Diet quality is a fundamental determinant of health across the life course, influencing risks of obesity, metabolic disorders, CVD and premature mortality(1,Reference Mozaffarian2) . Diet quality is not static; it changes in response to developmental processes, social environments and socio-economic conditions(Reference Chong3–Reference Darmon and Drewnowski5). Evidence from population-based studies – predominantly from Western populations – indicates that diet quality tends to improve in mid-to later adulthood(Reference Hiza, Casavale and Guenther6,Reference Robinson, Westbury and Cooper7) , whereas children, adolescents and young adults often show suboptimal dietary intake(Reference Hiza, Casavale and Guenther6,Reference Winpenny, Penney and Corder8–Reference Faught, McLaren and Kirkpatrick11) . However, the age groups with the poorest diet quality, the dietary components driving these patterns and how they vary by sex, lifestyle and socio-economic factors remain insufficiently characterised, particularly in non-Western settings.
Understanding age-and sex-specific patterns of diet quality is essential for effective public health action(Reference Chong3). Dietary behaviours established in childhood and adolescence shape future cardiometabolic risk(Reference Mikkilä, Räsänen and Raitakari12,Reference Buckland, Northstone and Emmett13) , while suboptimal nutrition among young women of reproductive age may have intergenerational consequences(Reference Fall14,Reference Wilkins, Wickramasinghe and Pullar15) . Identifying critical periods of decline in diet quality and their underlying determinants can guide early and targeted interventions, including preconception nutrition strategies, to improve both current and future population health(Reference Stephenson, Heslehurst and Hall16).
Japan provides a distinctive context in which to examine these issues. The country is experiencing rapid population ageing(17) alongside widening socio-economic disparities(Reference Kitao and Yamada18,Reference Nishinakagawa, Sakurai and Nemoto19) and ongoing shifts in dietary behaviours, including increased breakfast skipping among adolescents and young adults(20,Reference Murakami, Livingstone and Fujiwara21) . These changes are occurring in parallel with the gradual Westernization of traditional dietary patterns(Reference Murakami, Livingstone and Sasaki22). Given the strong cultural and social structuring of dietary habits in Japan, age-and sex-specific differences in diet quality may differ from those reported in Western populations, with implications for both national policy and international comparative research.
A further limitation of existing evidence is methodological. Many existing studies rely on dietary assessment methods with limited capacity to capture within-person and seasonal variation in intake, including single-day dietary recalls, short-term records and FFQ(Reference Hiza, Casavale and Guenther6–Reference Lipsky, Nansel and Haynie10). High-quality evidence covering the entire age spectrum using robust dietary assessment methods remains scarce.
To address these gaps, the objective of this study was to characterise age-and sex-specific patterns of diet quality among Japanese individuals aged 1–79 years using eight-day dietary records (DR) collected across four seasons. Diet quality was assessed using the HEI-2020, a food-based measure of overall dietary patterns(Reference Shams-White, Pannucci and Lerman23,Reference Pannucci, Lerman and Herrick24) , and the NRF9·3, a nutrient-based measure of overall nutrient adequacy(Reference Drewnowski and Fulgoni25). We also examined the contributions of individual dietary components and the associations of lifestyle and socio-economic factors with diet quality. By identifying age groups with particularly low diet quality, this study provides evidence to inform targeted nutritional interventions and serves as a benchmark for life-course nutrition research in rapidly ageing Asian populations.
Methods
Study procedure and participants
This cross-sectional analysis utilised data from the Ministry of Health, Labour and Welfare-sponsored Nationwide Study on Dietary Intake Evaluation (MINNADE study), originally designed to evaluate and validate methods for assessing dietary exposure to hazardous environmental chemicals in the Japanese population. The overall design and methodological framework of the study have been described elsewhere(Reference Murakami, Livingstone and Masayasu26–Reference Shinozaki, Murakami and Kimoto28). Briefly, healthy, community-dwelling Japanese individuals aged 1–79 years were recruited from thirty-two of Japan’s forty-seven prefectures, representing more than 85 % of the national population, by a nationwide network of 453 trained research dietitians. Eligible participants were required to be capable of completing DR independently or, in the case of children, with assistance from a parent or guardian. Individuals were excluded if they were dietitians or lived with a dietitian, were receiving dietary counselling from healthcare professionals, were undergoing insulin or dialysis treatment, were pregnant or lactating, or, in the case of infants, were exclusively consuming human milk. Only one individual per household was enrolled.
Data collection was conducted over three consecutive survey rounds spanning November 2016 to August 2020. The first round targeted 2304 participants, with equal numbers of participants (128 males and 128 females) recruited across nine predefined age groups: 1–6, 7–13, 14–19, 20–29, 30–39, 40–49, 50–59, 60–69 and 70–79 years. To compensate for attrition, additional participants were recruited during the second round, and a third round focused on supplementing the youngest age group (1–6 years). In total, 4736 individuals (first round: n 2263, second round: n 2036 and third round: n 437) consented to participate in the MINNADE study. Participants initially completed a brief-type self-administered diet history questionnaire(Reference Kobayashi, Murakami and Sasaki29–Reference Okubo, Tajima and Shinozaki31) and an age-specific questionnaire at the beginning of each survey round. Different versions of the age-specific questionnaire were used for children aged 1–5 years, children aged 6–17 years and adults aged 18 years or older, with questions adapted for age-appropriate reporting. Critically, participants provided eight days of weighed DR, comprising two non-consecutive days in each of the four seasons to capture habitual intake across the year. The protocol for the MINNADE study was in accordance with the guidelines of the Declaration of Helsinki and was reviewed and approved by the Ethics Committee of the University of Tokyo Faculty of Medicine (protocol codes: 11397 and 2019124NI; approval dates: 25 October 2016 and 17 October 2019, respectively). Written informed consent was obtained from all participants, or from parents or guardians in the case of participants under 18 years of age.
For the present analysis, we included children and adults aged 1–79 years who provided at least one valid day of DR data (n 4601, Supplementary Figure 1). Participants were excluded if they completed fewer than eight recording days (n 12), recorded intake on consecutive days (n 103), conducted dietary recording outside the prespecified seasonal windows (n 20) or were pregnant or lactating at the time of the survey (n 12); exclusion reasons were not mutually exclusive. Consequently, the final analytical sample comprised 1652 children (841 males and 811 females) and 2803 adults (1376 males and 1427 females).
