Use of a common food frequency questionnaire (FFQ) to assess dietary patterns and their relation to allergy and asthma in Europe: pilot study of the GA²LEN FFQ

Abstract

Background/Objectives:

Comparable international data on food and nutrient intake is often hindered by the lack of a common instrument to assess food intake. The objective of this study was within the Global Allergy and Asthma European Network of Excellence (GA²LEN), we developed and piloted a food frequency questionnaire (FFQ) to assess its validity in Europe.

Subjects/Methods:

Five countries participating in GA²LEN took part in the pilot study. A total of 200 adults aged 31–75 years were invited to complete a FFQ in two occasions and to give a blood sample. The intra-class correlation coefficient (ICC) was used to assess repeatability of the FFQ. Plasma phospholipid fatty acids (FAs) were analysed by gas chromatography. Pearson correlation was used to analyse the correlation between estimated dietary FA intake and plasma phospholipid FA levels.

Results:

A total of 177 participants (89%) had complete data on FFQ₁ and plasma phospholipid FAs. In all, 152 participants (76%) completed both FFQs. ICCs between macronutrients ranged from 0.70 (saturated FAs) to 0.78 (proteins) and between 0.70 (retinol) and 0.81 (vitamin D) for micronutrients. Dietary n-3 FAs showed a good correlation with total plasma phospholipid n-3 FAs and with docosahexaenoic acid in the whole sample (0.40) and in individual countries. Poor correlations were observed for other FAs.

Conclusions:

The GA²LEN FFQ is an appropriate tool to estimate dietary intake for a range of nutrients across Europe regardless of cultural and linguistic differences. The FFQ seems to be useful to estimate the intake of n-3 FAs but not other FAs.

Introduction

Diets are complex mixtures of foods providing a wide variety of nutrients. The most common methods for assessing food and nutrient intakes are through recall of all foods eaten over a fixed period, most commonly 24 h, or by estimates of the frequency with which specified foods are eaten over a longer period, usually days to a year (food frequency questionnaires (FFQ)) (Subar, 2004). From a list of foods eaten and information on the nutrient composition of foods listed in food composition tables, individual intake of nutrients can be estimated. Nutrients can also be measured in biological samples, such as blood and urine and such data may be more precisely related to nutrient intakes than information gained from dietary recall or the use of FFQs. However, blood concentrations of nutrients may not represent intake as they are also influenced by digestion, absorption, metabolism and excretion (Jenab et al., 2009).

Traditionally, diets have been assessed as individual foods and nutrients, but more recently there has been a trend to look at dietary patterns instead (Michels and Schulze, 2005). It has been argued that this may better represent the complexity of different diets and allows for the possibility that interactions among different foods may be of importance.

Internationally comparable data on the types of dietary patterns and their association with nutrition-related chronic diseases are hindered by the heterogeneity of the dietary surveys. Attempts have been made in few studies (Kuulasmaa et al., 2000; Tunstall-Pedoe et al., 2000; Slimani et al., 2002) to make international comparisons between dietary habits in relation to chronic diseases.

The Epidemiological Prospective Study in Cancer (EPIC) has reported dietary patterns in adults and elderly individuals in 10 European countries (Bamia et al., 2007; Masala et al., 2007; Touvier et al., 2009) using locally designed FFQs. Comparisons of dietary patterns were usually studied with a single 24-h recall questionnaire, standardised for all participant countries. To date, no data are available on dietary patterns across Europe using a common FFQ.

The Global Allergy and Asthma European Network of Excellence (GA²LEN) survey includes a nutritional component designed to study the relation between diet and allergy and asthma in adults across Europe. The GA²LEN FFQ pilot study developed and tested a method for determination of dietary intake patterns in Europe, targeting allergic disease and asthma. The aim of the pilot study was to evaluate whether the FFQ would yield measurements with a sufficient level of accuracy between very different countries; and to assess the validity and reproducibility of the FFQ.

Subjects and methods

Population and study design

Seven centres participating as partners or collaborators in the GA²LEN Network were invited to take part in the study, of which five centres took part: Poland (Lodz), Germany (Berlin), Portugal (Porto), Greece (Athens) and Finland (Helsinki).

Each centre invited 40 men and women aged 21–75 years to complete the FFQ and to provide a fasting blood sample. Exclusion criteria were minimal in order to allow for a sample as representative as possible of the general population: having a severe gastrointestinal disease or mental illness, having changed diet in the past 12 months or being pregnant.

Individuals were selected from hospital or university centres through posted invitations, and from direct recruitment from patient clinics. Selected participants answered the FFQ on two separate occasions within a time frame of 3 months and provided a blood sample to assess blood levels of fatty acids (FAs) after answering the first FFQ.

