nutrients
Article
Development of a Semi-Quantitative Food Frequency Questionnaire to Assess the Dietary Intake of a Multi-Ethnic Urban Asian Population
Nithya Neelakantan 1,†, Clare Whitton 1,†, Sharna Seah 1, Hiromi Koh 1, Salome A. Rebello 1, Jia Yi Lim 2, Shiqi Chen 2, Mei Fen Chan 2, Ling Chew 3 and Rob M. van Dam 1,4,5,*
1 Saw Swee Hock School of Public Health, National University of Singapore and National University Health System, Singapore 117549; nithya.neelakantan@u.nus.edu (N.N.); clare_whitton@nuhs.edu.sg (C.W.); sharna.seah@u.nus.edu (S.S.); e0013172@u.nus.edu (H.K.); salome_antonette_rebello@nuhs.edu.sg (S.A.R.)
2 Research and Evaluation Department, Research and Strategic Planning Division, Health Promotion Board, Singapore 168937; LIM_Jia_Yi@hpb.gov.sg (J.Y.L.); CHEN_Shiqi@hpb.gov.sg (S.C.); Chan_Mei_Fen@hpb.gov.sg (M.F.C.)
3 Research and Strategic Planning Division, Health Promotion Board, Singapore 168937; chew_ling@hpb.gov.sg
4 Department of Medicine, Yong Loo Lin School of Medicine, National University of Singapore and National University Health System, Singapore 117549
5 Department of Nutrition, Harvard T.H. Chan School of Public Health, Boston, MA 02115, USA * Correspondence: rob_martinus_van_dam@nuhs.edu.sg; Tel.: +65-6516-4980 † These authors contributed equally to this work.
Received: 7 July 2016; Accepted: 22 August 2016; Published: 27 August 2016
Abstract: Assessing habitual food consumption is challenging in multi-ethnic cosmopolitan settings. We systematically developed a semi-quantitative food frequency questionnaire (FFQ) in a multi-ethnic population in Singapore, using data from two 24-h dietary recalls from a nationally representative sample of 805 Singapore residents of Chinese, Malay and Indian ethnicity aged 18–79 years. Key steps included combining reported items on 24-h recalls into standardized food groups, developing a food list for the FFQ, pilot testing of different question formats, and cognitive interviews. Percentage contribution analysis and stepwise regression analysis were used to identify foods contributing cumulatively ≥90% to intakes and individually ≥1% to intake variance of key nutrients, for the total study population and for each ethnic group separately. Differences between ethnic groups were observed in proportions of consumers of certain foods (e.g., lentil stews, 1%–47%; and pork dishes, 0%–50%). The number of foods needed to explain variability in nutrient intakes differed substantially by ethnic groups and was substantially larger for the total population than for separate ethnic groups. A 163-item FFQ covered >95% of total population intake for all key nutrients. The methodological insights provided in this paper may be useful in developing similar FFQs in other multi-ethnic settings.
Keywords: food frequency questionnaire; FFQ; development; methodology; multi-ethnic; dietary assessment; nutrition

1. Introduction
Diet is an important modiﬁable risk factor for many diseases that are major contributors to morbidity and mortality worldwide [1]. Accurate information on dietary intakes is fundamental for various public health activities; for monitoring nutritional status, for prioritizing and assessing the impact of nutrition programs and interventions, and for identifying novel dietary risk factors for diseases [2]. Food frequency questionnaires (FFQ) are the most commonly used dietary instruments

Nutrients 2016, 8, 528; doi:10.3390/nu8090528

www.mdpi.com/journal/nutrients

Nutrients 2016, 8, 528

2 of 15

in epidemiological studies because they are more feasible to administer in large populations and are able to capture habitual dietary intake [3]. However, developing FFQs in the context of an increasingly globalized, cosmopolitan environment is challenging. Large diversity of food choices can lead to greater within-person and between-person variation in food intake and frequent out-of-home food consumption can limit knowledge of consumed food ingredients making recall more difﬁcult. Dishes in Asian cuisines tend to be particularly complex containing multiple mixed ingredients that can be difﬁcult to identify for persons that do not prepare the dishes.
Another consideration for dietary assessment in cosmopolitan settings is how to address ethnic differences in food consumption. In some multi-ethnic populations, several key ethnic-speciﬁc food items were added to existing generic FFQs [4,5]; in such cases, the validity of the amended FFQ needs to be assessed. In other populations, multiple ethnic-speciﬁc FFQs were used [6], which may perform well within an ethnic group but could result in comparability issues across groups. As it is common for individuals to consume foods from other ethnic cuisines in cosmopolitan settings, an alternative approach is the use of a single FFQ developed for coverage of a multi-ethnic population as a whole [7–9]. While this approach is likely to be preferable in terms of comparability, achieving a high level of nutrient coverage could potentially result in a lengthy FFQ food list.
Singapore is a multi-ethnic (74.3% Chinese, 13.3% Malay, 9.1% Indian and 3.2% others) [10], highly urbanized Asian country with a wide variety of traditional ethnic and international cuisines. Previously, an FFQ for Singaporeans was developed using food consumption data from 1993 (without ethnic group stratiﬁcation) and was primarily designed for measuring intake of fatty acids and cholesterol, in order to better understand cardiovascular risk factors in the population [11]. Food consumption habits have changed in recent decades and there is an increasing interest in food components other than lipids, given the continuing rise in chronic disease prevalence [12]. We therefore used a data-driven approach to develop and pilot test a new comprehensive FFQ that accounts for ethnic-speciﬁc foods and mixed dishes with the use of data from two 24-h dietary recalls collected as part of the Singapore National Nutrition Survey. Few descriptions of the systematic development of FFQs in multi-ethnic settings are available in the literature [6–9,13]; therefore, our procedures and results can be useful for the development of FFQs in similar settings.
2. Materials and Methods
2.1. 24-h Dietary Recall Data Collection
A 24-h dietary recall survey was conducted as part of the National Nutrition Survey 2010 in conjunction with the National Health Survey (NHS) 2010 in Singapore. In total, 1132 NHS-2010 participants aged 18–79 years were invited to take part in two 24-h dietary recalls and 805 participants provided their dietary recall for two intake days covering one weekday or working day and one weekend or non-working day within a month of each other between March and June 2010. Sampling was based on a selection matrix stratiﬁed by gender, ethnicity and age. Malay and Indian participants were over-sampled to provide adequate numbers for statistical comparisons between ethnic groups. Further details on the sampling methodology and sampling weights are published elsewhere [14].
Trained interviewers, using a multiple pass approach [15], conducted a face-to-face 24-h dietary recall interview at the participant’s household. Information regarding socio-demographic factors was also obtained during the interview. The 24-h dietary recall was used to obtain details of all foods and beverages consumed by the study participants over a 24-h period (midnight to midnight) of the day prior to the day of interview. Other information collected included time and venue of meal consumption, description of cooking methods and quantities of foods and beverages consumed. A compendium of food pictures and a set of standard household measures such as a bowl, plate, cup, glass and spoons of various sizes were used to aid the interview. Dietary data of the 24-h recalls were coded using the Food Information and Nutrient Database (FIND), Singapore, which housed Singapore

Nutrients 2016, 8, 528

3 of 15

food composition data [16]. The amount of energy and nutrients contributed by each food item was computed based on nutrient composition and weight.

