Nutr Hosp. 2015;31(Supl.

3):49-56
ISSN 0212-1611 • CODEN NUHOEQ
S.V.R. 318

Food Frequency Questionnaires

Carmen Pérez Rodrigo1, Javier Aranceta2, Gemma Salvador3 and Gregorio Varela-Moreiras4,5
Fundación FIDEC. Bilbao. 2Departamento de Medicina Preventiva y Salud Pública. Universidad de Navarra. 3Agència de Salud
1

Pública de Catalunya. Departamento de Salut. Generalitat de Catalunya. 4Department of Pharmaceutical and Health Sciences,
Faculty of Pharmacy, CEU San Pablo University; Madrid. 5Spanish Nutrition Foundation (FEN), Madrid (SPAIN).

Abstract MÉTODOS DE FRECUENCIA

DE CONSUMO ALIMENTARIO
Food Frequency Questionnaires are dietary assess-
ment tools widely used in epidemiological studies inves-
tigating the relationship between dietary intake and di- Resumen
sease or risk factors since the early ‘90s. The three main Los cuestionarios de frecuencia de consumo son herra-
components of these questionnaires are the list of foods, mientas ampliamente utilizadas en los estudios epidemio-
frequency of consumption and the portion size consu- lógicos que investigan la relación entre ingesta dietética
med. The food list should reflect the food habits of the y enfermedad o factores de riesgo desde comienzos de la
study population at the time the data is collected. The década de los 90. Los tres componentes principales de es-
frequency of consumption may be asked by open ended tos cuestionarios son la lista de alimentos, la frecuencia
questions or by presenting frequency categories. Quali- de consumo y el tamaño de la ración consumida. La lista
tative Food Frequency Questionnaires do not ask about de alimentos debe reflejar los hábitos de consumo de la
the consumed portions; semi-quantitative include stan- población de estudio en el momento en que se recogen
dard portions and quantitative questionnaires ask res- los datos. La frecuencia de consumo puede preguntarse
pondents to estimate the portion size consumed either de forma abierta u ofreciendo categorías de frecuencia
in household measures or grams. The latter implies a de consumo. Los cuestionarios cualitativos no preguntan
greater participant burden. Some versions include only por la ración consumida; los semi-cuantitativos presen-
close-ended questions in a standardized format, while tan raciones estándar y los cuestionarios cuantitativos
others add an open section with questions about some solicitan al encuestado que estime el tamaño de la ración
specific food habits and practices and admit additions consumida en medidas caseras o en gramos. Esta última
to the food list for foods and beverages consumed which opción supone un esfuerzo importante para el partici-
are not included. The method can be self-administered, pante. Existen versiones exclusivamente cerradas en un
on paper or web-based, or interview administered either formato estandarizado y otras que incorporan preguntas
face-to-face or by telephone. Due to the standard format, abiertas sobre algunos hábitos y prácticas alimentarias
especially closed-ended versions, and method of adminis- específicas y permiten añadir alimentos y bebidas con-
tration, FFQs are highly cost-effective thus encouraging sumidos que no están incluidos en la lista. Pueden ser
its widespread use in large scale epidemiological cohort auto-administrados, en papel o en soporte web, o bien
studies and also in other study designs. Coding and pro- mediante entrevista personal o telefónica. Por su for-
cessing data collected is also less costly and requires less mato estandarizado, especialmente los cerrados, y por
nutrition expertise compared to other dietary intake as- la forma de administración, son un método con un alto
sessment methods. However, the main limitations are sys- rendimiento en términos de coste-efectividad que ha fa-
tematic errors and biases in estimates. Important efforts vorecido su extendido uso en grandes estudios epidemio-
are being developed to improve the quality of the infor- lógicos de cohortes y también con otros diseños. También
mation. It has been recommended the use of FFQs with el coste de codificación y proceso de la información reco-
other methods thus enabling the adjustments required. gida resulta menos costoso y requiere menos experiencia
(Nutr Hosp 2015;31(Supl. 3):49-56) en temas nutricionales que otros métodos de evaluación
de la ingesta. Sin embargo, presenta el inconveniente de
DOI:10.3305/nh.2015.31.sup3.8751 incorporar errores sistemáticos y sesgos importantes, por
Key words: Food habits. Dietary assessment. Error vali- lo que en la actualidad se buscan procedimientos para
dity. Population studies. mejorar la calidad de la información y se recomienda uti-
lizarlos junto a otros métodos que permitan realizar los
ajustes necesarios.
(Nutr Hosp 2015;31(Supl. 3):49-56)
DOI:10.3305/nh.2015.31.sup3.8751
Palabras clave: Hábitos alimentarios. Evaluación ingesta
dietética. Error validez. Estudios poblacionales.
Correspondence: Carmen Pérez Rodrigo.
Fundación FIDEC.
Gurtubay s/n. 48010 Bilbao (Spain).
E-mail: carmenperezrodrigo@gmail.com

