Highly/ultra-processed foods (HPF), defined as multi-ingredient industrially formulated mixtures(Reference Poti, Mendez and Ng1), are increasingly consumed in many countries(Reference Baker, Machado and Santos2). In recent years, many epidemiological studies have focussed on the association between HPF consumption and health outcomes. Meta-analyses have shown positive associations between HPF consumption and overweight and obesity, CVD, cerebrovascular disease, metabolic syndrome, depression and mortality(Reference Pagliai, Dinu and Madarena3,Reference Taneri, Wehrli and Roa Diaz4) . Compared with non-HPF, HPF have an unhealthy nutrient profile with a higher energy density and higher contents of total fats, saturated fats, trans fats, free sugars and Na, as well as lower contents of protein, fibre, vitamins (e.g. vitamins A and C) and minerals (e.g. K and Fe)(Reference Batal, Johnson-Down and Moubarac5–Reference Marron-Ponce, Flores and Cediel7). In fact, an inverse association between HPF consumption and overall diet quality has been observed in many countries(Reference Batal, Johnson-Down and Moubarac5–Reference Slimani, Deharveng and Southgate20).
Meanwhile, little evidence is available regarding HPF consumption in Japan. Although the Japanese dietary pattern is considered to be based on dishes and meals made from a variety of unprocessed or less processed foods(Reference Monteiro, Cannon and Levy21), HPF consumption may not be low, as Japan ranks 10th out of 80 countries in annual per capita retail sales of HPF(22). To our knowledge, there have been only two studies on HPF consumption in Japan(Reference Koiwai, Takemi and Hayashi23,Reference Koiwai, Takemi and Hayashi24) . However, these studies have been conducted in a single prefecture in Japan, and HPF consumption in a diverse geographic population is unknown. Furthermore, despite the increasing consumption of alcoholic beverages and ready-made foods in Japan(25,Reference Murakami, Livingstone and Sasaki26) , these items were excluded from the previous studies, which may have led to the misestimation of HPF consumption. Moreover, obesity and depression, which have been reported to be positively associated with HPF consumption(Reference Pagliai, Dinu and Madarena3,Reference Taneri, Wehrli and Roa Diaz4) , are increasing in Japan(27,28) . Thus, HPF consumption by the Japanese should be carefully assessed in diverse geographic areas, including all the foods and beverages consumed.
The Food and Agricultural Organization recommends, in its guidelines for collecting information on food processing, distinguishing between food and dish items processed in industrial settings and those prepared by hand at home or in artisanal settings (e.g. street foods) and disaggregating recipes into their ingredients when possible(29). However, the distinction between artisanal and industrial foods is ambiguous(Reference Gibney30), and the types of dishes broken down vary among studies(Reference Slimani, Deharveng and Southgate20,Reference Monteiro, Levy and Claro31) . Specifically, there is no consensus on whether to disaggregate dishes prepared away from home (e.g. ready-made dishes bought from supermarkets and restaurant meals). Previous studies have indicated that the discrepancy between classification methods may lead to a different conclusion regarding HPF consumption and its relationship to diet quality(Reference Gibney30,Reference Bleiweiss-Sande, Bailey and Sacheck32) .
Therefore, the present study aimed to assess HPF consumption and its association with diet quality in Japanese adults. Moreover, we compared the results obtained when dishes prepared away from home were disaggregated into ingredients before classification by processing levels with the results when they were classified without disaggregation. Such investigations would be useful to understand the differences in the estimates of HPF consumption by classification methods. Based on existing literature, we hypothesised that the energy contribution of HPF would be inversely associated with overall dietary quality and that differences in classification methods might affect estimates of HPF contribution.
Methods
Data source
In this cross-sectional study, we used dietary data from a nationwide survey conducted at single point in time between February and March 2013 in twenty study areas, consisting of twenty-three prefectures in Japan. The original purpose of the survey was to evaluate the amount and source of Na intake in Japanese adults. The details of the survey have been described elsewhere(Reference Asakura, Uechi and Masayasu33). Briefly, 199 research dietitians working at separate welfare facilities invited their colleagues and family members of the colleagues to participate in the study. Approximately four apparently healthy subjects (two males and two females) were recruited from five 10-year age categories (20–29, 30–39, 40–49, 50–59 and 60–69 years) in each study area. Participation in the survey was limited to one participant per household. No participant was a dietitian or a medical professional, had received dietary therapy from a doctor or dietitian, had a history of educational hospitalisation for diabetes or was a pregnant or lactating woman. In total, 196 males and 196 females aged 20–69 years provided the dietary data. After excluding four participants with missing information on the variables of interest, 388 participants (196 males and 192 females) were included in the present analysis.
