Longitudinal association of ultra-processed food consumption with biological aging: the mediating role of adiposity measures

Abstract

Background & Aims: The consumption of ultra-processed foods (UPF) is rising in modern diets. However, the connection between UPF intake and biological aging still lacks research. This study aims to investigate the association between UPF consumption and biological aging and to explore the mediating effect of various adiposity measures. Methods: We performed a cross-sectional and prospective cohort study in 57,128 individuals included in the UK Biobank using 24-h dietary recall questionnaires. This research employed phenotypic age as biological age, with the age gap calculated as the difference between chronological and biological age. UPF was defined according to the NOVA classification. Linear regression models were employed to estimate the relationship between UPF consumption and biological aging. Mediation analysis was conducted to explore the mediating effect of various adiposity measures, including body mass index (BMI), waist circumference, hip circumference, body fat percentage, and waist-to-height ratio, in the observed associations. Results: Compared to the lowest quartile of UPF weight proportion consumption, the age gap increased by 0.378 years with a 95% confidence interval (CI) of (0.253, 0.504) for phenotypic age in the highest quartile in the cross-sectional study and 0.525 years (0.214, 0.836) in the longitudinal study. Higher UPF energy proportion consumption was associated with accelerated biological aging as well. Adiposity measures significantly mediated the association between UPF weight proportion and biological aging, with mediating proportions ranging from 25.80% to 36.76%. Conclusion: Higher consumption of UPFs is positively associated with accelerated biological aging, with adiposity measures serving as significant mediators.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11357-025-01865-6.

Introduction

Population aging has become a pressing social challenge. The number of people aged 65 years or older worldwide is expected to more than double, from 761 million in 2021 to 1.6 billion in 2050, which will make up about 16% of the global population [1]. While chronological aging is inevitable, researchers are increasingly focusing on biological aging, which uses a collection of aging-related biomarkers to indicate cellular age [2]. Biological aging serves as a key indicator of overall health and reflects the diversity of individual aging processes [3–5]. Additionally, certain controllable factors are believed to influence biological aging [6], making its investigation crucial for effective aging management.

Diet is widely recognized as an important and actionable factor for aging. A notable diet change in modern diets is the rising consumption of ultra-processed foods (UPF). UPF are industrially manufactured, ready-to-eat or ready-to-heat food products that are predominantly composed of substances extracted from foods, with little to no intact whole food ingredients [7]. The consumption of UPF has become increasingly prevalent in modern diets [8], with higher amounts and broader variety in developed countries and rapid growth in developing countries [9]. Examples of UPF include soft drinks, packaged snacks, confectionery, breakfast cereals, flavored yogurts, industrially processed bread and buns, and processed meat products, all classified according to the NOVA classification system [7, 10]. These items are often deficient in essential nutrients and excessively rich in sugars, saturated fats, and sodium, which may contribute to various adverse health issues [11].

Research has linked high UPF consumption to several adverse health outcomes. Studies have demonstrated significant associations between obesity and high UPF consumption [12–14]. Furthermore, high UPF consumption was associated with an increased risk of diabetes [13, 15–17], cardiovascular disease [18, 19], cancer risk and cancer mortality [20], and all-cause mortality [21–23]. Beyond these disease-specific risks, biological aging has also emerged as a critical health outcome associated with unhealthy diet [24, 25]. Recent studies demonstrated that higher UPF consumption was associated with a greater risk of having shorter telomeres [24] and an acceleration of biological aging [25]. However, the relationship between UPF consumption and accelerated biological aging remains inadequately understood, with limited evidence linking UPF intake to telomere length (TL) [24] and predictions of biological age from deep learning models [25]. Moreover, existing studies have established an association between high UPF intake and obesity [12–14], as well as between obesity and biological age [26, 27]. Consequently, it is essential to address this gap by examining the direct relationship between UPF consumption and biological aging and exploring the mediator role of adiposity measures through a longitudinal perspective.

This study aims to examine the association between UPF consumption and biological aging while also exploring the mediating effect of adiposity measures. Understanding this relationship is essential for developing public health strategies to promote healthier dietary choices and mitigate aging effects.

Materials and methods

Data source and study population

In this study, we analyzed data from the UK Biobank, a large prospective study with over 500,000 participants aged 40 to 69, recruited between 2006 and 2010. Participants completed a touch-screen questionnaire on demographics, health, and lifestyle, and some provided biological samples.

We extracted data from the UK Biobank (n = 502,244). For the cross-sectional study, we excluded participants (n = 291,706) who did not have a valid 24-h dietary recall questionnaire collected at baseline. Furthermore, we also excluded participants who did not finish the questionnaire on the same date as they attended the assessment center to meet the design of cross-sectional study (n = 141,280). Additionally, we excluded participants lacking blood biomarkers for calculating phenotypic age (n = 12,130). Ultimately, we extracted the study population for this cross-sectional study (n = 57,128). As for the longitudinal study, we first excluded these participants who did not do follow-up at the first-repeated assessment (n = 481,904). Moreover, participants without at least one valid 24-h dietary questionnaire (n = 7,300) and without the blood biomarkers for calculating phenotypic age (n = 4,932) were excluded. Finally, we got the study population (n = 8,108) for the longitudinal study. The study design flowchart is shown in Fig. 1.

Fig. 1.