49

006 Metodo de Frecuencia_Carmen Perez Rodrigo.indd 49 12/02/15 13:43

 Abbreviations may consist of an extensive or a relatively short list
of food items. The foods included should be major
FFQ: Food Frequency Querstionnaire. sources of a group of nutrients of particular interest
HDL: High Density Lipoprotein. for the purpose of the study in which the instrument is
EPIC: European Prospective Investigation into Can- to be used or foods which contribute to the variability
cer and Nutrition. in intake between individuals in the population. Addi-
NCI: National Cancer Institute. tionally, should be foods commonly consumed in the
DHQ: Diet History Questionnaire. study population2,3,7,8 and reflect the food habits and
common practices in that particular group. The length
of the food list can range from about 20 to 200 items.
Introduction FFQs should be developed specifically for each
study group and research purposes because diet may
Food Frequency Questionnaires (FFQ) are an ad- be influenced by ethnicity, culture, an individual’s pre-
vanced form of the diet history method aimed to assess ference, economic status, etc. and the appropriateness
habitual diet by asking how often and how much of of the food list is essential in this method of diet as-
selected food items or specific food groups included in sessment3,7-10.
a list are consumed over a reference period1,2. FFQs can ask the respondent to report either a com-
This method was originally designed to provide des- bined frequency for a particular food eaten both alone
criptive qualitative information about food-consump- and in mixtures or separate frequencies for each food
tion patterns and later developed to provide nutrient use3,4,10. However, people who is not used to cook or
information by specifying an average portion size. Ac- prepare meals may find it difficult to identify the in-
cording to the interests of the researchers, FFQs may gredients involved in mixed dishes and it is likely to
focus on the intake of specific nutrients, dietary expo- ignore how often consume certain food items.
sures related to a certain disease, or comprehensively Each quantitative FFQ must be associated with a
assess various nutrients1-3. food composition database to allow for the estimation
Instruments including about 100 to150 food items of nutrient intakes for the either assumed or reported
take 20-30 minutes to complete and can be comple- portion size of each food item included3,11,12. Such a
ted during an interview or can be self-administered. database is created using quantitative dietary intake
Thus, this method enables the assessment of long term information from the target population to estimate the
dietary intake in a relative simple, cost-effective and typical nutrient density of a particular food group ca-
time-efficient manner2-4. tegory. A mean or median nutrient composition is es-
FFQs have been widely used in epidemiological stu- timated based on the individual food codes reported
dies investigating links between diet and disease. For in a population survey. In this case, dietary analyses
this purpose it is more important to rank the intake of software, specific to each FFQ, is used to compute nu-
individuals in comparison to others in the population trient intakes for individual respondents11,12.
as high, medium, or low intake or as quantiles of the Comprehensive FFQ instruments designed to as-
distribution of intake, than to determine the absolute sess total diet generally list more than 100 food items,
intake3,5,6. In large epidemiological studies, data on many with additional portion size questions, requiring
diet from FFQs are compared with specific disease 30 to 60 minutes to complete. This raises concern
outcomes like cancer diagnosis, or risk factors for di- about length and its effect on response rates. Balan-
sease such as total or HDL-cholesterol levels. FFQs ce between length and specificity of the food list is
may also be used to identify food patterns associated required3,7-10,13. Optically scanned instruments require
with inadequate intakes of specific nutrients3. the use of closed-ended response categories forcing a
loss in specificity.

