Development and use of FFQ among adults in diverse settings across the globe

Abstract

In nutritional epidemiology, development of valid dietary assessment instruments specific to populations in diverse settings is of paramount importance. Such instruments are essential when trying to characterise dietary patterns and intake, investigate diet–disease associations, inform and evaluate nutrition interventions, assess nutrient–gene interactions, conduct cross-country comparison studies and monitor nutrition transitions. The FFQ is a relatively inexpensive tool for measuring long-term dietary intake for large populations and for allowing researchers to track dietary changes over time. However, FFQ must be population specific to capture the local diet and available foods. Collecting 24-h dietary recalls and utilising community feedback to build the FFQ ensures that a culturally appropriate instrument is developed. This article presents several examples describing FFQ development and utilisation in different settings globally. In the Canadian Arctic, FFQ were developed and utilised to inform and evaluate a community-based intervention programme, characterise the diet and track dietary changes occurring among Inuit and Inuvialuit, populations experiencing rising rates of chronic disease and likely to be extremely vulnerable to the potential effects of climate change. Another example is an FFQ developed to assess sodium intake and evaluate a sodium reduction trial in a high-risk population in Barbados. An example is provided from Brazil, where an FFQ was developed to assess associations between diet, heterocyclic aromatic amines and colorectal adenoma among Japanese Brazilians and to conduct cross-country comparisons. These and other case studies highlight the diversity in dietary intake between populations and the need for FFQ to be developed to capture this diversity.

Abbreviations:

BNCS

Barbados National Cancer Study

CRC

colorectal cancer

HAA

heterocyclic aromatic amine

NWT

Northwest Territories

QFFQ

quantitative FFQ

USDA

United States Department of Agriculture

Dietary assessment methodologies

In nutritional epidemiology, a variety of different approaches can be used to measure food and nutrient intakes in individuals and populations. These range from relatively simple techniques such as 24-h dietary recalls, estimated or weighed food records, narrative diet histories and FFQ, to the more complex biochemical approach of measuring static or functional markers of nutrient intake in blood, urine or other biological samples. The selection of a dietary assessment methodology depends on the population under investigation and typically involves a compromise between the desired level of accuracy and organisational or financial constraints⁽¹⁾. Biochemical markers of nutrient intake, while arguably more objective than other methods, are used infrequently outside of small validation or pilot studies due to their high cost and logistical requirements⁽²⁾. Food records and 24-h dietary recalls are expensive and tend to underestimate actual intake due to participants’ incomplete record keeping and fallible memories⁽³⁾. Although the accuracy of food records can be improved by using weighed-food records, this requires the participants to be literate and to have numerical skills. Moreover, it is time consuming and burdensome for participants. These negative characteristics may lead to modification of usual dietary habits⁽¹⁾. Interviewer-administered 24-h dietary recalls require much less time from participants, and, as a retrospective measure, cannot inadvertently influence participants to alter their food intake⁽¹⁾. However, dietary recalls cannot be used to rank participants’ intakes reliably; single 24-h dietary recalls cannot measure the day-to-day variation of dietary intake, while multiple recalls are expensive⁽⁴⁾. Narrative diet history, another dietary assessment tool that lends itself well to the clinical setting⁽⁵⁾, is particularly useful with populations that respond poorly to highly structured techniques and may be better able than other methods to capture infrequently consumed items. However, this approach tends to introduce greater heterogeneity in the results and has the same weaknesses of interview bias and imperfect memories^{(6, 7)}, which may limit their usability outside of clinical settings.

The drawbacks of these dietary assessment methodologies have spurred the development of new techniques intended to both increase objectivity and decrease participant burden. Much of the effort has focused on portable electronic systems for recording food intake, including mobile phones^{(8, 9)}, personal digital assistants⁽¹⁰⁾, and purpose-built devices⁽¹¹⁾. These technologies have the potential to significantly alter the way dietary assessments are performed; nevertheless, they are only in the initial stages of development and are yet to be validated against more traditional methodologies⁽¹²⁾.

