Skip to main content Accessibility help
×
Home
Hostname: page-component-558cb97cc8-6jfzc Total loading time: 1.375 Render date: 2022-10-07T01:06:27.868Z Has data issue: true Feature Flags: { "shouldUseShareProductTool": true, "shouldUseHypothesis": true, "isUnsiloEnabled": true, "useRatesEcommerce": false, "displayNetworkTab": true, "displayNetworkMapGraph": true, "useSa": true } hasContentIssue true

Comparison of dietary share of ultra-processed foods assessed with a FFQ against a 24-h dietary recall in adults: results from KNHANES 2016

Published online by Cambridge University Press:  19 January 2022

Sukyoung Jung
Affiliation:
Department of Epidemiology, Milken Institute School of Public Health, George Washington University, Washington, DC, USA
Sohyun Park*
Affiliation:
Department of Food Science and Nutrition, Hallym University, 24252Chuncheon-si, South Korea The Korean Institute of Nutrition, Hallym University, Chuncheon-si, South Korea
Jee Young Kim
Affiliation:
National Food Safety Information Service, Seoul, South Korea
*
*Corresponding author: Email sopark@hallym.ac.kr
Rights & Permissions[Opens in a new window]

Abstract

Objective:

To evaluate the performance of an FFQ for estimating dietary contributions of NOVA groups to individuals’ diets with a specific focus on ultra-processed foods (UPF) compared with a single 24-h dietary recall (24HR).

Design:

All consumed foods and beverages assess with both a 109-item FFQ and a single 24HR were classified into one of four NOVA groups: natural or minimally processed foods (MPF), processed culinary ingredients (PCI), processed foods (PF) and UPF. The contributions of each NOVA group to daily intakes of energy, protein, fat, saturated fat, carbohydrate, fibre and Na were expressed as crude intake, energy-adjusted intake and percentage intake. Mean differences, correlation coefficients and joint classification were calculated for intakes of energy and nutrients from each NOVA group between the FFQ and the 24HR.

Setting:

The Korea National Health and Nutrition Examination Survey 2016.

Participants:

Adults aged 19–64 years (n 3189).

Results:

The smallest group-mean differences between the two methods were observed in UPF (2–40 %). The greatest average Pearson’s correlation coefficients between the FFQ and 24HR were shown in dietary contributions of UPF (r = 0·22–0·25 for MPF; r = 0·02–0·05 for PCI; r = 0·11–0·18 for PF; r = 0·26–0·30 for UPF). The greatest agreement in quartile classification between the FFQ and the 24HR was observed in dietary contributions of UPF (70·0–71·5 % for MPF; 64·2–68·8 % for PCI; 66·9–69·2 % for PF; 71·8–73·9 % for UPF).

Conclusions:

The use of the FFQ for estimating absolute intake of UPF may not be encouraged in its current form, but it still may be used for relative comparisons such as quantile categorisation.

Type
Research Paper
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution, and reproduction in any medium, provided the original work is properly cited.
Copyright
© The Author(s), 2022. Published by Cambridge University Press on behalf of The Nutrition Society

Ultra-processed foods (UPF), which are foods and beverages with a high energy density and low nutritional quality, have become more dominant in the global diet over the past two decades(Reference Monteiro, Moubarac and Cannon1Reference Shim, Shim and Cha5). The most prominent method of food classification based on food processing levels is the NOVA classification, which has four classes of foods and beverages: natural or minimally processed foods (MPF), processed culinary ingredients (PCI), processed foods (PF) and UPF(Reference Monteiro, Cannon and Moubarac2). In NOVA classification, UPF are defined as ‘formulations of food substances often modified by chemical processes and then assembled into ready-to-consume hyper-palatable food and drink products using flavours, colours, emulsifiers and a myriad of other cosmetic additives(Reference Monteiro, Cannon and Levy6). Collective evidence exists that shows higher UPF consumption is linked with a 39–102 % increased risk of metabolic syndrome, high waist circumference (WC), low HDL-cholesterol levels and overweight/obesity from cross-sectional studies(Reference Askari, Heshmati and Shahinfar7Reference Pagliai, Dinu and Madarena9) and with a 1·20- to 1·34-fold greater risk of depression, cardiovascular diseases, cerebrovascular diseases and mortality from prospective cohort studies(Reference Lane, Davis and Beattie8,Reference Pagliai, Dinu and Madarena9) . As evidence is emerging, efforts to shift towards limiting UPF consumption have become a relevant and timely topic.