Components of FFQs
Frequency of consumption
The main components of FFQs are the food list, the
frequency of consumption in time units and the portion Frequency of consumption is assessed by a multiple
size consumed of each item. response grid or independent questions asking respon-
dents to estimate how often a particular food or bevera-
ge is consumed2,3. Frequency categories range from ne-
The food list ver or less than once per month up to 6 or more times per
day and respondents have to choose one of these options.
The food list should be clear, concise, structured Most FFQs collect data using nine possible responses.
and organized in a systematic way2,3. It can be newly Various answer choices have been used to improve data
designed specifically for the study or can be modified quality and reduce the burden on the subjects13.
from an existing instrument, but in that case it must be The reference period for which the frequency of
adapted and validated for the study population. FFQs consumption is asked can be variable, but usually fo-

50 FOOD AND BEVERAGE CONSUMPTION ASSESSMENT: METHODS AND CONTROVERSIES

006 Metodo de Frecuencia_Carmen Perez Rodrigo.indd 50 12/02/15 13:43

 cuses over the past six months or the past year, but it It is controversial whether FFQs should use portion
is possible to ask about the past week or month de- size questions, since reporting portion sizes poses ad-
pending on specific research situations. Even when ditional respondent burden3. Frequency has been re-
respondents are asked about intake over the past ported to be a greater contributor than typical serving
year, some studies indicate that the season in which size to the variance in intake of most foods, but other
the questionnaire is administered influences reporting authors cite small improvements in the performance of
over the entire year3,14. FFQs that ask the respondents to report a usual serving
The format of independent questions can be variable. size for each food13,15. Some research has been conduc-
Sometimes multiple choice questions are used, with an ted to determine the best ways to ask about portion size
optimal number of 5 to 10 closed, exhaustive and mu- on FFQs. Nevertheless the marginal benefit of such in-
tually exclusive answer options (Fig. 1). In other cases, formation in a particular study may depend on the ob-
partly open answer formats are used, requesting for the jectives of the study and population characteristics13-17.
frequency of consumption daily, weekly, monthly or Some FFQs may include supplementary questions
yearly (Fig. 2). This allows for more flexibility and can on cooking methods and specific types of fat, bread,
contribute to reduce misclassification error14,15. milk and condiments or additions to foods such as salt.
For foods eaten seasonally, subjects are typically Brand name information may also be requested8,13,15.
asked how frequently and over what duration they ate Other FFQs may have an open-ended section where
these seasonal foods. In some cases, the frequency of respondents may record consumption of other foods
consumption is averaged for the whole length of the not included on the food list. This ensures that the total
reference period. For frequently consumed foods such diet of the individual is captured and may also help to
as coffee, answers are collected directly as an open-en- identify those whose diet is very unusual, for whom
ded question in some FFQs. the FFQ may not be appropriate.
Processing of data from food frequency question-
naires require some analytical decisions, such as how
Portion sizes commonly eaten to handle missing data, either to assign null values or
to assign the median value from those who did provide
Qualitative FFQs do not ask respondents about por- valid answers. Both approaches have been found equi-
tions commonly used. Conversely, quantitative FFQs valent in terms of introducing bias into relative risk
try to estimate the portion size commonly used for estimates3,18,19.
each food item listed and then household measures,
food models or other aids may be required for that pur-
pose. Semi-quantitative FFQs include reference por- Uses
tion sizes and respondents are asked how often they
consume the specified portion of a particular food or FFQs can be self-administered using paper or
beverage or to assess their usual portion size based on web-based formats, or interviewer administered, ei-
a specified measure. ther face-to-face or telephone interview. Many food

Fig. 2.—Example of a semiquantitave Food Frequency and

Fig. 1.—Sample of a Qualitative Food Frequency Questionnaire propensity Questionnaire with open-ended options to answer
with close-ended frequency range answer options used in PER- frequency used in ENPE study. Data collected in a face-to-face
SEO Project to assess diet of school children. Parents or guar- interview. P: pequeña (small); M: Media (medium); G: grande
dians were invited to fill in the form. (large).

Food Frequency Questionnaires51

006 Metodo de Frecuencia_Carmen Perez Rodrigo.indd 51 12/02/15 13:43

 frequency instruments are designed to be self-admi- Despite FFQs can be designed to assess specific
nistered and are either optically scanned paper version nutrients such as calcium or vitamin D, it is often re-
or web based. Require 30 to 60 minutes to complete quired to assess the whole diet to obtain estimates of
depending on the instrument and the respondent and energy intake needed for adjustment and to estimate
use a standard format, which reduces the time required mis-reporting3,6,18,19.
for processing data. FFQs can also be used to identify food patterns as-
For these reasons FFQs are one of the most com- sociated with inadequate intakes of specific nutrients
monly used retrospective methods in nutritional epide- by using adequate statistical analysis.
miology, in a wide range of dietary study designs, as
a research tool in examining the relationship between
dietary intake and disease risk. Large-scale cohort stu- Validity
dies20-22 such as the European Prospective Investiga-
tion into Cancer and Nutrition (EPIC) have used the FFQs should be evaluated for their accuracy before
method23. In the EPIC study, the FFQ instrument was being used. Correlation coefficients ranging from 0.5
specifically adapted to each country24,25. However, lar- to 0.7 are considered moderate. Table I shows correla-
ge cohort studies can last for several decades and it tion coefficients ranges observed in a number of vali-
is likely that food habits and practices change or new dation studies of FFQs since de ‘80s and the reference
products enter within that period, and the food list may method used. Since FFQs are often designed to assess
need to be revised and updated to be useful. These the ranking of intakes within a population, they cannot
changes can make comparisons between observations be relied on to produce reliable estimates of absolute
at different time points in the study more difficult. intake. Over-estimation is common, particularly for