Age-specific patterns of diet quality among Japanese individuals aged 1–79 y. (A) Healthy Eating Index-2020 and (B) Nutrient-Rich Food Index 9.3. Solid lines indicate LOESS-smoothed mean scores across age, with shaded areas representing 95 % CI. Dashed vertical lines denote the ages at which the lowest predicted diet quality scores were observed for each sex.
Figure 1. Long description
Panel A: A line graph shows the Healthy Eating Index-2020 (HEI-2020) total score on the vertical axis and age in years on the horizontal axis. The graph includes data for males and females, with solid lines indicating LOESS-smoothed mean scores and shaded areas representing 95 percent confidence intervals. Dashed vertical lines denote the ages at which the lowest predicted diet quality scores were observed for each sex, with females at 16.1 years and males at 23.4 years. Panel B: Another line graph displays the Nutrient-Rich Food Index 9.3 (NRF9.3) total score on the vertical axis and age in years on the horizontal axis. Similar to Panel A, it includes data for males and females, with solid lines for LOESS-smoothed mean scores and shaded areas for 95 percent confidence intervals. Dashed vertical lines indicate the ages with the lowest predicted diet quality scores, which are 24.0 years for both sexes.
Eight-day weighed dietary records
Dietary intake data were collected using weighed DR on two non-consecutive days in each of the four seasons (eight days in total per participant)(Reference Murakami, Livingstone and Masayasu26–Reference Shinozaki, Murakami and Kimoto28). Recording days were allocated per season, such that approximately half of the participants completed two weekdays (Monday to Friday), whereas the remaining participants recorded intake on one weekday and one weekend (Saturday, Sunday or national holidays). Before data collection, research dietitians provided standardised verbal and written instructions, along with sample completed diary sheets. Participants were instructed to weigh and record all foods and beverages consumed, both at home and away from home, using a digital kitchen scale (KS-812WT; Tanita; 1-g increments up to 2 kg). The completed DR were collected by the research dietitian, usually on the day following recording, and reviewed for completeness and accuracy. Any missing or unclear information was clarified with the participant as needed. All DR then underwent centralised quality control by trained dietitians, including checks of food coding, recorded weights and item descriptions.
Estimates of daily intakes of energy and nutrients were calculated using the 2015 edition of the Standard Tables of Food Composition in Japan(32). For foods lacking information on added sugar content, values were assigned based on comparable items from the 2011–2012 Food Patterns Equivalents Database, with teaspoon equivalents converted to grams using a factor of 4·2 g per teaspoon(Reference Bowman, Clemens and Friday33). Dietary supplements were not included in nutrient intake calculations because a comprehensive supplement composition database is not available in Japan. To provide descriptive information on reported energy intake (EI) relative to estimated energy requirements (EER), the ratio of EI to EER (EI:EER) was calculated(Reference Livingstone and Black34). EER was estimated using equations from the dietary reference intakes (DRI) for the USA and Canada, based on sex, age, height and weight (measured as described below), assuming a low-active physical activity level, given the absence of empirical measures of physical activity(35). For all dietary variables, mean daily intakes across the eight recording days were used in the analysis.
Calculation of diet quality scores
Dietary intake data derived from the eight-day weighed DR were used to calculate diet quality scores using two complementary indices: HEI-2020(Reference Shams-White, Pannucci and Lerman23,Reference Pannucci, Lerman and Herrick24) and NRF9·3(Reference Drewnowski and Fulgoni25), with higher scores indicating better overall diet quality. The HEI-2020 primarily assesses overall diet quality based on adherence to food-based dietary recommendation(36), whereas the NRF9·3 evaluates nutrient adequacy by quantifying the density of beneficial nutrients relative to nutrients to limit(Reference Drewnowski and Fulgoni25,Reference Fulgoni, Keast and Drewnowski37) . Although HEI-2020 was developed on the basis of the Dietary Guidelines for Americans, it was selected to provide a common food-based framework applicable across the wide age range examined in the present study. To our knowledge, no validated Japanese food-based diet quality index is currently available for consistent application across this entire age range. NRF9·3 was also used as a complementary nutrient-based measure to assess diet quality from a different perspective. A previous study demonstrated the applicability of both HEI and NRF9·3 to Japanese populations(Reference Murakami, Livingstone and Fujiwara38). The calculation methods and components for each index are summarised in Supplementary Tables 1 and 2.
The HEI-2020 is a measure of overall diet quality that determines alignment with the 2020–2025 Dietary Guidelines for Americans(36). It comprises thirteen components: nine adequacy components (total fruits, whole fruits, total vegetables, greens and beans, whole grains, dairy, total protein foods, seafood and plant proteins and fatty acids) and four moderation components (refined grains, Na, added sugars and SFA) (Supplementary Table 1). Component scores are calculated per 1000 kcal, except for fatty acids, which are expressed as a ratio of unsaturated to SFA. Maximum component scores range from 5 to 10, with a total possible score of 100. Cup and ounce equivalents were assigned at the individual food-item level using a database developed specifically for Japanese foods, based on the US Food Patterns Ingredients Database and according to a previously published procedure(Reference Murakami, Livingstone and Fujiwara38). For children aged 1–2 years, scores were calculated using the HEI-Toddlers-2020 to reflect age-appropriate portion sizes and dietary patterns while maintaining alignment with the overall scoring framework(Reference Pannucci, Lerman and Herrick24).