Flowchart of study population extraction from the UK Biobank

Assessment of UPF consumption—exposure of interest

UPF consumption was assessed using the Oxford WebQ, a web-based 24-h dietary recall tool that has been validated in previous studies [14, 15, 20, 28]. This tool collects detailed dietary intake data on a range of food and beverage items consumed over the past 24 h. It includes over 200 food items and 30 drink categories, allowing for the calculation of total weight and nutrient intake for each participant. The 24-h dietary recall questionnaire was collected five times. The first cycle of dietary information started between April 2009 and September 2010, which was the initial assessment visit at the assessment center. Four additional cycles were gathered via email between February 2011 and June 2012. In our study, we only included the first cycle for cross-sectional study and utilized the average value of all the five cycles for longitudinal study. UPF consumption was determined using the NOVA classification system, which categorizes foods based on their level of processing. Foods were classified into four groups: 1) unprocessed or minimally processed foods (UNPFs), such as fruits, vegetables, milk, and meat; 2) processed culinary ingredients, comprising vegetable oils obtained by crushing seeds, nuts, and fruits; 3) processed foods, such as canned or bottled vegetables and legumes in brine; and 4) UPFs, including soft drinks, packaged snacks, confectionery, breakfast cereals, flavored yogurts, industrially processed bread and buns, and processed meat products.

Our primary focus was on the consumption of UPF, defined as the percentage of UPF content in the total diet. To estimate the proportion, we summed the consumption of each UPF item reported by participants and calculated its weight and energy as a percentage of total food intake. The UK Biobank provides portion sizes for each food and beverage to establish dietary intake. The total intake (grams per day or kcal per day) of UPF or any other food was the sum of the weights or the energy of each respective food item. The ratio of UPF intake to total food weight or energy was then determined for each participant. The methodology for assessing an individual’s dietary consumption based on the level of food processing has been meticulously documented elsewhere [29, 30]. Details of the food items used for estimating UPF intake are provided in Supplementary Table 1.

Assessment of biological aging—outcome of interest

Chronological age is defined as the duration of time since birth, while biological age serves as an integrated measure reflecting an individual’s aging pace [31]. We assessed biological aging by calculating the age gap, which is defined as the residual difference between biological age and chronological age in formula (1). If the age gap is greater than zero, the corresponding age gap group is defined as accelerated aging. Otherwise, it's defined as decelerated aging. This study utilized phenotypic age as the biological age.

$$agegap=biologicalage-chronologicalage$$ 1

Phenotypic age was calculated based on the original biomarker sets [32]. The formula (2–4) of phenotypic age was presented.

$$phenotypicage=141.50225+\frac{In[-0.00553\times In(1-\text{mortality risk})]}{0.090165}$$ 2

$$\text{mortality risk}=1-e^{-e^{\mathit{xb}}[exp(120\times \gamma )-1]/\gamma }$$ 3

where γ = 0.0076927,

$$\begin{array}{c}xb=-19.907-0.0336\times albumin+0.0095\times creatinine+0.1953\times glucose+\\ 0.0954\times ln(C-reactiveprotein)-0.012\times lymphocytepercentage+\\ 0.0268\times meancorpuscularvolume+0.3306\times redcelldistributionwidth+\\ 0.00188\times alkalinephosphatase+0.0554\times whitebloodcellcount+\\ 0.0804\times chronologicalage\end{array}$$ 4

Assessment of covariates

Detailed definitions of each covariate were provided in Supplementary Text 1.

Statistical analyses

Statistical analyses were provided in Supplementary Text 2.

Results

Baseline characteristics

The characteristics stratified by age gap group among the cross-sectional study population and the longitudinal study population are presented in Table 1. The cross-sectional study involved 57,128 participants from the UK Biobank, of whom 45.395% were male and 93.539% were white. The longitudinal study involved 8,108 participants, of whom 51.591% were male and 98.199% were white. Across the age gap group, the proportion of UPF weight consumption was 21.6% in the decelerated aging group and 24.4% in the accelerated aging group in the cross-sectional study. For the longitudinal study, the proportion of UPF weight consumption was 22.0% in the decelerated aging group and 24.9% in the accelerated aging group. For the proportion of UPF energy consumption, similar trends were shown. Characteristics between included and excluded participants can be seen in Supplementary Table 2.

Table 1.

