Abstract
OBJECTIVE
Prediabetes is an emerging public health concern among U.S. adolescents, yet nationally representative, up-to-date data on dietary contributors remain limited. Especially, the role of overall dietary intake and school meal participation (providing a substantial portion of daily calories in school children) has not been comprehensively evaluated. This study examined cross-sectional associations between school meal participation, overall dietary intake, and prediabetes (defined as having HbA1c and/or fasting glucose values within the prediabetes range) in the most recent nationally representative samples of U.S. adolescents, assessing differences by obesity status.
RESEARCH DESIGN AND METHODS
National Health and Nutrition Examination Survey data (2017–2020, 2021–2023) were analyzed, including adolescents (aged 12–18 years) with complete fasting glucose and HbA1c data (n = 802). Multivariable modified Poisson regression models evaluated associations between school meal participation (breakfast and lunch), dietary intake (food groups and nutrients), and prediabetes. Effect modification by obesity status was assessed.
RESULTS
Prediabetes prevalence was 33.8% (SE 3.0) and obesity prevalence was 20.96% (SE 1.44). School breakfast participation at least once a week was associated with higher prediabetes prevalence (prevalence ratio [PR] 1.3 (95% CI 1.01–1.66), while school lunch participation was not. Higher vitamin E and B6 intake was linked to lower prediabetes prevalence (PR 0.96, 0.94–0.99; 0.9, 0.81–0.99). Higher fruit intake was inversely (PR 0.96, 0.92–0.99) associated with prediabetes among adolescents with obesity.
CONCLUSIONS
School breakfast participation and specific dietary components were associated with prediabetes risk in U.S. adolescents. While confirmation is needed, the nutritional quality of school breakfasts may represent an important intervention target.
Graphical Abstract
Introduction
Prediabetes among adolescents is increasingly recognized as a critical public health issue (1,2). In 2023, an estimated 8.4 million adolescents aged 12–17 years, or 32.7% of the U.S. adolescent population, had prediabetes based on an analysis conducted with National Health and Nutrition Examination Survey (NHANES) data (3). Adolescence is a pivotal period for metabolic health, characterized by transient physiological insulin resistance driven most likely by pubertal hormonal changes. This period may represent a window of increased metabolic vulnerability during which adverse environmental and lifestyle factors can contribute to glycemic dysregulation and future risk of type 2 diabetes and related complications (2).
Amid these physiological changes, diet quality and eating patterns are established modifiable factors influencing metabolic indicators, such as glucose metabolism and insulin sensitivity. For instance, among youth, higher intakes of fruits, fibers, and vitamins were found to be associated with better metabolic parameters, whereas saturated fats were associated with suboptimal ones (4–6). Despite this, research specifically examining overall and comprehensive dietary intake in relation to prediabetes among adolescents in a nationally representative U.S. sample remains limited. Existing studies often rely on broad dietary quality indices, include a restricted range of dietary factors, or include earlier dietary data, which may not accurately capture current eating behaviors, food environments, or nutritional exposures (7–9). Moreover, the current literature has rarely considered whether associations between dietary factors and prediabetes differ according to obesity status in adolescents. This may be important because adolescents with obesity exhibit preexisting metabolic alterations, including greater insulin resistance and impaired glucose regulation, which could influence their susceptibility to dietary exposures (2,10,11).
When considering adolescent nutrition, it is crucial to focus on the context of food consumption. Within this landscape, schools play a pivotal role, supplying a substantial portion of daily calories, around half for many U.S. youths, particularly those from lower-income households (12). Although school nutrition standards have evolved, including recent efforts to limit added sugars, many school meals remain high in added sugars, ultraprocessed foods (UPFs), and nutrient-poor foods (13). Yet, to date, no U.S. studies have directly examined the relationship between school breakfast and lunch consumption and prediabetes in a nationally representative sample of adolescents.
Guided by a socioecological framework (14), this study aimed to cross-sectionally investigate the relationships between school meal participation (breakfast and lunch), overall dietary intake, and indicators of prediabetes in the most recent nationally representative samples of U.S. adolescents. Of note, although the term “prediabetes” is used throughout the manuscript, we define it as having HbA1c and/or fasting glucose values within the prediabetes range based on a single laboratory assessment, which should be interpreted as a marker of elevated glycemic risk rather than a clinical diagnosis. Importantly, we aimed to explore how these relationships differ by obesity status, given the association of excess adiposity with insulin resistance. By contextualizing overall individual dietary behaviors and adiposity within the broader school food environment, this study offers a more integrated understanding of how food-related factors may shape prediabetes risk in adolescents, while highlighting potential targets for future research, prevention, and remediation.
Research Design and Methods
Study
This cross-sectional study included the most recent data (2017–2020 and 2021–2023 cycles) from nationally representative samples of participants aged 12–18 years from NHANES (15). NHANES is a U.S. cross-sectional survey assessing health and nutrition through questionnaires, physical examinations, and laboratory tests. Details of the NHANES procedures and measurements have been described elsewhere (15).