Overall diet quality was also assessed using the NRF9·3, a composite measure of nutrient density(Reference Drewnowski and Fulgoni25,Reference Fulgoni, Keast and Drewnowski37) . The NRF9·3 was calculated as the sum of the percentages of reference daily values for nine qualifying nutrients – protein, dietary fibre, vitamin A, vitamin C, vitamin D, Ca, Fe, potassium and Mg – minus the sum of the percentages of reference daily values for three disqualifying nutrients – added sugars, saturated fats and Na. Reference daily values were determined by sex and age based on the DRI for Japanese, 2025(39): recommended dietary allowances for protein, vitamins A and C, Ca, Fe and Mg, and tentative dietary goal for preventing lifestyle-related diseases for dietary fibre, potassium, saturated fats and Na (Supplementary Table 2). For added sugars, the conditional recommendation of ≤5 % of EI from the WHO was used(40), as Japan lacks a formal guideline and intake levels are low. For children aged 1–2 years, reference daily values for dietary fibre, potassium and saturated fats were not defined; we therefore applied the values for the next older age group as a pragmatic approximation. Daily nutrient intakes for each participant were adjusted for energy using the density method and normalised to the sex-and age-specific EER for a moderate level of physical activity(39). These adjusted intakes were expressed as percentages of the corresponding reference daily values. For qualifying nutrients, percentages were capped at 100 so that excessive intake of one nutrient could not compensate for low intake of another. For disqualifying nutrients, only the proportion exceeding the recommended value was considered. The resulting percentages were summed (with disqualifying nutrient values subtracted) to obtain the total NRF9·3 score for each participant, with a theoretical maximum of 900.
Assessment of socio-demographic and lifestyle variables
An age-specific questionnaire was administered at the beginning of each survey round to collect information on basic characteristics of study participants. Age (years) at study entry was calculated from the date of birth and used as the age variable in all analyses. Annual household income was categorised as <4 000 000; 4 000 000–6 999 999 or ≥7 000 000 Japanese yen. Educational attainment was classified as <13 years (high school or less), 13–14 years (technical or professional school), ≥15 years (university or higher) and others. For adults, participants’ own educational attainment was used, whereas for children, both paternal and maternal education were considered. Employment status among adults was categorised as unemployed, student, part-time or full-time. Smoking status in adults was classified as current smoker, former smoker or never smoker. Breakfast consumption was assessed differently by age group. For children aged 1–5 years, caregivers reported the frequency of breakfast intake using six response categories ranging from ‘almost never’ to ‘almost every day.’ For participants aged ≥6 years, the number of breakfasts consumed per week was directly reported. Responses were then harmonised across age groups and categorised as 0, 1 or ≥2 skipped breakfasts per week.
At the beginning of each survey round, using standardised procedures, height and weight were measured with participants barefoot and wearing light indoor clothing, by family members or research dietitians. Height was recorded to the nearest 0·1 cm and weight to the nearest 0·1 kg. For participants without measured values, self-reported height and weight obtained from the BDHQ were used. BMI (kg/m2) was calculated as weight divided by height squared. For children, BMI z-scores were derived using sex-and age-specific Lambda-Mu-Sigma values from the International Obesity Task Force, and weight status was classified according to IOTF cut-offs(Reference Cole, Bellizzi and Flegal41). For adults, weight status was defined using WHO criteria(42).
Statistical analysis
All statistical analyses were conducted separately for males and females using R statistical software, version 4.5.1 (The R Foundation). Descriptive statistics for basic characteristics of the study participants were presented as mean (sd) for continuous variables and number (percentage) for categorical variables. Age-related patterns of diet quality (HEI-2020 and NRF9·3) were examined using locally weighted scatterplot smoothing (span = 0·30), specified a priori to capture non-linear age-related trends.
Contributions of individual dietary components to variation in total diet quality scores were evaluated within each age-and sex-specific stratum using multiple linear regression models that simultaneously included all component scores. Standardised regression coefficients (β) were used to quantify the relative contribution of variation in each component to variation in the total score. Components with no variability within a stratum were excluded from the model and assigned missing values. Age-and sex-specific patterns were visualised using heatmaps for illustrative purposes, with interpretation based on the magnitude of the estimated β coefficients rather than on formal significance testing.
Associations of diet quality with socio-demographic and lifestyle factors were assessed using multivariable linear regression models, adjusting for age. For children, only variables available in the age-specific questionnaire were included: weight status (underweight, normal, overweight or obese), annual household income (<4 000 000, 4 000 000–6 999 999 or ≥7 000 000 Japanese yen), paternal and maternal educational attainment (<13 years, 13–14 years, ≥15 years or others) and breakfast skipping (none, once a week or ≥2 times/week). For adults, participants’ own educational attainment, employment status (unemployed, student, part-time or full-time) and smoking status (non-smoker, past smoker or current smoker) were additionally considered. Missing values for covariates were imputed using multiple imputation by chained equations with random forests via the ‘miceRanger’ package in R (m = 10), assuming data were missing at random(Reference Aleryani, Wang and de la Iglesia43). In the main analyses, least-squares means (LSmeans) and standard errors (se) of total diet quality scores were estimated for each category of lifestyle and socio-economic factors, adjusting for age. Differences relative to reference categories were expressed as β coefficients with 95 % CI, and P for trend was calculated across ordinal categories, excluding ‘others’ in educational attainment due to unclear classification and small sample size. Sensitivity analyses were performed using mutually adjusted models, in which the association of each lifestyle or socio-demographic factor with diet quality was adjusted for age and all other lifestyle and socio-demographic factors included in the analysis. For socio-economic factors, because educational attainment, employment status and household income may act as confounders or mediators depending on the causal pathway, these analyses were considered exploratory and are presented in Supplemental Tables rather than as main results. Age interactions were tested by including interaction terms between age and each lifestyle or socio-economic variable. Statistical significance was set at a two-sided P value <0·05.
This study followed the Strengthening the Reporting of Observational Studies in Epidemiology reporting guidelines.
Results
Participant characteristics
Participant characteristics are shown in Table 1. Mean daily EI was 1949 (sd 738) kcal in male children, 1628 (sd 470) kcal in female children, 2246 (sd 486) kcal in adult males and 1822 (sd 338) kcal in adult females. The mean EI:EER ratio ranged from 0·92 to 1·16 across sex and age groups.
Basic characteristics of the study participants*
Table 1. Long description
A table comparing characteristics of children and adolescents aged 11-17 years with adults aged 18-79 years. The table has 18 rows and 16 columns. Column headers include Age, Body height, Body weight, BMI, EI, EER, EI:EER, HEI-2020, NRF9.3, Weight status, Annual household income, Employment status, Education, Paternal education, Maternal education, Smoking status, and Skipping breakfast. Row labels include categories such as Males, Females, and various age groups. Each cell contains mean values and standard deviations for different characteristics. Notable trends include variations in body height, weight, and energy intake across different age and sex groups. The table also shows differences in employment status, education levels, and smoking habits between children, adolescents, and adults.