Table I
Selected validation studies of FFQs: population, items, procedure and reference method

Number Reference Number of days Correlations

Authors Population FFQ
of items method repetition range
Bohlscheid-Thomas, 49 men, 55 women 104 Self 24-HR 12× in 1 y
1997
Boucher, 2006 166 women 126 Self 24-HR 2×
Flagg, 2000 216 men, 223 114 Self 24-HR 4× in 1 y
women
Jain, 2003 151 men, 159 166 Self 24-HR 3× in 1 y
women
Johansson, 2002 96 men, 99 women 84 Interview 24-HR 10× in 1 y
Katsouyanni, 1997 42 men, 38 women 190 Self 24-HR 12× in 1 y
Kroke, 1999 75 men, 59 women 146 Self DLW; 24-HR 14 d; 12× in 1 y
0,42 energy -
Ocke, 1997 63 men, 58 women 178 Self 24-HR 12× in 1 y 0,69 alcohol
Pisani, 1997 47 men, 150 women 47 Self 24-HR 8–14× in 1 y
Subar, 2001 247 men, 267 106 Self 24-HR 4× in 1 y
(Block FFQ) women
Subar, 2001 (DHQ) 501 men, 560 124 Self 24-HR 4× in 1 y
women
Subar, 2001 (Willett 254 men, 293 126 Self 24-HR 4× in 1 y
FFQ) women
Subar, 2003 261 men, 223 124 Self DLW; 24-HR 14 d; 2× in 3 m
women
van Liere, 1997 123 women 238 Self 24-HR 12× in 1 y
Feunekes, 1993 95 men, 96 women 104 Interview DH* 1×1m 0,77 energy
Jain et al, 1982 50 women 69 Self DH* DH 0,50 cholesterol -
0,64 energy
Andersen, 2003 17 women 180 Self DLW* 10 d

52 FOOD AND BEVERAGE CONSUMPTION ASSESSMENT: METHODS AND CONTROVERSIES

006 Metodo de Frecuencia_Carmen Perez Rodrigo.indd 52 12/02/15 13:43

 Table I (cont.)
Selected validation studies of FFQs: population, items, procedure and reference method

Number Reference Number of days Correlations

Authors Population FFQ
of items method repetition range
Andersen, 1999 125 men 180 Self FR* 14 d in 5 w
Barasi et al 1983 103 women 27 Self FR 4d
Bingham, 1997 156 women 130 Self FR 4 × 4 d in 1 y
Block, 1990 102 women 60 Self FR 2×7d
Block, 1992 85 men and women 98 Interview FR 4 × 3 d in 1 y
Brunner, 2001 457 men, 403 127 Self FR 7d
women
Callmer, 1993 57 men, 50 women 250 Self FR 6 × 3 d in 1 y
Engle, 1990 16 men, 34 women 120 Self FR 7d
Fidanza, 1995 11 men, 35 women 93 Self FR 7d
Goldbohm, 1994 59 men, 50 women 150 Self FR 3 × 3 d in 1 y
Hartwell, 2001 16 men, 9 women 162 Self FR 2×4d
Larkin, 1989 228 men and women 116 Interview FR 16 d in 1 y
Longnecker, 1993 64 men, 74 women 116 Self FR 3 × 2 d or 2 × 2 d
0,77 energy
Mannisto, 1996 152 women 110 Self FR 2 × 7 d in 3 m 0,50 cholesterol -
Martin-Moreno, 1993 147 women 118 Self FR 4 × 4 d in 1 y 0,64 energy
McKeown, 2001 58 men, 88 women 130 Self FR 2×7d
Munger, 1992 44 women 126 Self FR 3d
Patterson, 1999 113 women 122 Self FR 4×4d
Pietinen et al 1988 190 men 276 Self FR 12 × 2 d in 6 m
Pietinen et al 1988 190 men 44 Self FR 12 × 2 d in 6 m
Riboli, 1997 57 men, 50 women 350 Self FR 6 × 3 d in 1 y
Rimm, 1992 127 men and women 131 Self FR 2 × 7 d in 6 m
Schroder, 2001 44 men and women 157 Self FR 3d
Stuff et al, 1983 40 pregnant women 105 FR 7d
Tjonneland, 1991 59 men, 85 women 92 Self FR 2×7d
Willett, 1985 173 women 61 Self FR 4 × 7 d in 1 y
Willett, 1988 150 women 116 Self FR 4 × 7 d, 3–4 y before
Yarnell, 1983 119 men 54 Self FR 7d
24HR: 24 Hour recall; FR: food record; DLW; Doubly labelled water; d: day; y: year. Modified from: Molag ML et al. 200713.