Observational epidemiologic studies, especially prospective cohort studies, have merit in unravelling the association between dietary intake and health outcomes(Reference Willett10). Prior prospective studies that have examined the association of UPF consumption with disease outcomes among adults have used 24-h dietary recalls (24HR), diet/food records (DR) or FFQ as the main dietary assessment methods. In general, multiple 24HR or DR over a long period have been appreciated as the more appropriate way to estimate the usual intake of participants and to examine the association between diet and health outcomes than other methods, including FFQ; however, they are expensive and time-consuming(Reference Willett10,Reference Shim, Oh and Kim11) . Although FFQ have been administered as the most practical method in large-scale epidemiological settings since the 1990’s(Reference Willett10Reference Schatzkin, Kipnis and Carroll12), inevitable measurement errors would have attenuated measures of association even if measurement errors occurred randomly in most circumstances(Reference Willett10,Reference Dosemeci, Wacholder and Lubin13,Reference Freedman, Schatzkin and Midthune14) . Specific to UPF consumption, several issues related to the use of the FFQ have been raised. First, FFQ that have been used in previous studies may not cover the full spectrum of UPF consumption due to the limited number of predefined food lists and the lack of information on ingredients, cooking methods, eating place and the brand names of the packaged foods(Reference Khandpur, Rossato and Drouin-Chartier15). Second, although the FFQ were validated at the time of their inception as appropriate, they were not developed and/or validated in the explicit scope of UPF classification and/or their association with health outcomes. These concerns of FFQ use in UPF classification have become more evident in Korea because traditional Korean diets mainly consist of many dishes with combinations of individual ingredients and seasoning, which cannot be accurately classified using the food-based FFQ(Reference Shim, Oh and Kim11).

In Korea, UPF consumption gradually increased between 2010 and 2018(Reference Shim, Shim and Cha5) in parallel with the global trend; however, UPF and disease outcome associations have rarely been investigated. To date, no prospective study is available, and only one cross-sectional study has been conducted under the Korean scope, indicating the positive association between UPF and obesity in Korean women(Reference Sung, Park and Oh16). Since this may be due to the difficulty in the classification of foods assessed with the FFQ in Korean prospective cohort studies, the evaluation of the FFQ’s performance in UPF classification would provide new insights for future use of the FFQ.

With a nationally representative sample of Korean adults, we aimed to evaluate the performance of the FFQ to classify UPF consumption of individuals’ diets compared with the 24HR and examine the possibility of using FFQ in large-cohort studies that only measure dietary intake with FFQ.

Methods

Study population

The Korea National Health and Nutrition Examination Survey (KNHANES), an ongoing, cross-sectional and nationally representative survey, was established to monitor the health and nutritional status of the non-institutionalised civilian population in Korea(Reference Kweon, Kim and Jang17). The KNHANES has been conducted by the Korea Disease Control and Prevention Agency since 1998 and used a complex and multistage probability sampling design for representativeness of the non-institutionalised Korean population(Reference Kweon, Kim and Jang17). Since 2016, the KNHANES sample was selected from 192 primary sampling units (PSU), which are approximately 200 000 small geographical areas covering the whole country(18). Each PSU consisted of sixty households and twenty-three target households were selected using systematic sampling(18). All individuals aged ≥ 1 year within the selected households were considered eligible for further participation in the KNHNAES(18). In 2016, a total of 10 806 persons were screened and asked to participate in the KNHANES. Among them, 8150 persons completed at least one or more surveys among the three-component surveys (health interview, health examinations, and nutrition surveys), and thus response rates were 75·4 %(18). The present study was based on participants aged 19 years and older who completed the nutrition survey in 2016. More details of the KNHANES are available elsewhere(Reference Kweon, Kim and Jang17,18) .

In the current analysis, we considered the FFQ as the surrogate measure and the 24HR as the reference method. Among 4750 participants aged 19–64 years, individuals were excluded if they had the following conditions: missing weights for health interviews and nutrition surveys (n 947), incomplete information on the 24HR or the FFQ (n 545) and extreme FFQ-assessed energy intakes (< 800 kcal or > 4000 kcal for men, < 500 kcal or > 3500 kcal for women) (n 69). The final analysis sample included 3189 participants (1215 men and 1974 women) (see online supplementary material, Supplemental Fig. 1).

Dietary assessment

For the nutrition survey in the KNHANES, trained dieticians comprehensively collected dietary data, including two dietary assessment methods (1-day 24HR and FFQ). Eight dieticians were grouped into teams of four and conducted in-person interviews in participants’ homes using the computer-assisted personal interviewing system. Dietary information was collected 1 week after the health interviews and examinations.

FFQ

For the participants aged 19–64 years, the dish-based semiquantitative FFQ, newly developed in 2009(Reference Yun, Shim and Kweon19), was administered to collect information on the intakes of 112 food/dish items. The validity and reproducibility of the FFQ have been examined in detail elsewhere (reproducibility: correlation coefficients between the first and second FFQ, 0·54–0·61 (mean = 0·54) for nutrients, 0·33–0·87 (mean = 0·57) for food groups; validity: correlation coefficients between the first FFQ and the 12-day dietary records 0·29–0·45 (mean = 0·40) for nutrients)(Reference Kim, Song and Lee20). Participants reported how frequently they consumed the average portion sizes of 112 food/dish items over the preceding year. For the frequency of consumption, nine categories were generally available (never or less than once per month, once per month, 2–3 times per month, once per week, 2–4 times per week, 5–6 times per week, once per day, 2 times per day and 3 times per day), and as exceptions, milk items additionally asked types of milk (whole, low fat or similar); fruit items first asked if consumed seasonally or not and then frequencies accordingly; coffee items required further response if the respondent drank coffee more than 3 times per day. For the amount of consumption, three serving sizes (0·5, 1·0, or 1·5–2 times of standard) were available, and as exceptions, cooked rice, cooked mixed rice and kimbab (Korean rolls with dried seaweed) provided four serving size options because Koreans consumed those items as main dishes on a daily basis (0·5, 1·0, 1·5 or 2 times of standard); open-ended questions collected the amount of alcoholic beverages (soju, beer, makguli) if they consumed more than the greatest value of the given choice (more than 180 ml of soju, 200 ml of beer or 210 ml of makguli). In the NOVA system, alcoholic beverages can be classified as PF or UPF according to manufacturing procedures if those are considered foods(Reference Monteiro, Cannon and Levy21). Since alcohol is ‘a toxic and psychoactive substance with dependence producing properties’ and any alcohol use is related to some health risks(22), we excluded dietary intakes from alcoholic beverages. Thus, a 109-item was used in the analyses.