Dietary assessment
Dietary data were obtained using a 4-day weighed dietary record (DR). The details of the DR are provided elsewhere(Reference Asakura, Uechi and Masayasu33). Briefly, participants were asked to record all foods and beverages consumed on four non-consecutive days (three working days and one non-working day, excluding night-shift days and days before and after a night shift). Research dietitians explained how to keep the DR on the participants and requested them to weigh foods and beverages with a digital scale (KD-812WH) or measure with the spoon and cup provided.
The recording sheet included the following items: (i) dish names, (ii) whether dishes were prepared at home, away from home or other (foods eaten in a raw state, such as fresh fruits and vegetables), (iii) food names (ingredients included in dishes) and (iv) approximate amounts or measured weights of foods consumed. Participants were also asked to record the names of products and manufacturers for store-bought products and the names of menus and restaurants when they dined out. In addition, participants were asked to collect packaged food packages.
Recording sheets and packages were submitted to a research dietitian at each facility immediately after recording. The research dietitians reviewed the recording sheets as soon as possible and, if necessary, asked the participants to provide additional information to clarify the name or amount of food on the sheet. For packaged foods and dishes prepared away from home, each food ingredient and its consumed weight were estimated as precisely as possible based on the approximate amount of food, the website of the restaurant or manufacturer, ingredient labels, nutrition facts labels on food packages and cooking books. Each food item was assigned a food code from the Standard Tables of Food Composition in Japan (STFCJ)(34) in a uniform procedure.
All foods items recorded in the column ‘food names’ were then classified into three categories by the research dietitian as follows: (1) ‘home-made’: foods cooked at home (e.g. rice cooked at home and bread baked at home); (2) ‘store-bought’: ingredients of dishes prepared away from home and foods processed by the manufacturer (e.g. ready-to-heat curry, processed meat and chocolate) and (3) ‘other’: unprocessed ingredients before cooking at home (e.g. fresh vegetables, meats, fish and milk) or seasonings added during home cooking or used at the table at home (e.g. mayonnaise used to make sandwiches). In addition to this classification, all food codes and weights were reconfirmed by two other research dietitians at the central office of this study.
Classification of foods based on the degree of food processing
We used the framework developed by researchers at the University of North Carolina at Chapel Hill (UNC)(Reference Poti, Mendez and Ng1) to classify foods according to the level of processing. The UNC system, developed based on the most widely used classification system, NOVA(Reference Monteiro, Cannon and Levy21), provides enhanced definitions of food categories(Reference Poti, Mendez and Ng1) (online Supplementary Table 1). Although the UNC system classifies packaged food products with a barcode sold in the USA, it would be useful to classify unpackaged food products, as it provides broad and detailed examples of foods for each category, ranging from fresh vegetables (e.g. whole carrots) to refrigerated ready-made vegetable-based mixed dishes (e.g. coleslaw). A previous study showed that the UNC system had higher inter-rater reliability than the NOVA classification system(Reference Bleiweiss-Sande, Chui and Evans35).
According to the UNC system, the author classified all foods (except for dietary supplements) in the DR into one of the four groups:(1) unprocessed and minimally processed, (2) basic processed, (3) moderately processed and (4) highly processed. We decided to break down home-prepared dishes into component ingredients before classification, but not to break down industrial packaged food products, according to the guidelines of the Food and Agricultural Organization(29) and previous studies(Reference Moubarac, Batal and Louzada9,Reference Lauria, Dello Russo and Formisano10,Reference Liu, Steele and Li14,Reference Louzada, Ricardo and Steele17,Reference Monteiro, Levy and Claro31) . However, there is no consensus on whether dishes prepared away from home (e.g. ready-made dishes from the supermarket and restaurant meals) should be disaggregated. Therefore, we classified dishes prepared away from home at both the food level (classified after disaggregation into ingredients) and the dish level (classified without disaggregation). The food classification procedure is shown below and in Fig. 1.
Fig. 1 Flow chart of the classification of foods in the 4-day dietary record obtained from Japanese adults aged 20–69 years (n 388) in 2013
-
Step 1: Home-prepared dishes (including foods eaten raw) were disaggregated into component ingredients, each of which was classified based on its food code (e.g. miso, water, spinach and egg in miso soup cooked at home, fresh fruits and vegetables).
-
Step 2: For ready-made dishes consisting of a single food ingredient, each food ingredient was classified based on its food code (e.g. black tea, black coffee and candy).
-
Step 3: If dishes prepared away from home contained ‘home-made’ or ‘other’ food ingredients (mostly seasonings, such as soy sauce added to Chinese dumplings at home), that ingredient was individually classified based on its food code.