Outcome
For adolescents without a diagnosis of diabetes (self- reported from age 15 years and reported by parents before that), prediabetes was defined according to the 2020 American Diabetes Association (ADA) criteria as meeting any of the following conditions (16): 1) impaired fasting plasma glucose (FPG) between 100 mg/dL and 125 mg/dL only, 2) hemoglobin A1C (HbA1c) level between 5.7% and 6.4% only, or 3) both impaired FPG and an HbA1c level between 5.7% and 6.4%. Given that this definition was based on a single laboratory assessment, results should be interpreted as reflecting HbA1c and/or FPG values within the prediabetes range rather than a clinical diagnosis of prediabetes.
Exposures
School Meal Participation
Information on participation in school meals was collected using two parallel questions, one referring to school breakfast and the other to school lunch. Both questions followed the same format, asking participants (or their parents) how many times per week they (their child) usually consumed school meals during the academic year, with response options ranging from zero to five times per week. From these two questions, two binary variables were created. For school breakfast, participants were categorized as having no participation (0 times per week) or any participation (at least once per week). The same coding strategy was applied to school lunch, distinguishing adolescents with zero participation from those with at least one weekly participation.
Dietary Intake
Dietary intake data in NHANES were collected using two nonconsecutive 24-h recalls, using the five-step Automated Multiple-Pass Method (17). These data were gathered through private interviews, where participants were asked about the overall type and amount of food and beverages consumed in the previous 24 h. Children aged ≥12 years answered the questions for themselves. The first day of data was collected in a private room in the mobile examination centers, and the second day was collected by telephone 3–10 days later. This analysis combined dietary data from both recall days to estimate average usual intake (energy intake or milligrams per day, where relevant).
Nutrient intakes considered (14) included available and key nutrients of concern, such as protein, carbohydrates, total sugars, dietary fibers, total fat, saturated fat, monounsaturated fat, polyunsaturated fat, cholesterol, vitamin E (α-tocopherol), vitamin B6, vitamin C, vitamin D, and sodium. Total carbohydrate, sugar, and protein intakes were multiplied by 4 kcal/g, while total fat, saturated fat, polyunsaturated fat, and monounsaturated fat were multiplied by 9 kcal/g, to estimate the percentage of total energy intake from grams per day. Dietary fiber and micronutrients were analyzed as reported, in grams or milligrams per day, respectively.
In NHANES, each food item is assigned an eight-digit food code, with the first digit indicating one of the nine major U.S. Department of Agriculture (USDA) food groups (1 is milk and milk products; 2 is meat, poultry, fish, and mixtures; 3 is eggs; 4 is legumes, nuts and seeds; 5 is grain products; 6 is fruits; 7 is vegetables; 8 is fats, oils, and salad dressings; and 9 is sweets and beverages). This classification was used to define and aggregate major food groups in our analysis. Subsequently, intakes were summed within each food group and expressed as a percentage of total energy intake.
Potential Effect Modifier
BMI, from measured weight and height, was used to categorize adolescents into having obesity versus not having obesity. BMI values were compared with age- and sex-specific growth norms (18). Participants with a BMI at or above the 95th percentile were categorized as having obesity.
Analytical Approach
Main Analysis
NHANES uses a complex, multistage sampling design. The effects of this design (including sampling weights, stratification, and clustering) were accounted for in all analyses. The fasting subsample weight was applied to preserve national representativeness, as the fasting measurements were collected in a smaller, specific subsample of participants. Data from the two most recent consecutive NHANES cycles (2017–2020 and 2021–2023) were combined for analysis to increase sample size and statistical precision. To facilitate pooled analyses, sampling weights were combined using a fixed fraction of the constituent surveys as per the National Center for Health Statistics analytic guidelines (19). Figure 1 depicts the participant selection process, based on the availability of measurements used to define prediabetes and the presence of nonzero fasting weights (with zero indicating either missing laboratory results or fasting periods outside the 8- to <24-h window). Of note, adolescents without self-reported diabetes whose HbA1c (≥6.5%) or fasting glucose (≥126 mg/dL) met ADA thresholds suggestive of diabetes (two individuals) were excluded from the analysis. Missing values for exposures (school meal participation and overall dietary intake) and covariates (all with <30% missing values) were imputed using the missForest package in R (20), which implements a nonparametric random forest–based imputation method.
Figure 1.
Flow chart: participant selection. *Zero indicates either missing laboratory results or fasting periods outside the 8- to <24-h window. **ADA thresholds for diabetes (HbA1c ≥6.5% and/or fasting glucose ≥126 mg/dL).
We quantified associations between school meal participation (breakfast and lunch), overall dietary intake, and adolescent prediabetes using prevalence ratios (PRs), estimated via survey-weighted Poisson regression with robust standard errors. PRs offer a more interpretable measure than odds ratios when outcomes are common, reflecting the relative prevalence of prediabetes directly (21). Two adjustment strategies were applied in our analyses (7,22). For the associations between school meal participation (breakfast and lunch) and prediabetes, models were adjusted for adolescent age, sex, survey wave, household income-to-poverty ratio (a measure of family income relative to federal poverty thresholds), ethnicity, health insurance status, and averaged total energy intake. For analyses of dietary intake and prediabetes, models were adjusted for the same covariates except averaged total energy intake (as the dietary variables were already energy-adjusted, with the exception of fiber and micronutrients). These models were additionally adjusted for physical activity (number of days physically active for at least 60 min) and school meal participation (breakfast and lunch).