2.2. Development of the FFQ Nutrients 2016, 8, 528 

3 of 15 

Using 24-h dietary recall data, a data-driven approach [3,17] was adopted to decide on the foods
2.2. Development of the FFQ   
to be included in the FFQ. Cognitive interviews were used to inform decisions related to the wording
Using 24‐h dietary recall data, a data‐driven approach [3,17] was adopted to decide on the foods 
and layout of ttoh bee inFcFluQde,dp ino trhtei FoFnQ.s Ciozgenidtivees icnrteirpvtieiwosn wsearen udsedr etos ipnfoonrms edecciasitoengs orerlaietesd. toF thige wuorredi1ngi llustrates the

various steps inanvdo  llavyoeudt  oinf  tthhe eFFdQe, vpeorltoiopnm  siezen  dt eoscfritphtieonFsF  aQnd.  rEesapcohnsset  ceapteginoriFesi.g  Fuigruere1  1i silldusetrsactersi btheed  below; some steps were itervsataetrpiivos uews easrtnee pidtse irinanvtifovolev raemdnd ien i dnthfoepr dmreeevvde lipoorpeumvsieonouts no ofe ntshe.es.  FFQ. Each step in Figure 1 is described below; some 

 
Figure 1. Flowchart of the food frequency questionnaire (FFQ) development study. 
Figure 1. Flowchart of the food frequency questionnaire (FFQ) development study.
2.2.1. Standardization 
2.2.1. StandardizaWtioe nclassified  ~55,000  data  points  (i.e.,  consumption  of  various  food  items  reported  by  study 
participants  during  the  two  24‐h  dietary  recall  survey)  into  272  food  groups.  To  enable  this 
We classicﬁlaesdsifi~ca5ti5o,n0, w00e sdtaandtaardpizoeidn thtse in(id.iev.i,ducaol nfosoud imtempst i(eo.gn., roedf/gvraeernio apupsle fwoiothd/wiittheomut sskrine ipntoor  ted by study participants duionrnigenr egidteitemhn tes“.t aRwpepcoliep”2e)4  n-aahnmdde sai ewssteiagrern yesdtar nerdeccaairpldlei zsenudar mbvaeessey  df)o oirnn   etmaocah2in 7 ci2nomgfropedoosideitnegt  (rdeoi.sguh.,  pncsoo.omdTploersis)ei annngad  b mcloueolkttiihpnliges   classiﬁcation, we standardizmedethtohde  (ei.ng.d,  divryi,d  fruieadl,  ifno  gordavyit, eamnds  in( es.ogu.p,),r  aendd/  sgimrielaern  reacippeps lwe ewre igtrho/upwedi tthogoeuthters k(ei.gn., into one item “apple”) and afsisshigbanlle ndoordelec siopuep,n aandm inesstafnot nroeoadcleh socuopm). Wpeo csreitaetedd gisenhercaol fmoopd rgirsoiunpg nammeus lttoi pclalessiifny gforoedd  ients. Recipe names were sttiatheenm idsn duaenrpddeeinrz dtheeend tb irbdoeaandsteeirtyd f oooofd ns gtarmpolueap fsion (oed.igsn. ,a g“nbrdre eidatedim”e,s n“ rrteicp(eoe dr.tigesdh.,e bsn”y, o aao nladdr “glneeo spo)rdoalpen odrditsihocneos ”oo)f. k pWienoe prgeleta m(ien.eegd.t,h   od (e.g., dry, fried, in gravyw, ahinted ricien, nsooduleps), a, nadn rdoti spirmatai)l. aErthrneicc riipcee psrewpaerarteiongsr woiuthp deifdferteontg fleatvhoreinrg(se b.ugt .w, iﬁths shimbialalrl  noodle soup, and instant notnhouetd rciloeemntps dooesnuitsepi tdy)i. swhWeerse e( ec.ogcm.r, bebiianttteeedr .dg Fougurrtedhn ewre,i rtwha eml safpoylioto nodnra eigxsecr)lou. duepd onbascmuree sor tuonicdleantsifsiiafbyle iftoemosd friotme ms under the broader food groups (e.g., “bread”, “ri ce dishes”, and “noodle dishes”). We retained the independent identity of staple foods and items reported by a large proportion of people (e.g., white rice, noodles, and roti prata). Ethnic rice preparations with different ﬂavorings but with similar nutrient density were combined. Further, we split or excluded obscure or unidentiﬁable items from the composite dishes (e.g., bitter gourd with mayonnaise).
2.2.2. Pre-Testing of Vegetables Dish Questions
Unlike meat and ﬁsh dishes where the main ingredient was generally clear, the variability of vegetables dishes was vast, often with multiple vegetables mixed together. It was therefore

Nutrients 2016, 8, 528

4 of 15

unclear whether to standardize vegetables (and dishes) as discrete items, subsequently asking for example, “How often do you consume cabbage?” or to standardize as groups of vegetables (and dishes) subsequently asking for example, “How often do you consume pale green leafy vegetables (cabbage/lettuce/cauliﬂower)?” To inform handling of single vs. mixed vegetable dishes, a pre-test of the FFQ vegetable section was conducted using six question formats in a convenience sample of 10 participants from various age, education, occupation and ethnic backgrounds. Participants consistently preferred single item vegetable lists and reported cognitive difﬁculties when vegetables were grouped due to differing consumption frequencies of each individual vegetable. Therefore, vegetable dishes were standardized into groups based on one main vegetable.
2.2.3. Food List of the FFQ
In order to develop a food list for the FFQ, we focused on estimating total energy, carbohydrate, sugar, dietary ﬁber, protein, total fat, saturated fatty acid (SFA), mono-unsaturated fatty acid (MUFA), poly-unsaturated fatty acid (PUFA), calcium, vitamin A, vitamin C, iron and sodium. A separate food list was compiled for each of the three ethnic groups and then combined to ensure that the ﬁnal FFQ adequately represented foods consumed by the main ethnic groups in Singapore. Foods consumed by ≥2% people, contributing cumulatively to ≥90% of key nutrient intakes, or explaining ≥1% of between-person intake variance were considered for inclusion in the food lists. These cut-offs were chosen because they were considered achievable based on other FFQ development studies in the literature [6,11,17,18], yet enough to achieve comprehensive coverage. In addition, foods of particular public health interest such as whole grains and berries were included. More details on these analyses are given in the Statistical Analyses section.
2.2.4. FFQ Design
Layout of the FFQ
Basic, familiar food items consumed by large proportions of the population, such as breads, were placed at the beginning of the questionnaire. Given their major contribution to nutrient intakes, staple foods (e.g., rice, noodles) were placed close to the start to avoid these items being subject to participant fatigue at the end or errors during initial familiarization at the start as suggested by Cade et al. (2002) [19]. Cognitively similar items were listed consecutively (e.g., different types of rice and rice dishes). FFQ layout was also informed by pre-testing and cognitive interviews.
Sub-Type Questions
In order to reduce the length of the food list without losing key details on food intakes, we included sub-type questions, which could be skipped if the main food type was reportedly never consumed. We included questions on the cooking method (e.g., raw/steamed/boiled/stir-fried, etc. for vegetable dishes), type of oil used in food preparation and venue of obtaining food. Coffee, tea and malt beverages sections included sub-type questions for different types of milk and “less sugar” versions. We also asked sub-type questions on reﬁned grains and whole grains.
Frequency of Food Consumption
Food frequency categories “per day/per week/per month/never or rarely” were used in which participants would be asked to provide a number of times within one of the categories. Since self-administered comprehensive FFQs were considered unsuitable for the population, we opted for this open-ended frequency category format for our interviewer-administered FFQ to allow for greater precision. We tested various response category wordings of sub-type questions in our cognitive interviews and we included the optimal format in the FFQ. For example, in one format the FFQ question “how often do you consume vegetables in curry with coconut?” used the response category wording, “never/rarely, sometimes, half the time, mostly, or always”.