Characteristic by phenotypic age gap group among UK Biobank participants¹

Characteristics	Cross-sectional study				Longitudinal study
	Overall	Decelerated aging	Accelerated aging	P value	Overall	Decelerated aging	Accelerated aging	P value
Number of participants (%)	57,128	45,841	11,287		8,108	6,936	1,172
Sex = Male (%)	25,933 (45.395)	19,738 (43.058)	6,195 (54.886)	< 0.001	4,183 (51.591)	3,410 (49.164)	773 (65.956)	< 0.001
Age (mean (SD))	56.753 (8.144)	56.635 (8.109)	57.232 (8.272)	< 0.001	61.669 (7.196)	61.447 (7.139)	62.984 (7.395)	< 0.001
Phenotypic age (mean (SD))	53.204 (9.982)	51.117 (8.784)	61.682 (10.070)	< 0.001	57.133 (9.200)	55.433 (8.013)	67.193 (9.345)	< 0.001
Phenotypic age gap (mean (SD))	−3.549 (5.493)	−5.519 (3.210)	4.450 (5.578)	< 0.001	−4.536 (5.127)	−6.014 (3.267)	4.209 (5.416)	< 0.001
Phenotypic age gap group = accelerated aging (%)	11,287 (19.757)	0 (0.000)	11,287 (100.000)	< 0.001	1,172 (14.455)	0 (0.000)	1,172 (100.000)	< 0.001
UPF weight proportion (mean (SD))	0.221 (0.155)	0.216 (0.151)	0.244 (0.168)	< 0.001	0.224 (0.135)	0.220 (0.132)	0.249 (0.149)	< 0.001
UPF energy proportion (mean (SD))	0.461 (0.180)	0.456 (0.178)	0.479 (0.184)	< 0.001	0.470 (0.143)	0.465 (0.143)	0.497 (0.140)	< 0.001
Ethnicity group (%)				< 0.001				0.824
White	53,437 (93.539)	43,101 (94.023)	10,336 (91.574)		7,962 (98.199)	6,809 (98.169)	1,153 (98.379)
Asian or Asian British	1,279 (2.239)	971 (2.118)	308 (2.729)		36 (0.444)	32 (0.461)	4 (0.341)
Black or Black British	1,098 (1.922)	745 (1.625)	353 (3.127)		26 (0.321)	24 (0.346)	2 (0.171)
Chinese	168 (0.294)	148 (0.323)	20 (0.177)		10 (0.123)	8 (0.115)	2 (0.171)
Mixed	419 (0.733)	312 (0.681)	107 (0.948)		28 (0.345)	25 (0.360)	3 (0.256)
Other ethnic group	727 (1.273)	564 (1.230)	163 (1.444)		46 (0.567)	38 (0.548)	8 (0.683)
Education (%)				< 0.001				< 0.001
College or University degree	22,115 (38.711)	18,398 (40.134)	3,717 (32.932)		3,973 (49.001)	3,482 (50.202)	491 (41.894)
A levels/AS levels or equivalent	7,335 (12.840)	5,914 (12.901)	1,421 (12.590)		1,051 (12.963)	895 (12.904)	156 (13.311)
O levels/GCSEs or equivalent	12,613 (22.078)	10,093 (22.017)	2,520 (22.327)		1,525 (18.809)	1,285 (18.527)	240 (20.478)
CSEs or equivalent	2,981 (5.218)	2,313 (5.046)	668 (5.918)		226 (2.787)	189 (2.725)	37 (3.157)
NVQ or HND or HNC or equivalent	3,325 (5.820)	2,507 (5.469)	818 (7.247)		474 (5.846)	380 (5.479)	94 (8.020)
Other professional qualifications eg: nursing	2,815 (4.928)	2,260 (4.930)	555 (4.917)		397 (4.896)	337 (4.859)	60 (5.119)
None of the above	5,944 (10.405)	4,356 (9.502)	1,588 (14.069)		462 (5.698)	368 (5.306)	94 (8.020)
Income (%)				< 0.001				< 0.001
Less than 18,000 GBP	10,721 (18.767)	8,017 (17.489)	2,704 (23.957)		1,278 (15.762)	1,026 (14.792)	252 (21.502)
18,000 GBP to 30,999 GBP	14,408 (25.221)	11,409 (24.888)	2,999 (26.570)		2,510 (30.957)	2,117 (30.522)	393 (33.532)
31,000 GBP to 51,999 GBP	15,440 (27.027)	12,494 (27.255)	2,946 (26.101)		2,381 (29.366)	2,056 (29.642)	325 (27.730)
52,000 GBP to 100,000 GBP	12,693 (22.219)	10,613 (23.152)	2,080 (18.428)		1,578 (19.462)	1,409 (20.314)	169 (14.420)
Greater than 100,000 GBP	3,866 (6.767)	3,308 (7.216)	558 (4.944)		361 (4.452)	328 (4.729)	33 (2.816)
Townsend deprivation index (mean (SD))	−1.279 (2.835)	−1.381 (2.774)	−0.865 (3.032)	< 0.001	−2.199 (2.587)	−2.233 (2.547)	−1.994 (2.807)	0.003
Body mass index (mean (SD))	27.177 (4.728)	26.642 (4.281)	29.346 (5.730)	< 0.001	26.794 (4.439)	26.472 (4.205)	28.694 (5.241)	< 0.001
Body mass index category (%)				< 0.001				< 0.001
Underweight	300 (0.525)	257 (0.561)	43 (0.381)		46 (0.567)	44 (0.634)	2 (0.171)
Normal weight	19,911 (34.853)	17,470 (38.110)	2,441 (21.627)		3,008 (37.099)	2,731 (39.374)	277 (23.635)
Overweight	23,856 (41.759)	19,406 (42.333)	4,450 (39.426)		3,482 (42.945)	2,979 (42.950)	503 (42.918)
Obesity	13,061 (22.863)	8,708 (18.996)	4,353 (38.566)		1,572 (19.388)	1,182 (17.042)	390 (33.276)
Waist circumference (mean (SD))	89.672 (13.472)	88.079 (12.657)	96.141 (14.693)	< 0.001	90.536 (13.030)	89.428 (12.520)	97.094 (14.023)	< 0.001
Hip circumference (mean (SD))	102.950 (9.176)	102.116 (8.401)	106.335 (11.193)	< 0.001	102.928 (8.907)	102.413 (8.490)	105.978 (10.558)	< 0.001