-
Step 4: Packaged food products (i.e. dishes prepared away from home containing multiple food ingredients with the names of products, brands or manufacturers of packaged foods) were classified without disaggregation. For instance, we classified packaged sandwiches from convenience stores as single items.
-
Step 5: Other dishes prepared away from home (e.g. dishes purchased from supermarkets, restaurant meals and other miscellaneous dish items without information identifying them as packaged food) were classified at two levels: 1) the food level: all items were classified after disaggregation into food ingredients or 2) the dish level: all items were classified without disaggregation.
In Steps 1–3 and the food-level classification in Step 5, each food item was classified based on the food code of the STFCJ(34), considering whether each food item was categorised as ‘home-made’ or ‘store-bought’ by the research dietitian. For instance, among tea with food code 16 042 ‘oolong tea, infusion’, those considered ‘home-made’ were classified into the unprocessed and minimally processed category. By contrast, those considered ‘store-bought’ were classified into the basic processed category. For the dish-level classification in Step 5, we assumed that all dishes were ready-to-eat or ready-to-heat foods rather than frozen or shelf-stable foods.
Calculation of diet quality scores
The mean daily intake of energy and nutrients over 4 days was calculated for each participant based on the weight and nutrient content of each food item using the STFCJ(34). For foods with added sugar content not available in the STFCJ, added sugar values were calculated based on the same or similar food items in the 2011–2012 Food Patterns Equivalents Database(Reference Bowman, Clemens and Thoerig36). We converted the teaspoon equivalents in the Food Patterns Equivalents Database into grams by multiplying by 4·2 (grams of added sugar per teaspoon).
We assessed the diet quality of each participant using the Healthy Eating Index-2015 (HEI-2015)(Reference Krebs-Smith, Pannucci and Subar37) and Nutrient-Rich Food Index 9·3 (NRF9·3)(Reference Fulgoni, Keast and Drewnowski38). The usefulness of these indices has been verified in the Japanese population(Reference Murakami, Livingstone and Fujiwara39). The HEI-2015 evaluates a set of foods on a 100-point scale for compliance with the 2015–2020 Dietary Guidelines for Americans(40), with a higher score indicating better diet quality. The NRF9·3 is a composite measure of nutrient density calculated as the sum of the percentage of the reference daily value of nine qualifying nutrients minus the sum of the percentage of the reference daily values of three disqualifying nutrients. The maximum possible score was 900, with the higher NRF9·3 total score indicating a better diet quality. Details on the calculation of both indices are provided in Supplementary Text 1.
Assessment of basic characteristics
Body weight (in 0·1 kg) and height (in 0·1 cm) were measured by research dietitians or medical workers using standardised procedures while the participant was in light clothes without shoes. BMI was calculated as body weight (kg) divided by the square of height (m2). In addition, information on sex, age, education level (junior high school or high school, vocational school or junior college, or university or graduate school), self-reported hours spent per day or week on six activities (walking, cycling, standing, running, exercise causing sweating and sleeping) and smoking status (never, past or current) was collected using a questionnaire. Physical activity (total metabolic equivalents, h/d) was calculated by summing the product of the self-reported hours spent per day on each activity during the preceding month and the corresponding metabolic equivalent value(Reference Ainsworth, Haskell and Herrmann41,Reference Ainsworth, Haskell and Herrmann42) .
Data analysis
All analyses were conducted using the statistical software package SAS version 9.4 (SAS Institute Inc.). Two-sided P-values < 0·05 were considered statistically significant. First, all foods were categorised into food groups based on the similarity of nutrient composition or culinary use, mainly according to the STFCJ(34). The contribution (%) of each food group to the total energy intake of HPFs over 4 days was calculated for the whole population.
Next, the mean daily dietary intake over 4 days was calculated for each participant and used for all subsequent analyses. We calculated the mean energy contribution (%) of each of the four processing categories using the dish- and food-level classifications for dishes prepared away from home. A paired t test was used to compare the mean energy contributions for each food category between the dish- and food-level classification. Pearson correlation coefficients were calculated to examine the associations between energy contributions from each processed food category when the dish- and food-level classifications were used. Pearson’s correlation coefficients were also computed to evaluate the association between energy contributions from different processed food categories within the same classification level.
