Abstract
Background
The association between ultra-processed food (UPF) intake and markers of biological ageing has been scarcely investigated, despite the evident adverse health effects associated with UPF. This study aimed to test the association between UPF intake and biological ageing, and evaluate how much of this association is accounted for by overall diet quality.
Methods
This cross-sectional study assessed 16 055 participants aged 20–79 years (51% women, 46 ± 0.3 years) from the National Health and Nutrition Examination Survey (NHANES) 2003–2010. Dietary UPF intake was assessed using the Nova system. Values were expressed as % of total energy intake and were denominated as a continuous variable and in quintiles. Diet quality was assessed with the American Heart Association 2020 and the Healthy Eating Index 2015. Biological ageing was assessed using the PhenoAge algorithm.
Results
For each 10% of energy intake accounted for by UPF, participants were 0.21 (95%CI 0.16–0.26) years biologically older in terms of PhenoAge. As compared to participants in the lowest UPF quintile (≤39%), those in the highest UPF quintile (68–100%) were 0.86 (95% CI 0.55, 1.16) years older (P-for-trend across quintiles ≤0.001). Adherence to a healthy diet moderately attenuated the relationship between UPF and PhenoAge (adjusted β = 0.14 per 10% increment of UPF).
Conclusions
Adults with higher UPF tended to be biologically older. This association is partly independent of diet quality, suggesting that food processing may contribute to biological ageing acceleration. Our findings point to a compelling reason to target UPF consumption to promote healthier ageing.
Keywords: ultra-processed food, ageing, diets, biological ageing, older people
Key Points
Ultra-processed food (UPF) consumption is associated with signs of accelerated biological ageing.
Adherence to a healthy diet explained only part of the association of UPF intake with older biological age.
Other properties of UPF related to processing may contribute to an acceleration of biological processes of ageing.
Introduction
Life expectancy is increasing globally, with the population above 60 years projected to double from 2015 to 2050, reaching nearly 2.1 billion [1]. The global demographic shift towards an ageing population poses economic and societal challenges [2], in part because gains in lifespan are not being matched by gains in healthy years of life [3]. Understanding how dietary choices impact the ageing process may pave the way to promoting longevity and reducing the burden of age-related illnesses.
The emerging field of geroscience proposes that one strategy to increase healthy years of life or ‘healthspan’ is to intervene in the biological processes of ageing [4]. Biological ageing refers to the gradual accumulation of molecular and cellular damage over time, which leads to the decline in physiological function and increased vulnerability to diseases. The Geroscience Hypothesis is premised on evidence that a number of molecular changes (called ‘hallmarks’) are characteristic of ageing across mammalian species and mediate ageing-related risk for many different chronic diseases [5, 6]. In laboratory animals, interventions to slow or reverse the accumulation of ageing hallmarks can extend healthspan [7]. Translation of these therapies to treat human ageing is a key frontier in geroscience [8, 9]. In parallel, it may be possible to identify environments and behaviours that affect ageing biology, offering more immediate opportunities for intervention [10].
Diet is a key environmental/behavioural pathway with potential to affect healthy ageing [11]. People with healthier dietary patterns tend to exhibit a slower pace of biological ageing as compared to those with less healthy diets [12–15]. Healthy diet indices used in previous research mostly focus on the consumption of whole, plant-based foods, which results in high intake of fibre, and limited consumption of saturated fat, sodium and added sugar [16]. Although the importance of specific nutrients and food groups for health outcomes is indisputable, the impact of food processing on health is emerging as a dietary feature potentially relevant to healthy longevity.
Industrial food processing involves physical, chemical and biological processes used by food manufacturers to alter foods from their natural state before consumption or meal preparation. Based on the extent and purpose of these processes, the Nova classification system categorises foods and beverages into four groups: unprocessed or minimally foods, processed culinary ingredients, processed foods and ultra-processed foods (UPFs) [17]. UPFs are defined as industrial formulations of several ingredients including oils, fats and starch that typically contain cosmetic additives and/or substances of rare culinary use opposed to little (if any) whole foods [17]. UPFs account for >50% of energy intake in the usual diet of US and British populations [18], with a steady increase observed in Asian, African Middle Eastern and Latin-American countries [19]. Multiple studies found strong associations between UPF intake and cardiovascular diseases, type 2 diabetes, obesity, mental disorders, all-cause and heart-disease-related mortality [20]. In this study, we aimed to test if higher UPF consumption was associated with signs of accelerated biological ageing. Furthermore, we assessed how much of this association is accounted for by total energy intake and overall diet quality, assessed with the American Heart Association (AHA) 2020 continuous diet score [21] and the Healthy Eating Index 2015 (HEI-2015) [22, 23]. Our analysis from the US National Health and Nutrition Examination Surveys (NHANES) sheds new light on how dietary intake may contribute to healthy longevity.
Methods
Study population
This cross-sectional study included data from the 2003–2010 NHANES, given that only these sequenced cycles collected all data required for calculation of PhenoAge for ≥18-year-old participants (e.g. C-reactive protein (CRP) is not available for cycles 2011–12, 2013–14). The NHANES protocol was approved by the National Center for Health Statistics (NCHS) Research Ethics Review Board. All the NHANES participants provided informed consent. Given that questions about the history of CVD and diabetes, which are important factors to be accounted for because of potential reverse causation, were only asked to individuals aged 20 years and over, we excluded 18- and 19-year-olds from our analysis. Furthermore, individuals aged 85 and over were topcoded at 85 years of age in cycles 2003–04 and 2005–06; and individuals aged 80 and over were topcoded at 80 years of age in cycles 2007–08 and 2009–10. Thus, we limited our analysis to people aged between 20 and 79 years old with reliable data for at least the first of two 24-hour dietary recalls. We excluded all participants with missing data necessary for the PhenoAge calculation (n = 2096), pregnant women (n = 539) and people with missing data for physical activity, a continuous variable used as covariate (n = 3). From the 18 693 participants within the eligible age range in the 2003–2010 NHANES cycles and reliable data for at least the first of two 24 h dietary recalls, 16 055 also met the remaining eligibility criteria and were included in our analysis (Fig. S1).