Body fat percentage (mean (SD))	31.252 (8.532)	30.948 (8.325)	32.485 (9.227)	< 0.001	30.742 (8.253)	30.651 (8.199)	31.281 (8.548)	0.016
Waist-to-height ratio (mean (SD))	0.530 (0.075)	0.521 (0.069)	0.566 (0.084)	< 0.001	0.534 (0.072)	0.528 (0.069)	0.569 (0.080)	< 0.001
Smoking (%)				< 0.001				< 0.001
Never	32,213 (56.387)	26,580 (57.983)	5,633 (49.907)		4,852 (59.842)	4,245 (61.202)	607 (51.792)
Previous	20,053 (35.102)	16,022 (34.951)	4,031 (35.714)		2,948 (36.359)	2,477 (35.712)	471 (40.188)
Current	4,862 (8.511)	3,239 (7.066)	1,623 (14.379)		308 (3.799)	214 (3.085)	94 (8.020)
Alcohol (%)				< 0.001				< 0.001
Daily or almost daily	12,594 (22.045)	10,202 (22.255)	2,392 (21.193)		1,560 (19.240)	1,324 (19.089)	236 (20.137)
Three or four times a week	13,245 (23.185)	11,060 (24.127)	2,185 (19.359)		2,286 (28.194)	2,009 (28.965)	277 (23.635)
Once or twice a week	14,251 (24.946)	11,580 (25.261)	2,671 (23.664)		2,111 (26.036)	1,824 (26.298)	287 (24.488)
One to three times a month	6,606 (11.564)	5,175 (11.289)	1,431 (12.678)		906 (11.174)	756 (10.900)	150 (12.799)
Special occasions only	6,293 (11.016)	4,729 (10.316)	1,564 (13.857)		742 (9.151)	612 (8.824)	130 (11.092)
Never	4,139 (7.245)	3,095 (6.752)	1,044 (9.250)		503 (6.204)	411 (5.926)	92 (7.850)
Summed MET minutes per week (mean (SD))	2,683.628 (2,605.763)	2,718.600 (2,588.738)	2,541.596 (2,669.206)	< 0.001	2,337.092 (2,206.135)	2,376.186 (2,212.231)	2,105.729 (2,156.169)	< 0.001
Physical activity (MET-min per week) (%)				< 0.001				< 0.001
Low	9,451 (16.544)	7,070 (15.423)	2,381 (21.095)		1,490 (18.377)	1,231 (17.748)	259 (22.099)
Moderate	29,634 (51.873)	24,010 (52.377)	5,624 (49.827)		4,452 (54.909)	3,802 (54.815)	650 (55.461)
High	18,043 (31.583)	14,761 (32.200)	3,282 (29.078)		2,166 (26.714)	1,903 (27.437)	263 (22.440)
Total energy intake (mean (SD))	8,650.375 (2,981.149)	8,623.587 (2,936.358)	8,759.173 (3,154.339)	< 0.001	8,178.275 (2,197.694)	8,142.353 (2,160.223)	8,390.869 (2,397.537)	< 0.001
Healthy diet score (%)				< 0.001				< 0.001
0	1,107 (1.938)	799 (1.743)	308 (2.729)		168 (2.072)	134 (1.932)	34 (2.901)
1	5,545 (9.706)	4,129 (9.007)	1,416 (12.545)		798 (9.842)	645 (9.299)	153 (13.055)
2	11,847 (20.738)	9,130 (19.917)	2,717 (24.072)		1,662 (20.498)	1,379 (19.882)	283 (24.147)
3	15,850 (27.745)	12,651 (27.598)	3,199 (28.342)		2,121 (26.159)	1,823 (26.283)	298 (25.427)
4	14,760 (25.837)	12,234 (26.688)	2,526 (22.380)		2,153 (26.554)	1,881 (27.119)	272 (23.208)
5	8,019 (14.037)	6,898 (15.048)	1,121 (9.932)		1,206 (14.874)	1,074 (15.484)	132 (11.263)
Ischemic heart disease = 1 (%)	2,514 (4.401)	1,634 (3.564)	880 (7.797)	< 0.001	456 (5.624)	344 (4.960)	112 (9.556)	< 0.001
Hypertensive diseases = 1 (%)	14,824 (25.949)	10,642 (23.215)	4,182 (37.051)	< 0.001	2,150 (26.517)	1,696 (24.452)	454 (38.737)	< 0.001
Stroke = 1 (%)	673 (1.178)	443 (0.966)	230 (2.038)	< 0.001	111 (1.369)	78 (1.125)	33 (2.816)	< 0.001
COPD = 1 (%)	238 (0.417)	107 (0.233)	131 (1.161)	< 0.001	67 (0.826)	39 (0.562)	28 (2.389)	< 0.001
CKD = 1 (%)	128 (0.224)	35 (0.076)	93 (0.824)	< 0.001	19 (0.234)	8 (0.115)	11 (0.939)	< 0.001
Dementia = 1 (%)	14 (0.025)	13 (0.028)	1 (0.009)	0.395	3 (0.037)	2 (0.029)	1 (0.085)	0.913
Parkinsonism = 1 (%)	81 (0.142)	63 (0.137)	18 (0.159)	0.676	19 (0.234)	17 (0.245)	2 (0.171)	0.872
Multiple Sclerosis = 1 (%)	177 (0.310)	146 (0.318)	31 (0.275)	0.512	28 (0.345)	21 (0.303)	7 (0.597)	0.187
Diabetes = 1 (%)	2,649 (4.637)	1,188 (2.592)	1,461 (12.944)	< 0.001	332 (4.095)	173 (2.494)	159 (13.567)	< 0.001
Cirrhosis = 1 (%)	72 (0.126)	37 (0.081)	35 (0.310)	< 0.001	11 (0.136)	5 (0.072)	6 (0.512)	0.001
Osteoarthritis = 1 (%)	4,797 (8.397)	3,598 (7.849)	1,199 (10.623)	< 0.001	969 (11.951)	789 (11.375)	180 (15.358)	< 0.001
Osteoporosis = 1 (%)	899 (1.574)	745 (1.625)	154 (1.364)	0.051	104 (1.283)	94 (1.355)	10 (0.853)	0.203
Schizophrenia = 1 (%)	93 (0.163)	55 (0.120)	38 (0.337)	< 0.001	5 (0.062)	5 (0.072)	0 (0.000)	0.777
Depression = 1 (%)	3,183 (5.572)	2,365 (5.159)	818 (7.247)	< 0.001	512 (6.315)	430 (6.200)	82 (6.997)	0.331
Bipolar disorder = 1 (%)	177 (0.310)	116 (0.253)	61 (0.540)	< 0.001	17 (0.210)	10 (0.144)	7 (0.597)	0.005
Cancer = 1 (%)	4,960 (8.682)	3,862 (8.425)	1,098 (9.728)	< 0.001	992 (12.235)	798 (11.505)	194 (16.553)	< 0.001