In addition, the participants were divided into tertiles of the proportion of energy from HPF using both food- and dish-level classifications. We calculated the mean and sd or the number and percentage for each basic characteristic variable in each tertile of the proportion of energy from the HPF. Differences in basic characteristics across tertiles were assessed using the Mantel–Haenszel extension χ2 test for categorical variables (sex, education level and smoking status) and a linear trend test for continuous variables (age, height, body weight, BMI and physical activity). For linear regression, the median value in each tertile category of the energy contribution of the HPF was used as a continuous variable. Similarly, we calculated the average food group intake in each tertile group of the HPF energy contribution and tested the linear trend with increasing levels of food group intake using linear regression analysis. Finally, we calculated the adjusted means of the total and component scores of the HEI-2015 and NRF9·3 for each tertile of the energy contribution of the HPF. Potential confounding factors adjusted for were age, sex, BMI, physical activity, education level and smoking status(Reference Moubarac, Batal and Louzada9,Reference Marchese, Livingstone and Woods13,Reference Calixto Andrade, Julia and Deschamps15,Reference Adams and White19) . We tested for linear trends in diet quality scores across tertiles by assigning the median value of the energy contribution of HPF for each group as a continuous variable.
Results
The contribution of each food group to energy intake from HPFs
The mean age of the study participants was 44·5 years (sd 13·3) and the mean BMI was 23·3 kg/m2 (sd 3·7). Figure 2 shows the relative contribution of each food group to energy intake from HPF in the entire population. When dishes prepared away from home were classified at the dish level, the main food groups contributing to total energy intake from HPF were cereals and starchy foods (27·8 %), followed by meat, fish and eggs (16·2 %), confectionery (12·8 %), fats and oils (10·6 %) and alcoholic beverages (9·7 %). These food groups were also ranked in the top five when the food-level classification was used, albeit in a different order: cereals and starchy foods (23·1 %), followed by confectionery (18·7 %), alcoholic beverages (14·3 %), fats and oils (9·9 %) and meat, fish and eggs (9·0 %).
Fig. 2 Relative contribution of each food group to the total energy intake of highly processed foods in the entire population (n 388). In the dish-level classification, dishes prepared away from home (e.g. ready-made dishes from the supermarket and restaurant meals) were classified by level of processing without recipe disaggregation. In contrast, they were classified after disaggregation into component ingredients in the food-level classification
The energy contribution of foods classified by the level of food processing
Table 1 shows the descriptive statistics of the energy contribution of foods classified by the level of food processing. The energy contribution of HPF was significantly higher when dishes prepared away from home were classified at the dish level than the food level (48·3 % v. 32·9%, P < 0·0001). Conversely, the dish-level classification provided significantly lower estimates of energy contributions than the food-level classification for the other food categories, that is, unprocessed/minimally processed foods, basic processed foods and moderately processed foods (P < 0·0001 for all). The energy contributions from foods in each processing category between the dish- and food-level classifications were strongly correlated (r = 0·71–0·84). Moreover, for both dish- and food-level classifications, the energy contribution of HPF was significantly inversely correlated with unprocessed/minimally processed foods (r = –0·71 and –0·55, respectively) and basic processed foods (r = –0·78 and –0·70, respectively).
Table 1 Descriptive statistics of energy contribution (%) from foods classified by level of food processing; Japanese adults aged 20–69 years (n 388), 2013
UPF, unprocessed or minimally processed foods; BPF, basic processed foods; MPF, moderately processed foods; HPF, highly processed foods.
* Dishes prepared away from home (e.g. ready-made dishes from the supermarket and restaurant meals) were classified without recipe disaggregation.
† Dishes prepared away from home (e.g. ready-made dishes from the supermarket and restaurant meals) were classified after disaggregation into food ingredients.
‡ P < 0·0001 for all differences in the energy contribution of each food group between the dish- and food-based classification methods (paired t test).
Participant characteristics and energy and nutrient intake in relation to HPF consumption
Participant characteristics according to tertiles of the energy contribution of the HPF are shown in Table 2. In both classifications, the higher tertiles of energy contribution of HPF included younger participants and current smokers. Table 3 shows the energy and food group intake by tertiles of the energy contribution of the HPF. The results were similar for both the classifications. The mean energy intake did not differ across tertile groups, whereas the mean intake of some food groups differed. For instance, participants in higher tertiles had lower intakes of cereals and starchy foods; fruits, vegetables and pulses; meat, fish and eggs; and higher intakes of confectioneries; alcoholic beverages; fats and oils and seasoning and spices.
Table 2 Participant characteristics by tertiles (T) of energy contribution of highly processed foods*; Japanese adults aged 20–69 years (n 388), 2013
MET, metabolic equivalents.
* Values are means and standard deviations for continuous variables unless otherwise indicated.
† Dishes prepared away from home (e.g. ready-made dishes from the supermarket and restaurant meals) were classified without recipe disaggregation.
‡ Dishes prepared away from home (e.g. ready-made dishes from the supermarket and restaurant meals) were classified after disaggregation into food ingredients.
