Fast food, fast aging? A cross-sectional study in the UK

Abstract

Background

The rapid expansion of takeaway food consumption has raised concerns about its negative health effects, yet its relationship with biological aging remains unclear. This study aims to investigate the association between takeaway meal consumption and biological aging using data from the UK Biobank.

Methods

We performed a cross-sectional analysis from 43,478 participants in the UK Biobank. Takeaway meal consumption was assessed using the UK Biobank’s ‘Type of Meals Eaten’ survey question. Biological age was estimated using two validated metrics: Phenotypic Age (PhenoAge) and the Klemera-Doubal Method (KDM), both derived from clinical biomarkers. Generalized linear models assessed the association between takeaway consumption and biological aging, adjusting for demographic, socioeconomic, lifestyle, and health-related factors. Mediation analysis evaluated the potential mediating role of body mass index (BMI).

Results

The overall sample had a mean age of 56.01 years (standard deviation = 8.18), with 45.1% being male. Takeaway consumers were more likely to be younger, male, and less educated, with higher social deprivation, overweight prevalence, and unhealthy behaviors (smoking, alcohol use), lower physical activity, yet lower rates of hypertension and coronary heart disease compared to home-cooked meal consumers (reference group). Takeaway food consumption was significantly associated with accelerated biological aging, corresponding to 0.302 years (95% CI: 0.139–0.464, p < 0.001; ~3.6 months) and 0.240 years (95% CI: 0.114–0.366, p < 0.001; ~2.9 months) of additional aging based on PhenoAge and KDM, respectively, compared to home-cooked meals. BMI partially mediated this relationship, accounting for 30 to 37% of the total effect.

Conclusions

Takeaway meal consumption is linked to significant biological age acceleration, partly through increased BMI. The findings underscore the potential public health implications of dietary habits and support interventions targeting food quality and access in takeaway-dense environments.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-026-27534-7.

Background

Consumption of takeaway meals is associated with accelerated biological aging partially through increased body mass index in UK adults. The rapid proliferation of takeaway services in recent years, especially driven by mobile app technology, has drastically increased the convenience of food acquisition [1]. However, this trend has raised considerable public health concerns. Previous studies have shown that takeaway consumption is often associated with dietary patterns characterized by high energy density and low nutritional quality, potentially increasing the risk of obesity and chronic diseases [2–4]. Despite a growing body of literature on the relationship between diet and health, there is a lack of empirical research exploring the potential link between takeaway food consumption and biological aging.

Biological aging refers to the progressive decline in physiological function and increased vulnerability to disease and death, which may not align with chronological age [5]. Unlike chronological age, biological age captures the cumulative impact of genetics, lifestyle, and environmental exposures on an individual’s health trajectory. It serves as a more sensitive predictor of morbidity and mortality than chronological age alone. In recent years, various measures of biological aging have been developed. Among them, Phenotypic Age (PhenoAge), derived from a composite of clinical biomarkers (e.g., albumin, glucose, C-reactive protein), estimates biological age based on predicted mortality risk. The Klemera-Doubal Method (KDM), on the other hand, uses statistical modeling to integrate multiple physiological parameters to estimate the deviation from expected aging patterns. Biological aging has been linked to modifiable lifestyle factors such as smoking, physical inactivity, and poor diet. Takeaway food consumption, often associated with poor dietary quality and high calorie intake [6], may accelerate biological aging by increasing risks for obesity, insulin resistance, and systemic inflammation [7]. While these mechanisms have been hypothesized, empirical evidence linking takeaway consumption directly to biological aging remains limited.

This study aims to investigate the association between takeaway consumption and biological aging using data from the UK Biobank. Using PhenoAge and KDM as measures of biological age, we assess whether takeaway consumption is linked to accelerated biological aging. Additionally, we explore whether body mass index (BMI) mediates the relationship between takeaway food consumption and biological aging, in order to better understand potential biological pathways underlying this association. The findings may contribute to evidence-based dietary recommendations and targeted public health strategies, particularly in urban populations where takeaway consumption is prevalent.