Dietary intake
All NHANES participants were eligible for two 24-hour dietary recall interviews: the first in-person at the Mobile Examination Center (MEC) and the second by phone 3–10 days later [24, 25]. We utilised all available dietary intake data for each participant, using means of both recall days when available and one day otherwise.
Based on food code and SR code descriptions, all food items were assigned to one of the four Nova groups: unprocessed/minimally processed foods, processed culinary ingredients, processed foods and UPFs (Supplementary Methods 1) [17, 26]. The contribution (%) of each Nova group to total daily energy intake was calculated for each participant.
The overall diet quality was assessed using the AHA 2020 continuous diet score [21] and the HEI-2015 [22, 23]. The AHA diet score assigns points to ‘beneficial’ dietary components (fruits and vegetables; whole grains; fish and shellfish; nuts, seeds and legumes) while “harmful” components (sugar-sweetened beverages; sodium; processed meat; saturated fat) score inversely. HEI-2015, ranging 0–100, reflects 13 dietary components from 2015–2020 Dietary Guidelines for Americans [22] classified as “adequacy” (e.g. fruits, vegetables, whole grains) or “moderation” (e.g. refined grains, sodium). In both diet quality indices, higher scores represent a higher diet quality.
Biological ageing
In lack of a gold standard for the measurement of biological ageing [27], we used PhenoAge as this method can be applied to routinely collected clinical data with the best validation evidence for prediction of healthspan [28–32]. The PhenoAge was developed by modelling survival probabilities from blood chemistry data among participants in NHANES III [28] using an algorithm that includes chronological age and nine blood analytes. Importantly, prior studies established that PhenoAge is sensitive to lifestyle exposures, including nutrition [13, 33–35], and revealed slowed ageing in response to calorie restriction [36], an intervention established to slow biological ageing [37]. Values of the PhenoAge can be interpreted as the age at which an individual’s mortality risk would match the average in the NHANES III training sample.
We implemented a modified version of PhenoAge algorithm adapted for use in a randomised trial of calorie restriction [36] using the ‘BioAge’ R package (Table S1). We computed PhenoAge gap as the difference between predicted biological age and chronological age. A higher PhenoAge gap value indicates an advanced state of biological ageing and increased risk of diseases and mortality.
Covariates
Demographics (age, sex, ethnicity, education, poverty-income ratio) and smoking status were collected using a Computer-Assisted Personal Interviewing system by trained interviewers. Body measures (height and weight) were collected by trained health technicians. Physical activity was assessed with the Global Physical Activity Questionnaire. History of CVD was noted if participants reported a physician’s diagnosis of heart failure, coronary heart disease, angina, heart attack or stroke. History of diabetes was identified if participants reported a doctor’s diagnosis, use of hypoglycemic agents or insulin, or HbA1c levels ≥6.5% (Hb1Ac was measured in blood with Tosoh Automated Hb1Ac Analyser HLC-723G8). Further details of the covariates are presented in the Supplementary Methods 2.
Statistical analysis
Demographic and clinical characteristics were presented as mean with standard deviation (SD) for continuous variables, or % weighted (SD) for categorical variables. Data were compared across the population-stratified quintiles of the dietary contribution of UPF (% of total energy intake) using P-for-trend for continuous variables or Pearson’s chi-square for categorical variables.
We used regression models to test the association between UPF intake (as both a continuous variable and quintiles of intake) with biological ageing, including a progressively more comprehensive set of covariates in our models to account for potential confounding, reverse causality or test potential mediation. Tests of linear trend were carried out by treating quintiles as a single continuous ordinal variable. The models were adjusted as follows: Model 1: NHANES cycle, chronological age and gender; Model 2: additionally adjusted for ethnicity, education and poverty-income ratio; Model 3: additionally adjusted for physical activity and smoking status; and Model 4: additionally adjusted for body mass index (BMI) and history of CVD and diabetes. These latter covariates in model 4 were included to account for potential reverse causality that could dilute the studied associations towards the null when assessed through cross-sectional studies. We constructed three other models that were adjusted for total energy intake (Model 5), AHA diet score (Model 6) and HEI-15 diet score (Model 7) in addition to the covariates used in Model 4 to test the hypothesis that UPF intake affects biological ageing over and above total energy intake and overall diet quality.
We used a weighed restricted cubic spline in model 4 with five knots (5th, 27.5th, 50th, 72.5th, and 95th) as per Harrell’s recommendations [38] to examine the shape of the dose–response relationship curve between %kcal UPF (as a continuous variable) and PhenoAge gap.
To examine potential differences in the association between UPF and PhenoAge gap by gender, chronological age (continuous), BMI (continuous), smoking status, physical activity (continuous), history of CVD and diabetes, total energy intake and diet quality indices (AHA and HEI-15, continuous), Wald F tests were used to evaluate interaction terms in Model 4 using UPF as a continuous variable. Analyses were stratified according to statistically significant interaction variables. We further performed a sensitivity analysis where we ran all seven models (using UPF as both continuous and quintiles of intake) excluding participants with implausible energy intake (defined as consumption of <800 kcal/d or >4200 kcal/d in men and <500 kcal/d or >3500 kcal/d in women) for either day 1 or 2 as suggested by Banna et al. [39].