EI, energy intake; EER, estimated energy requirement; HEI-2020, Healthy Eating Index-2020; JPY, Japanese yen; NRF9.3, Nutrient-Rich Food Index 9.3.
* Values are presented as mean (sd) for continuous variables and number (percentage) for categorical variables.
† The number of subjects with missing values was 6 for body height (2 males and 4 females) and 8 for body weight (5 males and 3 females).
Age-and sex-specific variation in diet quality
Both HEI-2020 and NRF9·3 scores varied systematically with age across childhood and adulthood (Figure 1, Supplemental Tables 3 and 4). HEI-2020 scores were relatively high in early childhood, were lower through adolescence and reached estimated nadirs at 16·1 years in females and 23·4 years in males, with higher scores observed in older age groups (Figure 1A). NRF9·3 scores were higher in middle childhood, showed lower levels during adolescence, reached minima in young adulthood (approximately 24 years) and were subsequently higher in older age groups in both sexes (Figure 1B).
Contributions of dietary components
Figure 2 illustrates age-and sex-specific contributions of individual dietary components to variation in total diet quality scores. For HEI-2020, dairy contributed most prominently in children, while fatty acids and Na showed substantial contributions across age groups, particularly in early childhood (Figure 2A). Saturated fats contributed more prominently from middle childhood through adolescence, whereas refined grains showed greater contributions in older age groups. For NRF9·3, the largest absolute contributions to variation in total scores were seen for the disqualifying components, particularly Na and added sugars, whereas several qualifying nutrients, particularly vitamins A and D, showed moderate positive contributions (Figure 2B).
Age-and sex-specific contributions of individual dietary components to variation in total diet quality scores: (A) Healthy Eating Index-2020 (HEI-2020) and (B) Nutrient-Rich Food Index 9·3 (NRF9·3). Standardized regression coefficients (β) were estimated from multiple linear regression models including all component scores simultaneously within each age-and sex-specific stratum. The magnitude of β represents the relative contribution of variation in each component to variation in the corresponding total score. Blue and red shading indicate positive and negative β coefficients, respectively, with colour intensity reflecting the magnitude of β and white indicating values near zero. Red dashes indicate components with no variability within the corresponding stratum and therefore not included in the regression model.
Lifestyle and socio-demographic factors
Associations between diet quality and lifestyle or socio-economic factors varied by age, sex and diet quality index. In children (Table 2), HEI-2020 scores were associated with body weight status. Among males, household income was positively associated with HEI-2020, whereas parental educational attainment and breakfast skipping were positively and negatively associated with NRF9·3, respectively. Among females, higher household income and parental educational attainment were positively associated with both HEI-2020 and NRF9·3, and breakfast skipping was negatively associated with both indices.
Total diet quality scores of healthy eating index-2020 and nutrient-rich food index 9·3 according to socio-demographic and lifestyle characteristics among Japanese children aged 1–17 years
Table 2. Long description
The table presents diet quality scores for males and females using HEI-2020 and NRF9.3 indices, categorized by weight status, annual household income, paternal education, maternal education, and skipping breakfast. It includes columns for sample size, least squares mean, standard error, beta coefficient, 95 percent confidence interval, and p-values for trends and age interactions. The table has 10 rows and 20 columns, with detailed values for each category and subcategory.
HEI-2020, Healthy Eating Index-2020; IOTF, International Obesity Task Force; JPY, Japanese yen; LSMeans, least-squares means; NRF9.3, Nutrient-Rich Food Index 9·3.
* Missing values for covariates were imputed using multiple imputation prior to analysis.
† Values are presented as LSMean and se, estimated from linear models including the explanatory variable of interest and adjusting for age (years).
‡ β (95 % CI) indicates the difference between each category and the reference category (other − reference), estimated from the same model.
§ Weight status in children was classified according to age-and sex-specific IOTF BMI cut-offs(Reference Cole, Bellizzi and Flegal41).
|| P for trend was calculated for ordinal categories (excluding ‘Other’ levels where applicable) using age-adjusted linear models.
¶ Age interaction refers to the interaction term between age and the explanatory variable, tested in age-adjusted models.
** In December 2025, 100 Japanese yen was approximately equivalent to 0.65 USA dollars, 0.55 euros and 0.48 British pounds sterling.
In adults (Table 3), associations differed by sex and diet quality index. In males, higher household income and educational attainment were positively associated with both HEI-2020 and NRF9·3, while smoking and breakfast skipping were negatively associated; full-time employment (v. unemployment) was additionally associated with lower HEI-2020 scores. In females, higher household income and educational attainment were positively associated with NRF9·3 only. Smoking and higher body weight status were negatively associated with both HEI-2020 and NRF9·3, whereas breakfast skipping was negatively associated with NRF9·3 only.
Total diet quality scores of healthy eating index-2020 and nutrient-rich food index 9·3 according to socio-demographic and lifestyle characteristics among Japanese adults aged 18–79 years
Table 3. Long description
The table presents data on the healthy eating index-2020 (HEI-2020) and nutrient-rich food index 9.3 (NRF93) scores for males and females, categorized by various socio-demographic and lifestyle factors. The table has 43 rows and 21 columns. Column headers include variables such as weight status, annual household income, education, employment status, smoking status, and skipping breakfast. Row labels include categories like underweight, normal, overweight, obese, and different income and education levels. Each row provides the sample size (n), mean scores, standard deviations (SD), and confidence intervals (CI) for both HEI-2020 and NRF93 scores. Notable trends include positive associations of higher household income and educational attainment with both diet quality scores in males, and negative associations of smoking and breakfast skipping with these scores. In females, higher household income and educational attainment are positively associated with NRF93 only, while smoking and higher body weight status are negatively associated with both scores.
HEI-2020, Healthy Eating Index-2020; JPY, Japanese yen; LSMeans, least-squares means; NRF9.3, Nutrient-Rich Food Index 9.3.
* Missing values for covariates were imputed using multiple imputation prior to analysis.
† Values are presented as LSMean and se, estimated from linear models including the explanatory variable of interest and adjusting for age (years).
‡ β (95 % CI) indicates the difference between each category and the reference category (other − reference), estimated from the same model.