Methods

Study population

This cross-sectional study utilized data from the UK Biobank, a large-scale prospective cohort including over 500,000 participants recruited between 2006 and 2010. All participants completed detailed baseline assessments, including touchscreen questionnaires, in-person interviews, physical examinations, and biological sample collection. These assessments provided comprehensive data on dietary habits—including takeaway food consumption—demographics, lifestyle factors (e.g., smoking, alcohol intake, physical activity), and medical history. Biological aging was assessed using validated metrics derived from blood-based biomarkers. After excluding individuals with incomplete dietary data or missing covariates, a complete case records of 43,478 participants were used for the final analysis.

The UK Biobank study received ethical approval from the North West Multi-centre Research Ethics Committee. The present analysis was conducted using de-identified data under approved secondary use, and therefore did not require additional ethical review.

Dietary assessment

Dietary information was obtained from the “Type of Meals Eaten” survey, administered during the online follow-up, based on a 24-hour dietary recall of the previous day. This questionnaire consisted of five binary (yes/no) items, asking participants whether they typically consumed any of the following types of meals: (1)takeaway meals, (2) restaurant meals, (3)sandwiches, (4)ready meals, and (5) home-cooked meals. This classification allowed for the comparison of different food preparation methods in relation to health outcomes.

Since the questionnaire allowed for multiple responses, establishing mutually exclusive categories was necessary to avoid ambiguity while comparing distinct dietary patterns. Participants were thus categorized into three mutually exclusive groups: (1) Takeaway meals (participants reporting takeaway consumption, regardless of other concurrent selections); (2) Home-cooked meals (participants who exclusively reported consuming home-cooked meals, with no other prepared meals selected), which served as the reference group; and (3) Other meals (participants who did not select takeaway but reported consuming restaurant meals, ready meals, or sandwiches).”

Covariates

Covariates included demographic, socioeconomic, lifestyle, and health-related variables. Demographic variables comprised age, sex, ethnicity, and educational attainment. Socioeconomic status was assessed using the Townsend Deprivation Index, categorized into tertiles to represent different levels of material deprivation.

Lifestyle factors included BMI, smoking status, alcohol intake, and physical activity. BMI was calculated as weight in kilograms divided by the square of height in meters (kg/m²) based on measured anthropometric data. Physical activity was quantified using self-reported frequency, duration, and intensity of walking, moderate, and vigorous activities, converted into metabolic equivalent task (MET) minutes per week. For analysis, physical activity was dichotomized at a threshold of 600 MET-minutes per week, consistent with World Health Organization guidelines [8].

Health-related variables included medical history of hypertension, type 2 diabetes, coronary heart disease (CHD), and stroke. These variables were included to control for potential confounding effects on the association between dietary patterns and biological aging.

Outcome variables

Biological aging was assessed using PhenoAge and KDM. PhenoAge is derived from age and nine clinical biomarkers, with values above chronological age indicating accelerated aging [9]. KDM calculates biological age through weighted regression of multiple biomarkers, reflecting deviation from chronological age [10]. Both measures were standardized and calculated from clinical blood and physiological data. All biological samples were collected following standardized procedures and securely stored for analysis.

In this study, PhenoAge and KDM were calculated by the authors using raw biomarker data from the UK Biobank, applying the algorithms validated and documented in the aforementioned literature [9, 10]. Briefly, PhenoAge estimates the biological age at which an individual’s mortality risk matches that of the general population of the same chronological age. A PhenoAge exceeding chronological age indicates accelerated aging and increased risk of morbidity and mortality, while a lower PhenoAge suggests delayed aging. It is calculated using chronological age and nine clinical biomarkers reflecting multiple physiological systems: albumin (liver function), creatinine (kidney function), glucose (metabolic status), C-reactive protein (inflammation), lymphocyte percentage, mean cell volume, red cell distribution width (immune function), alkaline phosphatase (liver function), and white blood cell count (immune function). KDM biological age is derived from a weighted regression model integrating a set of biomarkers indicative of multi-system physiological aging. The biomarkers included in the KDM calculation for this study were: forced expiratory volume in one second (FEV₁, lung function), systolic blood pressure (SBP), total cholesterol, glycated hemoglobin (HbA1c, glucose metabolism), albumin, creatinine, natural log-transformed C-reactive protein (lnCRP, inflammation), alkaline phosphatase (ALP, liver function), and blood urea nitrogen (BUN, kidney function). This combination captures diverse aspects of biological aging and quantifies deviation from expected chronological age.