Nutrients 2016, 8, 528

5 of 15

FFQ Portion Size
Standard portions were based on conceptually meaningful amounts (e.g., 1 bowl) and researcher judgment. Reference was also made to median portion sizes reported in the 24-h recall data. Local nutrition experts were consulted to assess the face validity of these portion sizes and portion size descriptors were tested during cognitive interviews.
2.3. Pre-Testing and Pilot Testing of the FFQ
Several strategies were adopted to test the FFQ and to identify potential difﬁculties in responding to FFQ questions.
2.3.1. Feedback from Local Food Experts and Stakeholders
Researchers familiar with local food habits checked the face validity of FFQ, appropriateness of portion sizes, incorrect description of foods and advised if any ethnic-speciﬁc local foods were missing from the FFQ food item examples. This included nutrition experts with experience in conducting the National Nutrition Survey, cohort studies, and nutrition interventions in Singapore.
2.3.2. Cognitive Interviews
Cognitive interviews were conducted to understand the thought processes the FFQ questions led to. Thirty participants (10 Chinese, 10 Malays, and 10 Indians) aged 25–75 years who were recruited via local newspaper advertisement or word-of-mouth. Using a standardized protocol which was developed using the verbal probing method [20], researchers administered the FFQ to participants probing on the following aspects: selection of food items (single food vs. mixed dish), estimation of mixed vegetable intake, information on food items, wording of food items (e.g., ﬂavored rice), placement of food items, portion size, various formats of frequency of consumption, and FFQ layout. To ascertain how best to communicate standard serving sizes, participants were asked to compare various verbal descriptions and were also shown pictures of common household measurement tools such as dessert spoon, tea spoon, ladle, bowls, and cups. Each one-on-one cognitive interview took approximately 2 h. Written informed consent was obtained from all participants. This study was approved by the Institutional Review Board (reference code: B-14-082) at the National University of Singapore, Singapore.
2.4. Nutrient Database
A nutrient database for the FFQ was constructed using the 24-h dietary recall data. Each food or beverage was tagged to an FFQ item ID and subsequently data were aggregated to generate a weighted nutrient proﬁle for each FFQ item which reﬂected the relative consumption frequencies of each food sub-type included within each FFQ item.
2.5. Statistical Analysis
Descriptive statistics were used to describe socio-demographic characteristics of the study participants. Sampling weights were applied to the statistical analyses to account for unequal probability of selection and non-response rate. The weighted average intake of the two 24-h recalls was computed by summing 5/7ths of weekday intakes and 2/7ths of weekend intakes. After aggregation of all reported food items, the percentage of people consuming each of the 272 food items over the two-day period was calculated. Foods consumed by <2% people overall or within each ethnic group and contributing to intake of only one nutrient which could not be grouped elsewhere were excluded. Percentage contribution analysis was performed by calculating the contribution of each food item to total daily energy and selected key nutrients (carbohydrate, sugar, dietary ﬁber, protein, total fat, saturated fatty acid (SFA), mono-unsaturated fatty acid (MUFA), poly-unsaturated fatty acid (PUFA), calcium, vitamin A, vitamin C, iron and sodium). Foods that cumulatively contributed at least 90%

Nutrients 2016, 8, 528

6 of 15

to intakes were included in the food list. Foods that contributed at least 1% to intake variance were also considered for inclusion. Stepwise linear regression analyses were performed to assess the foods that explained at least 1% variance for each nutrient. This was done by modeling each key nutrient as a dependent variable and entering the 272 food items (grams per day) as independent variables. All analyses were conducted for the total study population and for each ethnic group separately. Statistical analyses were carried out in IBM SPSS version 23 (IBM Corp, Armonk, NY, USA) and the level of signiﬁcance was set at 5%.
3. Results
3.1. FFQ Development Phase
Table 1 shows characteristics of the 805 participants who took part in 24-h dietary recall survey for the total study sample and by ethnicity. The mean age was 44 (range: 18–79) years, 49.4% were females and the ethnic distribution was 39.5% Chinese, 30.3% Malay, and 30.2% Indians. The mean (standard deviation, SD) body mass index was 25.2 (4.2) kg/m2 for males and 25.3 (5.4) kg/m2 for females. In terms of age group distribution and educational and marital status, the sample reﬂected the general Singapore population [21].
Individual foods and mixed dishes reported in the dietary recall data were standardized into 272 food groups. The total number of food groups reported over the two-day dietary recall varied by ethnic group: Chinese, 207; Malay, 178; and Indian, 183. The key contributors to energy and nutrient intakes generally were food items consumed by large proportions of the population. White rice, fried noodles, poultry dishes, and white bread were commonly consumed in all ethnic groups and contributed to 13%–16%, 4%–7%, 4%–7% and 3%–5% of total energy intake, respectively (Table 2).

Table 1. Characteristics of the participants of the 24-h dietary recall survey.

Characteristics

Total Sample (n = 805)

Chinese (n = 318)

Malay (n = 244)

Indian (n = 243)

p-Value

Age (years) Females, n (%) Educational status, n (%) Primary education or less Secondary school Higher education including vocational
University Marital status, n (%)
Single Married Others † Working status, n (%) Employed Homemaker Retired Unemployed Student/National service Family income (S$ per month), n (%)
<2000 2000–3999 4000–5999
≥6000 Don’t know/refused Body mass index (kg/m2)

44.5 ± 16.0 398 (49.4)
161 (20.0) 309 (38.4) 198 (24.6) 135 (16.8)
173 (21.5) 555 (69.0)
76 (9.5)
519 (64.6) 144 (17.9)
64 (8.0) 17 (2.1) 60 (7.5)
174 (21.6) 222 (27.6) 157 (19.5) 138 (17.1) 113 (14.1) 25.2 ± 4.9

44.5 ± 16.4 43.9 ± 16.1 45.0 ± 15.5 161 (50.6) 122 (50.0) 115 (47.3)

56 (17.6) 109 (34.3) 79 (24.8) 74 (23.3)

54 (22.1) 124 (50.8) 59 (24.2)
6 (2.5)

51 (21.1) 76 (31.4) 60 (24.8) 55 (22.7)

80 (25.2) 213 (67.0) 25 (7.9)

51 (20.9) 172 (70.5)
21 (8.6)

42 (17.4) 170 (70.2) 30 (12.4)

207 (65.1) 49 (15.4) 26 (8.2) 1 (0.3) 35 (11.0)

155 (63.5) 44 (18.0) 24 (9.8)
6 (2.5) 15 (6.1)

157 (64.9) 51 (21.1) 14 (5.8) 10 (4.1) 10 (4.1)

47 (14.8) 67 (21.1) 67 (21.1) 79 (24.8) 58 (18.2) 23.6 ± 3.9

75 (30.7) 78 (32.0) 43 (17.6) 17 (6.9) 31 (12.7) 26.6 ± 5.6

52 (21.5) 77 (31.8) 47 (19.4) 42 (17.4) 24 (9.9) 26.1 ± 4.6

0.72 0.72 <0.001
0.11
0.002
<0.001 <0.001

Values are means ± SDs for continuous variables; number and percentages for categorical variable; † separated, divorced and widowed; S$, Singapore dollars.

Nutrients 2016, 8, 528

7 of 15

Table 2. Top 20 contributors according to percent contribution to total energy intake in the 24-h dietary recall study by ethnicity.