FFQ

Compared with other dietary assessment methodologies, FFQ are the most feasible and cost-effective means of assessing diet in large-scale nutritional epidemiology studies. In contrast to other methods, FFQ assess long-term usual dietary intake⁽¹⁾. Their low cost and ease of administration also make them viable for the repeated assessment of a population's diet over time to establish trends. Moreover, FFQ are one of the least invasive methods and incur low participant burden⁽¹³⁾. Although unable to assess absolute dietary intakes, FFQ are mainly used to categorise and rank individuals either by their usual frequency of consumption or nutrient intake, and, as a standardised dietary assessment methodology, allow for comparison of food and nutrient intake across cultures, populations and countries^(13–17). FFQ can be self- or interviewer administered, though interviewer administered is preferred to ensure adequate completion, and may take place over the telephone, via mail, on computers and through personal interviews, depending on the needs and funding of the study^{(1, 18)}. However, FFQ are subject to recall bias, and because of the large variety of foods often available, they are limited in their ability to capture the total diet for most populations. FFQ, like many other assessments, are limited by self-reported information. Other challenges include the need for validation and the possibility of participants over- or under-reporting consumption⁽⁷⁾. Because of time and funding constraints, researchers often use or adapt previously developed FFQ, such as the Harvard and National Cancer Institute questionnaires, for their study population^(18–20). However, due to the fact that food availability, accessibility and preferences can vary greatly between settings, research with ethnic/racial minority, multi-ethnic or culturally distinct populations and countries requires that questionnaires be developed specifically for that population in order to produce valid and reliable data^{(21, 22)}.

The objective of this article is to provide a guide to the development of population-specific, culturally appropriate FFQ for adults and to exemplify several uses for them once developed. To illustrate the process, examples will be taken from several research projects the author has carried out with multi-ethnic populations in a variety of developing or low-income settings in different parts of the world. These include the Apache and Navajo American Indian populations in the US^(23–25), low-income African-American adults in an inner city in the US⁽²⁶⁾, African-Barbadians in Barbados^(27–29), African-origin Caribbean migrants to the UK^{(17, 30–33)}, urban and rural Cameroonians^{(17, 34)}, African-origin Jamaicans⁽¹⁷⁾, First Nations Aboriginal populations in northwest Ontario, Canada⁽³⁵⁾, Inuit and Inuvialuit aboriginal populations in the Canadian Arctic^{(36, 37)}, and Japanese Brazilians in São Paulo, Brazil^{(16, 38)} (Table 1).

Table 1. Development of culturally competent, population-specific quantitative FFQ (QFFQ) through 24-h dietary recalls, food records and national databases by study

A comparison of the FFQ developed for each population illustrates diet diversity (Table 2). For example, the Cameroonian study population consumed snail soup and foo foo (cassava), compared with the consumption of polar bear and caribou by Inuit and Inuvialuit in Arctic Canada, coxinha and nishime by Japanese Brazilians, and tamales and menudo stew by Apache and Navajo in southwestern US. Dramatic differences between populations clearly demonstrate the necessity for researchers to develop their own population-specific FFQ.

Table 2. Quantitative FFQ developed by first author presented by study population

–, Not specified; WI, West Indian.

To develop a culturally relevant, locally appropriate and population-specific FFQ, four main steps are required if there are little or no data available on total food consumption: compilation of a complete and accurate food list, determination of culturally appropriate portion sizes, categorisation of frequencies of consumption and development of a food composition table^{(1, 18)} (Fig. 1). The food list is the most critical part of the development process because the omission of important items may lead to underestimation of nutrient intake, while the inclusion of irrelevant items unduly increases participant burden. For completeness, the food list must contain items that are consumed by a substantial proportion of the population, are frequently consumed, include a variety of foods consumed within the population, contain significant amounts of the nutrients or food constituents of interest and cumulatively provide at least 85% of the dietary intake^{(1, 18, 19)}. If the FFQ is developed for an intervention evaluation study, the foods promoted and de-promoted by the intervention should be added (even if not frequently consumed initially). Effective methods used to compile the food list include 24-h dietary recalls conducted with a population-based sample and dietary databases generated by national and community surveys^(39–41). The collection of 24-h dietary recalls to develop the FFQ has the additional advantage of providing preliminary characteristics on the dietary and nutrient intakes of understudied populations (Table 3).

Fig. 1. Diagram of the process for developing and using a validated quantitative food frequency questionnaire (QFFQ).

Table 3. Daily energy and select nutrient intake from the 24-h dietary recalls used to develop the quantitative food frequency questionnaires, presented by study population

* DRI, dietary reference intakes. Estimated amount of energies needed to maintain energy balance for men and women aged between 31 and 50 years at the level of very low physical activity-sedentary level.

† Acceptable macronutrient distribution ranges.