Statistical analysis

All statistical analyses were performed using R software (version 4.3.3). Descriptive statistics for continuous variables were presented as means ± standard deviations (SD) and compared across groups using one-way analysis of variance (ANOVA). Categorical variables were summarized as counts and percentages, and group differences were assessed using chi-square tests.

To evaluate the association between different meal types and biological aging markers (PhenoAge and KDM), generalized linear models (GLMs) with an identity link function were employed. Models were adjusted sequentially for potential confounders. Model 1 was unadjusted. Model 2 adjusted for demographic variables (age, sex, ethnicity, education) and socioeconomic status (Townsend Deprivation Index tertiles). Model 3 further adjusted for lifestyle factors (smoking status, alcohol intake, physical activity). Model 4 additionally adjusted for BMI to isolate its specific confounding effect. Model 5 was the fully adjusted model, incorporating health conditions (hypertension, diabetes, coronary heart disease, stroke). Regression coefficients (β) and 95% confidence intervals (CIs) were reported to quantify the strength and precision of associations. Since biological age metrics are calculated in years, the regression coefficients were multiplied by 12 to intuitively express the biological age acceleration in months for the discussion and interpretation of results.

Mediation analysis was conducted to explore whether BMI mediated the relationship between takeaway meal consumption and biological aging. The mediation models were adjusted for all covariates included in the fully adjusted model (i.e., demographic variables, socioeconomic status, lifestyle factors and health conditions).

Recognizing BMI categories as ordinal data, we fitted a proportional odds logistic regression (polr) for the mediator model. We then applied causal mediation analysis using the mediation R package to calculate the Average Causal Mediation Effect (ACME, representing the indirect effect) and the Average Direct Effect (ADE). Nonparametric bootstrapping with 1000 simulations was utilized to estimate the point estimates and 95% confidence intervals (CIs) for both direct and indirect effects. All tests were two-sided, and a p-value < 0.05 was considered statistically significant.

Results

In the UK Biobank cohort, a total of 66,876 participants completed the dietary habits questionnaire. After excluding individuals with incomplete data, 43,478 participants were included in the final analysis (Fig. 1).

Fig. 1.

Flowchart of participant selection from the UK Biobank, detailing exclusion criteria based on dietary data completeness and biological age marker availability

The overall sample had a mean age of 56.01 years (SD = 8.18), with 45.1% being male. The distribution of dietary patterns was as follows: takeaway meals (n = 1,874), home-cooked meals (n = 28,652), and others (n = 12,952).

Participants who consumed takeaway meals differed notably from those who primarily ate homecooked meals in several demographic, lifestyle, and health characteristics. Takeaway consumers were more likely to be younger, more likely male, had lower educational attainment, slightly higher levels of material deprivation, and showed a higher prevalence of overweight, unhealthy behaviors such as smoking and alcohol consumption, lower physical activity levels, and slightly lower rates of hypertension and coronary heart disease compared to the homecooked group. Importantly, despite being chronologically younger on average, the takeaway group exhibited higher unadjusted baseline values for both biological aging markers—PhenoAge and KDM—compared to the home-cooked group (Table 1).

Table 1.

Demographic and health characteristics by food consumption categories

Variable	Overall	Takeaway	Home-cooked	Others	P-value
N	43,478	1874	28,652	12,952
Age (mean (SD))	56.01 (8.18)	52.71 (7.94)	56.75 (8.08)	54.85 (8.16)	< 0.001
Male	19,616 (45.1)	987 (52.7)	12,383 (43.2)	6246 (48.2)	< 0.001
White	39,362 (90.5)	1684 (89.9)	25,967 (90.6)	11,711 (90.4)	0.736
College	17,435 (40.1)	619 (33.0)	11,014 (38.4)	5802 (44.8)	< 0.001
Body mass index					< 0.001
< 25 kg/m²	15,980 (36.8)	540 (28.8)	10,914 (38.1)	4526 (34.9)
25 kg/m²≤ BMI<30 kg/m²	18,239 (41.9)	780 (41.6)	11,893 (41.5)	5566 (43.0)
≥ 30 kg/m²	9259 (21.3)	554 (29.6)	5845 (20.4)	2860 (22.1)
Townsend Deprivation Index					< 0.001
Lowest	7603 (17.5)	299 (16.0)	5197 (18.1)	2107 (16.3)
Middle	28,264 (65.0)	1178 (62.9)	18,657 (65.1)	8429 (65.1)
Highest	7611 (17.5)	397 (21.2)	4798 (16.7)	2416 (18.7)
Current smoking	3568 (8.2)	232 (12.4)	2198 (7.7)	1138 (8.8)	< 0.001
Current drinking	40,629 (93.4)	1766 (94.2)	26,604 (92.9)	12,259 (94.6)	< 0.001
Healthy physical activity	24,656 (56.7)	1027 (54.8)	16,748 (58.5)	6881 (53.1)	< 0.001
Hypertension	10,455 (24.0)	445 (23.7)	7057 (24.6)	2953 (22.8)	< 0.001
Coronary Heart Disease	1330 (3.1)	45 (2.4)	952 (3.3)	333 (2.6)	< 0.001
Stroke	431 (1.0)	15 (0.8)	283 (1.0)	133 (1.0)	0.308
Diabetes	1626 (3.7)	77 (4.1)	1093 (3.8)	456 (3.5)	0.143
Phenoage	−4.03 (3.69)	−3.39 (3.65)	−4.12 (3.71)	−3.94 (3.65)	< 0.001
KDM	−1.35 (2.89)	−0.93 (2.86)	−1.36 (2.90)	−1.38 (2.87)	< 0.001