§ Weight status in adults was classified according to BMI cut-offs defined by the WHO: underweight (<18.5 kg/m²), normal weight (18.5–24.9 kg/m²), overweight (25.0–29.9 kg/m²) and obese (≥30.0 kg/m²)(42).
|| P for trend was calculated for ordinal categories (excluding ‘Other’ levels where applicable) using age-adjusted linear models.
¶ Age interaction refers to the interaction term between age and the explanatory variable, tested in age-adjusted models.
** In December 2025, 100 Japanese yen was approximately equivalent to 0.65 USA dollars, 0.55 euros and 0.48 British pounds sterling.
Several associations varied by age. In children, HEI-2020 scores differed across body weight categories (Figure 3A and 3B), and lower NRF9·3 scores were observed among males with lower maternal education from middle childhood onward (Figure 3C). Among adult males, income-related disparities in diet quality varied by age (Figure 3D), whereas among adult females, smoking was associated with lower NRF9·3 scores, particularly in younger and middle-aged adults (Figure 3E).
Age interactions in the associations of lifestyle and socio-demographic factors with diet quality scores. (A, B) Healthy Eating Index-2020 (HEI-2020) scores according to BMI-defined weight status across age groups among male children (A) and female children (B). (C) Nutrient-Rich Food Index 9.3 (NRF9.3) scores according to maternal educational attainment across age groups among male children. (D) NRF9.3 scores according to household income across age groups among adult males. (E) NRF9.3 scores according to smoking status across age groups among adult females. Lines represent estimated marginal means derived from linear regression models including age, each determinant and their interaction; shaded areas indicate 95 % confidence intervals. Age was treated as a continuous variable in the models and is presented by predefined age groups for visualisation.
Mutual adjustment for lifestyle and socio-demographic factors altered some associations (Supplementary Tables 5 and 6). Patterns remained largely unchanged in male children. Among females, associations between household income and diet quality were generally attenuated, although the association with HEI-2020 remained evident in children. In adult males, associations with educational attainment were attenuated, whereas full-time employment remained associated with lower HEI-2020 scores and showed a similar, although weaker, pattern for NRF9·3. Age interactions in mutually adjusted models were generally consistent with age-adjusted models, although interactions between age and body weight status became significant for both HEI-2020 and NRF9·3, whereas those between age and household income (both indices) and between age and employment status (NRF9·3 only) were no longer significant in adult males.
Discussion
Using nationwide, multi-season weighed DR spanning ages 1–79 years, this cross-sectional study provides a comprehensive characterisation of age-and sex-specific patterns of diet quality in Japan. Observed differences across both food-and nutrient-based indices indicate that diet quality was comparatively lower during adolescence and young adulthood, with the precise ages of lowest scores varying by sex and index. These periods exhibited distinct dietary component profiles and were associated with lifestyle and socio-economic factors.
Our findings are broadly consistent with previous studies from Western countries, while extending the existing literature in several important ways. In the US population, children and older adults had better-quality diets than younger and middle-aged adults(Reference Hiza, Casavale and Guenther6), broadly paralleling the age-related pattern observed in the present study. However, whereas income was associated with diet quality among US adults but not children(Reference Hiza, Casavale and Guenther6), socio-economic differences in diet quality were evident among Japanese children in our study. This is broadly in line with life-course evidence from Canadian young adults, in whom socio-economic disadvantage in childhood – assessed retrospectively – was associated with poorer diet quality in young adulthood(Reference Faught, McLaren and Kirkpatrick11). Longitudinal studies following adolescents into early adulthood have also shown suboptimal diet quality during this period, although some improvement occurred across early adulthood and dietary changes accompanied major life transitions(Reference Tao, Wall and Larson9,Reference Lipsky, Nansel and Haynie10) . Repeated assessments in a British birth cohort further demonstrated tracking of diet quality across adulthood and its relevance to physical function in later life(Reference Robinson, Westbury and Cooper7). Whereas these previous studies focused on specific life stages or age-restricted populations(Reference Robinson, Westbury and Cooper7,Reference Tao, Wall and Larson9–Reference Faught, McLaren and Kirkpatrick11) , the present study provides a unified depiction of diet quality from early childhood to older age within a single national context. Although the observations are cross-sectional and do not reflect within-individual trajectories, they offer valuable insight into age ranges that may warrant particular attention in population-based nutritional strategies.
A key contribution of this study is the parallel application of HEI-2020 and NRF9·3 to evaluate age-related variation in diet quality. Both indices consistently identified adolescence through young adulthood as periods of comparatively poor diet quality, underscoring the robustness of this life-course pattern. The timing and magnitude of the lowest scores differed by sex and index. HEI-2020 reached its lowest levels at 16·1 years in females, which may reflect the transition to high school, where discontinuation of school lunches and adolescent weight-loss dieting can compromise dietary balance(Reference Asakura and Sasaki44–Reference Kaneko, Kiriike and Ikenaga47). In contrast, the later nadirs around 24 years in both sexes likely reflect transitions to independent living and early career, characterised by time constraints and greater reliance on convenience foods(Reference Tao, Wall and Larson9). These differences reflect the complementary nature of the indices: HEI-2020 captures food-based dietary patterns, whereas NRF9·3 more directly captures nutrient adequacy using reference values based primarily on Japanese DRI. The transient peak in NRF9·3 during middle childhood may reflect caregiver-directed diets closely aligned with nutrient recommendations(Reference Ho and Birch48), illustrating how nutrient-based indices can reveal age-specific features not fully captured by food-based measures.
An additional insight from this study is the components contribution most to inter-individual variation in overall diet quality appeared to vary across age groups. For HEI-2020, dairy was particularly prominent in children, saturated fats from middle childhood through adolescents and refined grains in older age groups, while fatty acids and Na showed substantial contributions across the age range. For NRF9·3, the disqualifying components – particularly Na and added sugars – showed largest absolute contributions across age groups. These patterns are broadly consistent with previous evidence of age-related shifts towards lower dairy intake and higher consumption of processed and out-of-home foods(Reference Tao, Wall and Larson9,20,Reference Shinozaki, Murakami and Kimoto28,Reference Asakura and Sasaki44) . The prominence of Na across both indices underscores its central role in shaping diet quality in the Japanese context, where both traditional and contemporary dietary patterns are characterised by relatively high salt intake(20,Reference Murakami, Livingstone and Sasaki22,Reference Asakura, Uechi and Masayasu49) . Together, these results suggest that the components contributing most to inter-individual variation in diet quality differ across life stages and may help identify pragmatic targets for age-tailored nutritional interventions.