Rank 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Chinese (n = 318)

Malay (n = 244)

Indian (n = 243)

Food items White rice Fried noodles Noodles in soup Chicken dishes Pork dishes Dry noodles White ﬁsh dishes White bread
Coffee Flavored rice
Fried rice Oily ﬁsh dishes
Nuts Filled buns, savory
Dumplings Non-carbonated sweetened drinks 1
Tea Soybean curd dishes Flavored white rice porridge
Red/brown rice

Energy (%) 12.7 6.8 6.3 4.6 3.9 3.2 3.0 2.7 2.4 2.3 1.6 1.5 1.4 1.2 1.2 1.1 1.1 1.1 1.1 1.1

% consumers 81.1 42.1 37.1 60.1 50.3 22.0 46.5 42.5 56.9 22.6 11.0 28.6 4.7 12.6 13.5 29.2 44.0 28.3 12.9 7.5

Food items White rice Chicken dishes White bread Fried noodles White ﬁsh dishes Flavored rice Oily ﬁsh dishes
Coffee
Puffs and pies Tea
Beef dishes Non-carbonated sweetened drinks 1
Noodles in soup Roti prata 2 Fried rice
Malted drinks Noodles in gravy Fried chips and crackers, savory Processed chicken products Mutton and lamb dishes

Energy (%) 15.9 7.3 4.7 4.3 3.9 3.4 3.0 2.9 2.8 2.0 2.0 1.9 1.6 1.5 1.5 1.5 1.3 1.2 1.2 1.2

% consumers 88.1 64.3 54.5 30.3 44.3 33.6 43.0 57.4 20.9 58.6 24.6 31.1 15.2 19.7 10.2 29.9 10.2 20.5 15.6 11.5

Food items White rice White bread Fried noodles Chicken dishes
Coffee Thosai 3 Flavored rice Dhal 4 Chapati 5 White ﬁsh dishes Noodles in soup Wholemeal bread
Tea Oily ﬁsh dishes Mutton and lamb dishes Potato dishes
Roti prata Malted drinks Soybean curd dishes Non-carbonated sweetened drinks

Energy (%) 16.0 4.9 4.4 3.8 3.5 2.7 2.7 2.3 2.3 2.0 1.9 1.8 1.8 1.6 1.4 1.4 1.4 1.3 1.3 1.3

1 excluding juices and bottled tea/coffee; 2 ﬂour-based pancake; 3 fermented rice pancake/crepe; 4 lentil stew; 5 whole wheat Indian bread.

% consumers 85.6 52.7 30.0 43.6 62.1 25.1 23.5 46.5 17.3 26.3 11.5 23.0 51.0 23.9 11.9 37.9 16.5 23.9 18.1 18.5

Nutrients 2016, 8, 528

8 of 15

Some food groups were consumed by smaller proportions of the population but made signiﬁcant contributions to nutrient intakes because of high nutrient density, such as dhal (lentil stews) which was consumed by 47% of Indians and contributed to 9% of ﬁber intake, and puffs and pies which were consumed by 21% of Malays and contributed to 3% of energy intake. Although many food groups made important contributions to nutrient intakes in all ethnic groups such as white rice, white bread, poultry dishes, coffee and tea (Table 2), there were a number of foods for which contribution to nutrients, explanation of between-person variance (Table S1) and proportion of consumers varied greatly between ethnic groups. For example, the range of proportion of consumers across ethnic groups was 0%–25% for thosai (fermented rice pancake), 1%–47% for dhal, 0%–50% for pork dishes, 26%–47% for white ﬁsh dishes, 17%–38% for potato and 12%–37% for noodles in soup (Table 2).
For the selection of foods for inclusion in the FFQ, we examined both their contribution to absolute energy and nutrient intakes and to variation in energy and nutrient intakes. Here we will focus on the results for ﬁber intake to illustrate our ﬁndings. The food items with the greatest contribution to ﬁber intakes were white rice, white bread, noodles, wholegrain products, dhal, ﬁber-rich vegetables, and fruits (Table S2). However, for both contribution to absolute intakes and variation in intakes the rankings of foods differed between the ethnic groups. For instance, in the Indian ethnic group, dhal was the biggest contributor (9%) to ﬁber intake and was the top food item explaining most of the variation in ﬁber intake (R2 = 0.26) in the regression model, but it was not a substantial contributor in the Chinese or Malay ethnic group due to low consumption. Table 3 shows similarities (e.g., noodles) as well as differences (e.g., idli, chapati, and dumpling) in the selection of foods that explained at least 60% of variability in ﬁber intake and contributions to absolute ﬁber intake across ethnic groups. In the dhal example above for Indians, both percentage contribution and regression analysis indicated that dhal should be included in the FFQ food list. This was the case for a number of foods while for other foods, the results from the two types of analyses indicated different items for the food list. For example, white rice contributed 5%–6% of total ﬁber intake, across ethnic groups. However, due to the limited between-person variability, white rice did not emerge as a substantial contributor to variability in ﬁber intake. Conversely, innards contributed to <1% of absolute ﬁber intake, but due to higher between-person variability it was among the top contributors to ﬁber intake among ethnic Indians.
Figure 2 presents the number of foods needed to explain at least 60% variability for the total energy and selected nutrients by ethnicity. For carbohydrates, the highest number of foods was needed to explain the between-person variability, whereas the smallest number of foods was required for vitamin A and C. For most nutrients measured, fewer foods were needed to explain the between-person variability among Indians as compared to the other ethnic groups. Due to the consumption of a large variety of foods, a relatively higher number of foods were needed to explain the variability among Chinese.

Nutrients 2016, 8, 528

9 of 15

Table 3. Foods that explained at least 60% of variation in total ﬁber intake and their contribution to total ﬁber intake in the 24-h dietary recall survey by ethnicity.

Chinese (n = 318)

Malay (n = 244)

Indian (n = 243)

Rank
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27

Food items
Pears Chocolate Fried noodles Ice cream 1 Sweet desserts in soup Egg, fried/scrambled/braised French fries Red/brown rice
Nuts Olives Papaya Lettuce Fried rice with vegetables Noodles in soup 100% fruit or vegetable juice Flavored rice Oranges Potato dishes Dragon fruit Pasta with meat/ﬁsh/vegetables Non-carbonated sweetened drinks 2 Filled buns, savory Lotus root dishes Kaya 3 Dhal 4 Gourd dishes Canned meat

Cum R2 †
0.11 0.16 0.20 0.24 0.26 0.29 0.32 0.35 0.37 0.38 0.40 0.42 0.44 0.45 0.47 0.48 0.49 0.51 0.52 0.53 0.54 0.55 0.57 0.58 0.58 0.59 0.60

% of ﬁber intake ‡
2.78 0.55 7.05 0.49 1.32 0.02 1.17 1.30 1.58 0.01 1.72 0.12 0.39 5.29 0.44 3.09 1.07 1.38 0.57 0.90 0.28 1.63 0.13 0.06 0.21 0.53 0.03

Food items
Lontong dishes 5 Chicken dishes Puffs and pies
Roti prata 6 Dumplings Egg, fried/scrambled/braised Flavored rice Pancake/hotcake/wafﬂe
Oranges Fried chips and crackers, savory
Beef dishes Bean dishes Milkshake Cauliﬂower dishes Fried noodles Wholemeal bread Sweet desserts in soup (with coconut) Powdered nutrition drink White ﬁsh dishes Fried rice, plain or with meat/ﬁsh Mutton and lamb dishes Peanut butter
Pears Whole milk Soya milk, low sugar Preserved vegetables Cabbage dishes