Values are presented as mean (SD) for continuous variables or frequency (percentage) for categorical variables

Phenoage analysis

Regression

In unadjusted models (Model 1), takeaway food consumption was significantly associated with higher PhenoAge compared to home-cooked meals (β = 0.737, 95% CI: 0.569–0.904, p < 0.001). Participants consuming other meals (restaurant meals, ready meals, or sandwiches) also showed elevated PhenoAge (β = 0.174, 95% CI: 0.099–0.248, p < 0.001) (Table 2).

Table 2.

Association between meal types and PhenoAge

Characteristics	Model 1		Model 2		Model 3		Model 4		Model 5
	β (95%CI)	P Value	β (95%CI)	P Value	β (95%CI)	P Value	β (95%CI)	P Value	β (95%CI)	P Value
Mealtype
Home-cooked	REF		REF		REF		REF		REF
Takeaway	0.737(0.569, 0.904)	< 0.001	0.522(0.358,0.686)	< 0.001	0.468(0.302,0.634)	< 0.001	0.309(0.146,0.472)	< 0.001	0.302(0.139, 0.464)	< 0.001
Others	0.174(0.099, 0.248)	< 0.001	0.118(0.044,0.191)	0.002	0.093(0.019,0.167)	0.014	0.054(−0.019, 0.126)	0.149	0.062(−0.01, 0.134)	0.092
TDI
Lowest			REF		REF		REF		REF
Middle			0.180(0.091,0.269)	< 0.001	0.143(0.053,0.232)	0.002	0.110(0.022,0.198)	0.014	0.098(0.011, 0.186)	0.027
High			0.795(0.683,0.907)	< 0.001	0.610(0.496,0.724)	< 0.001	0.501(0.389,0.613)	< 0.001	0.437(0.325,0.548)	< 0.001
BMI
< 25 kg/m²							REF		REF
25 kg/m²≤ BMI<30 kg/m²							0.635(0.560,0.710)	< 0.001	0.579(0.504, 0.654)	< 0.001
≥ 30 kg/m²							1.887(1.796,1.977)	< 0.001	1.690(1.598, 1.783)	< 0.001
	0.002		0.061		0.083		0.117		0.127

Abbreviations: CI Confidence interval, REF Reference group, TDI Townsend Deprivation Index, BMI Body Mass Index

Data are presented as regression coefficients (β) with 95% confidence intervals

Model 1: Unadjusted

Model 2: Adjusted for age, sex, ethnicity, education, and socioeconomic status (Townsend Deprivation Index)

Model 3: Adjusted for Model 2 covariates plus lifestyle factors (smoking status, alcohol intake, and physical activity)

Model 4: Adjusted for Model 3 covariates plus BMI

Model 5: Fully adjusted for Model 4 covariates plus health conditions (hypertension, diabetes, coronary heart disease, and stroke)

The sample size for all models is N = 43,478, as participants with missing data on any covariates were excluded prior to analysis

Stepwise regression analysis revealed the influence of potential confounders. Adjusting for demographic factors and socioeconomic status (Model 2) attenuated the association for takeaway meals (β = 0.522, 95% CI: 0.358–0.686, p < 0.001). Notably, in this model, higher material deprivation (Townsend Index) was independently associated with accelerated PhenoAge (highest vs. lowest tertile: β = 0.795, 95% CI: 0.683–0.907, p < 0.001), confirming the role of socioeconomic status as a key predictor. Further adjustment for lifestyle factors (Model 3) partially attenuated the effect size (β = 0.468, 95% CI: 0.302–0.634, p < 0.001). Subsequent adjustment for BMI (Model 4) resulted in a substantial reduction in the effect size (β = 0.309, 95% CI: 0.146–0.472, p < 0.001), indicating that adiposity largely explains the observed association.