A levels, advanced levels; AS, advanced subsidiary levels; CKD, chronic kidney disease; CSEs, certificate of secondary education, national vocational qualifications; COPD, chronic obstructive pulmonary disease; GCSEs, general certificate of secondary education; GBP, Great Britain pound; HND, higher national diploma; MET, metabolic equivalent task; NC, higher national certificates; O levels, ordinary levels; SD, standard deviation; UPF, ultra-processed food

¹Values are numbers (%) for categorical variables and mean (SD) for continuous variables

Association between biological age and chronological age

We examined the relationships between phenotypic age and chronological age through histograms and scatter plots. In Supplementary Fig. 1A, the fitted line demonstrated a significant positive correlation between chronological age and phenotypic age (r = 0.835, MAE = 5.307) among the cross-sectional study population. According to Supplementary Fig. 1B, the distribution of chronological age and phenotypic age were highly overlapping, indicating a strong correlation among the cross-sectional study population. In addition, the longitudinal study population demonstrates a similar relationship in Supplementary Fig.  1C and D.

UPF consumption and age gap

The linear regression results of cross-sectional study in Table 2 showed that, compared to the first quartile (lowest UPF weight proportion consumption), the phenotypic age gap increased by 0.367 years (0.241, 0.492) in the second quartile, by 0.537 years (0.412, 0.663) in the third quartile, and by 1.208 years (1.082, 1.333) in the fourth quartile (highest UPF weight proportion consumption) in model 1. After adjusting for potential confounding factors in model 2 and model 3, the associations remained statistically significant. In detail, compared to the lowest quartile, the age gap in the fourth quartile of UPF weight proportion consumption increased by 0.531 years (0.407, 0.656) in model 2, and 0.378 years (0.253, 0.504) in model 3, respectively. As for the UPF energy proportion consumption for the cross-sectional study, similar trends were found among all models.

Table 2.

Association between the quartile of ultra-processed food consumption and age gap among UK Biobank participants

	Quartiles5 of the weight proportion of ultra-processed food consumption				P Trend7	Per 10% increase8
	First Quartile	Second Quartile	Third Quartile	Fourth Quartile
Results of the cross-sectional study
Median of UPF weight proportion	6.4%	14.3%	23.8%	40.5%
Model 11, β (95%CI)4	Reference	0.367 (0.241, 0.492)	0.537 (0.412, 0.663)	1.208 (1.082, 1.333)	< 0.001	0.326 (0.297, 0.354)
Model 22, β (95%CI)4	Reference	0.251 (0.130, 0.372)	0.280 (0.158, 0.403)	0.531 (0.407, 0.656)	< 0.001	0.143 (0.115, 0.172)
Model 33, β (95%CI)4	Reference	0.204 (0.083, 0.325)	0.194 (0.071, 0.316)	0.378 (0.253, 0.504)	< 0.001	0.104 (0.075, 0.133)
Results of the longitudinal study
Median of UPF weight proportion	8.7%	15.8%	24.1%	38.1%
Model 11, β (95%CI)4	Reference	0.522 (0.215, 0.829)	0.662 (0.355, 0.969)	1.170 (0.862, 1.477)	< 0.001	0.323 (0.242, 0.404)
Model 22, β (95%CI)4	Reference	0.398 (0.098, 0.698)	0.419 (0.116, 0.723)	0.664 (0.357, 0.971)	< 0.001	0.166 (0.085, 0.246)
Model 33, β (95%CI)4	Reference	0.350 (0.050, 0.651)	0.341 (0.036, 0.646)	0.525 (0.214, 0.836)	0.003	0.128 (0.046, 0.210)
	Quartiles6 of the energy proportion of ultra-processed food consumption				P Trend7	Per 10% increase8
	First Quartile	Second Quartile	Third Quartile	Fourth Quartile
Results of the cross-sectional study
Median of UPF energy proportion	24.6%	40.2%	52.1%	67.3%
Model 11, β (95%CI)4	Reference	0.189 (0.064, 0.314)	0.410 (0.285, 0.536)	0.991 (0.865, 1.116)	< 0.001	0.222 (0.197, 0.247)
Model 22, β (95%CI)4	Reference	0.163 (0.042, 0.284)	0.280 (0.157, 0.402)	0.570 (0.445, 0.696)	< 0.001	0.133 (0.108, 0.158)
Model 33, β (95%CI)4	Reference	0.099 (−0.022, 0.220)	0.165 (0.041, 0.288)	0.384 (0.256, 0.511)	< 0.001	0.093 (0.067, 0.118)
Results of the longitudinal study
Median of UPF energy proportion	30.5%	42.6%	51.9%	63.1%
Model 11, β (95%CI)4	Reference	0.272 (−0.035, 0.579)	0.647 (0.339, 0.955)	0.923 (0.614, 1.233)	< 0.001	0.260 (0.183, 0.337)
Model 22, β (95%CI)4	Reference	0.268 (−0.032, 0.568)	0.560 (0.255, 0.864)	0.639 (0.323, 0.955)	< 0.001	0.186 (0.106, 0.265)
Model 33, β (95%CI)4	Reference	0.216 (−0.085, 0.517)	0.459 (0.152, 0.766)	0.476 (0.155, 0.798)	0.001	0.142 (0.061, 0.223)