Cum R2 †
0.09 0.16 0.22 0.26 0.30 0.33 0.36 0.38 0.40 0.43 0.45 0.47 0.49 0.50 0.52 0.53 0.55 0.57 0.58 0.59 0.60 0.61 0.62 0.63 0.64 0.64 0.65

% of ﬁber intake ‡
1.30 6.04 2.77 1.76 0.58 0.03 3.69 0.88 0.63 1.20 1.93 1.33 0.20 0.22 4.25 2.25 0.61 0.00 3.29 0.85 0.89 0.49 0.93 0.01 0.29 0.05 1.50

Food items
Dhal Innards Chapati 7 Other grains 8
Idli 9 Red/brown rice Pasta with meat/ﬁsh/vegetables Flavored wholegrain breakfast cereal
Nuts Papaya Soya milk, low sugar Bean dishes Ice cream 1 Savory fried snack Chicken dishes Filled buns, sweet Guava Wholemeal bread Fried noodles Avocado Noodles in soup Plain wholegrain breakfast cereal Sweet desserts in soup Apples Dry noodles Cured pork products Banana ﬂower

Cum R2 †
0.26 0.39 0.46 0.54 0.61 0.64 0.67 0.70 0.72 0.74 0.75 0.77 0.78 0.79 0.79 0.80 0.81 0.81 0.82 0.83 0.83 0.84 0.84 0.85 0.85 0.86 0.86

% of ﬁber intake ‡
9.22 0.00 3.25 0.43 1.62 0.69 0.45 0.02 1.11 1.69 0.20 1.58 0.27 0.79 2.12 0.71 0.73 4.10 4.18 0.40 1.25 0.16 1.00 2.69 0.38 0.05 0.00

† Cumulative R2 of the stepwise regression model. The ordering shows the process and in the ﬁnal model, values may no longer be ordered according to size since addition of later items may reduce or increase values of earlier items; ‡ percent contribution to total ﬁber intake; 1 Ice cream with or without toppings; 2 excluding juices and bottled tea/coffee; 3 coconut jam; 4 lentil stew; 5 compressed rice cake; 6 ﬂour based pancake; 7 whole wheat Indian bread; 8 e.g., semolina, and millet; 9 savory steamed rice cake.

NNuuttrriieennttss 22001166,, 88,, 552288 

1100 ooff 1155 

 
Figuree 22..  Number  of  foods  that  explainedd  60%%  of  variatioonn  inn  eneerrggyy,  carbohydrate,  fﬁiber,  protein,  saturated faatt,, vviittaamminin AA anadnd vivtaitmamini nC Cintianktea kine einthentihc ngircougrposu apnsda tnhde otvheroalvl esrtauldl ys tpuodpyuplaotipounl.a Dtiaotna.  Daraet abaarseedb aosend  sotenpswteispew  riseegrreesgsrieosns iownitwh itnhuntruietnritesn  tass ads edpeepnednednetn  at nadn d227722  ssttaannddaarrddiizzeedd  foooddss  aas  indeppeennddeenntt vvaarriiaabblleess.. 

33..22.. FFFFQQ TTeessttiinngg PPhhaassee:: CCooggnniittiivvee IInntteerrvviieeww RReessuullttss 
MMoosstt ppaarrttiicciippaannttss ffeelltt aabbllee ttoo rreeppoorrtt wwiitthhiinn tthhee ssppeecciiﬁfieedd ttiimmeeffrraammee ooff tthhee ppaasstt yyeeaarr bbyy rreeffeerrrriinngg  ttoo hhaabbiittuuaall bbeehhaavviioorrss;; ((““IIff II wwoorrkk II wwiillll vviissiitt tthhaatt ssttaallll uussuuaallllyy tthhrreeee ttiimmeess ppeerr wweeeekk.. AAtt ttiimmeess wwhheenn II  vviissiitt tthhiiss hhaawwkkeerr.…. .I Iwwilill laalwlwaayyss aasskk ffoorr tthhee kkaannggkkoonngg…. .s.os othtahta’st’ shohwow I Ireremmeemmbbeer”r”. .mmaalele ((MM)), ,2255 yy,,  MMaallaayy));; ((““MMyy ppaatttteerrnn iiss tthhee ssaammee.…. .I Icacann rroouugghhlyly ssaayy tthhaatt”” MM,, 4400 yy,, IInnddiiaann)).. PPaarrttiicciippaannttss wwhhoo hhaadd  rreecceennttllyy cchhaannggeedd tthheeiirr ddiieettss wweerree uunnssuurree hhooww ttoo aannsswweerr,, ffoorr eexxaammppllee oonnee ppaarrttiicciippaanntt hhaadd ssttooppppeedd  ccoonnssuummiinngg ssuuggaarr-‐sswweeeetteenneedd bbeevveerraaggeess 66 mmoonntthhss pprreevviioouussllyy.. IItt wwaass nnootteedd tthhaatt iinntteerrvviieewweerrss wwoouulldd  nneeeedd ttoo rreecceeivivees pspeceicﬁifcict rtarianiinnigngo nonh ohwowto  thoa nhdanledclaes  ceasswesh  ewreheinreta  iknetsakheasd  hcahda ncgheadngoveder  otvheerp  tahset  ypeaasrt.  year.T  he preferred wording for frequency categories for food sub-type questions (such as how often wholTeghrea pinrevfearrrieedti ews oorfdiangfo foodr fwreeqrueenucsyed c)atwegaosri“eNs efovre rfo/oRda rseulby‐”t,yp“Se oqmueestitmioenss” (,su“Hcha alfs thhoewt iomftee”n,  “wMhoosltelgyr”a, iann  dva“rAieltmieoss  toaf lwa afyosodo rwalewrea yus”se. d)  was  “Never/Rarely”,  “Sometimes”,  “Half  the  time”,  “MosTthlye”re, awnedr e“Avalmrieodsto apliwnaioynss oorn altwheasytsa”n. d ard servings sizes used on the FFQ. For vegetables many particTihpearnet sweearsei vlyariideedn otipﬁiendiownsi tohn“ tshceo sotpan” daanrdd pserrevfeinrrgesd sitzheiss udseesdc roipn ttihoen FtFoQ“.c Fuopr” v(e“gNetoawblaedsa mysanayt   tphaertsitcaipllasnutss ueaalsliylyt hideyensttirfaieidgh wt iatwh a“yscjouospt ”s earnvde yporeufewrriethd tthhaist donesecrsicpotoiopn” tfoe m“caulep”(F ()“,N46owy,aMdaaylsa ya)t.  Dthees csrtiablilns gumsueaalltya sthae“yp  satlrmai-gshizte  adwpaieyc  eju”srt astehrevret hyaonua  wﬁiltlhe tt/hcahto  opnwe asscoporepf”e rfreemdablye  m(Fo),s  t46p ayr,t iMcipalaanyt)s.  (D“Iesccarnibﬁinggu rme eitaot uast baa “speadlmon‐stihzeedh apniedc”e.”F r,a5t1heyr, Cthhainn eas fei)l,lebtu/cthootph ewrsasfe pltreitfewrraesde absyi emr otosta pnasrwtiecripwanitths  r(e“fIe craenn cfeigtuorteh iet socuoto bpa,seesdp eocni atlhlye hwahnedn”m. Fe, a5t1w ya, sCchuint einsteo), sbmuat loltphieercse sfe. lt it was easier to answer with  referPenocste- ctoog tnhiet isvceoionpte, revsipeewciaamllye nwdhmeenn mtsetaot twheasF FcQut iinnctlou dsmeda:llc hpaiencgeisn. g the ordering of certain items, for exPaomspt‐lceopglnaictiinvge pinlatienrvpioerwri dagmeebnedfomreeﬂnatsv otore  tdhteo  FaFvoQi dindcoluubdleed-c: ocuhnatninggin; agd  tdhieti oonrdoefreixnagm  opfl ecserstuacihn  aitsefmors,“ ﬂfoarv oerxeadmripclee”  pfolar cwinhgic hpltahienn  puomrbriedrgoef ebxeafmorpe leflsagviovreendi ntcor eaavsoeiddf rdoomutbhlere‐ceotuontetinn;ga;n  addindcitliuodni nogf  mexualmtipplleesd seusccrhi patsi ofnosr o“fflsaevrovriendg rsiiczee”s fwohr ewrehiacphp trhoep rniuatme,bee.rg .o, f“ e1xﬁasmhpﬁllelse tg/i1vepna limnc-sreizaesdedp ifercoem”. three  to  ten;  and  including  multiple  descriptions  of  serving  sizes  where  appropriate,  e.g.,  “1  fish  fillet/1 
3p.a3l.mFi‐nsiazleFdF pQiece”. 