In the fully adjusted model (Model 5), takeaway consumption remained significantly associated with increased PhenoAge by approximately 3.6 months (β = 0.302, 95% CI: 0.139–0.464, p < 0.001). The initial association observed for other meals was fully attenuated and no longer statistically significant (β = 0.062, 95% CI: −0.010–0.134, p = 0.092). As shown in Table 2, the progressive addition of covariates increased the proportion of variance explained () from 0.002 in the unadjusted model to 0.127 in the fully adjusted model. Detailed regression coefficients for all adjusted covariates (including demographic, socioeconomic, lifestyle, and health-related factors) across Models 1 to 5 are provided in Supplementary Table S1.

Mediation

Causal mediation analysis demonstrated that BMI significantly mediated the relationship between takeaway food consumption and PhenoAge. The total effect of takeaway consumption on PhenoAge acceleration was estimated at 0.431 years. This effect was partitioned into a significant direct effect (ADE) of 0.302 (95% CI:0.139 to 0.464, p < 0.001) and a significant indirect effect (ACME) through BMI of 0.129 (95% CI: 0.093 to 0.156, p < 0.001). The estimated indirect effect accounted for approximately 29.9% of the total effect, indicating partial mediation by adiposity (Fig. 2).

Fig. 2.

Mediation model testing the association between takeaway food consumption and PhenoAge, with BMI as a mediator. * p < 0.05, ** p < 0.01, *** p < 0.001

KDM analysis

Regression

Through crude analysis (Model 1), takeaway consumption was strongly associated with increased KDM (β = 0.443, 95% CI: 0.312–0.575, p < 0.001). Other meal types showed no significant association in the unadjusted model (β = −0.030, 95% CI: −0.088–0.029, p = 0.317).

After adjusting for demographics and socioeconomic status (Model 2), the effect of takeaway consumption remained robust (β = 0.440, 95% CI: 0.311–0.569, p < 0.001), while socioeconomic deprivation was independently associated with increased KDM (highest vs. lowest tertile: β = 0.362, 95% CI: 0.274–0.450, p < 0.001). The addition of lifestyle factors(Model 3) slightly attenuated the takeaway coefficient (β = 0.420, 95% CI: 0.288–0.551, p < 0.001), whereas further adjustment for BMI (Model 4) markedly attenuated the takeaway coefficient (β = 0.253, 95% CI: 0.125–0.380, p < 0.001), aligning with the mediation role of adiposity.

In the fully adjusted model (Model 5), the effect of takeaway consumption remained significant, corresponding to an acceleration of about 2.9 months (β = 0.240, 95% CI: 0.114–0.366, p < 0.001). As shown in Table 3, the progressive addition of covariates increased the proportion of variance explained () from 0.001 in the unadjusted model to 0.144 in the fully adjusted model. Other meal types consistently showed no significant associations throughout all models (Table 3). Complete results for all sequential models, including the effect estimates of all adjusted covariates, are detailed in Supplementary Table S2.

Table 3.

