Early life sugar rationing and ageing related diseases, biological ageing and mortality Authors: Jiazhen Zheng, Zhen Zhou, Jinghan Huang, Qiang Tu, Haisheng Wu, Quan Yang, Qi Qi, Wenbo Huang, Qingwen Ren, Junchun Shen, Xinlong Shi, Can Chen, Chuang Yang, Daniel J Cuthbertson Published: 2026 ABSTRACT AbstractEarly-life nutrition may influence lifelong ageing, yet human evidence is scarce. Using Britain’s postwar sugar rationing as a natural experiment, we examine its long-term effects in 64,809 United Kingdom Biobank participants. Exposure to sugar rationing during the first 1,000 days of life is associated with a 9% lower incidence of hallmark-related disease, with a hazard ratio of 0.91 and a 95% confidence interval of 0.88–0.94, and a 19% lower risk of all-cause mortality, with a hazard ratio of 0.81 and a 95% confidence interval of 0.69–0.93. Mediation analysis indicates that the survival association is statistically mediated, by approximately 60%, through differences in incident hallmark-related disease. Rationed individuals show 1.0–1.2-year younger biological ages across multiple clocks and lower organ ages, particularly in the lung, heart, and liver. Proteomic profiling identifies 47 altered proteins, with enrichment of adenosine monophosphate-activated protein kinase and longevity pathways and suppression of mechanistic target of rapamycin signaling. These findings are consistent with international recommendations to limit free or added sugars from the World Health Organization, United States Dietary Guidelines, and American Heart Association, and may inform policy discussions related to sugar taxation and infant food and marketing policies under the United Nations 2030 Agenda.Subject terms: Risk factors, Health policy Introduction The Developmental Origins of Health and Disease (DOHaD) hypothesis has previously proposed that the first 1000 days of an individual’s life, from conception, intra-uterine exposure and early development through the second year of life, represents a critical time frame when a variety of environmental exposures (including dietary) profoundly and permanently influence the developmental health and well-being outcomes of that individual1. Such early exposures influence an array of metabolic processes and organ development with a resultant effect on chronic non-communicable disease trajectories across the lifespan and on long-term mortality2,3. As an extreme example, evidence from famine and undernutrition cohorts has consistently demonstrated that severe deprivation during this period predisposes individuals to cardiometabolic disease, impaired growth, and premature mortality4. Yet, far less is known about the long-term consequences of more moderate dietary restrictions, particularly regarding more selective macronutrient restriction, such as sugar restriction. Excessive intake of free or added sugars is strongly linked to childhood dental caries and is also associated with higher risks of incident obesity, type 2 diabetes (T2D), cardiovascular disease (CVD), obesity-related cancers, and dementia in adulthood5–7. Consequently, global dietary guidelines emphasize reducing free/added sugars: the World Health Organization (WHO) recommends keeping free sugars below 10% (and preferably below 5%) of total energy intake8, while the U.S. Dietary Guidelines advise limiting added sugars to below 10%9, and the UK Eatwell Guide promotes lower-sugar choices and reformulation10. Despite these policies, global sugar consumption remains persistently high, largely driven by sugar-sweetened beverage consumption11. Britain’s post–World War II sugar rationing provides a unique natural experiment to address this important question. Unlike famine, rationing constrained access to added sugars without broad caloric deprivation, offering a rare opportunity to isolate the long-term impact of early-life sugar reduction to determine the impact on a variety of ageing-related outcomes. The rationing system imposed strict limits on sugar and sweets consumption, distributing scientifically calculated weekly allowances12. While infants <2 years were excluded from sugar and sweet allocations13, all individuals (including pregnant women) received ~8 ounces of sugar/week and 12 ounces of sweets monthly14–16, resulting in daily sugar intake levels closely aligning with current dietary guidelines17–19. Thus, adults and children consumed no more than 40 and <15 grams/day. Following the end of sugar rationing in September 1953, sugar consumption rose sharply20 such that the average adult daily sugar intake for adults nearly doubled between early 1953 and late 1954 (from 41 to ~80 g) with other nutrient intake relatively stable21. This abrupt shift in daily sugar consumption offers a natural experiment by creating quasi-random early-life sugar exposure. Recent analyses using UK Biobank data indeed supported the DOHaD hypothesis by showing that in-utero/early-life exposure to sugar rationing was associated with substantially lower risks and a delayed onset of T2D and other cardiometabolic risk factors with stronger protective effects observed for longer exposure durations21. These findings suggest that moderate early-life sugar restriction specifically may reduce the risk of a variety of chronic long-term conditions, extending into later adulthood. Cellular ageing is driven by nine hallmark processes: altered intercellular communication, cellular senescence, deregulated nutrient sensing, epigenetic