i3n.3ta. kFTeinhsaeol ﬁfFneFnaQle fr ogoydalnisdt

comprised 163 items that contributed cumulatively to selected nutrients. There was little difference in the

at least 95% of percentage of

absolute variance

The final food list comprised 163 items that contributed cumulatively to at least 95% of absolute 

intakes  of  energy  and  selected  nutrients.  There  was  little  difference  in  the  percentage  of  variance 

Nutrients 2016, 8, 528

11 of 15

explained by the ﬁnal food list compared with that explained by all standardized food groups. For example, the ﬁnal food list accounted for 89.1% of between-person variation in energy intakes, while including all standardized food groups would account for 91.1%. For ﬁber the ﬁnal food list accounted for 66.0% variation, while including all standardized food groups would account for 73.8%. The FFQ covered 18 sections reﬂecting different types of foods (e.g., bread, rice, noodles, etc.) with ﬁxed serving sizes (e.g., 2 slices of bread, 1 rice bowl, and 1 plate of noodles) and four response categories to measure food frequency (per day, per week, per month, and never/rarely) over the past year. Administration time was approximately 45 min. The ﬁnal FFQ is shown in Figure S1.
4. Discussion
We developed a food frequency questionnaire for adults of Indian, Malay, and Chinese ethnicity residing in Singapore, using data-driven methods with an emphasis on ensuring good coverage of foods and nutrients for each of the ethnic groups. Using national level 24-h recall data, we ascertained both overall and by ethnic group the proportion of people consuming each food, which foods were key contributors to absolute intakes of energy and nutrients, and which foods were responsible for a large proportion of between-person variation in nutrient intakes. Pre-tests, cognitive interviews and stakeholder input allowed us to reﬁne the food list, wording and format of the questionnaire, resulting in a food list of 163 items which covers at least 95% of absolute intake and key sources of variability in intake of selected nutrients (energy, protein, carbohydrate, fat, MUFA, PUFA, SFA, sugar, ﬁber, vitamin A, vitamin C, iron, calcium and sodium), both in the general population and in major ethnic population sub-groups. Based on cognitive interviews, for all ethnic groups, the FFQ food item examples included items that appeared to be relevant and intuitive (for example, “ﬂavored rice”, e.g., chicken rice, briyani, nasi lemak, pilau, nasi minyak, yellow rice, olive rice, tomato rice, saffron rice, and yam rice), and the question format appeared to be comprehendible at all levels of education.
The variety of mixed dishes available in Singapore is wide, dishes generally have multiple minor ingredients, and eating out is common and as a result awareness of composition of dishes can be low. Therefore, the aim was for the FFQ to ask about these mixed dishes rather than the discrete items within a dish. In addition, the availability of recent population-level 24-h dietary recall data meant that an FFQ analysis database could be constructed containing nutrient proﬁles for each food list item that were weighted according to the relative consumption by the population of each variety of food within a food list item. The ﬁnal food list comprises a number of mixed dishes, as well as dishes that have been minimally separated. More standard cereal-based mixed dishes such as noodle dishes remained as a whole dish on the FFQ. Rice-based dishes where the rice was mixed with meat/ﬁsh/vegetables before serving also remained as a whole dish. For dishes that comprised rice that was not mixed with meat/ﬁsh/vegetables, the components were asked about separately. For example, steamed rice, black pepper chicken, and kai-lan in oyster sauce would be captured in three different sections of the FFQ, i.e., “Rice”, “Poultry dishes” and “Vegetable dishes”, despite being consumed as one meal. While a possible consequence of having a combination of mixed and discrete items on the food list is that there is potential for items to be counted twice, it was thought that a food list of only discrete items would result in under-reporting because participants may not be aware of the contents and quantities of items within complex mixed dishes.
The nutrient coverage of our FFQ food groups was higher than our original cut-off of 90% because of the combination of ethnic-speciﬁc lists which each individually covered 90% of nutrient intakes. Nutrient coverage was similar to that observed in other studies using similar data-driven methods [6,17,18], while the food list was longer than most of these. Willet [3] recommends that an FFQ list should contain a maximum of 130 food items to limit participant burden. Our list was slightly longer because of the inclusion of ethnic-speciﬁc items, inclusion of discrete rather than grouped items (e.g., “spinach”, “kai-lan”, and “cabbage” rather than “green leafy vegetables”), and inclusion of individual food items of speciﬁc public health interest, such as whole grains and berries. However, participant burden is not only determined by the number of questions, but also by the