24-hour dietary recall

For participants aged 1 year and older, a single 24HR instrument was administered to collect information on description, quantity, and time and place of eating for all foods and beverages that the participants had consumed during the past 24 h (from midnight to midnight) for each of the main meals and any other eating occasion. During all four seasons, the 24HR was administered on either weekdays or weekends. The respondents were provided the quantity of consumed foods and beverages in units of volume with the assistance of tools such as standard measuring tools and/or two-dimensional measuring guides. The person in charge of cooking was further provided unique recipes for the home-cooked dishes. The 24HR in the KNHANES used the multiple-pass approach to minimise respondent burden and to enhance complete and accurate food recall. A proxy person (parents or legal guardian) responded on behalf of children or participants who had difficulty reporting their diet. For this study, we only used the data from participants aged 19–64 years.

Energy and nutrients intake estimation

For the 24HR, daily intakes of energy and nutrients were estimated by multiplying the intakes of all foods and dishes that participants reported with their nutrient contents using the nutrient database. For FFQ, all frequencies were standardised into ‘times per day’ by using the conversion factors (1 month = 4·3 weeks = 30·4 d). Daily intakes of energy and nutrients were estimated as follows: the standardised frequency per day × the amount of food per standard unit × their nutrient contents for each 112-food item. To calculate total intakes of energy and nutrients for each NOVA group, the amount consumed of each individual food item in the same NOVA group was summed. For example, the estimated energy intake of each 32-food item classified into UPF was then summed as total energy intake from UPF. The nutrient database for both dietary assessment methods was the Ninth Edition of the Korean Food Composition Table of the Rural Development Administration(23). Our main interests were daily intakes of energy (kcal), protein (g), fat (g), saturated fat (g), carbohydrate (g), fibre (g) and Na (mg).

Categorisation of foods according to the NOVA classification

We used the NOVA classification system that enables us to evaluate diet according to the extent and purpose of food processing(Reference Monteiro, Cannon and Levy6). NOVA has four food categories according to the degree and purpose of food processing: unprocessed or MPF (group 1), PCI (group 2), PF (group 3) and UPF (group 4)(Reference Monteiro, Cannon and Levy6). In brief, the first group indicates natural or MPF without adding substances such as sugar, salt, oils or fats and with the rare use of additives for preservation purposes. Allowable processing limits the purpose of extending the life of unprocessed foods (chilling, freezing, drying and pasteurising) and of facilitating or diversifying food preparation (removing inedible parts, crushing, grinding, roasting and fermentation of milk to make yogurt). Fruits and vegetables, grains, nuts and seeds, tea and coffee, and eggs are included in NOVA group 1. The second group indicates the PCI that are directly obtained from the first NOVA group or from nature by processes of pressing, refining, grinding, milling and spray drying. They are mostly consumed in the presence of group 1 foods as a seasoning, and the use of additives is allowed when the purpose is to preserve the product’s original properties. Sugar, honey, salt, vinegar, vegetable oils and fats (butter, lard) are major examples of group 2 foods. The third group is PF in which substances such as sugar, salt, oils or other group 2 foods are added to group 1 foods and with rare use of additives for preservation and to resist microbial contamination. Processing for the purpose of longer durability or enhancing palatability is allowed. Canned vegetables, fruits or fish; bottled vegetables, fruits and legumes; salted or sugared nuts and seeds; smoked, salted or cured meats; fruits in syrup; cheeses and unpackaged freshly made bread are recognised as group 3 foods. The fourth group is UPF and beverages, which are defined as ‘formulations of food substances often modified by chemical processes and then assembled into ready-to-consume hyper-palatable food and drink products using flavours, colours, emulsifiers and a myriad of other cosmetic additives(Reference Monteiro, Cannon and Levy6). This group includes bread (mass-produced packaged); biscuits; cakes; cookies; chocolate; candies; ice cream, carbonated or fruit drinks; pre-prepared dishes; sausages; hog dogs; burgers; nuggets and instant soups, noodles or dumplings. The rationale and details of NOVA classification have been described elsewhere(Reference Monteiro, Cannon and Moubarac2). In the KNHANES 2016, a total of 653 food items were reported from the 24HR, and 109 food items (except for alcoholic beverages) were available from the FFQ. The final NOVA classification of food items in the FFQ is presented in Supplemental Table 1.