alterations, genomic instability, loss of proteostasis, mitochondrial dysfunction, stem cell exhaustion, and telomere attrition. These processes underlie many age-related diseases, and recent research has linked obesity to an increased risk of several hallmark-related conditions22. Understanding whether sugar restriction during the first 1000 days of life has chronic, longer-term consequences for slowing down biological aging and preventing/delaying onset of incident disease could inform nutrition policy from a life-course perspective. In this study, using UK Biobank data, we comprehensively examined the life-course association of early-life sugar rationing during the first 1000 days of life, across multiple biological layers, examining key hallmarks of ageing in later life. We explored the association of sugar rationing with ageing-related molecular/proteomic signatures, organ-specific biological ageing, risks of developing up to 83 diseases associated with the ageing-related hallmarks and all-cause mortality. Results Post-rationing dietary shifts in the UK Post-rationing increases in sugar intake in the UK have previously been reported by Tadeja Gracner et al.21. For clarity, we recompiled and visualized these dietary changes. NFS data show that sugar intake rose by more than 50% after rationing, fat intake increased modestly, and most other food groups changed little (Supplementary Fig. 1A). Overall calorie intake rose by only ~5% (Supplementary Fig. 1B), driven mainly by sugar and, to a lesser extent, fat (Supplementary Fig. 1C). Patterns of sugar intake were similar across social classes A–D (Supplementary Fig. 1D). Baseline characteristics of rationed and non-rationed participants Participant inclusion, exclusion, and grouping are shown in Fig. 1. Participants exposed to sugar rationing were slightly older at baseline (mean age 55.4 vs. 53.2 years, P < 0.001, SMD = 2.012) (Table 1). The rationed group had a lower household income, with 25.9% earning <£18,000 compared to 23.7% in the non-rationed group (P < 0.001, SMD = 0.079). They were also less likely to be current smokers (10.9% vs. 12.0%) and less likely to have parents still alive (13.1% vs. 19.6%). After excluding 2,499 individuals with prevalent diabetes, participants exposed to early-life sugar rationing showed modestly lower adult sugar intake. In fully adjusted models, total sugars were 4.14 g/day lower (95% CI, −8.15 to −0.14; p = 0.043) and non-milk extrinsic sugars were 3.12 g/day lower (95% CI, −6.04 to −0.20; p = 0.036). Glucose was also lower (β = −0.079 mmol/L; p < 0.001), whereas HbA1c did not differ significantly (Supplementary Table 1). Study flowchart and distribution of participants according to early-life sugar rationing exposure. A Flowchart depicting the study cohort; B Distribution of birth samples in the UK Biobank by calendar month and their exposure to sugar rationing. The sugar-rationed group is shown in blue, while the group unexposed to sugar rationing is shown in pink. The first group of individuals who were never exposed to sugar rationing has been labeled and used as the control group in Fig. 4. Table 1Baseline characteristics of participants by sugar rationing exposureTotalNon-RationedRationedP valueSMDNo. of participants64,80923,93740872Age at entry, mean (SD), years54.6 (1.6)53.2 (1.0)55.4 (1.2)<0.0012.012Sex Female36,655 (56.6)13,566 (56.7)23089 (56.5)0.6560.004 Male28,154 (43.4)10,371 (43.3)17783 (43.5)Place of birth England55,762 (86.0)20,490 (85.6)35,272 (86.3)0.0010.031 Scotland5,952 (9.2)2,274 (9.5)3,678 (9.0) Wales3,095 (4.8)1,173 (4.9)1,922 (4.7)Birth month Dec 1-Feb 2817,556 (27.1)6627 (27.7)10929 (26.7)<0.0010.206 Jun 1-Aug 3114,612 (22.5)5835 (24.4)8777 (21.5) Mar 1-May 3116,810 (25.9)4922 (20.6)11888 (29.1) Sep 1-Nov 3015,831 (24.4)6553 (27.4)9278 (22.7)Ethnic Non-white2519 (3.9)1006 (4.2)1513 (3.7)0.0040.025 White62290 (96.1)22931 (95.8)39359 (96.3)Education, % Below A levels26660 (41.1)9696 (40.5)16964 (41.5)<0.0010.041 A levels19559 (30.2)7456 (31.1)12103 (29.6) College/university8838 (13.6)3320 (13.9)5518 (13.5) Professional/other9752 (15.0)3465 (14.5)6287 (15.4)Household income, n (%), £ <18,00016267 (25.1)5683 (23.7)10584 (25.9)<0.0010.079 18,000-30,99911979 (18.5)4224 (17.6)7755 (19.0) 31,000-51,99916680 (25.7)6205 (25.9)10475 (25.6) 52,000-100,00015231 (23.5)5968 (24.9)9263 (22.7) >100,0004652 (7.2)1857 (7.8)2795 (6.8)Townsend deprivation index, mean (SD)−1.3 (3.1)−1.2 (3.1)−1.3 (3.1)<0.0010.035Smoking status, % Current7357 (11.4)2882 (12.0)4475 (10.9)<0.0010.058 Never36465 (56.3)13681 (57.2)22784 (55.7) Previous20987 (32.4)7374 (30.8)13613 (33.3)Alcohol consumption frequency, times/wk <336094 (55.7)13460 (56.2)22634 (55.4)0.0360.017 >=328715 (44.3)10477 (43.8)18238 (44.6)BMI, mean (SD), kg/m226.6 (2.0)26.6 (2.0)26.6 (2.0)0.8380.002Summed MET minutes per week for all activity2496.4 (2410.2)2510.5 (2441.4)2488.1 (2391.7)0.2530.009Parental medical conditions CVD36512 (56.3)13197 (55.1)23315 (57.0)<0.0010.039 Hypertension29805 (46.0)11434 (47.8)18371 (44.9)<0.0010.057 Diabetes12434 (19.2)4804 (20.1)7630 (18.7)<0.0010.035Parents still alive10043 (15.5)4684 (19.6)5359 (13.1)<0.0010.175Maternal smoking around birth18565 (28.6)6829 (28.5)11736 (28.7)0.6220.004Breastfed as a baby38152 (58.9)14014 (58.5)24138 (59.1)0.2040.01Birth weight, mean (SD), kg3.3 (0.5)3.3 (0.5)3.3 (0.5)0.9170.001Values are presented as mean (SD) for continuous variables and number (percentage) for