Nutrients 2016, 8, 528

12 of 15

amount of time taken to answer each question. Our FFQ contains a number of ethnic-speciﬁc items to which participants can immediately report non-consumption; for such questions the cognitive burden is minimal. Pre-testing of our questionnaire suggested that recalling intake frequency of grouped categories, such as “dark green leafy vegetables” was a greater cognitive challenge than to recall intake frequency of discrete items. This ﬁnding agrees with observations in previous studies [22]. Therefore, while our food list is relatively long, the burden and cognitive challenge of each individual question is relatively low compared to other FFQs.
The number of food items required to explain the majority of between-person variance in our population was similar to observations in several studies [9,18,23] but much higher than reported in other studies [6,24–28]. It should be noted, however, that there are large methodological differences between these studies. Several studies evaluated the proportion of variance in nutrient intakes assessed with a longer FFQ that a new shorter FFQ is able to detect [25,27,28]. This approach is expected to lead to higher R2 values than our approach, because the number of food items in the longer FFQ is already limited as compared with the large number of food items in 24-h recalls. In our study, we modeled nutrient intakes from 24-h dietary recalls as the dependent variable, using potential FFQ food groups as the independent variables, because this most closely represents the level of detail that will be collected by the FFQ. Our ﬁndings are similar to those from several previous studies that also used nutrient data from open-ended dietary assessment methods with predictors as grams of potential FFQ food groups [9,18,23]. In a few other studies, a relatively short food list explained a higher proportion of variance than in our study [24,26], which may reﬂect a lower level of dietary diversity within these populations.
Cognitive interviews led to reﬁnements in frequency category wording, food group ordering, descriptions and examples. They also showed that there is no “one-size-ﬁts-all” for certain FFQ aspects such as serving size descriptions. Rather than forcing participants to answer according to a measure that may not be relevant for them (e.g., palm-sized piece of chicken), we included two different but approximately equivalent standard servings sizes for a number of food items. For example, for ﬁsh, the standard serving is given as “1 ﬁsh ﬁllet/1 palm-sized piece”. Although our cognitive interview sample size was modest, it was diverse in terms of age, gender and ethnicity. Many of the participants had some interest in food and nutrition, and had higher than average educational attainment. However, we did not observe that participants with low educational attainment had more or less cognitive difﬁculty with the FFQ as compared with participants with higher educational attainment.
The aim was for the FFQ development to be data-driven and this was achieved to a large extent, but the process can never be completely objective. Standardization of foods/recipes into groups was conducted within a decision framework based on ingredients, nutrients and conceptual similarities supported by feedback from pre-tests and cognitive interviews. Creation of food groups, into which foods are standardized, however, is based on researchers’ judgment, for example, having a separate group for carbonated and non-carbonated soft drinks. Since the items on the ﬁnal food list were based on cut-offs such as contributing to 90% of intake of at least one nutrient of interest, a different grouping strategy may have resulted in different food list items on the questionnaire. For example, items that were not covered in our ﬁnal food list such as canned fruit and other grains would have been included if canned fruit had been grouped with fruit, and other grains had been grouped with rice at the initial standardization stage. This issue applies to all FFQs developed using data-driven methods.
Our FFQ was developed using two-day 24-h dietary recall data from a national sample of 805 participants. Using two days of dietary recall data rather than a single recall means that some intra-individual variation in food and nutrient intakes is accounted for, especially that related to day type (i.e., work day vs. non-work day). Although two days are unlikely to be enough to capture all of the food types eaten by an individual, this is compensated for by the size of the sample population. Thus values for percentage contribution of food to nutrient intakes should be fairly reliable. However, more caution is required when interpreting values such as proportion of consumers and between-person variation, which are more heavily affected by within-person variation. For these

Nutrients 2016, 8, 528

13 of 15

reasons, we believe that using both the contribution of foods to absolute energy and nutrient intakes and variability in these intakes is valuable. The relative validity of our new FFQ in comparison to other dietary assessment methods is currently being tested. The 24-h dietary recall data used also had some potential limitations. Firstly, the data were collected in 2010 but new food products and trends emerge regularly, so it is possible that certain important items are missing from the food list simply because they did not appear in the recall data and it was decided not to add items to the list based on anecdotal evidence. The two days of dietary recall were mostly collected within two weeks of each other, between March and June 2010. While seasonal variation was not well captured in this data, seasonal variation in Singapore is mostly limited to some fruits and foods associated with cultural/religious festivals, such as mooncakes and pineapple tarts. This was accounted for in part by adding common seasonal foods to the FFQ such as durian. Foods consumed at speciﬁc festivals may not be well captured, but the impact of these items on habitual intakes would be low. Some under-reporting in 24-h recalls is common due to incomplete recall of foods and desirability bias, but such under-reporting would probably have to be common and severe to lead to missing key foods from the FFQ list. Another limitation is that for sugar, but not for other nutrients, the food composition database contained missing values (12% of the food codes of which around 20% plausibly contain substantial amounts of sugar). Most of the local dishes in the food composition data had nutrient proﬁles derived from laboratory analysis of a number of samples purchased from food outlets, which is a major strength of the 24-h dietary recall data, and provides conﬁdence in the nutrient coverage estimates.
5. Conclusions
In the present study, an FFQ for the adult multi-ethnic Singapore population was developed with good coverage of foods and beverages contributing to intakes of energy, carbohydrate, protein, fat, SFA, MUFA, PUFA, sugar, ﬁber, vitamin A, vitamin C, calcium, iron and sodium. The intention is for this FFQ to be used in future surveys, for epidemiology and surveillance purposes, replacing the local FFQ developed in the 1990s. This should allow for more robust insights into diet-disease associations for a range of nutrients. An on-going validation study including repeat 24-h dietary recalls, collection of biomarkers and repeat administrations of the FFQ to assess reproducibility will provide insights into the performance of the FFQ. We have described our approach to developing this FFQ and associated nutrient analysis database, suitable for a diverse multi-ethnic population, and have outlined the challenges faced and solutions adopted. The experience of developing the FFQ provided us with several key lessons. Firstly, the percentage contribution and stepwise regression analyses were supplementary in that they highlighted different foods for inclusion in the food list. Secondly, a combination of mixed dishes and more discrete items were required on the FFQ for reporting ease and better accuracy, and this may be applicable to other Asian settings. Thirdly, in order to balance the length of the questionnaire with coverage, broad researcher-deﬁned categories were sometimes required, which may not be instantly familiar to participants. Therefore, several examples for each line item were often required to ensure variants relevant to every ethnic group were covered. In an increasingly globalized world, with widespread availability of international food choices, such methodological insights are relevant in a growing number of cosmopolitan settings, and the resulting dietary assessment tools will be instrumental for research to explain the role of ethnicity in diet-disease relationships.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6643/8/9/528/s1, Table S1: Foods that explained at least 60% of between-person variation in total energy intake in the 24-h dietary recall survey by ethnicity; Table S2: Top 20 contributors according to percent contribution to total ﬁber intake in the 24-h dietary recall study by ethnicity; Figure S1: The food frequency questionnaire
Acknowledgments: The authors would like to acknowledge participants who took part in this study. The current study was supported by funding from the Ministry of Education Academic Research Fund Tier 1 FRC (FY2013), Singapore. This fund covers the costs to publish in open access.
Author Contributions: N.N., S.A.R. and R.M.v.D. designed the study and developed the analytical plan; N.N., C.W. and S.S. executed the study; N.N. and C.W. conducted cognitive interviews; N.N., C.W. and

Nutrients 2016, 8, 528

14 of 15

H.K. analyzed the data; J.Y.L., S.C., M.F.C. and L.C. contributed data; N.N., C.W., S.A.R. and R.M.v.D. contributed to the interpretation of data and critically revised the manuscript for important intellectual content; N.N. and C.W. co-wrote the manuscript; and R.M.v.D. supervised the study. All authors reviewed and edited the manuscript. All authors read and approved the ﬁnal manuscript.
Conﬂicts of Interest: The authors declare no conﬂict of interest. The founding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.

Abbreviations

The following abbreviations are used in this manuscript:

F FFQ FIND M MUFA NHS PUFA SD SFA y

female food frequency questionnaire food information and nutrient database male mono-unsaturated fatty acid national health survey poly-unsaturated fatty acid standard deviation saturated fatty acid years