An investigator (S.J) classified all food and beverage items of the FFQ in the KNHANES 2016 into one of the four food groups in NOVA, and the whole classification was thoroughly cross-checked by another investigator (JYK). If there was disagreement in classifying such items, we then checked (1) product name and manufacturer information to identify raw materials and (2) contents of Na or sugar in such items based on the food code in the nutrient database. This is because that more PF are likely to contain higher levels of Na and/or sugars. For dishes with multiple ingredients, we used the ingredient database with the use of a standard recipe of Korean dishes (provided by the Korea Centres for Disease Control and Prevention) to check the amount of each ingredient in such dish. The database provides food name, food code and amount of ingredients (g) per dish. We classified mixed dishes based on the main ingredient, which can be assessed by the proportion of the amount of the main ingredient out of the total amount of dishes. All food items from both dietary assessment methods were mutually exclusively classified into one of the four NOVA groups. NOVA group 4, UPF, was the main interest in the present study.

Assessment of socio-demographic characteristics

To present the socio-demographic characteristics of the study participants, we included age, sex, residential area, educational level, income, occupation and marital status. Trained interviewers administered the health interview questionnaires to the participants aged 19 years and older. Information on demographics (age, sex, education, income and occupation) and the residential area of all members of the sampled household was collected from the household component in the health interview. The individual component questionnaire was used to collect marital status. Such variables were manipulated for the analyses as follows: age (continuous: years and categorical: 19–29, 30–39, 40–49, 50–59, and ≥ 60 years in the unit of years), sex (categorical: men and women), residential area (categorical: urban and rural), education level (categorical: elementary school graduate or less, middle school graduate, high school graduate, and college graduate or higher), monthly household income (categorical: ≤ 100, 100–200, 200–300, and > 300 in the unit of million Korean won), occupation (categorical: non-manual, manual and no job (including students and housewives)) and marital status (categorical: married or not).

Statistical analysis

As the measure of the consumption from each NOVA group, we presented three different types of dietary consumption of each NOVA group, expressed as: (1) crude quantity, (2) energy-adjusted quantity and (3) relative contribution to total intakes of energy and nutrients for each NOVA group (as a percentage). For the energy-adjusted quantity, six selected nutrients were adjusted for total energy intake using a residual method to adjust for measurement error due to underreporting and to make an isocaloric comparison(Reference Willett10). Energy-adjusted nutrient intakes were calculated as the sum of the residuals obtained from the regression model with total energy intake (kcal) as the independent variable and absolute intake of nutrient as the dependent variable and the expected nutrients intake for a person with mean energy intake(Reference Willett10).

All weights accounting for the complex sampling design of the KNHANES were applied to all further analyses. Nutrients were normalised by logarithmic transformation due to the skewed data distribution in most nutrients. The general characteristics of the study participants were described as the weighted means and their standard errors for continuous variables and as weighted prevalence and their se for categorical variables. We tested the mean difference in dietary contributions of each NOVA group to total intakes of energy and nutrients between the 24HR and the FFQ using a paired t test. The performance of the FFQ to apply the NOVA food classification was assessed by Pearson’s correlation coefficients between the 24HR and the FFQ. We also conducted a cross-classification analysis to compare the agreement of classification into quartiles between the two tools. For this cross-classification analysis, we classified dietary shares of NOVA groups to total intakes of energy and nutrients estimated by the 24HR and the FFQ into quartiles and determined the percentage of participants classified into the same, adjacent, or opposite quartiles. The opposite quartiles indicated gross misclassification, which ranked in the lowest or highest quartiles of the FFQ but in the highest or lowest quartiles of the 24HR as appropriate. As a sensitivity analysis, we repeated the main analyses by including those reported extreme energy intake (n 3258) (see online supplementary material, Supplemental Tables 25).

All analyses were performed using SAS software (Version 9.4, SAS Institute Inc.) while accounting for complex survey design effects. All tests were 2-sided, and significance was set at an α level of 0·05.

Results

Table 1 presents the general characteristics of the study participants in the KNHANES 2016. This study included 3189 Korean adults aged 19–64 years (mean 41·2 years, se 0·4). In our sample, 50·8 % were women (1215 men and 1974 women), 42·3 % were older than 45 years of age, 87·8 % were urban residents, 87·2 % were at least a high school graduate, 67·4 % had higher than 300 million won of monthly household income, 34·9 % had manual jobs and 72·8 % were married.

Table 1 General characteristics of the study participants in KNHANES 2016 (n 3189)

KNHANES, Korea National Health and Nutrition Examination Survey.

Note: Values are expressed as weighted means (se) for continuous variables and as weighted percentages (se) for categorical variables.

* No job included students and housewives.

Table 2 shows the mean daily intakes of energy and selected nutrients estimated by the 24HR and the FFQ. For crude values, total intakes of energy, protein, fat, saturated fat, carbohydrate, fibre and Na in the 24HR were greater than those in the FFQ. The intakes of the aforementioned energy and nutrients from the MPF, PCI and UPF in the 24HR were higher than those in the FFQ. The intakes of energy and nutrients from PF in the FFQ were lower than those in the 24HR. We observed similar patterns in both energy-adjusted nutrients and percentage intake from each food. Irrespective of the method of dietary assessment, MPF contributed the most to daily intakes of energy and carbohydrates (24HR: energy 55·6 %, carbohydrate 60·0 % v. FFQ: energy 50·1 %, carbohydrate 60·8 %). The percent intakes of protein, fat, saturated fat, fibre and Na differed by dietary assessment methods. MPF contributed the most to daily intakes of protein, fat, saturated fat and fibre, and PF contributed the most to daily intakes of Na in the 24HR, whereas PF contributed the most to daily intakes of protein, fat, saturated fat, fibre and Na in the FFQ.