The recalls can also highlight foods to target as part of a nutrition intervention programme. For example, Sharma et al. ⁽²³⁾ found that for Apaches in Arizona, non-nutrient-dense, high-fat, high-sugar foods were the major contributors to energy intake⁽²³⁾ (Table 4). Among the Apache and Navajo, five foods and beverages (i.e. crisps and popcorn, fried potato dishes, breads, tortillas and burritos, and sugary drinks) collectively contributed 30–35% of energy intake^{(23, 25)}. These foods can be substituted with nutrient-dense, low-fat, low-sugar foods, such as whole-wheat bread, diet carbonated drinks, and unsweetened 100% fruit juice. Substitution of preparation methods that reduce or have no impact on fat content (such as frying with cooking spray instead of oil or baking without added fat) can also significantly ameliorate nutrient intake, dietary adequacy, and chronic disease risk in these populations and, depending on the study objectives, can be added to an FFQ to evaluate intervention success.

Table 4. Highest five food sources of energy from 24-h dietary recalls used to develop the quantitative food frequency questionnaires, presented by study population

In addition, the preliminary characterisation of the diet can elucidate nutrition transitions occurring, and the FFQ can be used to monitor any changes in diet that occur. For example, the traditional foods of the Inuit and Inuvialuit study populations in the Canadian Arctic, such as caribou and locally gathered fish, were shown to be the highest contributors of protein intake (Table 5), but as traditional foods become less available, they will likely be replaced by high-fat, high-sugar store-bought foods^{(36, 37)}. Currently, traditional foods are not among the highest five sources of total fat among Inuit. In fact, similar to other populations, non-nutrient-dense, high-fat foods are the highest contributors to the total fat among Inuit (Table 6).

Table 5. Highest five food sources of protein from 24-h dietary recalls to develop the quantitative FFQ, presented by study population

Table 6. Highest five food sources of total fat from 24-h dietary recalls used to develop the quantitative FFQ, presented by study population

The collection of portion sizes is needed to assess the amount of food consumed^{(18, 31)}. The average portion sizes from one population are not always applicable to other populations, particularly other ethnic/racial groups. For example, Sharma et al. ⁽³¹⁾ found that African-Caribbean adults in the UK had average portion sizes of boiled potatoes and crisps that were 1·4–1·9 times (60–158 g) and 1·3–1·5 times (56–84 g), respectively, more than the Caucasian population⁽³¹⁾. Food models, photographs and measurements used to assist participants in accurately estimating portion sizes must be culturally and locally appropriate.

To assign valid nutrient values to each of the line items on the quantitative FFQ (QFFQ), the use of an appropriate dietary database is crucial and requires the creation of a food composition table and collection of portion size weights^{(18, 21, 30, 42)}. When the nutrient content of a food is unknown, a chemical analysis should be conducted or, for mixed or composite dishes, the nutrient content can be estimated based on established nutrient analyses of the ingredients using existing food composition databases and locally collected recipes^{(18, 21, 24, 27, 32, 34, 43)}. The author and her research team have collected 1172 weighed recipes for 262 local dishes for six unique populations (Table 7). Community involvement is an essential element in the FFQ development process for all studies, and particularly for culturally distinct populations. When available, local dietitians and public health nutritionists play a key role in bridging the gap between researchers and the local community. The community can be involved in the selection of appropriate food models and portion sizes, the ordering and grouping of foods, providing local terminology for foods (such as the term muktuk for whale skin and fat in Inuit communities), identifying important but easily forgotten foods (especially seasonally available foods), and providing feedback on the questionnaire^{(17, 21, 23, 25, 26, 28, 35–37, 45)}. This process can take many forms, including the involvement of local community members in the role of field staff, or through their participation in focus groups.

Table 7. Recipe collection by study population

Population-specific FFQ have great utility and can be constructed with relative ease in different settings around the world, and several examples will be presented to illustrate this diversity. First, an example is presented from the Arctic regions of Canada, where FFQ were developed to inform and evaluate a community-based nutrition and lifestyle intervention study and track the rapid dietary changes occurring among Inuit and Inuvialuit^{(36, 37, 46–60)}. Second, an example is presented from the Caribbean, where an FFQ has been developed⁽²⁸⁾ and is being utilised to monitor sodium intake and will evaluate a sodium reduction study aiming to reduce the risk of hypertension in Barbados, where the majority of the population is at increased risk of high blood pressure⁽⁶¹⁾. The final example is from Brazil, where an FFQ was developed to assess associations between diet (including heterocyclic aromatic amines (HAA) from well-done meat) and colorectal adenoma among Japanese Brazilians, which were then compared with Japanese living in Hawaii and Tokyo⁽¹⁶⁾.