Potential effect modification by obesity status was evaluated through the inclusion of an interaction term between the exposure (breakfast and lunch participation and dietary intakes) and obesity in survey-weighted regression models. In the absence of statistically significant interaction, associations were estimated from models excluding the interaction term to provide overall effect estimates. When significant interaction was detected, stratum- specific PRs were obtained from models stratified by obesity status (yes vs. no).
Sensitivity Analysis
For the primary (main) analysis, school meal participation (breakfast and lunch) was dichotomized as 0 vs. ≥1 day per week, allowing a clear comparison between nonparticipation versus any participation, which is directly relevant for policy and intervention interpretation. In sensitivity analyses, we further categorized meal frequency into three levels (0, 1–3, or 4–5 days per week) to explore whether associations with prediabetes vary across no, moderate, and high participation.
Statistical significance was P < 0.05. To account for multiple comparisons, false discovery rate (FDR) adjustment using the Benjamini-Hochberg procedure was applied within prespecified hypothesis families. These families were defined according to exposure domain and analysis type: school meal participation (overall effect analyses and interaction analyses, including both the two-level variable used in the primary analyses and the three-level variable used in sensitivity analyses) and overall dietary factors (overall effect analyses and interaction analyses). Given the hypothesis-driven nature of these analyses, results were primarily interpreted based on the original estimates, with FDR-adjusted values presented to assess robustness. Indeed, multiple-testing corrections may increase the risk of type II error and obscure potentially meaningful associations (23). As described above, overall effects and interaction analyses were considered distinctly, with overall effects interpreted only when no evidence of effect modification was observed. Accordingly, when a statistically significant interaction was identified (unadjusted P < 0.05), the corresponding overall effect was not included in the FDR adjustment for overall effect analyses. FDR adjustments were therefore performed across 6 tests for school meal participation overall effect analyses, 4 tests for school meal participation interaction analyses (no significant interactions with obesity status were identified; see RESULTS), 20 tests for overall dietary factor overall effect analyses, and 23 tests for overall dietary factor interaction analyses (three significant interactions with obesity status were identified; see RESULTS).
Statistical analyses were performed using R, version 4.3.3.
Data and Resource Availability
The combined data sets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Individually, each of the data sets used in this study is publicly available through this website: https://www.cdc.gov/nchs/nhanes/.
Results
Description
The prevalence of prediabetes was 33.8% (SE 3.0) and 20.96% (SE 1.44) for obesity. The average age of participants was 14.5 years, with ∼50% male and ∼51% non-White (Table 1).
Table 1.
Sample characteristics
| Mean or percentage1 (SE) | |
|---|---|
| Demographics | |
| Age (years) | 14.51 (0.06) |
| Household income-to-poverty ratio2 | 2.68 (0.08) |
| Male | 49.98% (1.77) |
| Non-White | 51.34% (1.76) |
| Health insurance (yes) | 94.53% (0.8) |
| Obesity (yes) | 20.96% (1.44) |
| School meal participation | |
| School breakfast frequency (at least once a week) | 51.86% (1.76) |
| School lunch frequency (at least once a week) | 75.71% (1.51) |
| Dietary intake | |
| Food groups | |
| Fats, oils, and salad dressings (% kcal) | 1.84 (0.14) |
| Grain products (% kcal) | 41.68 (0.92) |
| Meat, poultry, fish, and mixtures (% kcal) | 19.58 (0.74) |
| Milk and milk products (% kcal) | 10.81 (0.3) |
| Sugars, sweets, and beverages (% kcal) | 9.73 (0.32) |
| Vegetables (% kcal) | 6.92 (0.24) |
| Fruits (% kcal) | 4.06 (0.25) |
| Legumes, nuts, and seeds (% kcal) | 2.87 (0.27) |
| Eggs (% kcal) | 1.74 (0.15) |
| Nutrient intake | |
| Proteins (% kcal) | 14.42 (0.15) |
| Carbohydrate (% kcal) | 50.07 (0.33) |
| Total sugars (% kcal) | 21.46 (0.31) |
| Dietary fiber (g) | 14.88 (0.39) |
| Total fat (% kcal) | 36.37 (0.29) |
| Saturated fat (% kcal) | 12.39 (0.12) |
| Monounsaturated fat (% kcal) | 11.78 (0.12) |
| Polyunsaturated fat (% kcal) | 8.45 (0.14) |
| Cholesterol (mg) | 250.63 (8.3) |
| Vitamin E (mg) | 8.37 (0.36) |
| Vitamin B6 (mg) | 1.75 (0.05) |
| Vitamin C (mg) | 65.05 (3.29) |
| Vitamin D (μg) | 4.68 (0.21) |
| Sodium (mg) | 3,258.77 (66.61) |
Sample size n = 802. 1Weighted. 2Household income-to-poverty ratio was calculated by dividing family income by the poverty guidelines for the survey year. Poverty guidelines vary by family size and geographic location.