Table 2 Mean daily intakes of energy and nutrients from NOVA groups estimated from a 24HR and FFQ

24HR, 24-hour dietary recall; MPF, unprocessed and minimally processed foods; PCI, processed culinary ingredients; PF, processed foods; UPF, ultra-processed food and drink products.

Note: Values are expressed as weighted means (se).

* P > 0·05 for the differences between the 24HR and the FFQ using a paired t test.

The relative validity of the FFQ in terms of Pearson’s correlation coefficients is given in Table 3. For crude intake, the Pearson’s correlation coefficients between the 24HR and the FFQ were 0·33–0·42 for total intakes, 0·13–0·36 for intakes of those from MPF, –0·03–0·04 for intakes of those from PCI, 0·15–0·30 for intakes of those from PF and 0·18–0·35 for intakes of those from UPF. For energy-adjusted intake, the Pearson’s correlation coefficients between the 24HR and the FFQ were 0·29–0·47 for total intakes, 0·08–0·39 for intakes of those from MPF, –0·02–0·07 for intakes of those from PCI, 0·11–0·29 for intakes of those from PF and 0·18–0·34 for intakes of those from UPF. For percentage intake, the Pearson’s correlation coefficients between the 24HR and the FFQ were 0·09–0·36 for the MPF contribution, –0·02–0·08 for the PCI contribution, 0·04–0·25 for the PF contribution and 0·25–0·37 for the UPF contribution. Compared with the correlation coefficients between the two methods for total intakes of energy and nutrients, intakes from PCI and PF mostly showed lower correlations (≤ 0·20), and those from MPF and UPF showed relatively moderate correlations in all three values (range of average: 0·22–0·25 for MPF; 0·26–0·30 for UPF).

Table 3 Comparison of Pearson’s correlation coefficients of energy and nutrient intake from NOVA groups estimated from a 24-h dietary recall and FFQ

Abbreviations:MPF, unprocessed and minimally processed foods; PCI, processed culinary ingredients; PF, processed foods; UPF, ultra-processed food and drink products.

* Energy-adjusted intake was estimated using the residual method.

Table 4 presents the results of the cross-classification analysis. The average proportions of participants were classified into the same or adjacent quartiles as follows: 68·8 % (PCI) to 73·9 % (UPF) for crude intake; 64·2 % (PCI) to 71·8 % (UPF) for energy-adjusted intake and 66·9 % (PF) to 73·1 % (UPF) for percentage intake. The average proportions grossly misclassified ranged from 6·5 % to 12·1 %. Compared with the cross-classification agreement between the two methods for total intakes of energy and nutrients, UPF classifications showed the greatest agreement and the lowest misclassification. A sensitivity analysis using data including those reported extreme energy intake showed similar results to the main findings (n 3258) (see online supplementary material, Supplemental Tables 25).

Table 4 Proportion of agreement in quartile classification of energy and nutrient intake from NOVA groups estimated from a 24-h dietary recall and FFQ

Abbreviations: MPF, unprocessed and minimally processed foods; PCI, processed culinary ingredients; PF, processed foods; UPF, ultra-processed food and drink products.

* Nutrients estimated from FFQ were not categorised into quartiles.

Nutrients estimated from both 24-h recall and FFQ were categorised into two groups.

Discussion

In this study, using a nationally representative sample of Korean adults, we evaluated the performance of the FFQ compared with the 24HR in categorising foods and beverages according to the NOVA system with a specific focus on UPF. Group-mean differences between the FFQ and the 24HR in intakes of selected nutrients except for energy and carbohydrates from UPF ranged from 2 % to 40 %, which were the smallest compared with those from MPF, PCI or PF. Similarly, average Pearson’s correlation coefficients between the FFQ and the 24HR in dietary shares of UPF were approximately 0·30 (r = 0·28 in crude intake; r = 0·26 in energy-adjusted intake; r = 0·30 in percent intake), which were the greatest compared with those of MPF, PCI or PF (r = 0·22–0·25 for MPF; r = 0·02–0·05 for PCI; r = 0·11–0·18 for PF). We found acceptable agreement between the FFQ and the 24HR in quartiles of UPF contributions to diet, which were 71·8–73·9 %.