Association between meal types and KDM

Characteristics	Model 1		Model 2		Model 3		Model 4		Model 5
	β (95%CI)	P Value	β (95%CI)	P Value	β (95%CI)	P Value	β (95%CI)	P Value	β (95%CI)	P Value
Mealtype
Home-cooked	REF		REF		REF		REF		REF
Takeaway	0.443(0.312,0.575)	< 0.001	0.440(0.311,0.569)	< 0.001	0.420(0.288,0.551)	< 0.001	0.253(0.125,0.380)	< 0.001	0.240(0.114,0.366)	< 0.001
Others	−0.030(−0.088,0.029)	0.317	0.042(−0.016,0.099)	0.158	0.037(−0.022,0.096)	0.215	−0.010(−0.066,0.047)	0.738	−0.004(−0.060,0.052)	0.882
TDI
Lowest			REF		REF		REF		REF
Middle			0.033(−0.037,0.103)	0.36	0.020(−0.051,0.091)	0.572	−0.012(−0.081,0.056)	0.726	−0.014(−0.081,0.054)	0.694
High			0.362(0.274,0.450)	< 0.001	0.272(0.182,0.362)	< 0.001	0.168(0.081,0.255)	< 0.001	0.128(0.042,0.215)	0.004
BMI
< 25 kg/m²							REF		REF
25 kg/m²≤ BMI<30 kg/m²							0.950(0.892,1.009)	< 0.001	0.858(0.800,0.916)	< 0.001
≥ 30 kg/m²							2.023(1.952,2.094)	< 0.001	1.776(1.704,1.848)	< 0.001
	0.001		0.058		0.061		0.125		0.144

Abbreviations: CI Confidence interval, REF reference group, TDI Townsend Deprivation Index, BMI Body Mass Index

Data are presented as regression coefficients (β) with 95% confidence intervals

Model 1: Unadjusted

Model 2: Adjusted for age, sex, ethnicity, education, and socioeconomic status (Townsend Deprivation Index)

Model 3: Adjusted for Model 2 covariates plus lifestyle factors (smoking status, alcohol intake, and physical activity)

Model 4: Adjusted for Model 3 covariates plus BMI

Model 5: Fully adjusted for Model 4 covariates plus health conditions (hypertension, diabetes, coronary heart disease, and stroke)

The sample size for all models is N = 43,478, as participants with missing data on any covariates were excluded prior to analysis

Mediation

Similarly, causal mediation analysis for KDM confirmed that BMI partially mediated this association. The total effect of takeaway consumption on KDM was estimated at 0.381 years. The direct effect (ADE) of takeaway consumption, independent of BMI, was 0.240 (95% CI: 0.114 to 0.366, p < 0.001). Importantly, the analysis revealed a significant indirect effect (ACME) through BMI of 0.141 (95% CI: 0.103 to 0.168, p < 0.001). This indirect effect accounted for approximately 37.0% of the total effect (Fig. 3).

Fig. 3.

Mediation model testing the association between takeaway food consumption and KDM-based biological aging, with BMI as a mediator. * p < 0.05, ** p < 0.01, *** p < 0.001

Discussion

This study provides quantitative evidence that frequent takeaway food consumption is significantly associated with accelerated biological aging. On average, individuals who consumed takeaway meals exhibited 2.9 to 3.6 months (approximately 0.24 to 0.30 years) of biological age acceleration. From a strictly clinical perspective for an individual, a 3-month difference in biological age may appear modest. However, from a public health perspective, this effect size is substantial. Based on established validations where a one-year increase in PhenoAge corresponds to a 9% rise in all-cause mortality [9], this 3-month acceleration translates to an estimated 2.2% increase in mortality risk, assuming a linear relationship based on existing risk models. To accurately contextualize the magnitude of this dietary impact, we compared it against established lifestyle risk factors evaluated within our own fully adjusted models. For phenotypic aging, the burden of takeaway consumption (+ 0.302 years) effectively offsets the biological aging benefits gained from healthy physical activity (−0.242 years) and accounts for approximately 16% of the severe aging effect caused by current smoking (+ 1.865 years) (see full covariate results in Supplementary Table S1).

For KDM, takeaway consumption equivalent represents roughly half the detrimental impact of current smoking and acts to diminish the physiological gains derived from regular physical activity (Supplementary Table S2). These internal comparisons align well with previous independent studies demonstrating the profound impact of smoking and the mitigating effects of physical activity on biological clocks [11, 12]. This indicates that a single dietary habit—relying on takeaway food—carries a metabolic penalty comparable to a significant fraction of active smoking and can substantially attenuate the anti-aging benefits of regular exercise.