Regarding school meals, ∼52% of students participated in school breakfast at least once a week, with around 26% eating breakfast one to three times per week and 26% eating breakfast four or five times per week. Regarding school lunch, about 76% of students participated at least once a week, with 17% attending one to three times per week and 59% attending four or five times per week. Overall dietary intake (% calories) was primarily composed of 41.7% grain products; 19.6% meat, poultry, and fish; and 11% milk and milk products. Sugars, sweets, and beverages accounted for 9.7%, while vegetables and fruit contributed 6.9% and 4.1%, respectively. Smaller contributions came from legumes, nuts, and seeds (2.9%) and eggs (1.7%). Nutrient intake (% calories) included 14.4% protein, 50.1% carbohydrates, and 36.4% total fat, with 12.4% from saturated fat.
Association Analyses
School Meal Participation
Participation in school breakfast at least once a week was associated with a 30% higher prevalence of prediabetes in adolescents (PR 1.30 [95% CI 1.01–1.66], P = 0.039, FDR 0.119). No significant interaction was found between school breakfast participation and obesity (P = 0.40, FDR 0.54). No significant associations were found between school lunch participation (0 vs. ≥1 day per week) and prediabetes (PR 1.20 [0.89–1.62], P = 0.22, FDR 0.25), with no interaction observed with obesity (P = 0.40, FDR 0.54).
Sensitivity analyses suggested that modest participation in school breakfast (1–3 days/week) was associated with higher prediabetes prevalence (PR 1.37 [95% CI 1.03–1.83], P = 0.031, FDR 0.119), whereas higher participation (4–5 days/week) showed no significant association (PR = 1.2 [0.89–1.62], P = 0.223, FDR 0.25) compared with no participation. No association was observed between school lunch participation and prediabetes prevalence when the exposure variable was categorized into three groups (0, 1–3, 4–5 days per week).
Overall Dietary Intake
Table 2 summarizes the findings from the analyses examining the relationship between dietary intake (including food groups and macro- and micronutrient intake) and prediabetes. The Supplementary Material presents the complete set of unadjusted and FDR-adjusted P values for the analyses. No associations were observed for fats, oils, and salad dressings; grain products; meat, poultry, fish, and mixtures; milk and milk products; sweets and beverages; legumes, nuts, and seeds; or eggs. Similarly, macronutrient intake, including protein, carbohydrate, total sugars, dietary fiber, total fat, saturated fat, and monounsaturated fat, showed no associations with prediabetes. Among micronutrients, higher vitamin E and vitamin B6 intake was associated with a lower prevalence of prediabetes (respectively, PR = 0.96 [95% CI 0.94–0.99], P = 0.01, FDR 0.195; and 0.90 [0.81–0.99], P = 0.049, FDR 0.272). No associations were observed for cholesterol, vitamin C, vitamin D, or sodium intake.
Table 2.
Overall food group and nutrient intake in relation to prediabetes prevalence
| Prediabetes (defined as having HbA1c and/or fasting glucose values within the prediabetes range1) | ||||
|---|---|---|---|---|
| Overall | Adolescents without obesity PR (95% CI) | Adolescents with obesity | P values (noncorrected) interaction2 (exposure × obesity) | |
| Food groups | ||||
| Fats, oils, and salad dressings (% kcal) | 0.99 (0.96, 1.03) | 0.993 | ||
| Grain products (% kcal) | 1.01 (1.00, 1.01) | 0.468 | ||
| Meat, poultry, fish, and mixtures (% kcal) | 0.99 (0.98, 1.00) | 0.234 | ||
| Milk and milk products (% kcal) | 1.00 (0.99, 1.01) | 0.737 | ||
| Sugars, sweets, and beverages (% kcal) | 1.01 (1.00, 1.02) | 0.691 | ||
| Vegetables (% kcal) | 0.98 (0.95, 1.01) | 1.01 (0.99, 1.03) | 0.038 | |
| Fruits (% kcal) | 1.01 (0.98, 1.04) | 0.96 (0.92, 0.99) | 0.042 | |
| Legumes, nuts, and seeds (% kcal) | 1.00 (0.98, 1.02) | 0.493 | ||
| Eggs (% kcal) | 0.99 (0.95, 1.03) | 0.672 | ||
| Nutrient intake | ||||
| Protein (% kcal) | 0.97 (0.94, 1.01) | 0.708 | ||
| Carbohydrate (% kcal) | 1.01 (1.00, 1.03) | 0.253 | ||
| Total sugars (% kcal) | 1.01 (1.00, 1.03) | 0.118 | ||
| Dietary fiber (g) | 0.99 (0.97, 1.00) | 0.489 | ||
| Total fat (% kcal) | 0.99 (0.97, 1.02) | 0.35 | ||
| Saturated Fat (% kcal) | 1.01 (0.97, 1.05) | 0.286 | ||
| Monounsaturated fat (% kcal) | 0.98 (0.94, 1.02) | 0.223 | ||
| Polyunsaturated fat (% kcal) | 0.93 (0.87, 1.01) | 1.05 (1.00, 1.10) | 0.011 | |
| Cholesterol (mg) | 1.00 (1.00, 1.00) | 0.537 | ||
| Vitamin E (mg) | 0.96 (0.94, 0.99) | 0.117 | ||
| Vitamin B6 (mg) | 0.90 (0.81, 0.99) | 0.464 | ||
| Vitamin C (mg) | 1.00 (1.00, 1.00) | 0.212 | ||
| Vitamin D (μg) | 0.98 (0.96, 1.01) | 0.306 | ||
| Sodium (mg) | 1.00 (1.00, 1.00) | 0.056 | ||
Sample size n = 802. If the interaction between dietary factors and obesity status is significant, results are presented stratified by obesity status. Weighted estimates are displayed. Models were adjusted for age, sex, survey wave, household income-to-poverty ratio, ethnicity, health insurance, physical activity, and school meal participation (breakfast and lunch). Boldface denotes noncorrected P values < 0.05.