Based on our results, the use of the FFQ in the KNHANES may not be appropriate in PCI and PF classifications. Compared with 24HR, FFQ underestimated crude, energy-adjusted and percent intakes of energy and most selected nutrients from PCI and overestimated those of energy and all selected nutrients from PF. These may be attributed to the following speculations. First, the FFQ consisted of a limited number of predefined item lists with aggregated similar foods or dishes into one item, whereas the 24HR was administered in a completely open-ended format. Thus, possible gaps may exist in the coverage of all foods or dishes consumed between the two dietary assessment methods. In this study, only two (added sugar and butter/margarine) out of 109 items in the FFQ and 37 out of 653 items in the 24HR were available for PCI; 54 items in the FFQ and 128 items in the 24HR were available for PF. Second, the FFQ usually does not collect detailed information, such as food preparation methods, ingredients used in mixed dishes or places of eating; thus, cooked foods or dishes are difficult to correctly classify into each NOVA group. Since the 54 items classified as PF were mostly cooked dishes with various raw ingredients using diverse cooking methods (boiling, steaming, marinating, stir frying, grilling, frying), such as soups, stews, stir-fried or marinated vegetables, PF classification may not be accurate. Third, such substantial differences in means or correlations between the 24HR and the FFQ may be due to differences in the purpose and usage between the two methods. The 24HR is used to estimate the absolute mean value for population or true distribution of intakes for the population if 2 or more days of recalls per individual are available (e.g. comparison of nutrients intakes with dietary recommendations)(Reference Willett10). The FFQ is primarily used to estimate average long-term usual intake for individuals and is thus appropriate for most epidemiologic studies of diet–disease associations using correlations or relative risks based on relative rankings of dietary intakes(Reference Willett10). Thus, it is necessary to consider the primary purpose of the dietary measure.

Importantly, we observed an average of 74 % good agreement in quartile classification of UPF contribution to diets. This result indicated that UPF classified with the FFQ may be valid for the purpose of qualitative comparisons, such as ranking into quantiles and dietary patterns. Although the FFQ is not appropriate for the estimation of crude intake per se from UPF, its future use with categorical analysis can still be encouraged. As observed in many diet–disease association studies, previous studies of UPF consumption–disease association studies(Reference Askari, Heshmati and Shahinfar7Reference Pagliai, Dinu and Madarena9) have used dietary exposures as categorical, not continuous, for the following advantages. In nutritional epidemiology, the use of categorised intake can make the direct visualisation of actual counts of cases and non-cases according to intake level and compute relative risks. Additionally, it can minimise the possibility of a false linear trend because no assumptions are made about a dose–response relationship. Last, it can minimise the undue impacts of outliers(Reference Willett10). Taken together, the FFQ in this study may be appropriate to use to classify UPF in a qualitative comparison rather than in an absolute intake estimation.

To our knowledge, this is the first study on the performance of the FFQ in UPF classification compared with that of the 24HR in adults. We used the 24HR as the referent because no true gold standard dietary assessment tool is available to date. Two prior studies with a similar approach as our study are available. In a study with an Italian adult sample, the NOVA FFQ with ninety-four items aiming to estimate both the absolute (g/d) and relative (% of g) intake of the NOVA groups was developed and validated with a 7-day weighted DR(Reference Dinu, Bonaccio and Martini24). Given the correlation coefficients of 0·6 to 0·7 and Bland–Altman plots between the NOVA FFQ and the weighted DR, the NOVA FFQ has shown moderate to good validity. Another study with young children in New Zealand assessed the absolute (kcal/d) and relative (% of kcal) energy intake of the NOVA groups. Although a relatively low correlation coefficient (r = 0·3) was observed, agreements of correctly or adjacently classified into quartiles of energy intake from UPF between the two methods were acceptably high (73–74 %)(Reference Fangupo, Haszard and Leong25). Similarly, our data showed lower validity in the comparison of absolute intake and acceptably higher validity in the comparison of ranking of intake.

Several limitations in this study warrant attention. First, there is a possibility of misclassification of each NOVA group including UPF because the FFQ in this study was not explicitly designed to apply the NOVA system at the time of its inception. As such, UPF consumption assessed with the FFQ may be under- or overestimated, even though the FFQ was validated using DR(Reference Kim, Song and Lee20) and covered major foods and dishes frequently consumed by Koreans(Reference Yun, Shim and Kweon19). Second, our findings may not be applicable to the food-based FFQ because we assessed the whole process using a dish-based FFQ. However, food-based FFQ might be reasonable due to the limited use of MPF or UPF classification because most items classified as MPF or UPF in our FFQ were foods consumed independently (not as combinations or dishes) (see online supplementary material, Supplemental Table 1). Third, the applicability of our finding is limited to Korean adults aged 19–64 years. In addition, because those who participated in all study examinations but were excluded from the final analytic set were older, less likely to engage in higher education, higher income and non-manual job than study participants, interpretation should be made with caution. Studies targeting other groups with different demographic and/or socio-economic characteristics should be validated as appropriate. Fourth, although we used the single 24HR as the referent in this study, neither of the single 24HR and the FFQ are gold-standard measures of dietary intake. Multiple diet records or 24HR have been frequently used to test the validity of the FFQ, however, the collection of one or more diet records or 24HR is practically challenging in large-scale epidemiological studies(Reference Willett10). Although the single 24HR may poorly represent usual individual intake due to day-to-day variations in the foods or nutrients consumed, it has been used as a usual method for estimating the mean intake of a population(Reference Willett10). Future studies using more superior dietary assessment methods (e.g. multiple diet records or multiple 24HR) as the referent are warranted to examine the validity of the FFQ applying the NOVA classification system.