Statistical analysis was performed with STATA/SE 16.0 for Windows (StataCorp LLC) using the survey design (svy) and considering the sample weights provided by NHANES for the first day 24-h recall. A P-value at or below 0.05 was considered statistically significant.
Results
Table 1 shows characteristics of the study sample (n = 16 055) and their distribution across quintiles of UPF intake. On average, the participants in this study were 46 years old, and 51% were females. The mean %kcal UPF was 54.3% for the overall study population, ranging from 30.0% in the first quintile of intake (Q1) to 76.7% in the fifth quintile (Q5). Younger, Non-Hispanic White and Non-Hispanic Black, non-college graduates, lower income, current smokers, less-physically active participants tended to be concentrated among the highest quintiles of UPF intake. Overall, 34% of the study population was classified as having obesity, with increasing rates across the UPF intake quintiles. The study population presented similar demographic characteristics when compared with the 2003–2010 NHANES population at the same age bracket who completed at least the first of two 24 h dietary recalls (Table S2).
Table 1.
Characteristics and measured data from NHANES 2003–2010 participants across quintiles of UPFs intake (n = 16 055)
| Characteristics | Quintiles of UPF (% of total energy intake) | ||||||
|---|---|---|---|---|---|---|---|
| All participants | Q1 (0–39.1%) |
Q2 (39.2–49.5%) |
Q3 (49.6–57.9%) |
Q4 (60.0–67.5%) |
Q5 (67.6–100%) |
P-value | |
| NHANES cycle, n (%) | |||||||
| 2003–04 | 3398 (24.2) | 618 (16.2) | 760 (22.0) | 750 (21.9) | 674 (20.5) | 596 (19.4) | |
| 2005–06 | 3446 (24.5) | 678 (19.2) | 760 (21.8) | 703 (20.5) | 699 (20.1) | 606 (18.4) | |
| 2007–08 | 4459 (25.4) | 888 (17.5) | 847 (19.0) | 880 (19.7) | 892 (20.5) | 952 (23.3) | |
| 2009–10 | 4752 (25.9) | 1027 (19.5) | 844 (16.6) | 878 (18.7) | 946 (21.2) | 1057 (24.0) | |
| Chronological age (years)a | 45.6 (0.3) | 47.6 (0.4) | 47.2 (0.4) | 46.6 (0.4) | 45.4 (0.4) | 41.7 (0.4) | <0.001 |
| Female, % (SD)b | 50.7 (0) | 48.2 (1.1) | 51.2 (1.1) | 50.3 (1.0) | 51.6 (1.0) | 52.0 (1.1) | 0.116 |
| Race/Ethnicity, % (SD)b | <0.001 | ||||||
| Mexican American | 8.2 (0.9) | 8.4 (1.0) | 9.6 (1.1) | 9.0 (1.0) | 7.9 (0.9) | 6.2 (1.0) | |
| Non-Hispanic White | 72.0 (1.6) | 63.3 (2.4) | 71.9 (1.8) | 73.4 (1.6) | 74.3 (1.8) | 75.8 (1.8) | |
| Non-Hispanic Black | 10.5 (0.8) | 9.5 (0.8) | 9.1 (0.8) | 10.1 (0.9) | 11.3 (1.1) | 12.5 (1.2) | |
| Other/Multiracial | 9.3 (0.7) | 18.8 (1.7) | 9.4 (0.9) | 7.5 (0.8) | 6.5 (0.7) | 5.5 (0.6) | |
| Education, % (SD)b | <0.001 | ||||||
| Incomplete high school | 17.3 (0.7) | 20.0 (1.0) | 17.2 (1.1) | 16.4 (0.9) | 16.8 (1.1) | 16.5 (0.9) | |
| High school graduate | 24.7 (0.7) | 18.8 (1.1) | 22.5 (1.4) | 24.0 (1.0) | 25.9 (0.9) | 31.2 (1.4) | |
| Incomplete college | 31.3 (0.5) | 28.7 (1.1) | 30.0 (1.3) | 30.3 (1.1) | 34.4 (1.0) | 32.7 (1.1) | |
| College graduate | 26.6 (1.0) | 32.6 (1.4) | 30.2 (1.6) | 29.2 (1.4) | 22.8 (1.3) | 19.5 (1.4) | |
| Missing | 0.1 (0) | 0 | 0.1 (0) | 0.1 (0) | 0.1 (0) | 0.1 (0) | |
| Poverty-income ratio, % (SD)b | <0.001 | ||||||
| <1.30 | 18.7 (0.8) | 18.8 (1.0) | 15.6 (0.8) | 16.6 (0.9) | 18.1 (1.1) | 23.8 (1.6) | |
| ≥1.30 | 75.9 (0.9) | 75.0 (1.4) | 78.8 (0.9) | 78.6 (1.1) | 75.8 (1.2) | 71.6 (1.5) | |