categorical variables. P values were calculated using Mann-Whitney U test for continuous variables and chi-square tests for categorical variables. P values were two-sided.SMD standardized mean difference (an absolute SMD > 0.1 indicates a potentially meaningful difference between groups); CVD cardiovascular disease; BMI: body mass index; MET metabolic equivalent of task. Baseline characteristics of participants by sugar rationing exposure Values are presented as mean (SD) for continuous variables and number (percentage) for categorical variables. P values were calculated using Mann-Whitney U test for continuous variables and chi-square tests for categorical variables. P values were two-sided. SMD standardized mean difference (an absolute SMD > 0.1 indicates a potentially meaningful difference between groups); CVD cardiovascular disease; BMI: body mass index; MET metabolic equivalent of task. Associations of early-life sugar rationing with diseases related to the hallmarks of cellular ageing and all-cause mortality Involvement of hallmarks in incident disease There was extensive overlap in the involvement of the nine aging hallmarks in incident disease development, with many shared occurrences across the nine aging hallmarks (multi-hallmark intersections) (Supplementary Fig. 2). Composite incidence of all-hallmark related diseases Compared with the non-rationed group, the rationed group showed a lower cumulative incidence of a composite measure of all 83 hallmark-related diseases (HR 0.91, 95% CI 0.88–0.94) (Supplementary Fig. 3). In subgroup analyses examining the impact of sex, birthplace, ethnicity, and parental cardiometabolic history, early-life sugar rationing was consistently associated with an ~5–10% lower incidence of hallmark-related diseases (HRs≈0.89–0.94), with no significant interactions (Supplementary Fig. 4). Specifically, ethnicity-stratified analyses showed consistent associations across groups: the HRs were 0.92 (95% CI 0.89–0.95) in White participants and 0.90 (95% CI 0.82–0.99) in non-White participants, with no evidence of effect modification by ethnicity (P for interaction = 0.743; Supplementary Fig. 4). Incidence of hallmark-related diseases Examining each disease separately, early-life sugar rationing was associated with lower risks across a broad range of hallmark-related diseases in later life (Fig. 2). Diseases were broadly categorized by involvement of specific ageing hallmark(s), with many mapping to multiple hallmarks. Notable examples of the reduced risks include breast cancer (HR 0.82, 95% CI 0.70–0.95), relating to genomic instability; viral infection (HR 0.75, 95% CI 0.63–0.88), relating to cellular senescence and type 2 diabetes (T2D; HR 0.79, 95% CI 0.71–0.87) and fatty liver (HR 0.61, 95% CI 0.49–0.75) related to deregulated nutrient sensing and lower respiratory tract infection (HR 0.85, 95% CI 0.77–0.94) related to altered intercellular communication. The lowest HR was observed for polycythaemia vera (HR 0.43; 95% CI, 0.19–0.99). Associations between early-life rationing and incident diseases related to hallmarks of cellular ageing. Hazard ratios were obtained from Cox proportional hazards models. Data are presented as hazard ratio point estimates with 95% confidence intervals. Error bars represent 95% confidence intervals. P values were two-sided. To account for multiple testing across 83 disease outcomes, Bonferroni correction was applied, with a significance threshold of P ≤ 0.0006. Exact P values are shown in the figure where available. Model 1 adjusted for age, sex, ethnicity. Model 2 additionally adjusted for month of birth, place of birth, real food prices adjusted for consumer price index, parental disease history (CVD, diabetes, hypertension), maternal smoking around birth, whether breastfed as a baby and survey year. AIC altered intercellular communication; CS cellular senescence; DNS deregulated nutrient sensing; EA epigenetic alterations; GI genomic instability; HR hazard ratio; LOP loss of proteostasis; MD mitochondrial dysfunction. SCE stem cell exhaustion. TA telomere attrition. NHL non-hodgkin’s lymphoma, T2DM type 2 diabetes mellitus, SIADH syndrome of inappropriate antidiuretic hormone, MGUS monoclonal gammopathy of unknown significance, RA rheumatoid arthritis, CHD coronary heart disease (not otherwise specified), MI myocardial infarction, DCM dilated cardiomyopathy, HCM hypertrophic cardiomyopathy, AF atrial fibrillation, CKD chronic kidney disease, ESRD end stage renal disease, BPH benign prostatic hyperplasia, MND motor neurone disease, AMD macular degeneration, PAD peripheral arterial disease, COPD chronic obstructive pulmonary disease. Relationship of incident risk with duration of sugar rationing exposure Longer early-life exposure durations were associated with a clear protective effect in the risk of developing the nine hallmark-related disease categories (Fig. 3), although solely in utero exposure to sugar rationing did not show an association. Spline analyses confirmed such a dose–response pattern, with progressively lower disease risks for increasing duration of early-life sugar rationing exposure (Fig. 4). No significant differences in disease incidence were observed among the various non‑rationed groups. Associations between early-life rationing at distinct developmental stages and incident diseases related to hallmarks of cellular ageing. HRs were estimated using Cox proportional hazards models to assess