Importantly, the extent to which these key dietary components contribute to variation in overall diet quality appears to be strongly patterned by lifestyle and socio-economic factors, and these associations vary substantially by age and sex. In childhood, higher household income and parental educational attainment were generally associated with better diet quality, particularly among females and for the nutrient-based index. Lower maternal educational attainment became increasingly associated with poorer NRF9·3 scores among males from middle childhood onward, suggesting that socio-economic disparities in nutrient adequacy may widen as children gain dietary autonomy(Reference Lipsky, Nansel and Haynie10,Reference Faught, McLaren and Kirkpatrick11,Reference Ziegler, Kasprzak and Mansouri50,Reference Story, Neumark-Sztainer and French51) . In adulthood, higher income and educational attainment were positively associated with diet quality in age-adjusted models, whereas smoking and breakfast skipping were inversely associated, particularly for NRF9·3. However, mutual adjustment revealed important sex-specific differences. Among females, associations with household income were attenuated after accounting for other socio-demographic and lifestyle factors. In contrast, among males, the association of full-time employment with lower HEI-2020 scores, already evident in age-adjusted models, remained significant after mutual adjustment, with a similar, although weaker, pattern observed for NRF9·3. This pattern may reflect work-related constraints – such as long working hours, limited time for meal preparation and greater reliance on convenience foods(Reference Tanaka, Tsuji and Kusuhara52–Reference Oostenbach, Lamb and Crawford54) – which may act as more proximal determinants of diet quality among males than educational background per se. In the Japanese context, these findings highlight employment conditions as a potentially important structural factor to consider in dietary interventions and workplace-based nutrition initiatives.
Further analyses of age interactions provided additional insight into these sex-specific patterns. In children of both sexes and in adult females, age-modified associations observed in age-adjusted models were largely preserved after mutual adjustment, indicating relatively stable relationships between socio-economic factors and diet quality across age. In contrast, among adult males, the pattern of age interactions changed after mutual adjustment, with a significant interaction between age and body weight status observed for both HEI-2020 and NRF9·3, whereas age interactions for household income (both indices) and for employment status (NRF9·3 only) were no longer evident. This finding suggests that, in adult males, age-related differences in diet quality may increasingly reflect health-related factors rather than socio-economic position alone, pointing to a dynamic interplay between ageing, weight status and dietary behaviours.
This study has several strengths, including its large, nationwide sample spanning a wide age range, the use of multi-season weighed DR to capture habitual intake and the application of complementary diet quality indices. Nevertheless, several limitations should be considered. First, the cross-sectional design precludes inference about within-individual changes over time and may partly reflect cohort effects. Second, participants were volunteers, and the intensive nature of the dietary assessment may have resulted in a sample with relatively higher health consciousness; however, previous analyses of this dataset suggest that key demographic characteristics are broadly comparable to those of nationally representative samples(Reference Murakami, Livingstone and Masayasu26–Reference Shinozaki, Murakami and Kimoto28). Third, although weighed DR are considered a reference method, they remain subject to measurement error and potential behavioural reactivity. Fourth, the assessment of socio-economic and lifestyle factors was constrained by the available questionnaires, and residual confounding cannot be excluded; moreover, differences in questionnaires between children and adults may have affected comparability across age groups. Finally, the diet quality indices themselves have inherent limitations. HEI-2020 is based on USA dietary guidelines and therefore does not specifically measure adherence to Japanese dietary recommendations or fully capture qualitative aspects of Japanese diets. No validated Japanese food-based diet quality index is currently available for consistent application across the full age range examined here, highlighting the need for such an index in future research. In addition, the application of NRF9·3 required extrapolation of nutrient reference values for children aged 1–2 years.
In conclusion, diet quality among Japanese individuals aged 1–79 years is lowest during adolescence and young adulthood, with patterns differing by sex and by the diet quality index used. These findings identify critical life stages at which preventive nutritional strategies may be most effective, ideally implemented before diet quality declines to its lowest levels. The results further demonstrate that the lifestyle and socio-economic associations of diet quality vary substantially across age and sex, with employment-related factors playing a particularly prominent role among adult males. Future longitudinal studies incorporating more detailed social, occupational and behavioural information are warranted to elucidate within-individual trajectories of diet quality and their implications for long-term health.
Supplementary material
For supplementary material/s referred to in this article, please visit https://doi.org/10.1017/S0007114526108642
Data availability statement
Data are unsuitable for public deposition due to ethical restrictions and legal frameworks in Japan. It is prohibited by the Act on the Protection of Personal Information (Act No. 57 of 30th May 2003, amendment on 9th September 2015) to publicly deposit data containing personal information. Ethical Guidelines for Medical and Health Research Involving Human Subjects enforced by the Japan Ministry of Education, Culture, Sports, Science and Technology and the Ministry of Health, Labour and Welfare also restrict the open sharing of epidemiological data.