| Missing | 5.4 (0.4) | 6.2 (0.7) | 5.6 (0.5) | 4.8 (0.5) | 6.1 (0.7) | 4.6 (0.8) | |
| BMI category, % (SD)b | <0.001 | ||||||
| ≤24.9 kg/m2 | 31.2 (0.7) | 35.1 (1.4) | 31.4 (1.2) | 29.0 (1.0) | 30.1 (1.2) | 30.9 (1.4) | |
| 25–29.9 kg/m2 | 32.9 (0.6) | 34.0 (1.2) | 32.6 (1.1) | 35.5 (0.9) | 32.5 (1.2) | 30.2 (1.5) | |
| >30 kg/m2 | 34.2 (0.7) | 29.1 (1.2) | 34.1 (1.4) | 33.8 (1.3) | 35.1 (1.1) | 37.9 (1.2) | |
| Missing | 1.7 (0.1) | 1.8 (0.4) | 1.9 (0.3) | 1.7 (0.3) | 2.3 (0.3) | 1.0 (0.2) | |
| Smoking status, % (SD)b | <0.001 | ||||||
| Current smoker | 24.5 (0.6) | 21.6 (1.2) | 22.1 (1.2) | 21.9 (1.2) | 25.7 (1.1) | 30.4 (1.1) | |
| Non-smoker | 50.8 (0.8) | 51.2 (1.7) | 51.6 (1.3) | 52.8 (1.5) | 50.2 (1.4) | 48.3 (1.3) | |
| Former smoker | 24.7 (0.6) | 27.2 (1.3) | 26.2 (1.1) | 25.2 (1.4) | 24.1 (1.1) | 21.3 (0.9) | |
| Missing | 0.01 (0.01) | 0.02 (0.02) | 0.01 (0.01) | 0.04 (0.04) | 0.04 (0.04) | 0 | |
| Physical activity (z-score, min/day)a | 0.09 (0.01) | 0.21 (0.03) | 0.10 (0.02) | 0.09 (0.03) | 0.05 (0.03) | 0.03 (0.03) | <0.001 |
| History of CVD, % (SD)b | 7.3 (0.4) | 7.9 (0.7) | 7.3 (0.5) | 7.2 (0.6) | 7.2 (0.6) | 7.0 (0.7) | 0.964 |
| Missing | 0.4 (0.01) | 0.4 (0) | 0.3 (0.1) | 0.3 (0.1) | 0.4 (0.2) | 0.4 (0.1) | |
| History of diabetes, % (SD)b | 9.2 (0.3) | 9.1 (0.7) | 10.2 (0.6) | 9.9 (0.7) | 8.9 (0.5) | 8.0 (0.6) | 0.076 |
| Total energy intake (kcal/day)a | 2160 (11) | 2052 (24) | 2159 (22) | 2163 (20) | 2201 (22) | 2211 (20) | <0.001 |
| HEI scorea | 52.7 (0.3) | 60.4 (0.4) | 56.8 (0.3) | 53.4 (0.3) | 49.8 (0.3) | 44.7 (0.3) | <0.001 |
| AHA scorea | 35.4 (0.3) | 42.5 (0.4) | 38.5 (0.3) | 35.9 (0.4) | 32.4 (0.3) | 29.0 (0.3) | <0.001 |
| PhenoAge gap (years)a | −5.80 (0.07) | −6.50 (0.10) | −6.03 (0.11) | −5.88 (0.10) | −5.54 (0.10) | −5.19 (0.11) | <0.001 |
| Dietary %kcal UPFa | 54.3 (53.6, 55.1) | 30.0 (0.2) | 44.5 (0.0) | 53.7 (0.0) | 62.5 (0.0) | 76.7 (0.0) | <0.001 |
BMI, body mass index; HEI, Healthy Eating Index; AHA, American Heart Association index; UPF, ultra-processed food.
Values presented as mean (SD), unless stated otherwise. Comparisons across quintiles were performed using.
aLinear regression (P-for-trend).
bPearson chi-square.
There was a significant linear decrease in diet quality (assessed as HEI-15 and AHA scores) across the UPF intake quintiles. On the other hand, there was a significant linear increase in total energy intake and PhenoAge gap across the UPF intake quintiles. No differences regarding gender, history of CVD or diabetes were observed across UPF intake quintiles (Table 1).
The analysis revealed little evidence of non-linearity in the restricted cubic spline model (coefficient for linear term = 0.02 (95% CI 0.00, 0.04); Wald test for linear term P-value = 0.046; Wald test for all non-linear terms P-value = 0.550). Quintiles of UPF intake were positively associated with PhenoAge gap in all the adjusted models (P-for-trend ≤ 0.001), indicating accelerated biological ageing as UPF intake increased (Table 2). Based on the fully adjusted model (model 4), we observed that the highest UPF quintile presented a 0.9 (95% CI 0.60, 1.20) higher absolute PhenoAge gap compared to the lowest quintile, while a 0.21 (95% CI 0.16, 0.26) increase in PhenoAge gap was observed for every 10% increase in UPF intake (Table 2).
Table 2.