the association between exposure to food rationing during three critical developmental windows, in utero, in utero + (0, 1] year, and in utero + (1, 2] year, and the incidence of diseases reflecting hallmarks of cellular ageing. Data are presented as hazard ratio point estimates with 95% confidence intervals. Error bars represent 95% confidence intervals. P values were two-sided. Model 1 was adjusted for age, sex, and ethnicity. Model 2 was additionally adjusted for month of birth, place of birth, real food prices (adjusted for consumer price index), parental history of cardiovascular disease, diabetes, or hypertension, maternal smoking around birth, whether breastfed as a baby, and survey year. Exact P values are shown in the figure where available. To account for multiple testing across 27 category-level comparisons, including nine hallmark-related disease categories and three exposure windows, Bonferroni correction was applied, with a significance threshold of P ≤ 0.0019. Cubic spline models of the association between month of birth (as a proxy for early-life exposure to sugar rationing) and risk of diseases related to hallmarks of cellular ageing. Each panel displays the hazard ratio (HR) and 95% confidence interval (shaded area) for incident diseases corresponding to a specific hallmark of aging: A genomic instability; B telomere attrition; C epigenetic alterations; D loss of proteostasis; E deregulated nutrient sensing; F mitochondrial dysfunction; G cellular senescence; H stem cell exhaustion; and I altered intercellular communication. We classified exposure by birth month relative to the end of sugar rationing (September 1953) into nine groups: in-utero + 24 months (born Oct-1951–Mar-1952), in-utero + 18 months (Apr-1952–Sep-1952), in-utero + 12 months (Oct-1952–Mar-1953), in-utero + 6 months (Apr-1953–Sep-1953), in-utero only (Oct-1953–Jun-1954), and four no-exposure groups born ≥9 months (Jul-1954–Dec-1954) (reference group), ≥15 months (Jan-1955–Jun-1955), ≥21 months (Jul-1955–Dec-1955), and >21 months (Jan-1956–Mar-1956) after rationing ended. HRs were estimated from Cox proportional hazards models adjusted for age, sex, ethnicity, month of birth, place of birth, real food prices (adjusted for consumer price index), parental history of cardiovascular disease, diabetes, or hypertension, maternal smoking around birth, whether breastfed as a baby, and survey year. Solid lines represent adjusted hazard ratio estimates, and shaded bands represent 95% confidence intervals. P values for nonlinearity were two-sided and derived from restricted cubic spline models. Dashed vertical lines represent the end of sugar rationing in September 1953. Exact P values for nonlinearity are shown in each panel where available. All-cause mortality and mediation by hallmark-related diseases Participants in the rationed group had a lower risk of all-cause mortality compared with the non-rationed group (HR 0.78, 95% CI 0.66–0.90, basic model; HR 0.81, 95% CI 0.69–0.93, multivariable model). Mediation analysis indicated that the reduced incidence of hallmark-related diseases accounted for 57.6% (95% CI 51.7–63.9) and 60.4% (95% CI 54.3–66.1) of the survival benefit in the basic and multivariable models respectively (Supplementary Fig. 5). Association between early-life sugar rationing and biological age differences and telomere length Contour plots showed that rationed individuals, despite being chronologically older, had smaller biological age deviations (Supplementary Fig. 6). Compared to the non-rationed group, the rationed group exhibited larger negative biological age differences for PhenoAge (–1.16 years, 95% CI: –1.58 to –0.77, p < 0.001), KDM Biological Age (–1.03 years, 95% CI: –1.43 to –0.61, p < 0.001), and MileAge (–1.05 years, 95% CI: –1.42 to –0.57, p = 0.001). Additionally, the rationed group showed a higher Z-adjusted T/S log value (β = 0.15, 95% CI: 0.04 to 0.25, p = 0.003) (Table 2). Table 2Association of early-life sugar rationing with biological age differencesNon-rationed (Mean ± SD)Rationed (Mean ± SD)Rationed vs. non-rationed Model 1, β (95% CI)p value (Model 1)Rationed vs. non-rationed Model 2, β (95% CI)p value (Model 2)PhenoAge−5.82 ± 1.23−7.11 ± 1.21−1.25 (−1.64, −0.83)<0.001−1.16 (−1.58, −0.77)<0.001KDM-Biological Age0.58 ± 1.55−0.84 ± 1.51−1.11 (−1.53, −0.72)<0.001−1.03 (−1.43, −0.61)<0.001DNAm PhenoAge−1.40 ± 1.91−2.21 ± 1.78−0.77 (−1.63, 0.09)0.064−0.65 (−1.52, 0.20)0.076MileAge2.42 ± 1.571.11 ± 1.44−1.10 (−1.47, −0.73)<0.001−1.05 (−1.42, −0.57)0.001Proteomic Age−2.83 ± 1.95−3.37 ± 1.88−0.52 (−1.47, 0.45)0.088−0.44 (−1.37, 0.47)0.103Z-adjusted T/S log0.016 ± 1.00−0.038 ± 1.010.18 (0.08, 0.29)0.0010.15 (0.04, 0.25)0.003Estimates were derived from linear regression models. Biological age differences represent the deviation between estimated biological age and chronological age (Biological Age−Chronological Age); thus, β reflects the adjusted between-group difference in age gap (rationed vs non-rationed), where negative values indicate delayed biological ageing relative to chronological age. P values were two-sided.Model 1: Adjusted for age, sex, ethnicity. Model 2: Additionally adjusted for month of birth, place of birth, real food prices adjusted for consumer price index, parental disease history (CVD, diabetes, hypertension), maternal smoking around birth, whether breastfed as a baby and survey year.PhenoAge, Phenotypic Age; KDM Biological Age, Klemera-Doubal