References
1. World Health Organisation. Noncommunicable Disease Report-Chapter 1; Burden: Mortality, Morbidity and Risk Factors. Available online: http://www.who.int/nmh/publications/ncd_report_chapter1.pdf?ua=1 (accessed on 5 July 2016).
2. Thompson, F.E.; Subar, A.F. Nutrition in the Prevention and Treatment of Disease: Dietary Assessment Methodology, 3rd ed.; Elsevier Inc.: Oxford, UK, 2013. Available online: http://appliedresearch.cancer.gov/diet/adi/ thompson_subar_dietary_assessment_methodology.pdf (accessed on 11 August 2016).
3. Willett, W. Nutritional Epidemiology, 3rd ed.; Oxford University Press: New York, NY, USA, 2013; Volume 40. 4. Mayer-Davis, E.J.; Vitolins, M.Z.; Carmichael, S.L.; Hemphill, S.; Tsaroucha, G.; Rushing, J.; Levin, S. Validity
and reproducibility of a food frequency interview in a multi-cultural epidemiology study. Ann. Epidemiol. 1999, 9, 314–324. [CrossRef] 5. Nettleton, J.A.; Rock, C.L.; Wang, Y.; Jenny, N.S.; Jacobs, D.R. Associations between dietary macronutrient intake and plasma lipids demonstrate criterion performance of the multi-ethnic study of atherosclerosis (mesa) food-frequency questionnaire. Br. J. Nutr. 2009, 102, 1220–1227. [CrossRef] [PubMed] 6. Beukers, M.H.; Dekker, L.H.; de Boer, E.J.; Perenboom, C.W.; Meijboom, S.; Nicolaou, M.; de Vries, J.H.; Brants, H.A. Development of the helius food frequency questionnaires: Ethnic-speciﬁc questionnaires to assess the diet of a multiethnic population in the netherlands. Eur. J. Clin. Nutr. 2015, 69, 579–584. [CrossRef] [PubMed] 7. Ireland, P.; Jolley, D.; Giles, G.; O'Dea, K.; Powles, J.; Rutishauser, I.; Wahlqvist, M.L.; Williams, J. Development of the melbourne ffq: A food frequency questionnaire for use in an australian prospective study involving an ethnically diverse cohort. Asia Pac. J. Clin. Nutr. 1994, 3, 19–31. [PubMed] 8. Stram, D.O.; Hankin, J.H.; Wilkens, L.R.; Pike, M.C.; Monroe, K.R.; Park, S.; Henderson, B.E.; Nomura, A.M.; Earle, M.E.; Nagamine, F.S.; et al. Calibration of the dietary questionnaire for a multiethnic cohort in hawaii and los angeles. Am. J. Epidemiol. 2000, 151, 358–370. [CrossRef] [PubMed] 9. Shahar, D.; Shai, I.; Vardi, H.; Brener-Azrad, A.; Fraser, D. Development of a semi-quantitative food frequency questionnaire (ffq) to assess dietary intake of multiethnic populations. Eur. J. Epidemiol. 2003, 18, 855–861. [CrossRef] [PubMed] 10. Department of Statistics. Singapore in Figures 2016. Available online: https://www.singstat.gov.sg/docs/ default-source/default-document-library/publications/publications_and_papers/reference/sif2016.pdf (accessed on 4 August 2016). 11. Deurenberg-Yap, M.; Li, T.; Tan, W.L.; van Staveren, W.A.; Deurenberg, P. Validation of a semiquantitative food frequency questionnaire for estimation of intakes of energy, fats and cholesterol among singaporeans. Asia Pac. J. Clin. Nutr. 2000, 9, 282–288. [CrossRef] [PubMed]

Nutrients 2016, 8, 528

15 of 15

12. Ministry of Health, Epidemiology and Disease Control Division. Report of the National Health Survey 2010. 2011. Available online: https://www.moh.gov.sg/content/dam/moh_web/Publications/Reports/2011/ NHS2010%20-%20low%20res.pdf (accessed on 4 August 2016).
13. Jayawardena, R.; Swaminathan, S.; Byrne, N.M.; Soares, M.J.; Katulanda, P.; Hills, A.P. Development of a food frequency questionnaire for sri lankan adults. Nutr. J. 2012, 11, 63. [CrossRef] [PubMed]
14. Health Promotion Board, Research and Strategic Plannind Division. Report of the National Nutrition Survey 2010. Available online: http://www.hpb.gov.sg/HOPPortal/content/conn/HOPUCM/ path/Contribution%20Folders/uploadedFiles/HPB_Online/Publications/NNS-2010.pdf (accessed on 8 March 2016).
15. Conway, J.M.; Ingwersen, L.A.; Vinyard, B.T.; Moshfegh, A.J. Effectiveness of the us department of agriculture 5-step multiple-pass method in assessing food intake in obese and nonobese women. Am. J. Clin Nutr. 2003, 77, 1171–1178. [PubMed]
16. Health Promotion Board. Energy and Nutrient Composition of Food. 2011. Available online: http://focos. hpb.gov.sg/eservices/ENCF/ (accessed on 4 August 2016).
17. Block, G.; Hartman, A.M.; Dresser, C.M.; Carroll, M.D.; Gannon, J.; Gardner, L. A data-based approach to diet questionnaire design and testing. Am. J. Epidemiol. 1986, 124, 453–469. [PubMed]
18. Kobayashi, T.; Tanaka, S.; Toji, C.; Shinohara, H.; Kamimura, M.; Okamoto, N.; Imai, S.; Fukui, M.; Date, C. Development of a food frequency questionnaire to estimate habitual dietary intake in japanese children. Nutr. J. 2010, 9, 17. [CrossRef] [PubMed]
19. Cade, J.; Thompson, R.; Burley, V.; Warm, D. Development, validation and utilisation of food-frequency questionnaires—A review. Public Health Nutr. 2002, 5, 567–587. [CrossRef] [PubMed]
20. Willis, G.B. Cognitive Interviewing: A “How to” Guide; Research Triangle Institute: Research Triangle Park, NC, USA, 1999.
21. Department of Statistics. Census of Population 2010 Statistical Release 1: Demographic Characteristics, Education, Language and Religion. 2010. Available online: http://www.singstat.gov.sg/publications/ publications-and-papers/cop2010/census10_stat_release1#sthash.dawylkhf.dpuf (accessed on 26 April 2016).
22. Subar, A.F.; Thompson, F.E.; Smith, A.F.; Jobe, J.B.; Ziegler, R.G.; Potischman, N.; Schatzkin, A.; Hartman, A.; Swanson, C.; Kruse, L.; et al. Improving food frequency questionnaires: A qualitative approach using cognitive interviewing. J. Am. Diet. Assoc. 1995, 95, 781–788. [CrossRef]
23. Wakai, K.; Egami, I.; Kato, K.; Lin, Y.; Kawamura, T.; Tamakoshi, A.; Aoki, R.; Kojima, M.; Nakayama, T.; Wada, M.; et al. A simple food frequency questionnaire for japanese diet–Part i. Development of the questionnaire, and reproducibility and validity for food groups. J. Epidemiol. 1999, 9, 216–226. [CrossRef] [PubMed]
24. Bautista, L.E.; Herran, O.F.; Pryer, J.A. Development and simulated validation of a food-frequency questionnaire for the colombian population. Public Health Nutr. 2005, 8, 181–188. [CrossRef] [PubMed]
25. Byers, T.; Marshall, J.; Fiedler, R.; Zielezny, M.; Graham, S. Assessing nutrient intake with an abbreviated dietary interview. Am. J. Epidemiol. 1985, 122, 41–50. [PubMed]
26. Hamdan, M.; Monteagudo, C.; Lorenzo-Tovar, M.L.; Tur, J.A.; Olea-Serrano, F.; Mariscal-Arcas, M. Development and validation of a nutritional questionnaire for the palestine population. Public Health Nutr. 2014, 17, 2512–2518. [CrossRef] [PubMed]
27. Decarli, A.; Ferraroni, M.; Palli, D. A reduced questionnaire to investigate the mediterranean diet in epidemiologic studies. Epidemiology 1994, 5, 251–256. [CrossRef] [PubMed]
28. Overvad, K.; Tjonneland, A.; Haraldsdottir, J.; Ewertz, M.; Jensen, O.M. Development of a semiquantitative food frequency questionnaire to assess food, energy and nutrient intake in denmark. Int. J. Epidemiol. 1991, 20, 900–905. [CrossRef] [PubMed]
© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).