Association between UPF intake (% of total energy intake) and PhenoAge gap and in US adults from NHANES 2003–2010 (n = 16 055)
| Quintiles of UPF (% of total energy intake) | UPF (% of total energy intake) | |||||||
|---|---|---|---|---|---|---|---|---|
| β (95% CI) | P-trend | β (95% CI)a | P-value | |||||
| Q1 | Q2 | Q3 | Q4 | Q5 | ||||
| Model 1 | - | 0.48 (0.23, 0.73) | 0.63 (0.33, 0.90) | 1.03 (0.76, 1.30) | 1.44 (1.14, 1.75) | <0.001 | 0.33 (0.27, 0.38) | <0.001 |
| Model 2 | - | 0.47 (0.23, 0.72) | 0.60 (0.34, 0.86) | 0.90 (0.64, 1.16) | 1.21 (0.91, 1.51) | <0.001 | 0.28 (0.22, 0.33) | <0.001 |
| Model 3 | - | 0.46 (0.22, 0.70) | 0.60 (0.34, 0.85) | 0.86 (0.58, 1.13) | 1.14 (0.84, 1.44) | <0.001 | 0.26 (0.20, 0.31) | <0.001 |
| Model 4 | 0.32 (0.07, 0.56) | 0.44 (0.21, 0.67) | 0.71 (0.44, 0.98) | 0.90 (0.60, 1.20) | <0.001 | 0.21 (0.16, 0.26) | <0.001 | |
| Model 5 | - | 0.31 (0.07, 0.56) | 0.44 (0.21, 0.67) | 0.70 (0.43, 0.98) | 0.90 (0.60, 1.20) | <0.001 | 0.21 (0.16, 0.26) | <0.001 |
| Model 6 | - | 0.21 (−0.04, 0.46) | 0.26 (0.03, 0.50) | 0.45 (0.18, 0.71) | 0.57 (0.26, 0.88) | <0.001 | 0.14 (0.09, 0.20) | <0.001 |
| Model 7 | - | 0.24 (−0.01, 0.49) | 0.28 (0.05, 0.52) | 0.48 (0.20, 0.75) | 0.57 (0.24, 0.89) | <0.001 | 0.14 (0.09, 0.20) | <0.001 |
Model 1: adjusted for NHANES cycle, chronological age and gender
Model 2: adjusted for NHANES cycle, chronological age, gender, ethnicity, education and poverty-income ratio
Model 3: adjusted for NHANES cycle, chronological age, gender, ethnicity, education, poverty-income ratio, physical activity and smoking status
Model 4: adjusted for NHANES cycle, chronological age, gender, ethnicity, education, poverty-income ratio, physical activity, smoking status, BMI, history of CVD and diabetes
Model 5: adjusted for NHANES cycle, chronological age, gender, ethnicity, education, poverty-income ratio, physical activity, smoking status, BMI, history of CVD and diabetes, and energy intake
Model 6: adjusted for NHANES cycle, chronological age, gender, ethnicity, education, poverty-income ratio, physical activity, smoking status, BMI, history of CVD and diabetes, and AHA diet score
Model 7: adjusted for NHANES cycle, chronological age, gender, ethnicity, education, poverty-income ratio, physical activity, smoking status, BMI, history of CVD and diabetes, and HEI-15 diet score.
aPresented for 10% increase in UPF intake (as % kcal).
We further identified that adherence to a healthy diet, as indicated by higher AHA and HEI-15 scores, was associated with lower PhenoAge gap, while total energy intake had no association with PhenoAge gap (Table S3). Further adjustment for energy intake (model 5) resulted in no differences in the effect sizes between UPF and PhenoAge gap. Adherence to a healthy diet, assessed by both AHA and HEI-15 scores, attenuated the relationship between UPF and PhenoAge gap (from 0.21 to 0.14 years, models 6 and 7), although the association remained significant after adjusting for each of these two indicators of diet quality (Table 2).
Our analysis revealed no interaction of UPF intake with gender, BMI, physical activity, smoking status, history of CVD, diabetes, total energy intake and the two diet quality scores (HEI-15 and AHA) (Table S4). However, we observed an interaction between UPF intake and chronological age. Our stratified analysis by age group (20–39, 40–59 and ≥60 years) indicates a somewhat stronger association between UPF intake and PhenoAge gap among the ≥60-year-old group (Table 3). We further performed a sensitivity analysis excluding those with potential implausible energy intake. The results were based on data from 13 881 participants and did not meaningfully differ from those observed in the main analyses (Table S5).
Table 3.
Association between quintiles of UPF intake (% of total energy intake) and PhenoAge gap by age group in US adults from NHANES 2003–2010 (n = 16 055)
| Quintiles of UPF (% of total energy intake) | UPF (% of total energy intake) | |||||||
|---|---|---|---|---|---|---|---|---|
| β (95% CI) | P-trend | β (95% CI)a | P-value | |||||
| Q1 | Q2 | Q3 | Q4 | Q5 | ||||
| 20–39 years | - | 0.46 (0.05, 0.88) | 0.60 (0.21, 0.99) | 0.69 (0.22, 1.16) | 1.00 (0.55, 1.46) | <0.001 | 0.20 (0.12, 0.28) | <0.001 |
| 40–59 years | - | 0.02 (−0.40, 0.43) | 0.18 (−0.17, 0.52) | 0.52 (0.14, 0.90) | 0.64 (0.22, 1.07) | <0.001 | 0.19 (0.12, 0.27) | <0.001 |
| ≥ 60 years | - | 0.64 (0.19, 1.10) | 0.70 (0.25, 1.14) | 0.97 (0.54, 1.40) | 1.00 (0.52, 1.48) | <0.001 | 0.23 (0.14, 0.33) | <0.001 |
Models adjusted for NHANES cycle, chronological age, gender, ethnicity, education, poverty-income ratio, physical activity, smoking status, BMI, history of CVD and diabetes
20–39 years: n = 5632; 40–59 years: n = 5620; ≥ 60 years: n = 4803.
aPresented for 10% increase in UPF intake (as % kcal).
Discussion
In this analysis of a representative sample of US adults, we identified that UPF intake is associated with older biological age assessed by PhenoAge. Although lower diet quality, as assessed by standard indices, may partly explain the association between UPF intake and biological ageing, other mechanisms associated with food processing are also likely involved. Our results support earlier research linking UPF consumption to ageing markers such as telomere length [40], frailty [41], cognitive decline [42] and dementia [43]. Given the global demographic shift towards an ageing population, demonstrating UPF's adverse effects reinforces the need for dietary-focused public health strategies to prolong a healthy life span.