Method Biological Age; DNAm PhenoAge, DNA methylation PhenoAge; MileAge, Metabolomic Age. Association of early-life sugar rationing with biological age differences Estimates were derived from linear regression models. Biological age differences represent the deviation between estimated biological age and chronological age (Biological Age−Chronological Age); thus, β reflects the adjusted between-group difference in age gap (rationed vs non-rationed), where negative values indicate delayed biological ageing relative to chronological age. P values were two-sided. Model 1: Adjusted for age, sex, ethnicity. Model 2: Additionally adjusted for month of birth, place of birth, real food prices adjusted for consumer price index, parental disease history (CVD, diabetes, hypertension), maternal smoking around birth, whether breastfed as a baby and survey year. PhenoAge, Phenotypic Age; KDM Biological Age, Klemera-Doubal Method Biological Age; DNAm PhenoAge, DNA methylation PhenoAge; MileAge, Metabolomic Age. Organ-specific associations of lifestyle and early-life sugar rationing with multi-organ biological aging Proteomic analyses were performed on a subset with baseline characteristics of participants shown in Supplementary Table 2. Organ-specific aging measures were moderately and positively intercorrelated (Supplementary Fig. 7). The heatmap shows the associations between 20 lifestyle factors, both adverse (e.g., smoking, alcohol) and healthy (e.g., oily fish and vigorous exercise) factors and early-life sugar rationing with age gaps across 13 organ systems (Fig. 5, left). Early-life sugar rationing was associated with lower age gaps in 6 organs, a similar effect on organ health to vigorous exercise and oily fish intake. Smoking and alcohol were linked to higher age gaps in up to 9 organs (|log₁₀(q)| > 1.3). In a direct comparison, rationed individuals had consistently lower z-scored age gaps than non-rationed participants, with the largest differences seen in the lung (−0.7), heart (−0.8), and liver (−0.8) (Fig. 5, right). Associations of lifestyle and early-life sugar rationing with organ-specific biological aging. A Heatmap showing the associations between 20 lifestyle factors (including early-life sugar rationing) and age gaps across 13 organ systems. Associations were assessed using linear regression models with chronological age and sex as covariates. Non-organ-enriched proteins were exclusively used to train the organismal aging model, whereas the conventional model was developed using the complete proteomic profile. Education refers to years of schooling. Dietary variables (e.g., pork, tea, fish, meat, vegetables, bread) were coded as binary indicators (consumed vs. not consumed). Moderate and vigorous exercise represent the number of days per week with ≥10 minutes of activity. The heatmap is color-coded by signed log₁₀(q) values. Significance levels are indicated by letters: a (|log₁₀q | ≥ 1.3), b (|log₁₀q | ≥ 2), c (|log₁₀q | ≥ 3). B Comparison of z-scored age gaps between individuals who experienced early-life sugar rationing (n = 3758) and those who did not (n = 2269). Differences were analyzed across 13 organ systems. Age gaps adjusted for chronological age and z-scored within organ were used. Box plots indicate the median, interquartile range (IQR), and whiskers represent Q1 − 1.5×IQR and Q3 + 1.5×IQR. Proteomic signatures and pathway alterations associated with early-life sugar rationing Differential plasma proteomic analysis identified 47 significantly altered proteins between the rationed and non-rationed groups (FDR < 0.05, |log₂FC | > 0.25), including 24 upregulated and 23 downregulated proteins in the rationed group. KEGG pathway enrichment revealed enhanced activity in the MAPK, AMPK, and longevity regulating pathways, while aging-associated pathways such as mTOR and FoxO signaling were significantly suppressed (FDR < 0.05) (Supplementary Fig. 8). Discussion In this quasi-experimental analysis of UK Biobank adults born around the end of post-war sugar rationing, we found that limited sugar exposure in the first 1000 days after conception was associated with lower incidence of diseases linked to the hallmarks of cellular ageing, slower biological ageing across multiple omics-based clocks, and reduced all-cause mortality. The mortality advantage was substantially mediated (≈58–60%) by lower risks of hallmark-related disease, supporting a life-course model in which early-life sugar restriction shapes cellular ageing processes that modulate ageing trajectories. Our findings extended and connected several strands of prior evidence, from rationing-era natural experiments to contemporary ageing biology and are entirely consistent with the DOHaD model. Early-life nutrition has long-term consequences for adult health, as shown in extreme examples from famine and undernutrition cohorts; however, such severe deprivation typically causes harm with significant morbidity and mortality23,24. By (natural) contrast, Britain’s sugar rationing limited added sugars without broad caloric deprivation. Gracner and colleagues, using an event-study design in the UK Biobank, reported exposure to sugar rationing in the first 1000 days reduced later-life type 2 diabetes and hypertension by ~35% and ~20% respectively21, delaying disease onset by several years, effects potentiated by longer postnatal exposure. Our results provide similar evidence of chronic disease protection but extend these findings