§ Derived from a Mantel–Haenszel extension χ2 test for categorical variables and linear trend test for continuous variables. For linear regression, the median value in each tertile category of the energy contribution of highly processed foods was used as a continuous variable. The median (range) energy contribution (%) of highly processed foods per tertile were as follows: 1st = 32·5 (14·8–40·9), 2nd = 47·7 (41·2–54·2) and 3rd = 62·7 (54·5–100) for the dish-level classification method; and 1st = 21·9 (9·6–26·7), 2nd = 31·1 (26·9–37·4) and 3rd = 43·6 (37·5–79·6) for the food-level classification method.
Table 3 Intakes of energy (kJ/d) and food groups (g/d) according to the tertiles (T) of the energy contribution of highly processed foods; Japanese adults aged 20–69 years (n 388), 2013
* Dishes prepared away from home (e.g. ready-made dishes from the supermarket and restaurant meals) were classified without recipe disaggregation.
† Dishes prepared away from home (e.g. ready-made dishes from the supermarket and restaurant meals) were classified after disaggregation into food ingredients.
‡ Least square means adjusted for age, BMI, physical activity, sex, education level and smoking status.
§ Derived from linear regression using the median value in each tertile category of the energy contribution of highly processed food as a continuous variable. The median (range) energy contribution (%) of highly processed foods per tertile were as follows: 1st = 32·5 (14·8–40·9), 2nd = 47·7 (41·2–54·2) and 3rd = 62·7 (54·5–100) for the dish-level classification method; and 1st = 21·9 (9·6–26·7), 2nd = 31·1 (26·9–37·4) and 3rd = 43·6 (37·5–79·6) for the food-level classification method. The models were adjusted for age, BMI, physical activity, sex, education level and smoking status.
‖ Including ready-to-eat or ready-to-heat foods such as curry, Chinese dumplings, Hamburg steak, pilaf, meatballs and croquettes.
Diet quality and HPF consumption
The mean (sd) scores of HEI-2015 and NRF9·3 scores were 51·1 (7·5) and 621·9 (116), respectively (Table 4). For both dish-level and food-level classifications, participants in higher tertiles of energy contribution of HPF had lower HEI-2015 total scores and component scores for total vegetables, greens and beans, total protein foods and added sugars and a higher score for refined grain. Moreover, participants in the higher tertiles had a lower component score of saturated fat only when the dish-level classification was used. Similarly, participants in higher tertiles of energy contribution of HPFs had lower NRF9·3 total scores and component scores for dietary fibre, vitamin A, vitamin C, vitamin D, Fe, K and Mg, and a higher score for added sugars, regardless of the classification levels. The component score for saturated fat was higher in participants in higher tertiles only when the dish-level classification was used.
Table 4 Diet quality scores according to the tertiles (T) of the energy contribution of highly processed foods; Japanese adults aged 20–69 years (n 388), 2013
HEI-2015, Healthy Eating Index-2015; NRF9·3, Nutrient-Rich Food Index 9·3.
* Dishes prepared away from home (e.g. ready-made dishes from the supermarket and restaurant meals) were classified without recipe disaggregation.
† Dishes prepared away from home (e.g. ready-made dishes from the supermarket and restaurant meals) were classified after disaggregation into food ingredients.
‡ Least square means adjusted for age, BMI, physical activity, sex, education level and smoking status.
§ Derived from linear regression using the median value in each tertile category of the energy contribution of highly processed food as a continuous variable. The median (range) energy contribution (%) of highly processed foods per tertile were as follows: 1st = 32·5 (14·8–40·9), 2nd = 47·7 (41·2–54·2) and 3rd = 62·7 (54·5–100) for the dish-level classification method; and 1st = 21·9 (9·6–26·7), 2nd = 31·1 (26·9–37·4) and 3rd = 43·6 (37·5–79·6) for the food-level classification method. The models were adjusted for age, BMI, physical activity, sex, education level and smoking status.
‖ Calculated as the sum of all component scores. The maximum scores are shown in parentheses. A higher score indicates higher diet quality.
¶ Defined as the ratio of the sum of polyunsaturated and MUFA to SFA.
** Calculated as the sum of scores for nine nutrients to encourage (i.e. protein, dietary fibre, vitamins A, C and D, Ca, Fe, K and Mg) minus the sum of scores for three nutrients to limit (i.e. added sugars, saturated fats and sodium). The maximum scores are shown in parentheses. For added sugars, saturated fats and sodium components, the maximum score was infinite, depending on the amount. A higher score indicates higher diet quality, except for added sugars, saturated fats and Na components, for which a higher score indicates a lower diet quality.