This analysis showed that for every 10% increase in UPF consumption, there is a 2.4-month increase in the gap between biological and chronological age, evidence that UPF intake may accelerate biological ageing. If we assume a standard diet of 2000 kcal/day, adding an extra 200 calories of UPF daily (roughly equivalent to an 80 g serving of chicken bites or a small chocolate bar) could lead to the biological ageing process advancing by >2 months compared to chronological ageing. Despite a seemingly small effect size, it holds significant public health implications. Based on previous research on PhenoAge's links to morbidity, disability and mortality [32], this rise predicts nearly 2% more mortality, 0.75% more incident disability and 0.5% more incident chronic disease over 2 years. The association between UPF and PhenoAge gap was to some extent stronger in older adults, which highlights the potential role of nutrition in promoting healthy ageing and preventing age-related diseases.
The association between UPF intake and biological ageing remained significant after adjusting for diet quality and total energy intake, suggesting that other factors such as lower flavonoid or phytoestrogen intake [44, 45], or higher content of package materials such as bisphenol or phthalates [46] or compounds formed during processing such as acrylamides [47] might also explain the association [16]. Previous research indicates that caloric restriction slows biological ageing [36] and increases healthy lifespan [48, 49]. While we found an association between UPF and energy intake (as reported in previous research [50], our findings indicate caloric intake does not affect the link between UPF intake and biological ageing.
Our results extend an earlier observation in a Spanish cohort linking high UPF intake with shorter leukocyte telomere length, a candidate biomarker of ageing [40]. Several dietary guidelines advocate for limiting UPF intake [51] due to their connection with over 30 adverse health outcomes [20], and our findings provide another reason to target UPF consumption to promote healthier ageing.
We acknowledge limitations. There is no gold standard for assessing biological ageing. We focused on PhenoAge based on its predictive value for ageing-related health outcomes and sensitivity to nutritional exposure [29, 30, 32, 52]. Dietary intake was assessed using self-reported 24-hour recalls, the least biased self-report instrument available [53]. We employed a standardised system to classify foods, but NHANES data may lack brand-specific details crucial for accurate classification, as it focuses on nutrient concentrations rather than food processing [26, 54]. Social desirability bias may lead to UPF underreporting, while food intake estimates may not reflect the usual diet, potentially biasing associations towards null. Furthermore, despite standardised NHANES methods [55–57], previous research indicates that over 25% of energy intake may be misreported [58]. However, our sensitivity analysis revealed no impact of misreporting on results, as effect sizes remained similar when excluding individuals with extreme energy intake. Our study focused on ages 20–79 due to NHANES design, limiting findings' extrapolation to younger and older populations. The cross-sectional design of this study limits causal inference due to the lack of event temporality and residual confounding. Thus longitudinal studies tracking diet and lifestyle changes are needed to solidify our findings. Additionally, given prior research linking UPF intake to mortality, investigating biological ageing as a potential mediator of the association is recommended.
Conclusions
Intake of UPF was associated with older biological age in US adults aged 20 to 79 years. Adherence to a healthy diet explained only part of the association of UPF intake with older biological age, suggesting that other properties of UPF related to processing may contribute to an acceleration of biological processes of ageing.
Supplementary Material
Acknowledgements
The authors thank Dianne Cook for providing the initial support for PhenoAge calculation.
Contributor Information
Barbara Rita Cardoso, Department of Nutrition, Dietetics and Food, Monash University, Notting Hill, VIC 3168, Australia; Victorian Heart Institute, Monash University, Clayton, VIC 3168, Australia.
Junxiu Liu, Department of Population Health Science and Policy, Icahn School of Medicine at Mount Sinai, New York, NY, USA.
Priscila Machado, Institute for Physical Activity and Nutrition, Deakin University, Geelong, VIC, Australia.
Dayoon Kwon, Fielding School of Public Health, University of California Los Angeles, Los Angeles, CA, USA.
Daniel W Belsky, Robert N Butler Columbia Aging Center, Columbia University Mailman School of Public Health, New York, NY, USA.
Euridice Martinez Steele, Center for Epidemiological Studies in Health and Nutrition, University of Sao Paulo, Sao Paulo, Brazil.
Declaration of Conflicts of Interest
DWB is the consulting CSO and SAB chair of BellSant and a SAB member of the Hooke Clinic.
Declaration of Sources of Funding
E.M.S. received funding from Fundacão de Amparo à Pesquisa do Estado de São Paulo (Processo n° 2023/16144-3). P.M is funded by a Deakin University Postdoctoral Research Fellowship.
Data availability
Code will be made available on request by contacting the corresponding author.