by demonstrating a significant reduction in a myriad of incident ageing-hallmark disease clusters and increased longevity. From a disease biology perspective, the study by Mika Kivimäki et al. mapped the risk of common age-related long-term medical conditions with characteristics hallmarks of cellular aging, revealing that obesity was similarly associated with multiple hallmark-specific pathological clusters22. We extend this paradigm by highlighting early-life sugar rationing being uniformly associated with protective effects across nine hallmark-related disease processes and a lower mortality mediated by fewer hallmark-mediated chronic diseases. Global dietary guidance increasingly prioritizes lowering free/added sugars to curb the rising and alarming tsunamis of dental caries, obesity and type 2 diabetes6. During the rationing period, sugar allocations for all individuals, including pregnant women and children, were strikingly consistent with today’s dietary guideline recommendations. Adults were limited to <40 g of sugar per day, aligning closely with WHO recommendations of limiting free sugar intake to below 10% of daily energy intake (≈50 g in a 2,000-kcal diet)17. Importantly, infants under two years of age were not allowed any added sugars, consistent with current guidance emphasizing complete avoidance in this age group18. By unintentionally providing a parallel in terms of the quantity of sugar consumption, sugar rationing has provided a natural experiment whereby the clinical outcomes of those exposed provide confirmatory support of the quantities proposed in contemporary recommendations from the WHO, the U.S. Dietary Guidelines, and the AHA17–19. In an era of pervasive ultra-processed, sugar-dense products seemingly throughout the lifespan, aligning maternal/early-life diets with modern nutrition guidance could contribute to healthier ageing. Ageing accelerates at differential rates across the multiple organs, tissues and/or physiological systems25. Recent UK Biobank proteomic clocks and emerging organ-age framework data have demonstrated that accelerated organ age (e.g., heart, brain, liver) confers elevated risk independent of chronological age25. In our study, early-life sugar rationing was associated with smaller (younger) age gaps across multiple organs, with the largest biological differences in cardiac, lung and liver health, patterning that is directionally consistent with proteomic clocks and metabolomic age models linking organ-targeted ageing to disease outcomes. These findings are particularly relevant to incident cardiovascular and respiratory diseases and to major adverse liver outcomes, notably metabolic dysfunction–associated steatotic liver disease (MASLD). Together, these data suggest that early-life sugar rationing profoundly influences organ trajectories, tipping them towards healthy tissues characterised by less biologically aged proteomic/metabolomic signatures. Biological pathways are likely to differ across early-life stages. In utero, fetal development may be influenced by maternal insulin and broader metabolic states, with downstream effects mediated through epigenetic regulation, hormonal signaling, and developmental reprogramming26. During breastfeeding, milk glucose is generally stable and not strongly driven by maternal glycemia27; however, maternal dietary changes that reduce energy/carbohydrate/sugar intake during lactation have been associated with lower insulin concentrations in human milk, supporting the plausibility that maternal diet, including added sugars, may influence milk insulin28. After weaning, complementary feeding may increase exposure to processed foods high in added sugars, potentially shaping early metabolic set points and long-term cardiometabolic risk. Although these pathways are distinct and not all directly measured in our data, they are conceptually consistent with an early-life sugar environment framework, which is the scope of our natural experiment. Another plausible pathway is that early-life sugar restriction may shift dietary trajectories across the life course. Humans show a strong innate liking for sweetness, particularly in childhood, and early experience can shape the intensity of sweet preference and sweet-food liking in the first years of life29,30. Developmental research further suggests that children’s food preferences and eating patterns are modifiable through early exposure and the family food environment, with associations between sweet taste perception/liking and sugar intake31,32. Prospective evidence also indicates that early consumption of sweetened beverages can predict higher intake later in childhood/adolescence33. Consistent with this plausibility, we observed directionally lower adult sugar intake measures and a slightly lower glucose level among individuals exposed to early-life sugar rationing; however, these differences were small in absolute magnitude and are unlikely to be clinically meaningful or to explain the observed long-term associations. Accordingly, these adult phenotypes should be interpreted as downstream, supportive context rather than evidence for a specific mechanistic pathway. While the cellular and molecular mechanisms remain incompletely understood, high sugar exposure may accelerate hallmarks such as genomic instability34, telomere attrition35, and impaired intercellular communication36. Conversely, lower early-life sugar intake could plausibly reduce oxidative