Use of quantitative FFQ in the Arctic to track the nutrition transition and develop and evaluate an intervention programme

Within the last century, aboriginal populations in the Canadian Arctic have undergone rapid changes in diet and lifestyle from nomadic hunter-gatherers with a largely protein-based diet to decreased physical activity, greater amounts of market foods and fewer traditional foods procured from hunting, fishing and gathering^{(46, 62–67)}. This nutrition transition is associated with greater, often excessive, intake of refined carbohydrate and fat and insufficient intake of many nutrients (e.g. dietary fibre, folate, potassium, vitamins A, B6, and D), leading to a double burden of under-nutrition and over-nutrition^{(46, 63, 68–78)}. Moreover, food security presents a growing issue, particularly for traditional foods, as climate change alters the physical landscape and poverty rates among the Aboriginal populations remain high^{(46, 79–88)}. The rapid lifestyle and nutrition transition may explain the rising prevalence of chronic diseases and higher rates of mortality among aboriginal populations in Canada compared to the general population^(89–93). Recent data from Inuit and Inuvialuit living in the Arctic regions of Nunavut and the Northwest Territories (NWT) showed that 24% were overweight and 43% were obese; 5% had type 2 diabetes and 22% were hypertensive^{(48, 49, 55)}. Clearly, there is a need to monitor the changes in diet over time among these populations and develop an informed nutrition and lifestyle intervention programme to improve dietary adequacy and reduce the risk of chronic disease.

Researchers have conducted dietary assessments on Inuit, Inuvialuit and other Aboriginal Canadian populations using various methodologies, including FFQ (with and without portion sizes) and single 24-h dietary recalls^{(35, 62, 66, 70, 94–96)}. Methods for FFQ development were consultations with nutrition experts, researchers, community members and local government representatives and community workshops. Until now, no comprehensive, population-specific, culturally competent and cost-effective dietary assessment instrument has been developed that can monitor changes in these populations over time. Of the FFQ that were developed, none have been developed using an evidence-based food list generated from up-to-date dietary data collected using scientific research methodology, combined with community consultations and including all foods consumed and with proper validation.

The author and her research team therefore developed two culturally competent, evidence-based, interviewer-administered QFFQ to fulfil these needs and to evaluate a nutritional and lifestyle intervention programme, one for Inuvialuit in the Beaufort Delta region in the NWT and one for Inuit in the Kitikmeot region in Nunavut^{(36, 37)}. To develop the QFFQ, single 24-h dietary recalls were conducted by trained staff in two communities in each territory to generate a complete and accurate food list, and items recalled more than once were included on the questionnaire, except for items very low in energy and nutrients, such as condiments and spices^{(36, 37)}. Three-dimensional food models, local household utensils, standard units (e.g. slice of bread) and packaging from foods commonly purchased in local stores (e.g. crisps packets, sweet wrappers) were used to assess portion size in the recalls, and the models were later assigned to QFFQ line items based on portion sizes reported in the recalls and consultation with community staff. The recalls were analysed using the Canadian Nutrient File database (10th edition) in NutriBase Clinical Nutrition Manager version 5.18 to provide an initial overview of dietary patterns and to identify the major contributors to energy and selected nutrients. The recalls showed inadequate intakes of dietary fibre, calcium, vitamins A, C, D, and E and folate among Inuit and Inuvialuit^{(36, 37)}.

Focus groups were held with local staff and residents to list all the traditional foods to ensure that foods consumed in other seasons were not inadvertently omitted. A manual of procedures was developed for the QFFQ, and the data collectors piloted the QFFQ with eighteen to twenty participants in the four communities. In consultation with community staff and stakeholders after piloting, the QFFQ were refined, and the foods were grouped and ordered to fit into the local conceptual framework⁽²¹⁾. For example, traditional meats were separated from store-bought meats and were placed first in the food group order.

The final NWT QFFQ contained 144 line items while the final Nunavut QFFQ contained 153 line items. Both questionnaires had eight frequency categories ranging from ‘never’ to ‘two times per day or more,’ and additional questions concerning supplement use and smoking habits. Although they contain many of the same food and beverage items, the two questionnaires were developed separately so as to be tailored for each region during the intervention study. Sample pages from the QFFQ are presented in Appendices 1 and 2.