Food frequency questionnaire

The GA²LEN FFQ was designed to include a wide range of foods and nutrients that could be representative of the average intake in national populations. Foods were also included to allow the investigation of potential associations between intake of specific food items or nutrients and allergic diseases across Europe.

In order to create a representative dietary questionnaire for most countries in Europe, information was collected from national dietary surveys that were used as part of the EPIC study (Slimani et al., 2002) and the FFQ used in the European Community Respiratory Health Survey II (ECRHS II). After combining items that had a very similar nutritional composition, a final list of 200 foods was included.

Classification into food groups

Food groups were classified using the European Food Group (EFG) classification system (Brussaard et al., 2002). This is an international classification based on the European Food Consumption Survey Method designed to ‘define a method for the monitoring of food consumption in nationally representative samples of all age-sex categories in Europe in a comparable way’.

The EFG classification seems sufficiently extensive and clear to allow international comparisons as well as between national nutritional surveys (Ireland et al., 2002), but it has not been implemented in standard international nutritional studies until now. The GA²LEN FFQ was therefore designed with 32 food sections. To improve the accuracy in the collection on data on fat, animal and vegetable spreads were asked separately. Following the example in the Dutch EPIC FFQ, we included various percentage of fat to ascertain the different types of spreads. Local brand names were also included wherever possible in order to facilitate their identification by the participants. Similarly, a broad range of oils was included in the FFQ, and an additional section specifically asked for types of oils or fats used for cooking.

Translation of the FFQ

The standard operational procedure for translation of documents as advised by the World Health Organisation was used.

Nutritional composition and portion sizes

The GA²LEN FFQ comprises seven options to report frequency of food consumption, namely 2+ per day; once per day; 5–6 per week; 2–4 per week; once per week; 1–3 times per week; and rarely or never. Dietary intake reported in the FFQ was registered in standard portion sizes following the Food Standard Agency Food Portion Sizes Guidelines (Food Standard Agency, 2006). The GA²LEN FFQ includes these recommended portion sizes across all the food sections. Nutritional composition was calculated using the sixth edition of the McCance and Widdowson's Composition of Foods (McCance and Widdowson, 2006).

Biomarkers of FA intake: determination of FA composition of plasma phosphatidylcholine

Samples were analysed at the University of Southampton. Blood (2 ml) was taken into lithium heparin, and centrifuged to separate cells from plasma (3000 r.p.m. for 10 min in a standard bench top centrifuge). Plasma was stored at minus 70 °C until shipping on dry ice. The FA composition of plasma phosphatidylcholine (PC), the major circulating phospholipid, was determined using gas chromatography. In brief, total lipid was extracted from plasma using chloroform/methanol (2:1 vol/vol). PC was isolated from the total lipid extract using aminopropylsilica solid phase extraction columns; PC was eluted with chloroform/methanol (3:2 vol/vol). FA methyl esters were prepared from the isolated PC by incubation with methanol containing 2% sulphuric acid for 2 h at 50 °C. FA methyl esters were extracted using hexane and then separated on a Hewlett Packard 6890 gas chromatograph (Hewlett Packard, Avondale, PA, USA). Running conditions were as described elsewhere (Thies et al., 2001). The FA methyl esters were identified by comparison with standards run previously and data are expressed as percentage of total FAs present in the PC fraction.

Statistical analyses

Agreement between nutrient intakes in FFQ₁ and FFQ₂ was assessed using the intra-class correlation coefficient (ICC). Nutrient intakes were log transformed to make their distributions less skewed (a small constant was added to each intake so that zero values could be transformed). Agreement was also calculated using the log transformed nutrient density (nutrient intake in milligrams divided by total energy intake (TEI) in kcal) as a way of adjusting for energy intake. Dietary intake of total saturated FAs (SFAs), monounsaturated FAs (MUFAs), polyunsaturated FAs (PUFAs), n-3 PUFAs, n-6 PUFAs and trans FAs was validated against the corresponding levels in plasma PC using the Pearson correlation coefficient. Strength and direction of association between intake of specific foods rich in FAs with individual and total FAs in plasma PC was measured as the Pearson correlation coefficient. Statistical analyses were performed using Stata 10.1 (STATA StataCorp, College Station, TX, USA)

Ethics

Each participant gave written informed consent and ethical approval was obtained by each participant centre.

Results

A total of 177 participants took part in the study. Mean age was 36.7 years (range 21–75) (Table 1). The distribution of dietary intakes in the participating countries is described in Table 2, which includes data for those that completed the FFQ on both occasions (FFQ_t1 and FFQ_t2) (n=162). In Greece, Portugal, Poland and Germany, the TEI was higher for FFQ_t1 than FFQ_t2, and macro-nutrient intakes were typically higher for FFQ_t1 than FFQ_t2, while in Finland the reverse was true. Intake of micro-nutrients was also slightly higher in FFQ_t1 than FFQ_t2 (data not shown).

