Park, NC: Re-
search Triangle Institute; 2001.
18. Schatzkin A, Subar AF, Thompson FE, Harlan LC,
Tangrea J, Hollenbeck AR, Hurwitz PE, Coyle L,
Schussler N, Michaud DS, Freedman LS, Brown CC,
Midthune D, Kipnis V. Design and serendipity in
establishing a large cohort with wide dietary intake
distributions: The National Institutes of Health–
American Association of Retired Persons Diet and
Health Study. Am J Epidemiol. 2001;154:1119-1125.
19. Thompson FE, Kipnis V, Midthune D, Freedman LS,
Carroll RJ, Subar AF, Brown CC, Butcher MS, Mouw
T, Leitzmann M, Schatzkin A. Performance of a food
frequency questionnaire in the U.S. National Insti-
tutes of Health-AARP Diet and Health Study. Pub
Health Nutr. In press.
20. Thompson FE, Kipnis V, Subar AF, Krebs-Smith SM,
Kahle LL, Midthune D, Potischman N, Schatzkin A.
Evaluation of 2 brief instruments and a food-fre-
quency questionnaire to estimate daily number of
servings of fruit and vegetables. Am J Clin Nutr.
21. Douglass D. Food Intake Analysis System (FIAS),
Version 2.3. Houston, TX: The University of Texas;
22. Subar AF, Midthune D, Kulldorff M, Brown CC,
Thompson FE, Kipnis V, Schatzkin A. Evaluation of
alternative approaches to assign nutrient values to
food groups in food frequency questionnaires. Am J
23. Freedman LS, Carroll RJ, Wax Y. Estimating the
relation between dietary intake obtained from a food
frequency questionnaire and true average intake.
Am J Epidemiol. 1991;134:310-320.
24. Thompson FE, Midthune D, Subar AF, McNeel T,
Berrigan D, Kipnis V. Dietary intake estimates in the
National Health Interview Survey, 2000: Methodol-
ogy, results, and interpretation. J Am Diet Assoc.
25. Block G, Gillespie C, Rosenbaum EH, Jenson C. A
rapid food screener to assess fat and fruit and vege-
table intake. Am J Prev Med. 2000;18:284-288.
26. Spencer EH, Elon LK, Hertzberg VS, Stein AD,
Frank E. Validation of a brief diet survey instrument
among medical students. J Am Diet Assoc. 2005;
27. Rohrmann S, Klein G. Development and validation of
a short food list to assess the intake of total fat,
saturated, mono-unsaturated, polyunsaturated fatty
acids and cholesterol. Eur J Public Health. 2003;13:
28. Rohrmann S, Klein G. Validation of a short question-
naire to qualitatively assess the intake of total fat,
saturated, monounsaturated, polyunsaturated fatty
acids, and cholesterol. J Hum Nutr Diet. 2003;16:111-
29. Kaaks R, Riboli E, van Staveren W. Calibration of
dietary intake measurements in prospective cohort
studies. Am J Epidemiol. 1995;142:548-556.
30. Johnson RK, Soultanakis RP, Matthews DE. Literacy
and body fatness are associated with underreporting
of energy intake in US low-income women using the
multiple-pass 24-hour recall: A doubly labeled water
study. J Am Diet Assoc. 1998;98:1136-1140.
31. Kroke A, Klipstein-Grobusch K, Voss S, Moseneder J,
Thielecke F, Noack R, Boeing H. Validation of a self-
administered food-frequency questionnaire adminis-
tered in the European Prospective Investigation into
Cancer and Nutrition (EPIC) Study: Comparison of
energy, protein, and macronutrient intakes esti-
mated with the doubly labeled water, urinary nitro-
gen, and repeated 24-h dietary recall methods. Am J
Clin Nutr. 1999;70:439-447.
32. Bathalon GP, Tucker KL, Hays NP, Vinken AG,
Greenberg AS, McCrory MA, Roberts SB. Psycholog-
ical measures of eating behavior and the accuracy of
3 common dietary assessment methods in healthy
postmenopausal women. Am J Clin Nutr. 2000;71:
33. Tran KM, Johnson RK, Soultanakis RP, Matthews
DE. In-person vs telephone-administered multiple-
pass 24-hour recalls in women: Validation with dou-
bly labeled water. J Am Diet Assoc. 2000;100:777-
34. Hill RJ, Davies PS. The validity of self-reported en-
May 2007 Volume 107 Number 5
ergy intake as determined using the doubly labelled Download full-text
water technique. Br J Nutr. 2001;85:415-430.
35. Livingstone MB, Prentice AM, Strain JJ, Coward
WA, Black AE, Barker ME, McKenna PG, Whitehead
RG. Accuracy of weighed dietary records in studies of
diet and health. BMJ. 1990;300:708-712.
36. Heitmann BL. The influence of fatness, weight
change, slimming history and other lifestyle vari-
ables on diet reporting in Danish men and women
aged 35-65 years. Int J Obes Relat Metab Disord.
37. Martin LJ, Su W, Jones PJ, Lockwood GA, Tritcher
DL, Boyd NF. Comparison of energy intakes deter-
mined by food records and doubly labeled water in
women participating in a dietary-intervention trial.
Am J Clin Nutr. 1996;63:483-490.
38. Sawaya AL, Tucker K, Tsay R, Willett W, Saltzman
E, Dallal GE, Roberts SB. Evaluation of four methods
for determining energy intake in young and older
women: Comparison with doubly labeled water mea-
surements of total energy expenditure. Am J Clin
39. Black AE, Bingham SA, Johansson G, Coward WA.
Validation of dietary intakes of protein and energy
against 24 urinary N and DLW energy expenditure in
middle-aged women, retired men and post-obese sub-
jects: Comparisons with validation against presumed
energy requirements. Eur J Clin Nutr. 1997;51:
40. Kipnis V, Carroll RJ, Freedman LS, Li L. Implica-
tions of a new dietary measurement error model for
estimation of relative risk: Application to four cali-
bration studies. Am J Epidemiol. 1999;150:642-651.
41. Schatzkin A, Kipnis V, Carroll RJ, Midthune D, Su-
bar AF, Bingham S, Schoeller DA, Troiano RP, Freed-
man LS. A comparison of a food frequency quest-
ionnaire with a 24-hour recall for use in an
epidemiological cohort study: Results from the bio-
marker-based Observing Protein and Energy Nutri-
tion (OPEN) study. Int J Epidemiol. 2003;32:1054-
42. National Institutes of Health. Behavior Change
Consortium. Available at: http://www1.od.nih.gov/
behaviorchange. Accessed January 20, 2006.
43. Thompson FE, Midthune D, Subar AF, Kahle LL,
Schatzkin A, Kipnis V. Performance of a short tool to
assess dietary intakes of fruits and vegetables, per-
centage energy from fat and fibre. Public Health
May 2007 ● Journal of the AMERICAN DIETETIC ASSOCIATION