Acknowledgements
We thank all the children and their parents/guardians who took part in this study. We also thank the following research dietitians and staff who collected and processed the data: Tamotsu Noshiro*, Ikuko Kato, Yoshie Awa, Erika Takayama, Mari Sakurai, Mihoko Yanase, Masae Kato, Mihoko Furukawa, Yuna Nodera, Kazue Fukushi, Miwako Onodera, Yoshie Sato, Megumi Yoshida, Masako Shimooka, Kaori Takahashi and Fuki Kudo (Hokkaido); Yumiko Sato, Yutaka Shojiguchi, Kazunori Kimoto, Saori Kikuchi, Megumi Maeta, Mayumi Sugawara, Shinogu Muraoka, Kanako Takahashi, Noriko Suzuki, Yoko Fujihira and Megumi Onodera (Iwate); Yukiko Takahashi, Kaoru Honda, Chie Yamada, Miki Sato, Katsue Watanabe, Akemi Konno and Reina Kato (Miyagi); Akiko Sato, Hiromi Kawaguchi, Miyuki Awano, Chigusa Miyake, Ayako Konno, Ayumi Goto, Shizuko Taira, Yuka Takeda, Akiko Matsunaga, Nao Konta, Yumi Miura, Satoshi Numazawa, Chiemi Ito, Sachie Yokosawa, Manam Endo and Hiromi Seki (Yamagata); Yoko Tsukada, Tomoko Oga, Satoko Fujita, Hitomi Sonobe, Hanayo Kadoi, Toshie Nakayama, Hiromi Takasawa, Yoko Ichikawa, Yuko Takano, Junko Hanzawa, Kiyomi Seki, Emi Kamoshita, Yuri Kawakami and Hisako Watahiki (Ibaraki); Reiko Ishii, Yoshiyuki Tatsuki, Daisuke Mogi, Akiko Nakamura, Suguru Yagi, Yumiko Furushima, Noriko Ogiwara, Kiyomi Kimura, Kinue Takahashi, Kikue Tomaru, Kana Tsukagoshi, Fumiko Fujiu, Kyoko Maehara and Yuki Kobayashi (Gunma); Kaoru Goto, Yuka Inaba, Michiko Koresawa, Tomoko Tsuchida, Naoko Sakakibara, Fumika Shimoyama, Akiko Kato, Miki Hori, Rika Kurosaki, Hiroko Yamada, Hitomi Sasaki, Keiko Arai, Yuka Arai, Manami Honda, Akiko Utsumi, Asako Hamada, Keiko Sekine, Akiko Yamada, Mami Ono, Satoko Maruyama, Emiko Kajiwara, Taeko Takahashi, Hitomi Kawata, Satoru Arai, Ryoko Hirose, Madoka Ono and Mihoko Ainai (Saitama); Masako Shinohara, Noriko Nakamura, Mitsuko Ito, Yuka Takahara, Minako Fukuda, Masae Ito, Yayoi Sueyoshi, Hiroko Shigeno, Tomohiro Murakami, Masako Kametani, Kyoko Wada, Mika Ueda, Jun Kouno, Hiroyo Yamaguchi, Mariko Oya, Junko Suegane, Yumiko Asai, Miyuki Ono, Mitsuko Uekusa, Chieko Sunada, Yumi Tanada, Mariko Shibata, Emi Tsukii, Kae Terayama, Hiroko Iwasaki, Keiko Yokoyama, Haruna Kayano, Kazuyo Shimota, Keiko Ifuku and Keiko Honma (Chiba); Yoshiko Saito, Megumi Suzuki, Eiko Kobayashi, Yoshiko Katayama, Sonoko Yamaguchi, Tomoko Kita, Naoko Yuasa, Hitomi Okahashi, Shinobu Matsui, Yurina Arai, Sanae Togo, Eiko Horiguchi, Juri Sato, Takehiro Komatsu, Yumi Matsuura, Junko Higashi, Ayaka Nakashita, Takako Sakanashi, Yoko Kono, Naomi Nakazawa and Yukiko Shibata (Tokyo); Machiko Tanaka, Ikue Sahara, Yasue Watanabe, Kanako Yoshijima, Yuko Harashima, Yoshiko Iba, Haruko Irisawa, Junko Inose, Reiko Okui, Taeko Endo, Mayuko Sakitsu, Ikuko Endo, Haruko Terada, Chiaki Nishikawa, Ai Yasudomi, Suzuyo Takeda, Kaori Shimizu, Mari Ikeda, Yuko Okamoto, Keiko Yamada, Fumiko Nemoto, Shinobu Katayama, Yuki Takakuwa, Michiru Goukon, Megumi Koike, Masae Kamiya, Takako Okada, Yayoi Hayashi, Etsuko Abe, Akiko Hamamoto, Kumiko Ono, Kazumi Takagi, Sachiko Ito, Yuki Kumagai, Noriko Ozaki, Haruka Sato, Hisae Takahashi, Masuko Komazaki, Akiko Nako, Tatehiro Inamoto and Kimiyo Matsumoto (Kanagawa); Masako Koike, Reiko Kunimatsu, Keiko Kuribayashi, Hiroko Adachi, Yuri Shikama, Yurika Seida, Ryouko Ito, Satoko Kimura, Yoko Sato, Michiyo Nakamura, Hisako Kaneko, Hatsuyo Ikarashi, Mamiko Karo, Keiko Hirayama, Ikumi Torigoe and Fumiko Gumizawa (Niigata); Natsuko Mizuguchi, Aki Sakai, Hisako Noguchi, Chie Tanabe, Yokako Osaki, Ikiko Kawasaki, Nobuko Yoshii, Yumiko Nishihara, Izumi Takahara, Mika Minato, Yuuki Okamoto, Yukiyo Sakai, Shitomi Nakamura, Kanako Kobayashi, Emi Hano and Megumi Emori (Toyama); Kazumi Horiguchi, Michiyo Kubota, Naoya Mochizuki, Miyuki Yokokoji, Kazuko Koizumi, Megumi Ariizumi, Hozumi Kakishima, Mayumi Kawakubo, Chisato Nakajima, Yasuko Ishii, Yukie Shiogami, Yukiko Uchida, Ikuko Kayanuma and Kikuyo Moriya (Yamanashi); Noriko Sumi, Noriko Takahashi, Kuniko Watanabe, Yoko Ido, Akiko Adachi and Manami Tauchi (Gifu); Naoya Terada, Chisato Suwa, Toshihiro Tamori, Natsuko Osakabe, Toshiyuki Serizawa, Akiko Seki, Izumi Mochizuki, Nagako Matsui, Eiko Watanabe, Kyoko Yui, Yuki Murakami, Tomomi Iwasaki and Tomoko Sugiyama (Shizuoka); Keiko Kawasumi, Masako Tanaka, Kayoko Ishida, Megumi Yamatani, Shihoko Yama, Miyuki Otono, Mie Kojima, Tamaki Kobayashi, Hiroe Komaki, Miki Yanagida, Yumiko Fukaya, Syoko Sawaki, Tomomi Ota, Yasuko Kito, Mei Tobinaga, Takashi Yasue, Kuniko Hatamoto, Toru Ono, Takako Minami, Akemi Kumazawa, Masami Kato, Miyuki Kondo, Kyoko Shimizu, Sayoko Tanaka and Shizue Masuda (Aichi); Hiromi