Discussion
Main findings
To the best of our knowledge, this is the first study to evaluate the consumption of HPF and their association with diet quality using data from a nationwide dietary survey in Japan. In Japan, dietary patterns have recently become continuously westernised(Reference Murakami, Livingstone and Sasaki26), and obesity (in males) and depression, both of which may be associated with HPF consumption(Reference Pagliai, Dinu and Madarena3,Reference Taneri, Wehrli and Roa Diaz4) , are increasing(27,28) . Therefore, applying the classification method by the level of processing to foods consumed by the Japanese and thereby clarifying HPF consumption is important for future research on HPF and related health outcome, as well as for setting a public nutrition policy in Japan. In this study, we found that at least one-third of the total energy intake accounted for HPF, of which cereals and starchy foods were the main contributors. Participants in higher tertiles of the energy contribution of HPFs had lower total scores for the HEI-2015 and NRF9·3. Thus, as hypothesised, the dietary share of HPFs is inversely associated with the overall dietary quality. These results were consistent, regardless of whether dishes prepared away from home were broken down into ingredients before being categorised by the food processing level.
Differences in food classification methods
There is no consensus regarding which dishes should be disaggregated into their ingredients before categorising them by processing levels(Reference Gibney30,Reference Bleiweiss-Sande, Bailey and Sacheck32) . For instance, the classification system used in the European Prospective Investigation into Cancer and Nutrition (EPIC) study breaks down all recipes(Reference Slimani, Deharveng and Southgate20). On the other hand, a widely used classification system, NOVA, proposes to break down the recipes of dishes and meals made in traditional restaurants(Reference Monteiro, Levy and Claro31). However, it is difficult to determine which restaurants are traditional and how they process their food(Reference Gibney30). Some previous studies disaggregated dishes prepared away from home, such as restaurant meals(Reference Salome, Arrazat and Wang8,Reference Moubarac, Batal and Louzada9) , whereas others did not disaggregate(Reference Calixto Andrade, Julia and Deschamps15), excluded(Reference Koiwai, Takemi and Hayashi23) or did not mention how to treat them. A previous study has also highlighted inconsistencies in classification methods and indicated that the misclassification of foods may contribute to the misrepresentation of the consumption, nutrient profile and associated health outcomes of HPF(Reference Bleiweiss-Sande, Bailey and Sacheck32). In this study, the energy contribution of HPF was higher when dishes prepared away from home were classified without recipe disaggregation (48·3 %) than when they were disaggregated (32·9 %), whereas both estimates were strongly correlated. This result is reasonable because most dishes were categorised as HPF when classified as a single item. It should be noted, however, that there was about a 15 % difference in the estimated energy contribution from HPF depending on whether the food prepared outside the home was broken down or not.
The energy contribution of HPFs
Our results showed that at least one-third of the total energy intake accounted for energy from the HPF. The energy contribution of HPF was similar to that in previous studies in Japan (38 %(Reference Koiwai, Takemi and Hayashi23) and 30 %(Reference Koiwai, Takemi and Hayashi24)). Meanwhile, the energy contribution of HPF in this study was lower than those in Canada (54 %)(Reference Batal, Johnson-Down and Moubarac5), the USA (58 %)(Reference Martinez Steele, Popkin and Swinburn18) and ten European countries attending the EPIC study (61—79 %)(Reference Rauber, da Costa Louzada and Steele16,Reference Adams and White19,Reference Slimani, Deharveng and Southgate20) , but higher than or equivalent to those in other areas or countries, such as Australia (39 %(Reference Marchese, Livingstone and Woods13)), France (31 %)(Reference Salome, Arrazat and Wang8,Reference Calixto Andrade, Julia and Deschamps15) , Mexico (30 %)(Reference Marron-Ponce, Flores and Cediel7), Belgium (30 %)(Reference Vandevijvere, De Ridder and Fiolet11), Chile (29 %)(Reference Cediel, Reyes and Corvalan6), South Korea (25 %)(Reference Shim, Shim and Cha12) and Brazil (20 %(Reference Louzada, Ricardo and Steele17)). However, it should be noted that direct comparison of the results may be difficult because of the differences in classification systems or the choice of food items to disaggregate or omit.
Participant characteristics and HPF consumption
Although associations between the energy contribution of HPFs and education level(Reference Marchese, Livingstone and Woods13,Reference Calixto Andrade, Julia and Deschamps15) and sex(Reference Adams and White19) have been reported previously, no association was observed in this study. However, the higher tertiles of the energy contribution of the HPF included younger participants and current smokers. This is consistent with the results of previous studies(Reference Salome, Arrazat and Wang8–Reference Vandevijvere, De Ridder and Fiolet11,Reference Marchese, Livingstone and Woods13,Reference Calixto Andrade, Julia and Deschamps15,Reference Adams and White19) . The inverse association between the energy contribution of HPF and age may be attributed to the fact that younger people tend to emphasise the convenience of food(Reference Murakami, Shinozaki and Yuan43), which is one of the facilitators of HPF consumption(Reference Almeida, Scagliusi and Duran44).