References
- 1. United Nations, Department of Economic and Social Affairs, Population Division . World Population Prospects 2022: Ten Key Messages. 2022, New York: United Nations. 10.18356/9789210014380. [DOI] [Google Scholar]
- 2. Scott AJ. The longevity economy. Lancet Healthy Longev. 2021;2:e828–35. [DOI] [PubMed] [Google Scholar]
- 3. World Health Organization . Progress Report on the United Nations Decade of Healthy Ageing, 2021–2023. Geneva: World Health Organization, 2023. [Google Scholar]
- 4. Kennedy BK, Berger SL, Brunet Aet al. Geroscience: linking aging to chronic disease. Cell. 2014;159:709–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. López-Otín C, Blasco MA, Partridge Let al. The hallmarks of aging. Cell. 2013;153:1194–217. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. López-Otín C, Blasco MA, Partridge Let al. Hallmarks of aging: an expanding universe. Cell. 2023;186:243–78. [DOI] [PubMed] [Google Scholar]
- 7. Campisi J, Kapahi P, Lithgow GJet al. From discoveries in ageing research to therapeutics for healthy ageing. Nature. 2019;571:183–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Kaeberlein M, Rabinovitch PS, Martin GM. Healthy aging: the ultimate preventative medicine. Science. 2015;350:1191–3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Barzilai N, Cuervo AM, Austad S. Aging as a biological target for prevention and therapy. JAMA. 2018;320:1321–2. [DOI] [PubMed] [Google Scholar]
- 10. Belsky DW, Baccarelli AA. To promote healthy aging, focus on the environment. Nat Aging. 2023;3:1334–44. [DOI] [PubMed] [Google Scholar]
- 11. Al-Naggar IM, Newman JC, Kuchel GA. Healthy eating patterns: a stealthy geroscience-guided approach to enhancing the human healthspan. J Nutr Health Aging. 2023;27:238–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Wang S, Li W, Li Set al. Association between plant-based dietary pattern and biological aging trajectory in a large prospective cohort. BMC Med. 2023;21:310. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Thomas A, Belsky DW, Gu Y. Healthy lifestyle behaviors and biological aging in the U.S. National Health and Nutrition Examination Surveys 1999-2018. J Gerontol A Biol Sci Med Sci. 2023;78:1535–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Thomas A, Ryan CP, Caspi Aet al. Diet, pace of biological aging, and risk of dementia in the Framingham Heart Study. Ann Neurol. 2024;95:1069–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Gensous N, Garagnani P, Santoro Aet al. One-year Mediterranean diet promotes epigenetic rejuvenation with country- and sex-specific effects: a pilot study from the NU-AGE project. Geroscience. 2020;42:687–701. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Dicken SJ, Batterham RL. The role of diet quality in mediating the association between ultra-processed food intake, obesity and health-related outcomes: a review of prospective cohort studies. Nutrients. 2022;14:23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Monteiro CA, Cannon G, Levy RBet al. Ultra-processed foods: what they are and how to identify them. Public Health Nutr. 2019;22:936–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Srour B, Kordahi MC, Bonazzi Eet al. Ultra-processed foods and human health: from epidemiological evidence to mechanistic insights. Lancet Gastroenterol Hepatol. 2022;7:1128–40. [DOI] [PubMed] [Google Scholar]
- 19. Baker P, Machado P, Santos Tet al. Ultra-processed foods and the nutrition transition: global, regional and national trends, food systems transformations and political economy drivers. Obes Rev. 2020;21:e13126. [DOI] [PubMed] [Google Scholar]
- 20. Lane MM, Gamage E, Du Set al. Ultra-processed food exposure and adverse health outcomes: umbrella review of epidemiological meta-analyses. BMJ. 2024;384:e077310. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Lloyd-Jones DM, Hong Y, Labarthe Det al. Defining and setting national goals for cardiovascular health promotion and disease reduction: the American Heart Association's strategic Impact Goal through 2020 and beyond. Circulation. 2010;121:586–613. [DOI] [PubMed] [Google Scholar]
- 22. Krebs-Smith SM, Pannucci TE, Subar AFet al. Update of the Healthy Eating Index: HEI-2015. J Acad Nutr Diet. 2018;118:1591–602. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Reedy J, Lerman JL, Krebs-Smith SMet al. Evaluation of the Healthy Eating Index-2015. J Acad Nutr Diet. 2018;118:1622–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Centers for Disease Control and Prevention (CDC). National Center for Health Statistics (NCHS) . National Health and Nutrition Examination Survey Questionnaire. MEC In-Person Dietary Interviewers Procedures Manual. Hyattsville, MD: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention. March, 2010.
- 25. Centers for Disease Control and Prevention (CDC). National Center for Health Statistics (NCHS) . National Health and Nutrition Examination Survey (NHANES). Phone Follow-Up Dietary Interviewer Procedures Manual. Hyattsville, MD: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention. March, 2001.