stress and inflammation and influence epigenetic programming37,38, thereby modulating multi-system aging. Our proteomic findings, showing relative activation of AMP-activated protein kinase (AMPK)/longevity signaling and suppression of mTOR-related pathways, are broadly consistent with findings from caloric restriction biology and AMPK–FOXO axes described in model organisms39. These shifts may provide a mechanistic basis for the observed reductions in hallmark-related disease and their contribution to survival benefits. The strengths of this study include a quasi-experimental design (birth-year assignment around the end of rationing), a large sample size, the duration of follow-up, linkage to hospital and mortality records, and convergence across phenotypic, epigenetic, metabolomic and proteomic ageing clocks. The hallmark-based disease taxonomy offers biological interpretability that typical organ-system ICD groupings lack, and mediation analysis situates mortality benefits within an ageing-disease pathway. We must acknowledge several limitations. First, although birth timing around policy change approximates randomization, residual confounding by secular trends (diagnostics, healthcare access) cannot be fully excluded despite our 6-month cohort comparisons. Second, our exposure is defined by the sugar-rationing policy environment and eligibility rules rather than direct measurement of infant (0–2 years) sugar intake, and within-household sharing could have reduced the true between-cohort contrast. However, such non-differential exposure misclassification would be expected to attenuate associations toward the null rather than generate a spurious protective pattern aligned with exposure duration, and the consistent dose–response gradients and falsification check against smooth secular trends support the robustness of our main conclusions. Third, UK Biobank’s volunteer bias and demographic structure may limit generalizability beyond predominantly white UK adults; findings may be different considering other ethnic groups’ differential susceptibility to a variety of long-term medical conditions. Although effect estimates were directionally consistent in White and non-White participants, UK Biobank is predominantly of European ancestry and the non-White subgroup is substantially smaller, which limits precision and may constrain generalizability to specific non-European ancestry populations. Fourth, we acknowledge that national or household averages may obscure meaningful heterogeneity in sugar consumption across individuals. To strengthen inference despite this limitation, we conducted subgroup analyses by sex, race/ethnicity, place of birth, and parental health, observing broadly consistent directional associations across groups. Moreover, after extensive covariate adjustment, the association between exposure timing and outcomes remained coherent, supporting an interpretation more consistent with a policy-environment effect than with idiosyncratic individual differences. Fifth, mediation analyses relied on observed mediators and strong assumptions; unmeasured mediator–outcome confounding or alternative pathways (e.g., growth trajectories, later-life behaviors, microbiome) may contribute, so the mediation proportions should be interpreted as statistical rather than definitive mechanistic mediation. Sixth, although several biological-age metrics showed statistically significant differences in age gap (≈1–2 years), statistical significance does not necessarily translate into clinically meaningful differences at the individual level given the inherent prediction uncertainty of many biological-age clocks; thus, these findings are best interpreted as modest, directional population-level shifts. Seventh, biological-age clocks capture risk gradients but are proxies; causal mechanisms require experimental corroboration. These limitations are common to life-course natural experiments and are unlikely to negate the coherent pattern across multiple endpoints and data layers. Eighth, adult diet (Oxford WebQ) and glycaemic markers were assessed only once in adulthood and were not repeatedly assessed across the life course. Accordingly, these analyses are intended as supportive plausibility checks and should not be interpreted as definitive evidence of a single mechanism (e.g., lifelong preference), particularly given measurement variability and selection into WebQ completion. We found that rationing of sugar during the first 1,000 days of life was associated with broad and sustained benefits across organ-level aging, reducing aging-related disease clusters and improving survival, predominantly (60%) mediated by the reduced risk of these incident aging-related disease clusters. Mechanistic work and clinical trials are needed to test biological causality and assess feasibility. Our findings also provide population-level evidence base for sugar reduction not only within modern adult dietary recommendations but also for integrating early-life sugar reduction into early years (from conception/pregnancy and beyond) public-health nutritional recommendations. Policy levers include clearer feeding guidance for infants and toddlers, reformulation and marketing restrictions to reduce added sugars in early-life foods, and rigorous evaluation of sugar-focused fiscal and labeling policies throughout life. Methods Study participants The design, participant characteristics, and data collection procedures of the