Table 1 General characteristics of the participants

Table 2 Daily intake of macronutrients and selected micronutrients by country in participants who responded to the FFQ in both occasions (Median (IQR))

Repeatability of the FFQ at the nutrient level

The ICCs for TEI as well as macro- and micro-nutrient intakes for FFQ_t1 compared with FFQ_t2 are shown in Table 3. ICCs for macronutrients in all participants (that is, all countries) ranged from 0.65 (n-3 FAs) to 0.78 (proteins and trans FAs). ICCs for micronutrients ranged from 0.70 (retinol) to 0.80 (vitamin D). Analyses within individual countries showed ICCs in a similar range. Agreements were generally lower after energy adjustment (using nutrient densities).

Table 3 ICCs for log-transformed nutrient intake (1) and log-transformed nutrient density (2) in FFQ₁ and FFQ₂

A Bland Altman plot (on a log scale) of n-3 FA intake is shown in Figure 1. The 95% limits of agreement for the ratio of intakes in FFQ_t2 and FFQ_t2 were 0.41 to 2.33

Figure 1

Bland Altman plot of n-3 PUFA intake (mg/day) (on a log scale) in FFQ_t2 compared with FFQ_t1 (n=162).

Validation against plasma PC FAs

Total dietary intake (FFQ_t2) of SFAs, MUFAs, PUFAs, n-6 PUFAs and n-3 PUFAs was compared against plasma PC levels of the corresponding FAs. Table 4 shows the correlations between FA estimates from FFQ_t2 and plasma PC levels of those FAs for all participants (all countries). Total dietary PUFAs had a correlation of 0.11 with total plasma PC PUFAs across all participants (all countries) and ranged between −0.22 and 0.16 within individual countries.

Table 4 Correlations between fatty acid intakes reported from FFQ_t1 and plasma PC levels of fatty acids in participant countries

Dietary n-6 PUFAs showed negative correlations with plasma PC n-6 PUFAs in all countries. Dietary n-3 PUFAs showed a correlation of 0.40 with the corresponding plasma PC level in the whole sample and ranged between 0.16 and 0.45 across countries. Good correlations were also observed between dietary n-3 PUFA intake and several individual n-3 PUFAs in plasma (data not shown). Table 5 shows the correlation between estimated dietary intakes of n-6 and n-3 PUFAs and plasma PC levels of individual FAs of potential interest for asthma and allergy. Dietary n-3 PUFAs and plasma PC eicosapentaenoic acid showed a correlation of 0.31 in the total sample, ranging from 0.21 to 0.43 across countries. Dietary n-3 PUFAs showed a correlation of 0.54 with plasma PC docosahexaenoic acid in the total sample, which ranged between 0.19 and 0.54 across countries. The correlations for dietary SFAs and MUFAs and the corresponding plasma PC FAs showed generally poor correlations in all participant countries.

Table 5 Correlations between fatty acid intakes reported from FFQ_t1 and plasma PC levels of fatty acids in participant countries

Discussion

The objective of this pilot study was to assess the repeatability and validity of a FFQ designed to collect typical dietary intake information using a common instrument in different countries across Europe. The study aimed specifically at validating the intake of n-3 FAs, as these have a potential role in asthma and allergic diseases (Kremmyda et al., 2009). We found a good level of repeatability in the FFQ, although we found evidence that intakes from FFQ_t1 were systematically and significantly higher than intakes from FFQ_t2 in most centres and in the total sample. There was generally poorer agreement after adjusting for TEI. Nevertheless, the 95% limits of agreement were still fairly wide: n-3 FA intake estimated from FFQ_t1 could still be half or double the intake estimated from FFQ_t2. Intake of major food groups such as meats, cereals, vegetables, fruits and fish showed a high repeatability (data not shown).

We included a number of food items that were only relevant to one country but not to others. This was intended to ascertain local consumption as well as general intake of foods. The average TEI reported from the FFQ indicates that the questionnaire is able to capture general nutritional intake within the usual range of energy consumed by adults.