foods eaten less often or for foods perceived as ‘heal- Various statistical methods employing measure-
thy’ such as fruit and vegetables. There is some evi- ment error models and energy adjustment are used to
dence that over-estimation increase with the length of assess the validity of FFQs but also to adjust estima-
the food list3,26-28. tes of relative risks for disease outcomes6,19. Analy-
Validation studies of various FFQs using biomar- ses comparing relative risk estimation from FFQs to
kers have found large underestimates of self-repor- dietary records in prospective cohort studies indicate
ted energy intake and some underestimation of pro- that observed relationships using an FFQ are severely
tein intake. Correlations of FFQs and the biomarkers attenuated, thereby obscuring associations that might
ranged from 0.1 to 0.5 for energy and from 0.2 to 0.7 exist29. Controversy about error in FFQs has raised a
for protein26,28. These results suggest that the measu- debate considering alternative methods of collecting
rement error associated with FFQs is larger than was dietary data in large-scale prospective studies. In parti-
previously estimated. cular, the association between dietary fat consumption

Food Frequency Questionnaires53

006 Metodo de Frecuencia_Carmen Perez Rodrigo.indd 53 12/02/15 13:43

 and breast cancer is controversial and limitations of Advantages and limitations of the method
FFQs have been questioned30-32.
For any study, the advantages and disadvantages Table II summarizes main pros and cons of Food
of using FFQs compared to other dietary assessment Frequency Questionnaires. FFQs can be self-adminis-
methods should be carefully considered. It has been tered using paper or web-based formats thus reducing
suggested that using a combination of methods, such data collection costs. Paper forms are often designed
as FFQ with dietary records or 24 hour recalls, or FFQ to be optically scanned so that data can be entered and
with biomarker levels provide more accurate estimates analyzed in a comparatively short period of time, often
of dietary intakes than individual methods4,33. in an automated process, allowing dietary data on a
Several researchers have explored the use of cog- large number of people to be collected relatively inex-
nitive interviewing techniques to increase the validity pensively. There is also less need for nutritional exper-
of self-reported data. Respondents are encouraged to tise in data entry.
verbalize their thought processes as they retrieve infor- Additional advantages of FFQs include low respon-
mation from long-term memory to answer questions dent burden compared to other methods. Frequency
on the FFQ and then identify difficulties in formula- questionnaires assess habitual consumption over an ex-
ting answers to specific questions, such as intake of tended period of time. More complete data may be co-
seasonal foods or estimating portion sizes34,35. llected if the FFQ is interviewer administered, but res-
Qualitative study methods inform that several fac- pondent bias may be less if self-administered. FFQs can
tors can influence individuals’ perception of portion be designed to focus on a particular group of foods or to
size, including; the type of food being considered, the assess the whole diet, including portion size estimates
role of a given food item in the meal (i.e. as a main or that can be used to obtain absolute nutrient intakes.
a side dish) and personal preference for the food. Cog- When an open section is included respondents can
nitive research has also indicated that the level of grou- record consumption of foods which are not included
ping of foods can affect the recall of food intake and on the food list. Sometimes include separate sections
that respondents find it easier to respond to items when asking about consumption of seasonal foods, cooking
disaggregated, but this needs to be balanced with the procedures or additions to foods such as sauces and
disadvantages of longer food lists and the likelihood condiments.
of over-estimation of intake and additional respondent The major limitation of the food frequency method
burden34,35. is that it contains a substantial amount of measurement