Diet quality and HPF consumption
Reference Murakami, Livingstone and Fujiwara Consistent with previous studies(Reference Batal, Johnson-Down and Moubarac5,Reference Cediel, Reyes and Corvalan6,Reference Salome, Arrazat and Wang8–Reference Lauria, Dello Russo and Formisano10,Reference Shim, Shim and Cha12–Reference Martinez Steele, Popkin and Swinburn18,Reference Koiwai, Takemi and Hayashi23) , the high-energy contribution of HPF was associated with a lower overall diet quality, regardless of the classification method of dishes prepared away from home. This may be due to the nutritional quality of HPF itself, food combinations, or dietary patterns in relation to HPF consumption. In addition, participants with higher energy intake from HPF had unfavourable intakes of total vegetables, greens and beans, total protein foods, added sugars, suggesting that increased HPF consumption is associated with a decrease in healthy food intake and an increase in unhealthy food intake. This finding is consistent with the results of previous studies(Reference Moubarac, Batal and Louzada9,Reference Liu, Steele and Li14,Reference Adams and White19) . However, contrary to a previous study in the USA(Reference Liu, Steele and Li14), refined grain intake was more favourable in the higher tertile group for the energy contribution of HPF. This is possibly due to the decrease in white rice in the higher tertile groups in this study.
Similar to a previous study(Reference Calixto Andrade, Julia and Deschamps15), participants with higher energy intake from HPF had unfavourable intakes of dietary fibre, vitamins, minerals and added or free sugars. However, in HEI-2015 and NRF9·3, the component score of sodium was not associated with the energy contribution of HPFs and was relatively low in all groups. This may be because unfavourable Na intake may be caused not only by HPF consumption, but also by other non-HPF sources, such as salt added to meals. Salt as a seasoning, which is categorised as a basic processed food, is the top contributor to sodium intake in this population(28). Previous studies have reported inconsistent results regarding the association between the energy contribution of HPF and inadequate Na intake(Reference Cediel, Reyes and Corvalan6,Reference Liu, Steele and Li14,Reference Calixto Andrade, Julia and Deschamps15) or Na density(Reference Batal, Johnson-Down and Moubarac5,Reference Cediel, Reyes and Corvalan6,Reference Moubarac, Batal and Louzada9,Reference Shim, Shim and Cha12,Reference Liu, Steele and Li14–Reference Rauber, da Costa Louzada and Steele16,Reference Martinez Steele, Popkin and Swinburn18) . This may be explained by differences in the types or sodium contents of HPF consumed, eating habits in each country or food classification systems used. Indeed, the top food group contributing to the total energy intake from HPF differed among countries: soft and fruit drinks(Reference Moubarac, Batal and Louzada9) or fast food and ready-to-eat dishes(Reference Batal, Johnson-Down and Moubarac5) in Canada, carbonated soft drinks in Chile(Reference Cediel, Reyes and Corvalan6), processed meat in Belgium(Reference Vandevijvere, De Ridder and Fiolet11), cookies and sweets in Brazil(Reference Louzada, Ricardo and Steele17), cookies, pastries and sweet bread in Mexico(Reference Marron-Ponce, Flores and Cediel7), ready-to-eat meals in France(Reference Calixto Andrade, Julia and Deschamps15) and cereals and starchy foods (e.g. rice, bread and noodles) in Japan(Reference Koiwai, Takemi and Hayashi24), South Korea(Reference Shim, Shim and Cha12), the USA(Reference Martinez Steele, Popkin and Swinburn18), the UK(Reference Rauber, da Costa Louzada and Steele16) and ten countries participating in the EPIC study(Reference Slimani, Deharveng and Southgate20).
Implications for public health nutrition research and practice
This study’s findings have several implications for future research and practice. Given that the energy contribution of HPF is associated with low intake of unprocessed/minimally processed foods and low diet quality, reducing HPF may lead to an increase in non-HPF foods with a better nutritional profile(Reference Batal, Johnson-Down and Moubarac5–Reference Marron-Ponce, Flores and Cediel7). The concept of HPF has been incorporated into official guidelines in several countries, including Brazil and Canada(Reference Quinn, Jordan and Lacy-Nichols45). The American Heart Association also recommends choosing minimally processed foods over HPF(Reference Lichtenstein, Appel and Vadiveloo46). However, some processed foods provide key nutrients, such as potassium and vitamin C, and have beneficial effects on food availability, convenience and safety(Reference Dwyer, Fulgoni and Clemens47). Thus, uniformly reducing all HPF in heterogeneous foods may not be an appropriate public health goal(Reference Tobias and Hall48). Therefore, consideration is necessary about which HPF should be reduced or reformulated to improve diet quality. Moreover, the differences in HPF consumption among participants’ characteristics or eating occasions may provide valuable information on efficient intervention strategies to promote healthy eating habits. In addition, since the estimates of HPF contribution differ widely depending on whether dishes prepared away from home are disaggregated, it is necessary to confirm the process of food classification when comparing the results across studies. In addition, future research should clearly explain what types of foods are broken down and could benefit from the standard guidelines for recipe disaggregation.