1Ranges: impaired FPG between 100 mg/dL and 125 mg/dL; HbA1c level between 5.7% and 6.4%.
2Complete results, including both unadjusted and adjusted P values (for multiple comparisons) for all analyses, are provided in the Supplementary Material.
Evidence of effect modification by obesity was observed for vegetable intake (interaction P = 0.038, FDR 0.32), fruit intake (interaction P = 0.042, FDR 0.32), and polyunsaturated fat intake (interaction P = 0.011, FDR 0.244). For vegetables and polyunsaturated fats, stratum- specific estimates suggested that the interaction was not meaningful, that is, not statistically significant (Table 2). For fruits, the association was protective among adolescents with obesity only (PR = 0.96, 95% CI 0.92–0.99).
Conclusions
Leveraging the most recent nationally representative data, this cross-sectional study examined the interplay between school meal participation (breakfast and lunch), overall dietary intake, and prediabetes in adolescents, with a focus on potential modification by obesity. Participation in school breakfast at least once a week was positively associated with prediabetes, whereas school lunch participation showed no significant relationship. Regarding overall dietary intake, higher vitamin E and B6 consumption appeared to confer a protective effect against prediabetes. Evidence of effect modification by obesity (significant interaction terms) was observed for vegetable, fruit, and polyunsaturated fat intake. Although tests for interaction suggested potential effect modification by obesity status for vegetable and polyunsaturated fat intake, stratified analyses did not reveal clear associations within either obesity subgroup. In contrast, higher fruit intake was associated with a lower prevalence of prediabetes among adolescents with obesity. Although these associations met conventional significance thresholds, they did not remain significant after FDR correction; accordingly, these findings should be viewed as hypothesis-generating and require confirmation.
Comparison With Other Studies
Although detailed nutrient information for school meals was not available, nor were specific and comprehensive nutrient intakes by children based on participation in these meals, the link between eating school breakfasts and higher rates of prediabetes may partly be due to the limited availability of minimally processed foods in schools. In 2023, across eight states, only 17% of schools served scratch or modified-scratch cooked meals (minimally processed) daily, and just 3% served only scratch or modified-scratch foods (13). For adolescents, even a modest increase in proinflammatory compounds from UPFs may exacerbate metabolic dysfunction and heighten the risk of prediabetes (24). One possible explanation for the lack of association between school lunch participation and prediabetes as compared with breakfast is that breakfast offerings in schools may contain a higher proportion of added sugars, juice, and UPFs than lunch items. Of note, the positive association between school breakfast participation and prediabetes was most pronounced among adolescents who consumed breakfast at school only one to three times per week, whereas those attending four or five times per week exhibited a weaker, close to significance association. Importantly, these patterns persisted even after adjustment for total energy intake and household income-to-poverty ratio, indicating that overall caloric intake or socioeconomic factors alone may not explain the observed trends. This pattern may instead reflect differences in school breakfast quality: sporadic participants are likely to rely on a limited selection of less balanced or more ultraprocessed items, whereas regular participants have exposure to a broader variety of foods across the week. Averaged over multiple days, habitual consumption may provide a relatively more balanced nutrient profile (although still not optimal, as suggested by the borderline association for those attending four or five times per week), thereby mitigating potential metabolic consequences. Alternatively, the attenuation of the association among more frequent participants may indicate that unmeasured factors related to school breakfast participation patterns contribute to these relationships. For example, adolescents who participate sporadically may differ from regular participants in ways that were not fully captured in our analyses, including household routines or other social and environmental factors. Although differences in the nutritional quality or composition of foods consumed through school breakfast programs could also play a role, our data do not allow us to evaluate these potential mechanisms directly. Therefore, these findings should be interpreted cautiously and warrant further investigation in studies incorporating more detailed information on school meal content and consumption patterns.
Our findings also suggest a potential protective role of overall intake of vitamin E and B6 against prediabetes in adolescents. As a fat-soluble antioxidant, vitamin E may improve insulin sensitivity by reducing oxidative stress and inflammation (25,26), highlighting the importance of including vitamin E–rich foods, such as nuts, seeds, and green leafy vegetables, in adolescents’ diets. Consistent with adult studies (27), higher vitamin B6 (abundant in poultry, fish, and bananas) was associated with lower prediabetes prevalence in adolescents, also because it supports insulin sensitivity.