Table II
Advantages and limitations of Food Frequency Methodology

Advantages Limitations
Can be self-administered Marked frequency of consumption and portion size
may not represent usual intake of respondent
Requires certain literacy and cognitive skills
Often incomplete data is collected
Can be optically scanned Effort and time consuming instrument design
Modest respondent burden Depends on the respondents ability to describe diet
Relatively low cost for large scale studies Relatively low cost for large scale studies
May be a better representation of usual dietary patterns than only Particularly complex for children and elderly people
a few days of observation
Instrument design can be based on population data Memory of diet in the past may be biased by present diet
Does not influence dietary behaviour Sometimes limited precision in estimates and
quantifying food portion sizes
Classify individuals in food consumption categories
Does not require deeply trained interviewers
Easy to code and viable for automated processing if closed-ended
Web-based administered improve the quality of collected data. Requires computer and internet access
Can add help aids, additional models and information. Requires web navigation skills
Persist systematic errors inherent to the method
Response bias
Security risk for study data

54 FOOD AND BEVERAGE CONSUMPTION ASSESSMENT: METHODS AND CONTROVERSIES

006 Metodo de Frecuencia_Carmen Perez Rodrigo.indd 54 12/02/15 13:43

 error. Many details of dietary intake are not measu- research about conventional paper-based FFQs. Im-
red, and the quantification of intake is not as accura- plementing digital pictures, for example, may improve
te as with recalls or records. Inaccuracies result from the food identification. Also, certain technical func-
an incomplete listing of all possible foods and from tionalities can facilitate the skipping to more tailored
errors in frequency and usual serving size estima- questions at individual level or provide explanations
tions. A comprehensive list of all foods eaten cannot on how to answer questions appropriately3,4,15.
be included and reported intake is limited to the foods It has been suggested that in addition to focusing
contained in the food list. Accurate reporting relies on attention on trying to measure nutrients, also formu-
respondent memory. Bias may be introduced with res- late hypotheses in terms of dietary behaviors inclu-
pondents reporting eating according to social desirabi- ding questions about usual dietary practices. Such
lity, thus resulting in over-estimation of certain foods questions may be more easily and accurately recalled
and under-estimation of other items. A relatively high than the frequencies and portion sizes of a long list of
degree of literacy and numeracy skills are required foods34,35.
if self-administered. Interviewers can help overcome When FFQs are interviewer administered, inter-
this problem3,4. viewers should be properly trained to conduct the
The serving size of foods consumed is difficult for interview and to provide adequate information to
respondents to evaluate in all assessment instruments, participants either if the interviews are conducted fa-
but attempting to estimate usual serving size in FFQs ce-to-face or by telephone. Appropriate ways of asking
may be even more complex because a respondent is the questions and recording the answers need to be
asked to estimate an average for foods that may have considered, particularly for telephone interviews that
highly variable portion sizes depending on eating oc- require full attention through the whole duration of
casions. The use of small, medium and large to des- the questionnaire. When the respondents are children,
cribe portion size may not have a commonly accepted elder or handicapped people, parents, guardians or ca-
meaning34,35. rers have to be present.
FFQs developed in one country or for a specific Standardised operating procedures are required for
subpopulation are unlikely to be appropriate for use data checking, cleaning and analysis.
in another, since dietary habits differ. The same pro- Some large epidemiological studies have imple-
blem arises due to ethnical and cultural differences in mented short-term dietary assessment methods, either
a population. as reference calibration method in a sub-sample or as
Pre-prepared meals such as ready meals or take- main dietary assessment method for the entire popula-
away foods may not be easy for respondents to classify tion, given the debate about the accuracy of FFQs in
if the food list is based on more basic food catego- ranking the individuals according to their usual dietary
ries, but grouping of foods into individual items may intake because methodological limitations. There is
make answering some questions problematic. Validity cumulated evidence that repeated open-ended quan-
can vary widely between foods and nutrients from the titative 24-hour dietary recalls or food records may
same FFQ outperform the FFQ in assessing accurately individual
usual intake. The complementation of the repeated
short-term quantitative measures with the non-quan-
Recommendations and suggested improvements titative information on usual consumption (e.g. from
non-quantitative FFQs, defined as Food Propensity
Innovative FFQs are web-based. However, measu- Questionnaires or FPQs) and/or biomarkers and inte-
rement errors in such instruments are most likely si- grated with statistical modelling, may yield more ac-
milar to those in conventional paper-based Food Fre- curate individual usual dietary intake estimates15. This
quency Questionnaires, suggesting that the underlying may particularly provide less biased estimates for the
methodology is unchanged by the technology. New intake of infrequently consumed foods that are often
instruments such as the web-based version of the Na- missed in the short-term dietary measures.
tional Cancer Institute’s (NCI) 124-item diet history
questionnaire (Web-DHQ) include digital photographs
to estimate portion sizes. Moderate-to-a-very good co- References
rrelation has been assessed indicating reproducibility
was observed between the NCI’s Web-DHQ and its 1. Serra Majem Ll, Aranceta Bartrina J (eds). Nutrición y salud
original paper version (Paper-DHQ)15. pública: Métodos, bases científicas y aplicaciones. 2 ed. Bar-
celona: Elsevier- Masson 2006.
The strengths of these innovative alternatives inclu- 2. Martin-Moreno JM, Gorgojo L.Valoración de la ingesta dieté-
de data consistency and completeness through tech- tica a nivel poblacional mediante cuestionarios individuales:
nical requisites, particularly in larger, geographically sombras y luces metodológicas. Rev Esp Salud Pública 2007;
dispersed or multi-centric study populations. . Innova- 81: 507-518.
3. Thompson FE, Subar AF. Dietary Assessment Methodology.
tive technological alternatives of the FFQs may have In: Coulston A, Boushey C (eds). Nutrition in the Preven-
greater potential to overcome some of the completion tion and Treatment of Disease, 2nd ed. Academic Press 2008:
difficulties that have been reported in the cognitive pp 3-38.