In addition to being a useful tool for QFFQ development, the recalls highlighted foods for a nutrition intervention programme. For example, in the NWT, sugar (added to tea or coffee), sweetened juice/drinks, carbonated drinks, white bread and crisps were the greatest contributors to energy intake, contributing to nearly a quarter of total energy intake (Table 4)⁽³⁶⁾. Clearly, replacing these non-nutrient-dense items with nutrient-rich alternatives, such as unsweetened juice, whole-wheat bread, or low-sugar low-fat alternatives, such as artificial sweeteners or diet carbonated drinks, would significantly improve dietary quality and prevent excessive energy intake. These results were presented back to the communities in workshops, and the workshop participants used these results to identify the ‘problem’ foods in their communities, acceptable healthier alternatives, dietary behaviours to target for a behaviour change strategy and key messages for a nutrition intervention⁽⁹⁷⁾. The foods selected by the community for promotion and demotion aligned very well with the findings from the dietary recalls. These foods were included in the QFFQ to construct a comprehensive instrument for an intervention evaluation.

The data from the dietary recalls, community workshops and qualitative in-depth interviews were used to develop the Healthy Foods North programme. This is a community-based, multi-institutional, multi-level nutrition and lifestyle intervention programme that aims to reduce the risk of chronic disease and improve dietary adequacy among Inuit and Inuvialuit⁽⁹⁷⁾. Healthy Foods North has been described previously in detail⁽⁴⁷⁾. An example of culturally appropriate media, developed from the workshops and consultations with community partners, is presented in Appendix 3. The pilot programme was conducted in 2008–2009 in the two NWT and two Nunavut communities that participated in the formative phase. The QFFQ were administered prior to Healthy Foods North implementation in the four intervention communities plus two delayed intervention communities (who would receive the intervention after the evaluation was completed). Data collection for a QFFQ validation study was additionally undertaken in one intervention community in each territory with the baseline QFFQ collection. A series of up to three 24-h dietary recalls were conducted with seventy-one Inuit and sixty Inuvialuit adults in a 7-d period on non-consecutive days, capturing both weekend and weekday consumption^{(50, 56)}. The validation analysis for the NWT and Nunavut QFFQ, respectively, showed that 76% and 80% of the observations of the estimation of nutrient intake by the QFFQ and validation recalls were in the same or adjacent quartiles; consequently, the QFFQ were determined to be valid dietary assessment tools for these populations^{(58, 59)}. The QFFQ were re-administered after the intervention was completed. The data collectors consisted of local community members and Community Health Representatives, research assistants, and university students, and all were trained and certified by the author.

To accurately characterise the diet in these populations, the analysis of the QFFQ also needed to be population specific and culturally appropriate. A food composition table was developed for the NWT QFFQ and for the Nunavut QFFQ, using the Canadian Nutrient File database (10th edition) in NutriBase as the basis, or the United States Department of Agriculture (USDA) database for commercial food products (USDA SR 20 Search and What's In The Foods You Eat Search Tool, 3.0) if a particular item was unavailable in Canadian Nutrient File⁽⁴⁶⁾. For items that were the same between the two questionnaires, the same NutriBase entries were selected for inclusion in the food composition tables to ensure comparability and consistency between the two QFFQ. Seventeen recipes for nine different traditional dishes were collected by trained data collectors in the Nunavut field sites using a standardised protocol and were used for both food composition tables. Each ingredient was recorded and weighed before cooking, and the final weight of the dish was recorded to capture moisture loss during cooking. Portion weights were obtained for both the QFFQ and the 24-h dietary recalls for every food model and household utensil. The average weight for the foods in the models was calculated using ten consecutive weightings, and similar foods were substituted when a food item was unavailable. For items reported in standard units (e.g. slice of bread), the weight from the Canadian Nutrient File or USDA was used.

This setting presents a unique case in that, unlike other populations, QFFQ developed specifically for these populations may be able capture the total diet because of the limited variety in the diet and available food sources⁽⁴⁶⁾. Analysis of the baseline QFFQ is currently underway to estimate nutrient intake and dietary adequacy from the questionnaires. In addition, dietary adequacy was calculated from the validation 24-h dietary recalls^{(50, 56)}. The changes in consumption from baseline to post intervention captured by the QFFQ are also currently being analysed, and preliminary results indicate that the instrument is sensitive enough to evaluate the impact of Healthy Foods North on the diet of Inuit and Inuvialuit. Moreover, these QFFQ will be implemented over time to capture long-term changes in dietary intake.