Although interaction terms pointed to potential effect heterogeneity across several dietary factors (vegetables, fruits, and polyunsaturated fats), associations for vegetables and polyunsaturated fats did not reach statistical significance within strata, likely reflecting comparatively weaker differential signals. In contrast, fruit intake exhibited a differential association, with a protective effect observed exclusively among adolescents with obesity. This pattern may reflect underlying metabolic differences, as adolescents with obesity frequently present with impaired insulin sensitivity and chronic low-grade inflammation, conditions that may enhance responsiveness to beneficial dietary exposures. The fiber and bioactive compounds in fruits may therefore exert a more substantial influence on glycemic regulation in this subgroup (28,29). These potential differential effects underscore that the metabolic consequences of specific nutrients can vary depending on obesity status, highlighting the importance of considering both diet quality and metabolic health when evaluating prediabetes risk in adolescents.
For other food groups and nutrients, no associations were found. Notably, two previous NHANES-based studies in adolescents, relying on older data and more limited dietary measures, similarly found no association between sugar intake and prediabetes, in line with our results (7,9). Authors have suggested that the near- universal overall consumption of sugar from sources such as ready-to-eat cereals and sugar-sweetened beverages may mask differences by prediabetes status. Comparable observations have been made recently in adults (NHANES) (30), and the broader literature shows mixed findings, likely reflecting differences in study design, population characteristics, and how sugar is quantified.
Limitations and Strengths
There are several limitations that should be considered when interpreting these results. A cross-sectional study can only establish associations, not causality, meaning that the observed relationships between school meal participation, overall dietary intake, and prediabetes cannot be definitively attributed to the exposures studied. Additionally, although we use the term “prediabetes” throughout the manuscript for readability, our classification was based on a single HbA1c and/or fasting glucose assessment and should therefore be interpreted as reflecting glycemic values within the prediabetes range rather than a clinical diagnosis of prediabetes. The reliance on self-reported data for school meal participation and dietary intake may introduce recall bias or inaccuracies in exposure assessment. Of note, some diet-related variables such as added sugars and whole grains were not available for the 2021–2023 NHANES wave, because the USDA Food Patterns Equivalents Database (FPED), which translates reported food intake into food groups and nutrients consistent with the USDA dietary guidelines, currently (at the time of the analysis) only covers 2017–2020. As a result, these components could not be included in our combined analysis of overall dietary intake. Additionally, the 2021–2023 NHANES cycle was conducted using modified sampling and data collection procedures following the coronavirus disease 2019 pandemic. Although the National Center for Health Statistics concluded that no major sources of bias remained that were not adequately addressed through revised survey weighting methods (31), the pandemic substantially disrupted school attendance and school meal program operations. Therefore, school lunch participation and related responses collected during this survey cycle may not fully reflect typical school-year patterns and should be considered when interpreting the findings. Although we included survey cycle as a covariate in all analyses to account for potential differences across NHANES cycles and partially address temporal variations, this adjustment may not fully account for all pandemic-related changes in school meal access and participation patterns. Another limitation is that NHANES sampling and weighting procedures are designed to achieve national representativeness of individuals rather than schools or school districts. While NHANES weights account for the complex survey design and improve population-level representativeness, they may not fully capture clustering or contextual differences in school food environments across schools and districts. Despite controlling for a range of covariates, there may still be unmeasured confounders that could influence the studied relationships, such as genetics, hormones, or unaccounted-for lifestyle factors. Finally, although our study highlights associations between school breakfast participation and prediabetes, it lacks detailed information on the specific foods served and consumed in schools. In addition, because NHANES does not include school-level information on meal preparation or implementation, we were unable to account for potential variation in school meal quality across settings. This limits our ability to determine how the nutritional quality of school meals may influence adolescents’ metabolic health. Future research incorporating precise school- and student-level dietary data could help clarify these relationships more definitively.
The strengths of this study lie in its use of nationally representative data from the NHANES data set, which provides a broad, diverse, and recent sample of participants aged 12–18 years. This enables the findings to be generalized to the larger U.S. adolescent population. Additionally, the study offers a well-defined approach to assessing prediabetes, using objective measurements and the ADA’s criteria to identify adolescents with impaired glucose metabolism. Moreover, the study’s comprehensive approach to assessing both school meal participation and overall dietary intake provides valuable insight into how specific food-related factors may influence adolescent prediabetes. Finally, accounting for potential interactions with obesity allowed us to capture any combined effects of obesity and diet-related factors on prediabetes risk.
Public Health and Research Implications
The presence of UPFs in school meals is a potential concern for adolescent health (13), as these meals provide a substantial portion of daily nutrition, particularly for youth from lower-income households (12), and may influence overall metabolic outcomes, as suggested by our findings. In this study, school breakfast participation was examined and associated with a higher prevalence of prediabetes, although detailed nutrient-level data from these meals were not available. Thus, while the mechanisms underlying this association remain unclear, differences in the nutritional composition of school meals, including the presence of UPFs, may represent one potential contributing factor, given prior evidence linking higher UPF intake with altered glucose regulation and insulin sensitivity (13). Importantly, skipping breakfast has been associated with increased insulin resistance and other suboptimal cardiometabolic markers in children and adolescents (32). While school breakfast programs were initially implemented to address barriers to adequate nutrition, our findings highlight the need for further research to better understand the factors underlying the observed association between school breakfast participation and prediabetes. Although we did not assess the nutritional composition of school meals or student-level dietary intake during school breakfast, ensuring access to balanced and nutrient-dense meal options remains an important consideration for school nutrition programs. Complementing these recommendations, our broader dietary analyses suggested that overall intake of nutrient-rich foods may be protective, highlighting additional avenues for interventions targeting adolescent metabolic health.