Food Frequency Questionnaires55

006 Metodo de Frecuencia_Carmen Perez Rodrigo.indd 55 12/02/15 13:43

 4. Shim JS, Oh K, Kim HC. Dietary assessment methods in epi- llett, and National Cancer Institute food frequency question-
demiologic studies. Epidemiol Health 2014; 36: e2014009. naires: the Eating at America’s Table Study. Am J Epidemiol
5. Willett WC, Hu FB. The food frequency questionnaire. Cancer 2001;154: 1089–1099.
Epi­demiol Biomarkers Prev 2007;16:182-183. 21. Estruch R, Martinez-Gonzalez MA, Corella D, Salas-Salvado
6. Willett WC. Nutritional Epidemiology. 2ª ed. Nueva York: J, Ruiz-Gutierrez V, Covas MI, et al. Effects of a Mediterra-
Oxford University Press 1998. nean-style diet on cardiovascular risk factors: a randomized
7. Block G, Hartman AM, Dresser CM, Carroll MD, Gannon J, trial. Ann Intern Med 2006; 145:1-11.
Gardner L. A data-based approach to diet questionnaire design 22. Martínez-González MA. The SUN cohort study (Seguimiento
and testing. Am J Epidemiol 1986;124:453-469. University of Navarra). Public Health Nutr 2006; 9:127-31.
8. Cade JE, Burley VJ, Warm DL, Thompson RL, Margetts BM 23. Bingham SA. Dietary assessments in the European prospective
Food-frequency questionnaires: a review of their design, vali- study of diet and cancer (EPIC). Eur J Cancer Prev 1997; 6:
dation and utilization. Nutr Res Rev 2004; 17:5-22. 118–124.
9. Martin-Moreno JM, Boyle P, Gorgojo L, Maisonneuve P, Fer- 24. Bingham SA, Welch AA, McTaggart A, Mulligan AA,
nandez-Rodriguez JC, Salvini S, Willett WC. Development Runswick SA, Luben R et al. Nutritional Methods in the Eu-
and validation of a food frequency questionnaire in Spain. Int J ropean Prospective Investigation of Cancer in Norfolk. Public
Epidemiol 1993; 22:512-9. Health Nutr 2001; 4(3), 847-858.
10. Shai I, Shahar DR, Vardi H, Fraser D. Selection of food items 25. Ocke MC, Bueno-de-Mesquita HB, Goddijn HE, Jansen A,
for inclusion in a newly developed food-frequency question- Pols MA, van Staveren WA, Kromhout D. The Dutch EPIC
naire. Public Health Nutr 2004; 7:745-9. food frequency questionnaire. I. Description of the question-
11. Welch AA, Luben R, Khaw KT, Bingham SA. The CAFE naire, and relative validity and reproducibility for food groups.
computer program for nutritional analysis of the EPIC-Norfolk Int J Epidemiol 1997; 26(suppl 1): S37–S48.
food frequency questionnaire and identification of extreme nu- 26. Bingham SA, Gill C, Welch A, Cassidy A, Runswick SA,
trient values. J Hum Nutr Diet 2005;18:99–116. Oakes S et al. Validation of dietary assessment methods in the
12. Mulligan AA, Luben RN, Bhaniani A, Parry-Smith DJ, O’Con- UK arm of EPIC using weighed records, and 24-hour urinary
nor L, Khawaja AP, Forouhi NG et al. A new tool for conver- nitrogen and potassium and serum vitamin C and carotenoids
ting food frequency questionnaire data into nutrient and food as biomarkers. Int J Epidemiol 1997;26(suppl 1):S137–51.
group values: FETA research methods and availability. BMJ 27. Block G, Wakimoto P, Jensen C, Mandel S, Green RR. Vali-
Open 2014;4:e004503. dation of a food frequency questionnaire for Hispanics. Prev