Despite these limitations, this study has considerable strengths. Unlike the Western diet, the Korean diet is characterised as dish based with the combination of multiple individual ingredients(Reference Shim, Oh and Kim11). Given this, our finding is informative to general Korean adults because the performance of the FFQ in UPF classification has been tested under the nationally representative sample of Korean adults who completed both the 24HR and the FFQ. Furthermore, previous studies mostly focussed solely on energy intake (kcal) or food quantity (g) from UPF; however, we expanded the range of dietary share from energy to several major nutrients, including fat and Na.

In conclusion, the FFQ in the current form may not be encouraged for estimating absolute intake of UPF. Rather, its use in quantiles may be possible for investigating the association between UPF consumption and various disease outcomes based on an epidemiological study design. Further validation studies in other study populations and new development of an FFQ specifically designed to apply the NOVA classification system in the Asian population, especially in Korea, are needed to better capture UPF consumption and for future investigations on UPF and disease associations.

Acknowledgements

Acknowledgements: The authors would like to thank all the participants and colleagues for their contribution to this study. Financial support: This work was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (No. 2021R1A6A1A03044501) and Hallym University (HRF-202003–008). The funders had no role in the design, analysis or writing of this article. Conflict of interest: There are no conflicts of interest. Authorship: S. J.: Data curation, formal analysis, writing – original draft preparation, writing – reviewing and editing; S. P.: conceptualisation, methodology, writing – reviewing and editing, supervision; J. Y. K.: data curation and methodology. All gave final approval and agreed to be accountable for all aspects of ensuring integrity and accuracy. Ethics of human subject participation: Based on the Bioethics and Safety Act, the 2016 KNHANES was exempted from a review of research ethics between 2014 and 2017. All participants gave written informed consent before participation in the survey since 1998, and all survey data were anonymised before analysis.

Supplementary material

For supplementary material/s referred to in this article, please visit https://doi.org/10.1017/S1368980022000179

References

Monteiro, CA, Moubarac, JC, Cannon, G et al. (2013) Ultra-processed products are becoming dominant in the global food system. Obes Rev 14, 2128.CrossRefGoogle ScholarPubMed
Monteiro, CA, Cannon, G, Moubarac, JC et al. (2018) The UN Decade of Nutrition, the NOVA food classification and the trouble with ultra-processing. Public Health Nutr 21, 517.CrossRefGoogle ScholarPubMed
Juul, F & Hemmingsson, E (2015) Trends in consumption of ultra-processed foods and obesity in Sweden between 1960 and 2010. Public Health Nutr 18, 30963107.CrossRefGoogle ScholarPubMed
Baraldi, LG, Martinez Steele, E, Canella, DS et al. (2018) Consumption of ultra-processed foods and associated sociodemographic factors in the USA between 2007 and 2012: evidence from a nationally representative cross-sectional study. BMJ Open 8, e020574.CrossRefGoogle ScholarPubMed
Shim, JS, Shim, SY, Cha, HJ et al. (2021) Socioeconomic characteristics and trends in the consumption of ultra-processed foods in Korea from 2010 to 2018. Nutrients 13, 1120.CrossRefGoogle ScholarPubMed
Monteiro, CA, Cannon, G, Levy, RB et al. (2019) Ultra-processed foods: what they are and how to identify them. Public Health Nutr 22, 936941.CrossRefGoogle Scholar
Askari, M, Heshmati, J, Shahinfar, H et al. (2020) Ultra-processed food and the risk of overweight and obesity: a systematic review and meta-analysis of observational studies. Int J Obes 44, 20802091.CrossRefGoogle ScholarPubMed
Lane, MM, Davis, JA, Beattie, S et al. (2021) Ultraprocessed food and chronic noncommunicable diseases: a systematic review and meta-analysis of 43 observational studies. Obes Rev 22, e13146.CrossRefGoogle ScholarPubMed
Pagliai, G, Dinu, M, Madarena, MP et al. (2021) Consumption of ultra-processed foods and health status: a systematic review and meta-analysis. Br J Nutr 125, 308318.CrossRefGoogle ScholarPubMed
Willett, W (2013) Nutritional Epidemiology, 3rd ed. NY: Oxford University Press.Google Scholar
Shim, JS, Oh, K & Kim, HC (2014) Dietary assessment methods in epidemiologic studies. Epidemiol Health 36, e2014009.CrossRefGoogle ScholarPubMed
Schatzkin, A, Kipnis, V, Carroll, RJ et al. (2003) A comparison of a food frequency questionnaire with a 24-hour recall for use in an epidemiological cohort study: results from the biomarker-based Observing Protein and Energy Nutrition (OPEN) study. Int J Epidemiol 32, 10541062.CrossRefGoogle Scholar
Dosemeci, M, Wacholder, S & Lubin, JH (1990) Does nondifferential misclassification of exposure always bias a true effect toward the null value? Am J Epidemiol 132, 746748.CrossRefGoogle Scholar
Freedman, LS, Schatzkin, A, Midthune, D et al. (2011) Dealing with dietary measurement error in nutritional cohort studies. J Natl Cancer Inst 103, 10861092.CrossRefGoogle ScholarPubMed
Khandpur, N, Rossato, S, Drouin-Chartier, JP et al. (2021) Categorising ultra-processed foods in large-scale cohort studies: evidence from the Nurses’ Health Studies, the Health Professionals Follow-up Study, and the Growing Up Today Study. J Nutr Sci 10, e77.CrossRefGoogle ScholarPubMed
Sung, H, Park, JM, Oh, SU et al. (2021) Consumption of Ultra-processed foods increases the likelihood of having obesity in Korean Women. Nutrients 13, 698.CrossRefGoogle ScholarPubMed
Kweon, S, Kim, Y, Jang, MJ et al. (2014) Data resource profile: the Korea National Health and Nutrition Examination Survey (KNHANES). Int J Epidemiol 43, 6977.CrossRefGoogle Scholar
Ministry of Health and Welfare, Korea Centers for Disease Control and Prevention & Korea Health Statistics (2016) (2017) Korea National Health and Nutrition Examination Survey (KNHANES VII-1). Sejong: Korea Centers for Disease Control and Prevention.Google Scholar
Yun, SH, Shim, J-S, Kweon, S et al. (2013) Development of a food frequency questionnaire for the Korea National Health and Nutrition Examination Survey: data from the fourth Korea National Health and Nutrition Examination Survey (KNHANES IV). J Nutr Health 46, 186196.Google Scholar
Kim, DW, Song, S, Lee, JE et al. (2015) Reproducibility and validity of an FFQ developed for the Korea National Health and Nutrition Examination Survey (KNHANES). Public Health Nutr 18, 13691377.CrossRefGoogle Scholar
Monteiro, CA, Cannon, G, Levy, R et al. (2016) NOVA. The star shines bright. World Nutr 7, 2838.Google Scholar
World Health Organization (2021) Health Topics. https://www.who.int/health-topics/alcohol#tab=tab_2 (accessed December 2021).Google Scholar
Rural Development Administration, National Institute of Agriculatural Sciences (KR) (2016) Standard Food Composition Table. 9th rev.ed. Suwon: Rural development Administration, National Institute of Agriculatural Sciences.Google Scholar
Dinu, M, Bonaccio, M, Martini, D et al. (2021) Reproducibility and validity of a food-frequency questionnaire (NFFQ) to assess food consumption based on the NOVA classification in adults. Int J Food Sci Nutr 72, 861869.CrossRefGoogle ScholarPubMed
Fangupo, LJ, Haszard, JJ, Leong, C et al. (2019) Relative validity and reproducibility of a food frequency questionnaire to assess energy intake from minimally processed and ultra-processed foods in young children. Nutrients 11, 1290.CrossRefGoogle ScholarPubMed
Figure 0