Overall, while these findings are preliminary, and further longitudinal and interventional research are needed to clarify the impact of breakfast quality on prediabetes risk, they suggest that providing nutrient-rich breakfast options, such as fruits and vitamins E– and B6–rich foods like nuts, seeds, poultry, and leafy greens, alongside nutrition education may offer a valuable opportunity to promote healthier dietary habits early in life (33).
Conclusion
In this nationally representative sample of U.S. adolescents and cross-sectional analysis, prediabetes was highly prevalent and associated with both school breakfast participation and specific dietary factors, with some differences by obesity status. These findings seem to underscore the importance of moving beyond participation alone to consider the nutritional quality of school meals, particularly breakfast, as a potential lever for early metabolic risk reduction. The observed associations with vitamins E and B6 and fruit intake highlight both the complexity of dietary influences on glycemic health during adolescence and potential avenues for interventions. Although causal inferences are limited by the cross- sectional design, this study nonetheless provides timely evidence to inform school nutrition policies and targeted dietary interventions. Future longitudinal and interventional studies are needed to clarify underlying mechanisms and to determine whether optimizing school meal composition with specific dietary strategies can meaningfully influence metabolic health at the population level.
This article contains supplementary material online at https://doi.org/10.2337/figshare.33155177.
Acknowledgments
Article Information
Acknowledgments
The authors thank the NHANES team for their extensive work to create, document, and improve the studied variables.
During the course of preparing this work, the authors used ChatGPT for the purpose of grammar and language refinement. Following the use of this tool, the authors formally reviewed the content for accuracy and edited it as necessary. The authors take full responsibility for all the content of this publication.
Duality of Interest
No potential conflicts of interest relevant to this article were reported.
Author Contributions
A.D. conducted the analyses and wrote the original manuscript. S.E. reviewed the analysis code. All authors edited, reviewed, and approved the final version of the manuscript. M.I.G. supervised this work. A.D. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Funding Statement
Funding was provided by the Southern California Center for Chronic Disease Research and Prevention from the National Institute on Minority Health and Health Disparities P50 MD017344. T.L.A. received funding from the National Institutes of Health (R01ES035035, R01ES035056, P50MD017344, and P30ES032756).
Supporting information
References
- 1. Han C, Song Q, Ren Y, Chen X, Jiang X, Hu D.. Global prevalence of prediabetes in children and adolescents: a systematic review and meta-analysis. J Diabetes 2022;14: 434–441 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Ng HY, Chan LTW.. Prediabetes in children and adolescents: an updated review. World J Clin Pediatr 2023;12:263–272 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Centers for Disease Control and Prevention . Prediabetes in US adolescents, 2025. Accessed 20 August 2025. Available from https://gis.cdc.gov/grasp/diabetes/diabetesatlas-spotlight.html
- 4. Van Hulst A, Paradis G, Harnois-Leblanc S, Benedetti A, Drapeau V, Henderson M.. Lowering saturated fat and increasing vegetable and fruit intake may increase insulin sensitivity 2 years later in children with a family history of obesity. J Nutr 2018;148:1838–1844 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Batista CC, Nascimento LM, Lustosa LCRdS, Rodrigues BGM, Campelo V, Frota KdMG. Metabolic syndrome in adolescents and antioxidant nutrient intake: a cross-sectional study. Rev Assoc Med Bras (1992) 2021;67:918–925 [DOI] [PubMed] [Google Scholar]
- 6. Dong Y, Chen L, Gutin B, Zhu H.. Total, insoluble, and soluble dietary fiber intake and insulin resistance and blood pressure in adolescents. Eur J Clin Nutr 2019;73:1172–1178 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Leak TM, Overcash F, Reicks M, Leung CW, Barr-Anderson DJ, Censani M.. Diet and physical activity by prediabetes status among U.S. Adolescents: National Health and Nutrition Examination Survey, 2007–2018. Child Obes 2024;20:543–548 [DOI] [PubMed] [Google Scholar]
- 8. Chen Z, Wu J, Ai K, Bu Z, Niu W, Li M.. Trends in Children’s Dietary Inflammatory Index and association with prediabetes in U.S. adolescents. Nutr Diabetes 2024;14:94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Peprah Osei E. Prevalence and predictors of prediabetes/type 2 diabetes mellitus among adolescents in the United States: NHANES (2021–2023). PLOS Glob Public Health 2026;6:e0005596. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Liu C, Wu S, Pan X.. Clustering of cardio-metabolic risk factors and pre-diabetes among U.S. adolescents. Sci Rep 2021;11:5015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Thota P, Perez-Lopez FR, Benites-Zapata VA, Pasupuleti V, Hernandez AV.. Obesity-related insulin resistance in adolescents: a systematic review and meta-analysis of observational studies. Gynecol Endocrinol 2017;33:179–184 [DOI] [PubMed] [Google Scholar]