As illustrated in this example, QFFQ are a valuable approach for informing nutrition interventions and relatively easy to construct. Comprehensive, feasible, population-specific and culturally competent instruments are needed for vulnerable populations undergoing drastic changes in diet and lifestyle.

Development of a quantitative FFQ for dietary salt evaluation in Barbados

CVD is a leading cause of death in the Caribbean, and hypertension has been identified as the primary risk factor^{(98, 99)}. In Barbados, an estimated 55·4% of the African-origin population aged 40–84 years has hypertension, with moderate rates of treatment and low rates of control⁽⁶¹⁾. The relationship between excess dietary sodium and high blood pressure is well documented^(100–105), and sodium reduction has been shown to improve hypertension, cardiovascular and stroke health outcomes^(106–109). To reduce the burden of cardiovascular and other hypertension-related chronic disease in Barbados, the Chronic Disease Research Centre of the University of the West Indies, Cave Hill, is collaborating with the author in the Barbados Salt Intake Survey to design a comprehensive, locally appropriate dietary assessment instrument that is able to monitor salt and sodium intakes in the Barbadian population over time, assess both dietary intake and adequacy and evaluate the efficacy of a planned intervention. As part of the Barbados National Cancer Study (BNCS) in 2005, a QFFQ was developed and validated^{(28, 110)} to assess total dietary intake in the Barbadian population, and local recipes for mixed dishes were collected to aid in the analysis⁽²⁷⁾. A sample page from the QFFQ is presented in Appendix 4. The results of BNCS highlighted the need for a nutrition intervention to reduce the risk of chronic disease in this population⁽²⁹⁾. Because of the elapsed time and different aims of the two studies, the original QFFQ requires updating to maintain relevance for the population as well as modification to capture dietary sodium intake and to be used as an intervention evaluation tool.

To update and modify the BNCS QFFQ, pilot work was undertaken to identify important sources of sodium in the current Barbadian diet. Up to three 24-h dietary recalls were collected from fifty randomly selected Barbadian adults in March and April of 2010 and were analysed using NutriBase Nutrition and Fitness Manager version 8.3.8 (CyberSoft, Inc., Phoenix, AZ) and USDA SR 20 Search. Portion size was assessed using a collection of food models from the BNCS. The primary contributors to sodium intake were identified for inclusion in the adapted QFFQ. New foods reported by more than one person in the multiple 24-h recalls were added to the BNCS QFFQ food list, as were additional foods identified by a focus group of Barbadian nurses, a dietitian and other public health researchers as important contributors to the diet. These foods included energy drinks, cereal bars, fruit smoothies and other products that were not widely available during the development of the BNCS QFFQ.

The Barbados Salt Intake Survey QFFQ has been pilot tested among Barbadian adults. It will be used to identify significant foods contributing to sodium intake in Barbados that may be targeted in a sodium reduction intervention. The QFFQ is being used to measure baseline sodium intake and overall dietary intake, as well as to track intakes and food consumption patterns post intervention and over time. The QFFQ developed for a previous study among Barbadian adults was culturally competent and was, therefore, an ideal tool for adaptation for the Barbados Salt Intake Survey.

Assessment of dietary risk factors for colorectal adenoma using a quantitative FFQ specifically developed for Japanese Brazilians

Brazil has the largest population of Japanese living outside Japan, and in 1998 São Paulo, Brazil had a population of 350 000 Japanese⁽¹¹¹⁾. Colorectal cancer (CRC) is the fourth commonest cancer in São Paulo, Brazil, and its incidence doubled between 1969 and 1993⁽¹¹²⁾. CRC incidence among Japanese populations in Japan has considerably increased over the past 30 years, and the incidence in Japanese migrants to Hawaii, US, markedly increased upon migration. It is likely that this is in part due to Westernisation of the diet characterised by low intakes of calcium and dietary fibre and high consumption of animal fat and meat^(113–115). However, CRC incidence rates did not increase among Japanese populations upon migration to São Paulo, despite high meat intake, a relatively affluent urban lifestyle, and higher BMI compared to Japanese in Japan⁽¹¹⁶⁾. Further research is needed to improve our understanding of CRC aetiology and to develop new strategies for its prevention among Japanese populations.