Strengths and limitations
The strength of this study is the use of 4-day DR obtained from a nationwide sample of Japanese adults. For the purpose of the survey, detailed information on foods was collected in the DR, such as the names of products, manufacturers and menus, as well as whether the food was prepared at home, away from home, or elsewhere. This allowed for the detailed classification of foods by processing level. In addition, comparing the results using two different classification methods for foods prepared away from home would be helpful in future studies on how to classify processed foods.
However, this study had some limitations. First, the participants voluntarily participated in the study and most worked at welfare facilities; therefore, the generalisability of our results may be limited. The proportions of current smokers and graduates from university or graduate school in this study were higher than those in the national survey (19·3 %(49) and 19·9 %(50), respectively). Nevertheless, participants were selected from all over Japan, and their mean height, weight and BMI did not materially differ from those of the general adult population (160 cm, 58·8 kg and 22·9 kg/m2 in 2013, respectively(49)). Second, the UNC system may be suboptimal for classifying foods sold in Japan because it was developed using the US food supply. Nevertheless, the UNC system provides detailed definitions and examples for foods in each processing category to help classify a variety of foods and has been applied to several countries outside the US, such as Portugal(Reference de Araujo, de Moraes and Afonso51) and Spain(Reference Martinez-Perez, San-Cristobal and Guallar-Castillon52). Third, the food classification was conducted by a single author and was not double-checked. Although the UNC system was found to have the highest inter-rater reliability among the three processing classification systems (including NOVA) used in the USA(Reference Bleiweiss-Sande, Chui and Evans35), some foods may have been misclassified in this study. Moreover, the description of some foods and beverages in the DR may not have been sufficient to correctly classify foods, which could have resulted in inaccurate categorisation of, for example, packaged food products. Fourth, disaggregating dishes into food ingredients may lead to a misestimation of the type or amount of ingredients. To estimate the type or amount of food items as accurately as possible, we referred to various information on the food item, including the approximate total amounts, the website of the restaurant or manufacturer, ingredient and nutrition facts labels and typical recipes. Fifth, the classification process was determined based on the format of the recording sheets used in this study, which cannot be applied in other studies. Nevertheless, the methodology demonstrated in this study would be informative for future studies on processed foods, as detailed descriptions of the classification process, particularly for foods away from home, have not always been provided in previous studies. Finally, self-reported dietary data are subject to social desirability bias(Reference Subar, Freedman and Tooze53), which may lead to a lower intake of unhealthy foods, potentially underestimating the contribution of HPF.
Conclusion
HPF accounted for at least one-third of the total energy intake among Japanese adults. Moreover, the higher energy contribution of HPFs was associated with lower diet quality, regardless of whether dishes prepared away from home were broken down into ingredients before being categorised by the food processing level. Therefore, decreasing the consumption of HPF would potentially be an effective public strategy to improve the diet quality in Japanese adults. Further research is needed to develop successful policies and programmes to improve the diet quality related to HPF reduction in the Japanese context.
Acknowledgements
Acknowledgements: The authors are grateful to the participants and local staff for their participation in this study. Authorship: N.S. formulated the research question, analyzed the data, interpreted the findings and wrote the first draft of the manuscript. K.M. formulated the research question, interpreted the findings and provided critical input for the final draft of the manuscript. K.A. collected and arranged the data. S.M. managed field establishment and recruitment for this study. S.S. designed and directed the survey. All authors have read and approved the final manuscript. Ethics of human subject participation: The study was conducted in accordance with the guidelines laid down in the Declaration of Helsinki. All procedure was approved by the Ethics Committee at the University of Tokyo, Faculty of Medicine (No. 10005). Written informed consent was obtained from all participants.
Financial support:
This study was financially supported by the Ministry of Health, Labour and Welfare, Japan (Nos. H23-Jyunkankitou(seishuu)-ippan-001 and 20KA1009).
Conflict of interest:
There are no conflicts of interest.
Supplementary material
For supplementary material/s referred to in this article, please visit https://doi.org/10.1017/S1368980023000721
Open access