- 12. Centers for Disease Control and Prevention . School nutrition, 2024. Accessed 5 March 2026. Available from https://www.cdc.gov/school-nutrition/about/index.html
- 13. Cohen J, Chapman L, Gombi-Vaca M, et al. Research brief: current state of processed foods in schools, 2025. Accessed 5 March 2026. Available from https://www.childnourishlab.org/healthy-school-meals-for-all
- 14. Bronfenbrenner U. The Ecology of Human Development: Experiments by Nature and Design. Cambridge, MA, Harvard University Press, 1981 [Google Scholar]
- 15. Centers for Disease Control and Prevention . NHANES questionnaires, datasets, and related documentation. Accessed 20 August 2025. Available from https://wwwn.cdc.gov/nchs/nhanes/
- 16. ElSayed NA, Aleppo G, Aroda VR, et al. Introduction and methodology: Standards of Care in Diabetes—2023. Diabetes Care 2023;46(Suppl. 1):S1–S4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Blanton CA, Moshfegh AJ, Baer DJ, Kretsch MJ.. The USDA automated multiple-pass method accurately estimates group total energy and nutrient intake. J Nutr 2006;136:2594–2599 [DOI] [PubMed] [Google Scholar]
- 18. Kuczmarski RJ, Ogden CL, Guo SS, et al. 2000 CDC growth charts for the United States: methods and development. Vital Health Stat 11 2002;May:1–190 [PubMed] [Google Scholar]
- 19. Centers for Disease Control and Prevention . NHANES tutorial - weighting module, 2026. Accessed 5 March 2026. Available from https://wwwn.cdc.gov/nchs/nhanes/tutorials/weighting.aspx
- 20. Stekhoven DJ, Bühlmann P.. MissForest—non-parametric missing value imputation for mixed-type data. Bioinformatics 2012;28:112–118 [DOI] [PubMed] [Google Scholar]
- 21. Zou G. A modified Poisson regression approach to prospective studies with binary data. Am J Epidemiol 2004;159:702–706 [DOI] [PubMed] [Google Scholar]
- 22. Harrison C, Peyyety V, Rodriguez Gonzalez A, et al. Prediabetes prevalence by adverse social determinants of health in adolescents. JAMA Netw Open 2024;7:e2416088. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Rothman KJ. No adjustments are needed for multiple comparisons. Epidemiology 1990;1:43–46 [PubMed] [Google Scholar]
- 24. Ciaffi J, Mancarella L, Ripamonti C, et al. Ultra-processed food consumption and systemic inflammatory biomarkers: a scoping review. Nutrients 2025;17:3012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Asbaghi O, Nazarian B, Yousefi M, Anjom-Shoae J, Rasekhi H, Sadeghi O.. Effect of vitamin E intake on glycemic control and insulin resistance in diabetic patients: an updated systematic review and meta-analysis of randomized controlled trials. Nutr J 2023;22:10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Lee S, Kim HK.. α-Tocopherol and γ-tocopherol decrease inflammatory response and insulin resistance during the interaction of adipocytes and macrophages. Nutr Res Pract 2024;18:761–773 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Mohammed H, M Mustafa M.. Vitamin B6 and glucose control: a forgotten role in prediabetics? Narrative review. Libyan J Med Appl Sci 2026:81–92 [Google Scholar]
- 28. Takey M, Giannini DT, Kuschnir MCC, Bloch KV, Szklo M.. Association between polyunsaturated fatty acids intake and insulin resistance in Brazilian adolescents (ERICA Study). Nutrition 2023;111:112051. [DOI] [PubMed] [Google Scholar]
- 29. Ren Y, Sun S, Su Y, Ying C, Luo H.. Effect of fruit on glucose control in diabetes mellitus: a meta-analysis of nineteen randomized controlled trials. Front Endocrinol (Lausanne) 2023;14:1174545. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Sneed NM, Azuero A, Moss J, Goss AM, Morrison SA.. Total added sugar consumption is not significantly associated with risk for prediabetes among U.S. adults: National Health and Nutrition Examination Survey, 2013–2018. PLoS One 2023;18: e0286759. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Centers for Disease Control and Prevention . Brief overview of sample design, nonresponse bias assessment, and analytic guidelines for NHANES August 2021-August 2023, 2024. Accessed 5 March 2026. Available from https://wwwn.cdc.gov/nchs/nhanes/continuousnhanes/OverviewBrief.aspx?Cycle=2021-2023&utm_source=chatgpt.com
- 32. Monzani A, Ricotti R, Caputo M, et al. A Systematic Review of the Association of Skipping Breakfast with Weight and Cardiometabolic Risk Factors in Children and Adolescents. What Should We Better Investigate in the Future? Nutrients 2019;11:387. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Contento IR. Nutrition education: linking research, theory, and practice. Asia Pac J Clin Nutr 2008;17(Suppl. 1):176–179 [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.