- 26. Steele EM, O’Connor LE, Juul Fet al. Identifying and estimating ultraprocessed food intake in the US NHANES according to the Nova classification system of food processing. J Nutr. 2023;153:225–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Ferrucci L, Gonzalez-Freire M, Fabbri Eet al. Measuring biological aging in humans: a quest. Aging Cell. 2020;19:e13080. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Levine ME, Lu AT, Quach Aet al. An epigenetic biomarker of aging for lifespan and healthspan. Aging (Albany NY). 2018;10:573–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Liu Z, Kuo PL, Horvath Set al. A new aging measure captures morbidity and mortality risk across diverse subpopulations from NHANES IV: a cohort study. PLoS Med. 2018;15:e1002718. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Hastings WJ, Shalev I, Belsky DW. Comparability of biological aging measures in the National Health and Nutrition Examination Study, 1999-2002. Psychoneuroendocrinology. 2019;106:171–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Parker DC, Bartlett BN, Cohen HJet al. Association of blood chemistry quantifications of biological aging with disability and mortality in older adults. J Gerontol A Biol Sci Med Sci. 2019;75:1671–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Graf GH, Crowe CL, Kothari Met al. Testing black-white disparities in biological aging among older adults in the United States: analysis of DNA-methylation and blood-chemistry methods. Am J Epidemiol. 2022;191:613–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Graf GH, Zhang Y, Domingue BWet al. Social mobility and biological aging among older adults in the United States. PNAS Nexus. 2022;1:1–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Cao X, Zhang J, Ma Cet al. Life course traumas and cardiovascular disease-the mediating role of accelerated aging. Ann N Y Acad Sci. 2022;1515:208–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Yang G, Cao X, Li Xet al. Association of unhealthy lifestyle and childhood adversity with acceleration of aging among UK biobank participants. JAMA Netw Open. 2022;5:e2230690. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Kwon D, Belsky DW. A toolkit for quantification of biological age from blood chemistry and organ function test data: BioAge. Geroscience. 2021;43:2795–808. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Flanagan EW, Most J, Mey JTet al. Calorie restriction and aging in humans. Annu Rev Nutr. 2020;40:105–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Harrell F. Regression Modeling Strategies: With Applications to Linear Models, Logistic Regression, and Survival Analysis. New York: Springer, 2001. 10.1007/978-1-4757-3462-1. [DOI] [Google Scholar]
- 39. Banna JC, McCrory MA, Fialkowski MKet al. Examining plausibility of self-reported energy intake data: considerations for method selection. Front Nutr. 2017;4:45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Alonso-Pedrero L, Ojeda-Rodríguez A, Martínez-González MAet al. Ultra-processed food consumption and the risk of short telomeres in an elderly population of the Seguimiento Universidad de Navarra (SUN) project. Am J Clin Nutr. 2020;111:1259–66. [DOI] [PubMed] [Google Scholar]
- 41. Sandoval-Insausti H, Blanco-Rojo R, Graciani Aet al. Ultra-processed food consumption and incident frailty: a prospective cohort study of older adults. J Gerontol A Biol Sci Med Sci. 2020;75:1126–33. [DOI] [PubMed] [Google Scholar]
- 42. Cardoso BR, Machado P, Steele EM. Association between ultra-processed food consumption and cognitive performance in US older adults: a cross-sectional analysis of the NHANES 2011-2014. Eur J Nutr. 2022;61:3975–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Li H, Li S, Yang Het al. Association of ultraprocessed food consumption with risk of dementia: a prospective cohort study. Neurology. 2022;99:e1056–66. [DOI] [PubMed] [Google Scholar]
- 44. Leitão AE, Roschel H, Oliveira-Júnior Get al. Association between ultra-processed food and flavonoid intakes in a nationally representative sample of the US population. Br J Nutr. 2024;131:1074–83. [DOI] [PubMed] [Google Scholar]
- 45. Martínez, Steele E, Monteiro CA. Association between dietary share of ultra-processed foods and urinary concentrations of phytoestrogens in the US. Nutrients. 2017;9:1–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. Martínez Steele E, Khandpur N, Costa Louzada MLet al. Association between dietary contribution of ultra-processed foods and urinary concentrations of phthalates and bisphenol in a nationally representative sample of the US population aged 6 years and older. PloS One. 2020;15:1–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47. Martínez Steele E, Buckley JP, Monteiro CA. Ultra-processed food consumption and exposure to acrylamide in a nationally representative sample of the US population aged 6 years and older. Prev Med. 2023;174:107598. [DOI] [PubMed] [Google Scholar]
- 48. Waziry R, Ryan CP, Corcoran DLet al. Effect of long-term caloric restriction on DNA methylation measures of biological aging in healthy adults from the CALERIE trial. Nature Aging. 2023;3:248–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49. Kraus WE, Bhapkar M, Huffman KMet al. 2 years of calorie restriction and cardiometabolic risk (CALERIE): exploratory outcomes of a multicentre, phase 2, randomised controlled trial. Lancet Diabetes Endocrinol. 2019;7:673–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50. Hall KD, Ayuketah A, Brychta Ret al. Ultra-processed diets cause excess calorie intake and weight gain: an inpatient randomized controlled trial of ad libitum food intake. Cell Metab. 2019;30:67–77.e3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51. Koios D, Machado P, Lacy-Nichols J. Representations of ultra-processed foods: a global analysis of how dietary guidelines refer to levels of food processing. Int J Health Policy Manag. 2022;11:2588–99. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52. Belsky DW, Moffitt TE, Cohen AAet al. Eleven telomere, epigenetic clock, and biomarker-composite quantifications of biological aging: do they measure the same thing? Am J Epidemiol. 2018;187:1220–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53. Prentice RL, Mossavar-Rahmani Y, Huang Yet al. Evaluation and comparison of food records, recalls, and frequencies for energy and protein assessment by using recovery biomarkers. Am J Epidemiol. 2011;174:591–603. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54. Slining MM, Yoon EF, Davis Jet al. An approach to monitor food and nutrition from “factory to fork”. J Acad Nutr Diet. 2015;115:40–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55. Blanton CA, Moshfegh AJ, Baer DJet al. The USDA automated multiple-pass method accurately estimates group total energy and nutrient intake. J Nutr. 2006;136:2594–9. [DOI] [PubMed] [Google Scholar]
- 56. Moshfegh AJ, Rhodes DG, Baer DJet al. The US Department of Agriculture Automated Multiple-Pass Method reduces bias in the collection of energy intakes. Am J Clin Nutr. 2008;88:324–32. [DOI] [PubMed] [Google Scholar]
- 57. Rumpler WV, Kramer M, Rhodes DGet al. Identifying sources of reporting error using measured food intake. Eur J Clin Nutr. 2008;62:544–52. [DOI] [PubMed] [Google Scholar]
- 58. Murakami K, Livingstone MBE. Prevalence and characteristics of misreporting of energy intake in US adults: NHANES 2003–2012. Br J Nutr. 2015;114:1294–303. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Code will be made available on request by contacting the corresponding author.