CI, confidence interval; SD, standard deviation; UPF, ultra-processed food; β, the coefficient of linear regression

1 Model 1 was adjusted for age and sex

2 Model 2 was adjusted for Model 1 variables plus ethnicity group, education, income, Townsend deprivation index, body mass index category, smoking, alcohol, physical activity, and total energy intake

3 Model 3 was adjusted for Model 2 variables plus healthy diet score

4 β and 95% CI were derived from linear regression models

5 Cut-offs for quartiles of UPF weight proportion were 10.3%, 18.7%, and 30.4% for the cross-sectional study, and 12.5%, 19.6%, and 29.5% for the longitudinal study

6 Cut-offs for quartiles of UPF energy proportion were 33.4%, 46.1%, and 58.7% for the cross-sectional study, and 37.5%, 47.1%, and 56.6% for the longitudinal study

7 P trend was derived from linear regression models by using the median of UPF weight or energy proportion

8 Per 10% increase was calculated by multiplying the weight or energy proportion of UPF consumption by 10

The results of the longitudinal study, presented in Table 2, exhibit trends similar to results observed in the cross-sectional study. Both the UPF weight proportion and UPF energy proportion consumption are associated with accelerated biological aging. Specifically, higher UPF intake levels are correlated with an increase in the age gap of 0.525 years (95% CI: 0.214, 0.836) for UPF weight proportion, and 0.476 years (95% CI: 0.155, 0.798) for UPF energy proportion, as indicated in Model 3.

Furthermore, the multivariable-adjusted RCS, employing the same adjustment variables as in model 3, revealed non-linear relationships between UPF consumption and the age gap in the cross-sectional study. P non-linear values were 0.040 for the UPF weight proportion and 0.003 for the UPF energy proportion, as shown in Figs. 2A and 2B. In contrast, the longitudinal study demonstrated linear relationships between UPF consumption and the age gap, as illustrated in Figs. 2C and 2D, with p non-linear values of 0.054 for the UPF weight proportion and 0.714 for the UPF energy proportion. In conclusion, there is a positive relation between UPF intake and age gap.

Fig. 2.

RCS between UPF consumption and the phenotypic age gap. A, RCS between UPF weight proportion and phenotypic age gap among the cross-sectional study population. B, RCS between UPF energy proportion and phenotypic age gap among the cross-sectional study population. C, RCS between UPF weight proportion and phenotypic age gap among the longitudinal study population. D, RCS between UPF energy proportion and phenotypic age gap among the longitudinal study population. Abbreviations: RCS, restricted cubic spline; β, the coefficient of linear regression; CI, confidence interval; UPF, ultra-processed food. Regression coefficients β and 95% CIs obtained from a linear regression model adjusted for age, sex, ethnicity group, education, income, Townsend deprivation index, body mass index category, smoking, alcohol, physical activity, total energy intake, and healthy diet score. The reference value of UPF weight or energy proportion consumption is set 0

Mediation analyses

The result of the association between adiposity measures, including BMI, waist circumference, hip circumference, body fat percentage, and waist-to-height ratio, and the age gap can be seen in Supplementary Table 3, which indicates a strong association between obesity and the age gap.

As shown in Table 3, the analysis explored associations of UPF weight proportion consumption on phenotypic age in the longitudinal study population, focusing on BMI, waist circumference, hip circumference, body fat percentage, and waist-to-height ratio as mediating factors. For the mediator BMI, the total effect was statistically significant, with an estimate and 95% CI of 1.7641 (0.8673,2.5816). The average causal mediation effect (ACME) was 0.6484 (0.4562,0.8625), while the average direct effect (ADE) was 1.1156 (0.2473,1.9327). Notably, the average proportion of the total effect mediated by BMI in the pathway between UPF weight proportion consumption and phenotypic age accelerated aging was 30.76% (22.72%, 72.78%). As for other mediators, average mediated proportions were 33.58% (20.92%, 61.83%) for the waist circumference, 26.40% (15.76%, 56.62%) for the hip circumference, 25.80% (15.14%, 51.21%) for the body fat percentage, and 34.89% (20.96%, 70.33%) for the waist-to-height ratio. Additionally, the adiposity measures do not show the mediator effects between the UPF energy proportion consumption and phenotypic age.