Substantial evidence suggests that dietary factors play an important role in the development of CRC and colorectal adenoma, a precursor lesion for most CRC^(117–120). Recent studies have emphasised high intakes of red meat and processed meat as likely risk factors for colorectal neoplasia^(121–123). Research has implicated this association to be at least partly mediated by the known chemical carcinogens HAA and polycyclic aromatic hydrocarbons. HAA are formed when meat is cooked at a high temperature for a long duration, and polycyclic aromatic hydrocarbons are formed when meat is cooked directly above a heat source^{(121, 122, 124, 125)}. Conversely, fruit and vegetables may have a protective effect against CRC⁽¹²⁶⁾. Despite high exposure to HAA and polycyclic aromatic hydrocarbons in churrasco meat, a commonly consumed barbecued meat in Brazil, Japanese in São Paulo appear to have a relatively low risk of CRC compared to Japanese populations in Japan or Hawaii. Low risk in this group may be attributed, at least in part, to protective factors such as high intakes of fruits and vegetables (including legumes)⁽¹²⁷⁾. Further research is needed to characterise the diet of Japanese Brazilians living in São Paulo, including their HAA intake, and subsequent risk of colorectal neoplasia, and to compare these with Japanese populations in Hawaii and Japan.

A colonoscopy-based, case–control study of adenoma was initiated to investigate dietary intake, nutrient–gene interactions and the risk for adenoma among Japanese Brazilians in São Paulo, and also to compare findings with those of parallel studies among Japanese in Hawaii and Japan. Using single 24-h dietary recalls from sixty Japanese Brazilian outpatients men (n 29) and women (n 31) (mean ages 58 years and 57 years, respectively) in a hospital in São Paulo, a culturally competent QFFQ was developed to assess food, nutrient and HAA intake⁽¹⁶⁾. Recall data showed that salad (e.g. vegetable salad and tomato salad) was the most commonly reported food, with rice second and oranges and tangerines the third most commonly reported food items. Seventy percent of respondents reported consuming a red-meat-based dish or sausages, and more than one third of the items reported were vegetables (including legumes), fruit or fruit juices.

The QFFQ contains 161 food and drink items, put into fifteen food groups. A local dietitian consulted on the most appropriate method of assessing portion size to determine amounts consumed for each item listed on the QFFQ. Specific meat items were listed individually on the QFFQ to capture differences in HAA content depending on cooking method, level of doneness, use of marinades and type of meat. Coloured photographs were added with meat cooked at several ‘doneness’ levels (rare; medium rare; medium; well cooked; very well cooked) for fourteen grilled, churrasco or pan-fried foods (four beef, one pork, six chicken and three fish items) to obtain accurate preparation method information and to determine the food sources of HAA. In addition, questions were included on the frequency and amount of consumption of gravy made with drippings from pan-fried and roasted meats and poultry to capture the intake of HAA from all the possible sources. A sample page from the QFFQ is presented in Appendix 5. A validation of the QFFQ using 4-d food diaries is currently under review. Some 387 weighed recipes for seventy-six local mixed dishes commonly consumed by Japanese Brazilians in São Paulo, Brazil, were collected and analysed for nutrition composition using NutriBase Clinical Nutrition Manager⁽³⁸⁾. Additionally, a laboratory analysis of the concentrations of HAA in meat and fish prepared in accordance with the cooking practices of Japanese Brazilians was conducted to allow estimation of dietary HAA exposure using the QFFQ and to study the association between HAA and cancer risk⁽⁴³⁾.

This comprehensive, study-specific and up-to-date QFFQ is being used to estimate HAA intake, assess dietary adequacy and investigate diet–gene interactions in colorectal neoplasia among Japanese Brazilians and to compare the findings with Japanese in Hawaii and Japan.

Conclusion

Despite their inherent limitations and especially regarding the ‘self-report’ nature of the collected data, FFQ are the most feasible and cost-effective dietary assessment tool currently available for large epidemiological studies. They allow for the characterisation of dietary patterns and intake, investigation of diet–disease associations, intervention evaluation, assessment of nutrient–gene interactions, cross-population comparison and monitoring of the nutrition transition taking place globally. The examples presented here illustrate the QFFQ development process and demonstrate the use of QFFQ for multiple purposes in diverse settings with multi-ethnic populations. It is of utmost importance that an FFQ is developed and validated specifically for the study population in nutritional epidemiology and intervention studies.