The assignment of standard portion sizes was preferred in order to facilitate the interpretation and completion of the FFQ. In large epidemiological studies that involve many different geographical and cultural areas a questionnaire should be kept simple (Noethlings et al., 2003). Studies looking at the effect of including portion size questions have shown small changes in estimated nutrient intakes (Block et al., 1990). Tjonneland et al. (1992) reported that correlation coefficients for food groups and nutrients changed only slightly, indicating that little extra information could be obtained by additional questions about portion size. Cade et al. (2004) observed that correlation coefficients of nutrient intakes tended to be higher in subjects who reported their own portion sizes compared with standard portion size specified in the questionnaire. Bearing in mind the objective of our FFQ, which is to ascertain food intake and variations in dietary habits across countries, we opted to include standard food portion sizes.

We used a nutrient density approach to adjust for energy intake, rather than looking at residuals from a regression adjusting for energy intake. Both approaches are in common use, and a high degree of correlation between intakes adjusted by these two methods (Willet and Stampfer, 1986). Calculating nutrient density meant that we were using the same adjustment for each FFQ assessment, and thus comparing the same quantity each time.

A major element of interest in the design of the FFQ was to include a representative variety of foods rich in n-3 and n-6 FAs because of their potential roles in allergic disease risk (Sala-Vila et al., 2008). Measurement of plasma levels of FAs is considered a good biomarker of the intake of some FAs (Ashley, 1996; Hodson et al., 2008). Although all diet-related biomarkers carry a measurement error and may be poor indicators of intake because of variations in digestion, absorption and metabolic changes after consumption, it is suggested that PUFA status in biological samples including plasma is a useful marker of their regular dietary intake (Arab, 2003; Hodson et al., 2008).

The GA²LEN FFQ includes food groups with separate questions for the varieties of foods with a high content of n-3 and n-6 FAs. These include fish, shellfish, eggs and egg-based dishes, margarine, meats and cereals (Meyer et al., 2003). Our study found a correlation between dietary intake of n-3 PUFAs and plasma PC levels of n-3 PUFAs of 0.40 for all participants and a range between 0.16 and 0.45 across countries. Similar ICCs were observed between dietary n-3 PUFAs and plasma PC docosahexaenoic acid in all the participants. Our results are in agreement with those reported by other studies validating dietary n-3 PUFA estimated intakes from FFQs with corresponding plasma levels (Ma et al., 1995; Woods et al., 2002; Sullivan et al., 2006; McNaughton et al., 2007). In the GA²LEN FFQ, fatty fish showed the best correlation with plasma docosahexaenoic acid and eicosapentaenoic acid in analyses per country and with the total sample (data not shown). Correlations between lean fish and plasma FAs were less strong, while shellfish and other seafood showed poor correlations.

In our study, correlations between dietary intake of SFAs and MUFAs and plasma levels of these FAs were poor. This is likely to be explained by the capacity for endogenous synthesis of SFAs and MUFAs. It is also plausible that differences might be due in part to the fact that plasma reflects more recently intake of FAs than other tissues, such as red blood cells and adipose tissue (Hodson et al., 2008).

The overall aim of the GA²LEN FFQ is the assessment of the relation between dietary habits, food intake and nutrient intakes with asthma and allergy in adults across Europe, using a single common dietary instrument. The pilot study indicates that the FFQ can be answered without major difficulties by individuals from very different countries and cultures. We chose two Mediterranean countries (Greece and Portugal), a central European country (Germany), an Eastern European country (Poland) and a Nordic country (Finland) in an attempt to represent the diversity of the people that will participate in the survey where the FFQ will be used.

There is little evidence so far accumulated in international comparisons of dietary patterns in relation to chronic disease, and none using a common dietary questionnaire or a common food classification system that facilitates comparisons. With few exceptions (Bamia et al., 2005), most of the data published on comparisons of dietary patterns across the participant countries in EPIC have been analysed using a standardised 24-h recall questionnaire (Slimani et al., 2002), which gave a generally accurate measurement of intake in the short term (Kaaks et al., 1997).

The correlations between foods rich in FAs or total dietary n-3 and n-6 PUFAs with plasma PC levels of arachidonic acid, eicosapentaenoic acid, docosahexaenoic acid and total plasma PC levels of n-3 and n-6 PUFAs found in this study indicate that the FFQ captures moderately well dietary intake of these FAs. We are aware of the limitation that using a country-specific table in this validation study with several European countries might not always reflect nutrient composition relevant to each of these countries. An advantage of using the British Table is that it is one of the most comprehensive Food Composition Tables and is being used as a reference method by countries where National Tables are less complete. We are currently working with the EuroFir Network to use the updated nutritional composition databases from all the available tables in Europe to analyse the nutritional information obtained from the GA²LEN follow-up.

In summary, the GA²LEN FFQ is an appropriate tool to estimate dietary intake of a varied range of foods across Europe regardless of cultural and idiomatic differences. The FFQ is a reasonably good indicator of the intake of n-3 FAs, but seems not to be useful for estimating intakes of other FAs.