Table 3.

Mediation analysis to evaluate whether adiposity measures mediated the association ultra-processed food consumption and biological age¹

Mediator	Effects	Estimate	95% CI2	P value
Mediation results of the association between UPF weight proportion and phenotypic age
BMI	Total Effect	1.7641	(0.8673,2.5816)	< 0.001
	ACME (average)	0.6484	(0.4562,0.8625)	< 0.001
	ADE (average)	1.1156	(0.2473,1.9327)	0.008
	Prop. Mediated (average)	0.3676	(0.2272,0.7278)	< 0.001
Waist circumference	Total Effect	1.7641	(0.8563,2.6531)	< 0.001
	ACME (average)	0.5924	(0.3866,0.7697)	< 0.001
	ADE (average)	1.1717	(0.3092,2.0259)	< 0.001
	Prop. Mediated (average)	0.3358	(0.2092,0.6183)	< 0.001
Hip circumference	Total Effect	1.7641	(0.8871,2.6107)	< 0.001
	ACME (average)	0.4657	(0.3150,0.6395)	< 0.001
	ADE (average)	1.2983	(0.4003,2.1167)	0.004
	Prop. Mediated (average)	0.2640	(0.1576,0.5662)	< 0.001
Body fat percentage	Total Effect	1.7641	(0.9611,2.5882)	< 0.001
	ACME (average)	0.4551	(0.3050,0.5997)	< 0.001
	ADE (average)	1.3090	(0.5057,2.2057)	0.004
	Prop. Mediated (average)	0.2580	(0.1514,0.5121)	< 0.001
Waist-to-height ratio	Total Effect	1.7641	(0.8989,2.6480)	< 0.001
	ACME (average)	0.6155	(0.4291,0.8248)	< 0.001
	ADE (average)	1.1485	(0.2713,2.0073)	0.004
	Prop. Mediated (average)	0.3489	(0.2096,0.7033)	< 0.001
Mediation results of the association between UPF energy proportion and phenotypic age
BMI	Total Effect	1.2508	(0.4365,2.0795)	0.004
	ACME (average)	−0.1843	(−0.3667,0.0070)	0.060
	ADE (average)	1.4350	(0.6515,2.2621)	< 0.001
	Prop. Mediated (average)	−0.1473	(−0.5687,0.0076)	0.064
Waist circumference	Total Effect	1.2508	(0.3974,2.0378)	0.004
	ACME (average)	−0.0205	(−0.2041,0.1453)	0.708
	ADE (average)	1.2712	(0.4305,2.0724)	0.004
	Prop. Mediated (average)	−0.0164	(−0.2797,0.1179)	0.712
Hip circumference	Total Effect	1.2508	(0.4886,2.0808)	0.004
	ACME (average)	0.0078	(−0.1396,0.1624)	0.876
	ADE (average)	1.2429	(0.4588,2.0421)	0.004
	Prop. Mediated (average)	0.0063	(−0.1482,0.1449)	0.880
Body fat percentage	Total Effect	1.2508	(0.3563,2.0337)	0.008
	ACME (average)	−0.0173	(−0.1805,0.1096)	0.736
	ADE (average)	1.2681	(0.3716,2.0341)	0.008
	Prop. Mediated (average)	−0.0139	(−0.2381,0.0940)	0.744
Waist-to-height ratio	Total Effect	1.2508	(0.4468,2.0482)	< 0.001
	ACME (average)	−0.0179	(−0.2070,0.1589)	0.776
	ADE (average)	1.2687	(0.4704,2.0678)	< 0.001
	Prop. Mediated (average)	−0.0143	(−0.2594,0.1382)	0.776

ACME, average causal mediation effect; ADE, average direct effect; BMI, body mass index; CI, confidence interval; Prop, proportion

¹Mediation analyses were adjusted for age, sex, ethnicity group, education, income, TDI, BMI category, smoking, alcohol, physical activity, total energy intake, and healthy diet score

²Confidence intervals were obtained using 500 bootstrap repetitions by the R package “mediation”

Stratified analyses

Supplementary Fig. 2 presents the results of a stratified analysis examining the relationship between the weight proportion of UPF consumption and the age gap for cross-sectional study. Across various subgroups stratified by sex, age, BMI, smoking, education, TDI, and income, higher UPF intake was significantly associated with a greater likelihood of accelerated aging in phenotypic age, particularly in the fourth quartile. While the second and third quartiles showed heterogeneous patterns across subgroups. Significant modification effect (P < 0.05) emerged for most stratification variables except age, smoking, TDI, and income. The other subgroup results were in Supplementary Fig. 3, 4 and 5.

Sensitivity analyses

Sensitivity analyses were further adjusted for 16 chronic diseases in both cross-sectional and longitudinal studies for the association between UPF consumption and phenotypic aging (Supplementary Table 4) and the mediation analysis result (Supplementary Table 5). Overall, the results from the sensitivity analyses did not materially alter our primary findings.