Ashida, Shintaro Hinaga, Yoshiko Shoji, Ryusuke Yamaguchi, Hiroka Morita and Atsuko Nakabayashi (Mie); Erika Shioi, Sawa Mizukawa, Miwako Ohashi, Eriko Taniguchi, Yuri Mitsushima, Mariko Teraya, Kazuko Ogawa, Yoko Minami, Megumi Ito, Yasuhiro Morimoto, Shizuka Kurokawa and Manami Hayashi (Kyoto); Yumiko Noutomi, Yoshiko Iwamoto, Junko Ikukawa, Shinobu Fujiwara, Tami Irei, Keiko Takata, Yasuka Tabuchi, Naoko Murayama, Kaori Maruyama, Hiromi Tashiro, Miki Tanaka, Miho Nomura, Shizuyo Umezawa, Minori Shintani, IkuyoMaruishi, Atsuko Toyokawa, Rumi Kitada, Yuka Takashima, Eriko Nakatani, Wakana Tsujimoto, Yumi Koori, Emi Iwamoto and Masumi Yamada (Osaka); Atsuko Konishi, Yoshiko Nakamori, Yumi Ikawa, Junko Shimizu, Mie Atsumi, Atsuko Fukuzaki, Akemi Yamamoto, Yasuko Inoue, Miyuki Nakahara, Reiko Fujii, Yumi Tanaka, Rika Miyachi, Mari Matsuyama, Ayana Honda, Tomoka Nakata, Miho Inagaki, Mikiyo Ueno, Mami Kamei, Kiyomi Kawachi, Yasuyo Hasegawa and Masayo Fukumoto (Hyogo); Sachiyo Otani, Tomomi Sugimoto, Kanako Mizoguchi, Tomomi Shimada, Shima Takahashi, Yoshiko Okuno, Takahide Kijima, Masayo Ueda, Yuko Sakamoto, Hitoshi Matsuda, Yumie Shimizu, Rie Hataguchi, Junko Nohara, Yuko Sakakitani, Hatsuki Matsumoto, Sakiko Tanaka, Moe Yoshikawa and Rena Yuki (Nara); Yurika Adachi, Akiko Notsu, Keiko Uzuki, Atsuko Umeda, Hiroko Nishio, Chikako Takeshita and Sayuri Omoso (Tottori); Hiromi Watanabe, Nami Sakane, Nao Hino, Sakiko Sonoyama, Yukiko Katagiri, Kaori Nagami, Tsunemi Moriyama and Yoshiko Kirihara (Shimane); Sachiko Terao, Akemi Inoue, Mieko Imanaka, Noriyuki Kubota, Sachiko Sugii, Yuri Fujiwara and Tomoko Miyake (Okayama); Youko Fujii, Hiroko Tamura, Kimie Tanaka, Izumi Hase, Etsuko Kimura, Akiko Hamada, Tomoko Kawai, Masako Ogusu and Emi Isomichi (Hiroshima); Kyoko Ueda, Atsuko Nakanishi, Tomoko Ishida, Nobuko Morishita, Hitomi Nakata and Takako Yagi (Yamaguchi); Yuriko Doihara, Toyoko Kitadai, Hideyo Yamada, Mariko Nakamura, Nanako Honda, Eri Ikeuchi, Kayo Hashimoto, Azusa Onishi and Manami Iwase (Tokushima); Machiko Ueda, Ayano Kamei, Reiko Motoie, Yoko Moriki, Nobumi Yoshida and Kiyoko Kubo (Ehime); Kyoko Kaku, Emi Ibushi, Miho Otsuka, Kiyoko Katayama, Hisami Kumagai, Chizuru Shibata, Miki Hamachi, Yuko Hayashida, Akiko Matsuzaki, Mika Yoshioka, Yoshie Yanase, Yoshiko Yahagi and Tomomi Ota (Fukuoka); Junko Kiyota, Hiromi Ide, Takako Fukushima, Shiho Tominaga, Satsuki Miyama, Yoko Okamura, Kayoko Kurahara, Tomomi Nagamori, Chika Shino, Akiko Taira, Yuki Kuwajima and Miyuki Matsushita (Kumamoto); Miki Hamada, Kiyomi Aso, Toshie Eto, Mitsue Kodama, Miyoko Sato, Mutsuko Shuto, Yuko Soga, Taeko Nagami, Machiko Hirayama, Mika Moribe, Junko Yamamoto, Hideko Yoshioka, Yuko Kawano, Rika Matsuoka and Satomi Sato (Oita); and Hisami Yamauchi, Satomi Moromi, Satoko Tomari, Kaoru Miyara, Chikako Murahama, Yukiko Furugen, Kana Awano, Hiromi Arakaki, Suzumi Uema, Yasuko Tomori, Nariko Mori, Ayako Iho, Michiru Hokama, Yasue Higa and Chigusa Chibana (Okinawa). The authors would also like to thank the following research team staff at the survey centre (University of Tokyo): Tomoko Doi, Hitomi Fujihashi, Akiko Hara, Nana Kimoto, Nanako Koe, Eri Kudo, Fumie Maeda, Keika Mine, Akemi Nakahara, Hiroko Onodera, Hiroko Sato, Chifumi Shimomura and Fusako Takahashi. *Deceased.
Financial support
This research was funded by the Ministry of Health, Labour and Welfare, Japan (grant number 24KA2004). The Ministry of Health, Labour and Welfare had no role in the design, analysis or writing of this article.
Authorship
H.O.: Conceptualisation, Methodology, Software, Formal analysis, Data curation, Visualization, Writing – original draft, Writing – review & editing. N.S.: Software, Formal analysis, Data curation, Writing – review & editing. R.T.: Investigation, Data curation, Writing – review & editing. S.M.: Investigation, Data curation, Resources, Writing – review & editing. S.S.: Methodology, Software, Data curation, Resources, Funding acquisition, Writing – review & editing. K.M.: Methodology, Software, Investigation, Resources, Funding acquisition, Project administration, Writing – review & editing.
Competing interests
H.O. and R.T. belong to the Department of Nutritional Epidemiology and Behavioural Nutrition, Graduate School of Medicine, The University of Tokyo, which is a social cooperation program with Ajinomoto, Co., Inc. The authors declare that Ajinomoto, Co., Inc. had no role in the design of the study; in the collection, analyses or interpretation of data and in the writing of the manuscript or in the decision to publish the results. The remaining authors declare that they have no competing interests.
Open access