Table 1 General characteristics of the study participants in KNHANES 2016 (n 3189)

Figure 1

Table 2 Mean daily intakes of energy and nutrients from NOVA groups estimated from a 24HR and FFQ

Figure 2

Table 3 Comparison of Pearson’s correlation coefficients of energy and nutrient intake from NOVA groups estimated from a 24-h dietary recall and FFQ

Figure 3

Table 4 Proportion of agreement in quartile classification of energy and nutrient intake from NOVA groups estimated from a 24-h dietary recall and FFQ

Supplementary material: File

Jung et al. supplementary material

Jung et al. supplementary material

Download Jung et al. supplementary material(File)
File 85 KB
You have Access Open access
1
Cited by

Save article to Kindle

To save this article to your Kindle, first ensure coreplatform@cambridge.org is added to your Approved Personal Document E-mail List under your Personal Document Settings on the Manage Your Content and Devices page of your Amazon account. Then enter the ‘name’ part of your Kindle email address below. Find out more about saving to your Kindle.

Note you can select to save to either the @free.kindle.com or @kindle.com variations. ‘@free.kindle.com’ emails are free but can only be saved to your device when it is connected to wi-fi. ‘@kindle.com’ emails can be delivered even when you are not connected to wi-fi, but note that service fees apply.

Find out more about the Kindle Personal Document Service.

Comparison of dietary share of ultra-processed foods assessed with a FFQ against a 24-h dietary recall in adults: results from KNHANES 2016
Available formats
×

Save article to Dropbox

To save this article to your Dropbox account, please select one or more formats and confirm that you agree to abide by our usage policies. If this is the first time you used this feature, you will be asked to authorise Cambridge Core to connect with your Dropbox account. Find out more about saving content to Dropbox.

Comparison of dietary share of ultra-processed foods assessed with a FFQ against a 24-h dietary recall in adults: results from KNHANES 2016
Available formats
×

Save article to Google Drive

To save this article to your Google Drive account, please select one or more formats and confirm that you agree to abide by our usage policies. If this is the first time you used this feature, you will be asked to authorise Cambridge Core to connect with your Google Drive account. Find out more about saving content to Google Drive.

Comparison of dietary share of ultra-processed foods assessed with a FFQ against a 24-h dietary recall in adults: results from KNHANES 2016
Available formats
×
×

Reply to: Submit a response

Please enter your response.

Your details

Please enter a valid email address.

Conflicting interests

Do you have any conflicting interests? *