13. Molag ML, de Vries JH, Ocke MC, Dagnelie PC, van den Chronic Dis 2006;3: A77.
Brandt PA, Jansen MC, et al. Design characteristics of food 28. Subar AF, Kipnis V, Troiano RP, Midthune D, Schoeller DA,
frequency questionnaires in relation to their validity. Am J Epi- Bingham S et al. Using intake biomarkers to evaluate the ex-
demiol 2007 15;166(12):1468-78 Epub 2007 Sep 18. tent of dietary misreporting in a large sample of adults: the
14. Marks GC, Hughes MC, van der Pols JC. Relative validity of OPEN study. Am J Epidemiol 2003;158:1-13.
food intake estimates using a food frequency questionnaire is 29. Day NE, Wong MY, Bingham S, Khaw KT, Luben R, Miche-
associated with sex, age, and other personal characteristics. J ls KB, et al. Correlated measurement error--implications for
Nutr 2006; 136:459-65. nutritional epidemiology. Int J Epidemiol 2004; 33:1373–81.
15. Illner AK, Freisling H, Boeing H, Huybrechts I, Crispim SP, 30. Bingham SA, Day N. Commentary: fat and breast cancer: time
Slimani N. Review and evaluation of innovative technologies to re-evaluate both methods and results? Int J Epidemiol 2006;
for measuring diet in nutritional epidemiology. International 35:1022-4.
Journal of Epidemiology 2012;41:1187–1203. 31. Kristal AR, Peters U, Potter JD. Is It Time to Abandon the Food
16. Forster H, Fallaize R, Gallagher C, O’Donovan CB, Woolhead Frequency Questionnaire? Cancer Epidemiol Biomarkers Prev
C, Walsh MC, et al. Online Dietary Intake Estimation: The 2005;14(12):2826 – 8.
Food4Me Food Frequency Questionnaire. J Med Internet Res 32. Willet WC, Hu FB. Not the time to abandon the Food Frequen-
2014; 16(6): e150. cy Questionnaire: Point. Cancer Epidemiol Biomarkers Prev
17. Quandt SA, Vitolins MZ, Smith SL, Tooze JA, Bell RA, Da- 2006; 15:1757-8.
vis CC, et al. Comparative validation of standard, picture-sort 33. Bingham SA, Gill C,Welch A, Day K, Cassidy A, Khaw KT, et
and meal-based food-frequency questionnaires adapted for an al. Comparison of dietary assessment methods in nutritional epi-
elderly population of low socio-economic status. Public Health demiology: weighed records v. 24 h recalls, food-frequency ques-
Nutr 2007; 10:524-32. tionnaires and estimated-diet records. Br J Nutr 1994; 72:619-43.
18. Kushi LH: Gaps in epidemiology research methods: design 34. Chambers E, Godwin SL, Vecchio FA. Cognitive strategies for
considerations for studies that use food-frequency questionnai- reporting portion sizes using dietary recall procedures. J Am
res. Am J Clin Nutr 1994, 59(Suppl 1):180S-184S. Diet Assoc 2000; 100, 891-7.
19. Liu K: Statistical issues related to semiquantitative food-frequen- 35. Thompson FE, Subar AF, Brown CC, Smith AF, Sharbaugh
cy questionnaires. Am J Clin Nutr 1994, 59 (Suppl 1):262S-265S. CO et al Cognitive research enhances accuracy of food fre-
20. Subar AF, Thompson FE, Kipnis V, Midthune D, Hurwitz P, quency questionnaire reports: results of an experimental vali-
McNutt S, et al. Comparative validation of the Block, Wi- dation study. J Am Diet Assoc 2002; 102:212-25.

56 FOOD AND BEVERAGE CONSUMPTION ASSESSMENT: METHODS AND CONTROVERSIES

006 Metodo de Frecuencia_Carmen Perez Rodrigo.indd 56 12/02/15 13:43