Given the ubiquitous nature of food delivery services, such a rightward shift in the population’s biological age distribution could significantly expand the absolute number of individuals reaching the threshold for metabolic diseases, echoing Geoffrey Rose’s population strategy paradigm [13]. Consequently, our findings necessitate translation into targeted public health policies. First, in terms of risk communication, framing the adverse effects of highly processed takeaway foods in terms of “accelerated aging”—demonstrating that regular consumption cancels out the benefits of exercise—provides a highly tangible behavioral nudge for consumers compared to abstract, long-term disease risks. Second, at the platform level, policymakers should mandate nutritional transparency on digital delivery platforms, such as implementing a traffic-light labeling system for meals exceeding optimal sodium, sugar, or saturated fat thresholds. Finally, structural interventions, such as subsidizing workplace canteens or providing tax incentives to restaurants offering whole-food, minimally processed options, are urgently needed to reduce the occupational reliance on takeaway platforms and mitigate this widespread environmental driver of accelerated aging.

The biological plausibility of these findings lies in the characteristic nutritional profile of takeaway meals. It is important to distinguish takeaway meals from the strictly defined ultra-processed foods (UPFs) based on the NOVA classification, which focuses on the extent of industrial processing. Takeaway meals encompass a wide variety of culinary preparations made outside the home; however, extensive nutritional analyses have consistently demonstrated that typical takeaway and fast foods share an adverse nutrient profile. Compared to home-cooked meals, takeaway foods are generally more energy-dense, significantly higher in saturated fats, sodium, and refined carbohydrates, and substantially lower in protective nutrients such as dietary fiber, vitamins, and phytochemicals [6, 7, 14].

Diets characterized by such macronutrient imbalances and micronutrient deficiencies are established drivers of chronic low-grade inflammation and oxidative stress, which are core mechanistic pillars of biological aging [15]. High intake of sodium and saturated fats can induce endothelial dysfunction and metabolic dysregulation, directly elevating clinical biomarkers such as blood pressure, C-reactive protein, and fasting glucose—key components of the PhenoAge and KDM algorithms. Furthermore, frequent reliance on takeaway meals often displaces the intake of nutrient-dense whole foods, thereby exacerbating systemic inflammation and accelerating physiological decline [16]. Therefore, the accelerated biological aging observed in frequent takeaway consumers is primarily attributable to the cumulative physiological burden of poor dietary composition and high energy density inherent in most takeaway food environments [17].

Mediation analyses revealed that BMI accounted for approximately 30 to 37% of the relationship between takeaway consumption and biological aging, suggesting that increased adiposity partially mediates this association. Obesity has been recognized as a driver of aging through chronic low-grade inflammation, altered insulin signaling, and mitochondrial stress [18, 19].Our finding provides quantitative evidence that BMI partly mediates the link between takeaway consumption and biological aging, while also suggesting the existence of additional pathways beyond BMI.

Beyond adiposity, other mechanisms may contribute to the observed associations. Takeaway meals are frequently characterized by lower nutritional quality, including reduced fiber and phytochemical content compared to home-cooked options [20, 21]. Such dietary patterns have been linked to gut microbiota dysbiosis and systemic inflammation, which are known drivers of biological aging [22, 23]. The observed biological aging acceleration may be further exacerbated by the consumption patterns associated with takeaway meals, which are often characterized by hurried or solitary eating, potentially leading to less mindful food intake [24, 25]. Moreover, the widespread use of digital ordering platforms facilitates frequent takeaway consumption, which may amplify the associated health risks [1, 2]. In addition, the packaging materials used for takeaway meals, such as certain plastics or Per- and Polyfluoroalkyl Substances, may introduce harmful chemicals, contributing to the overall adverse health impact; for instance, substances associated with food packaging have been hypothesized to disrupt metabolic and endocrine function in experimental models [26–28]. These nutritional and environmental factors may act in concert to accelerate aging processes, independent of body weight.

Importantly, the observed associations were consistent across two independent aging indices—PhenoAge and KDM—each capturing distinct dimensions of biological aging. PhenoAge incorporates inflammatory and metabolic biomarkers to estimate mortality risk, whereas KDM reflects deviation from physiological aging trajectories [29, 30]. The agreement between these complementary indices supports the robustness of our findings and reduces the likelihood that results are driven by model-specific bias. Additionally, the use of validated composite aging measures enhances the biological relevance of our outcome variables and offers a comprehensive picture of systemic aging.