Discussion

This study aimed to find the association between UPF consumption and biological aging, as well as the mediating effect of adiposity measures. We found that participants in the highest quartile of UPF weight proportion consumption exhibited an average acceleration of biological aging for phenotypic age compared to those in the lowest quartile of UPF consumption. Additionally, adiposity measures accounted for significant proportions of the mediation effect in these associations.

Our findings indicated that biological aging was associated with higher consumption of UPF, which is consistent with previous studies. Earlier studies have demonstrated that UPF is associated with shorter TL [24] and accelerated machine learning-based biological aging [25]. Additionally, studies focusing on specific food items have yielded similar results. For example, the consumption of sugary beverages has been linked to shorter TL [33], and butter intake has also been associated with shorter TL [34]. Furthermore, processed meat consumption has been shown to negatively impact TL [35]. Collectively, these studies [33–37] indicated that the components of UPF had an association with accelerated biological aging, which aligns with our findings.

Adiposity measures accounted for substantial proportions of the mediation effect, ranging from 25.80% to 36.76%, between UPF weight proportion consumption and biological aging. This finding is consistent with previous results showing that BMI status partly mediates healthy aging [38]. In addition, several studies have shown a strong connection between higher UPF consumption and obesity [12–14, 39], and obesity is regarded as a sign of biological aging [40]. The results of our mediation analysis further elucidate the relationship between these findings, suggesting that high UPF consumption contributes to biological aging through its association with adiposity measures.

In our stratified analyses, although age, smoking, TDI, and income subgroups showed statistical significance in phenotypic age, the coefficients and 95% CIs only showed differences in the second or third quartile, which might be caused by insufficient UPF intake. These results were consistent with the finding that higher UPF consumption is associated with biological aging [24, 25].

Several mechanisms may explain the association between higher UPF consumption and accelerated biological aging mediated by obesity. First, UPFs are often high in sugar and saturated fats, and low in nutrients. Long-term consumption may lead to weight gain and increased BMI. Furthermore, the mechanism linking high UPF consumption and obesity is well-established [12–14]. High BMI is used to assess obesity, and obesity is the main risk factor for metabolic diseases, including type 2 diabetes [41], liver diseases [42], and cardiovascular diseases [43]. These metabolic diseases may accelerate biological aging [44]. In addition, obesity is associated with chronic inflammation [45] and metabolic disorders [46], which can lead to cell damage and senescence, further contributing to accelerated aging [47].

To the best of our knowledge, this is the first longitudinal study to explore the mediating effect of various adiposity measures on the association between UPF consumption and biological aging. Using phenotypic age as a measure of biological age offers a comprehensive assessment of an individual's health by integrating various physiological and biochemical indicators. Additionally, our study utilized a large sample size of participants based on the UK Biobank, enhancing the reliability of our findings regarding the connection between UPF consumption and biological aging.

However, this study has some limitations. First, biological aging is a holistic estimation, and an individual’s biological aging may vary among different organs. Further research on organ-specific biological aging is necessary. Second, our sample may not fully represent the broader population, as included participants exhibited higher socioeconomic status and healthier baseline profiles than excluded individuals, potentially limiting generalizability [48]. Nevertheless, this does not undermine the validity of the findings when the focus is on understanding the relationship between exposure and a specific health outcome [49, 50]. Third, genetic, environmental [51], and psychological factors [52] may also influence biological aging. Finally, additional validations are necessary, particularly in diverse cohorts encompassing various demographic and healthy characteristics.

Conclusion

This study found that higher consumption of UPF is positively associated with accelerated biological aging in phenotypic age. adiposity measures were mediating factors between UPF consumption and biological aging. Therefore, public health initiatives aimed at reducing UPF consumption should be prioritized to help slow biological aging. Further experimental studies are necessary to validate our conclusions.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

A levels

Advanced levels

ACME

Average causal mediation effect

ADE

Average direct effect

AS

Advanced subsidiary levels

BMI

Body mass index

CI

Confidence interval

CKD

Chronic kidney disease

CSEs

Certificate of secondary education, national vocational qualifications

GCSEs

General certificate of secondary education

GBP

Great Britain pound

HND

Higher national diploma

ICD

International classification of diseases

MAE

Mean absolute error

MET

Metabolic equivalent task

NC

Higher national certificates

O levels

Ordinary levels

PA

Physical activity

Prop

Proportion

r

Correlation coefficient

RCS

Restricted cubic spline

SD

Standard deviation

TDI

Townsend deprivation index

TL

Telomere length

UPF

Ultra-processed food

β

Coefficient of linear regression

Authors’ contributions

The authors’ contributions were as follows – JL, LL: Conceptualization; JL, QH: Investigation, Software, Validation, Writing – Original Draft; LL: Resources, Supervision; JL: Formal analysis, Visualization; JL, QH, and LL: Writing – Review & Editing; and all authors have read and approved the final manuscript.

Funding

The work described in this paper was supported by grants from the City University of Hong Kong start-up grant (Project No. 9610576).

Data availability

Data are available with the permission of UK Biobank (https://www.ukbiobank.ac.uk/).

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval

This is an observational study. UK Biobank has approval from the North West Multi-centre Research Ethics Committee as a Research Tissue Bank approval (11/NW/0382, 16/NW/0274, 21/NW/0157). In addition, all participants provided informed consent when collecting data (available at https://www.ukbiobank.ac.uk/explore-your-participation/basis-of-your-participation).