Notably, we observed divergent attenuation patterns between the two biological aging metrics: the reduction in the takeaway effect size from the unadjusted model (Model 1) to the sociodemographic-adjusted model (Model 2) was less pronounced for KDM compared to PhenoAge. This phenomenon likely stems from the distinct biomarker compositions of the two indices. PhenoAge is heavily weighted toward inflammatory and immune markers (e.g., C-reactive protein, white blood cell count), which are profoundly sensitive to psychosocial stress and socioeconomic deprivation. Consequently, adjusting for socioeconomic status (Townsend Deprivation Index) absorbed a considerable portion of the variance in PhenoAge. In contrast, KDM incorporates broader functional and structural parameters (e.g., forced expiratory volume, systolic blood pressure, glycated hemoglobin) that may more directly capture the cumulative physiological and metabolic burden of poor dietary exposures, rendering the KDM models more robust to sociodemographic adjustments.

Despite these strengths, several limitations warrant acknowledgment. First, the cross-sectional nature of the analysis precludes causal inference, and unmeasured confounding cannot be ruled out. Furthermore, the possibility of reverse causality should be acknowledged; distinct health conditions or lifestyle constraints (e.g., time poverty) might influence dietary choices, potentially leading individuals with poorer health status to rely more frequently on convenience foods like takeaways. Therefore, the observed relationships should be interpreted as statistical associations rather than definitive causal pathways.

Second, regarding exposure classification, participants reporting “Takeaway meals” were assigned to the Takeaway group irrespective of other selections. We acknowledge the inherent limitations of recoding multiple responses into a single mutually exclusive variable. Specifically, the takeaway group likely includes individuals with mixed dietary patterns (e.g., consuming takeaway alongside home-cooked meals or other prepared foods like ready meals). Forcing these multiple responses into mutually exclusive categories carries the risk of diluting the observed effects in either direction. Therefore, residual confounding from overlapping dietary behaviors cannot be entirely ruled out. Notably, the observed effects are based on a 24-hour recall and likely reflect short-term takeaway consumption. Further research is needed to explore the long-term impact on aging and health outcomes.

Finally, the UK Biobank is primarily composed of individuals of European ancestry, which may limit generalizability to more diverse populations. Future research should consider longitudinal designs to better assess causality. From a policy perspective, these findings highlight the need to promote healthy food environments in urban settings where takeaway reliance is highest. Interventions may include subsidizing fresh ingredients, regulating ultra-processed food advertising, and incorporating nutrition labeling into digital platforms. Ultimately, population-level strategies targeting dietary behaviors could play a role in mitigating premature aging.

Conclusions

In summary, takeaway food consumption was significantly associated with accelerated biological aging in this large UK Biobank cohort, with effects consistently observed across both PhenoAge and KDM measures. Approximately one-third of this association was mediated by higher body mass index, indicating that adiposity plays an important but partial role. These results suggest that consumption on takeaway meals may contribute to premature biological decline through both adiposity-dependent and additional nutritional or environmental pathways. Given the rising prevalence of take-away consumption worldwide, public health strategies promoting healthier dietary patterns and improving the nutritional quality of takeaway options may be critical for fostering healthy aging.

Abbreviations

PhenoAge

Phenotypic Age

KDM

Klemera-Doubal Method

BMI

Body Mass Index

CHD

Coronary Heart Disease

FEV1

Forced Expiratory Volume in One Second

SBP

Systolic Blood Pressure

HbA1c

Glycated Hemoglobin

lnCRP

Natural Log-transformed C-Reactive Protein

ALP

Alkaline Phosphatase

BUN

Blood Urea Nitrogen

TDI

Townsend Deprivation Index

MET

Metabolic Equivalent Task

ANOVA

Analysis of Variance

GLM

Generalized Linear Model

CI

Confidence Interval

SD

Standard Deviation

UPF(s)

Ultra-Processed Food(s)

Authors’ contributions

Conceptualization, G.L. and X.W.; methodology, Y.C. and X.W.; software, Y.C.; validation, G.L., X.W. and N.L.; formal analysis, Y.C.; investigation, G.L. and W.H.; resources, N.L.; data curation, N.L. and Y.C.; writing—original draft preparation, G.L.; writing—review and editing, G.L., X.W. and N.L.; visualization, G.L.; supervision, X.W. and N.L.; project administration, X.W.; funding acquisition, X.W. All authors have read and agreed to the published version of the manuscript.

Funding

This research was supported by the National Natural Science Foundation of China (82101264, 81701067).

Data availability

The data that support the findings of this study are available from the UK Biobank (https://www.ukbiobank.ac.uk), but restrictions apply to the availability of these data. Data are therefore not publicly available.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.