UK Biobank have been reported previously40. In brief, over 500,000 adults aged 40–70 years were recruited between 2006 and 2010 from the general population through 22 assessment centers in England, Scotland, and Wales. Data were collected via questionnaires, interviews, physical assessments, and linkage to electronic health records, including physiological, psychological, social, and health-related domains40. The UK Biobank received ethical approval from the North West Multi-Centre Research Ethics Committee (REC reference 21/NW/0157), and all participants provided written informed consent. This research was conducted using the UK Biobank Resource under Application Number 97089. Of the 74,213 eligible UK Biobank participants born between October 1951 and March 1956, we excluded 9223 individuals who were born outside the UK, from multiple births, or adopted. After further excluding 181 withdrawals and participants with missing biomarker data, 64,809 individuals remained for the main analyses (Fig. 1A). Ascertainment of early-life exposure to sugar rationing We applied an event study design to examine the long-term associations of limited sugar intake during the first 1,000 days after conception with hallmark related diseases and biological-age measures (organ-specific and multi-omic clocks). Birth month and year determined exposure, with individuals quasi-randomly assigned to the rationed (low sugar) or non-rationed (high sugar) group (Fig. 1B). This quasi-experimental approach has been used previously by Gracner et al.21, who also detail the 1000-day window and regression discontinuity rationale. We analyzed the National Food Survey (NFS) reports to examine quarterly dietary trends during and after the rationing period21,41,42, spanning 1950 to 1960. Detailed annual data on food intake, Food Price Index, consumer price index (CPI), and related socioeconomic indicators are available in Supplementary Material 2 or via (https://www.gov.uk/government/statistics/family-food-historic-reports). Exposure to sugar rationing was classified by duration as: no exposure, in utero only, or in utero plus up to 6, 12, 18, or 24 months postnatally. Broader categories were also used: in utero only, in utero plus the first year, in utero plus years 1–2, or no exposure. Figure 1B shows the exposure timeline and birth distribution across the study period. Assessments of covariables At baseline (i.e., at the time of UK Biobank recruitment), participants completed an interactive questionnaire collecting information on sex, age, race, place of birth (England, Wales, Scotland, or outside the UK), household income, education, Townsend deprivation index, smoking habits, alcohol consumption, physical activity, parental health history, early life conditions, number of days per week of moderate physical activity 10+ minutes, number of days per week of vigorous physical activity 10+ minutes, oily fish intake, beef intake, cooked vegetable intake, tea intake, pork intake, processed meat intake, salad raw vegetable intake, bread intake, fresh fruit intake, non-oil fish intake and poultry intake. Birth weight and maternal smoking status around birth were collected through interviews conducted by trained personnel. Below is detailed information regarding covariates: body mass index (BMI; kg/m²) is determined by dividing body weight (kilograms) by square of height (metres). The Townsend Deprivation Index is an indicator of socio-economic hardship43, based on factors such as employment status, car ownership, housing conditions and household overcrowding. Physical activity was collected via an electronic questionnaire and defined by Total Metabolic Equivalent Task (MET) minutes per week for all activities, including walking, moderate and vigorous physical activities. The MET score was derived based on the guidelines of the International Physical Activity Questionnaire (IPAQ). Additional details about these measurements are available on the UK Biobank website (www.ukbiobank.ac.uk). Hallmark-related diseases and all-cause mortality UK Biobank participants were linked to the NHS Hospital Episode Statistics for hospital admissions and the NHS Central Registry for mortality, using unique NHS identifiers assigned to all UK residents. These electronic health records provided dates and causes of hospitalization and death, with diseases coded according to ICD-10. Hallmark-related diseases Diseases associated with the hallmark of ageing were defined following the framework of Kuan et al.44. and Fraser et al.45, including 83 medical conditions (listed in Fig. 2) across nine hallmark categories (Fig. 2, first column): altered intercellular communication, cellular senescence, deregulated nutrient sensing, epigenetic alterations, genomic instability, loss of proteostasis, mitochondrial dysfunction, stem cell exhaustion, and telomere attrition. For each hallmark category, 27–30 associated diseases were identified, including both hallmark-specific and overlapping conditions (Supplementary Table 3). Participants with a baseline diagnosis of a given disease were excluded from the corresponding analysis. For analysis on each hallmark category, the time to the earliest occurring disease within that category was used as the event time. The follow-up period for all participants began at recruitment (between March 2006 and July 2010) and ended at the time of outcome diagnosis, death, loss to follow-up or study completion (March 25, 2023), whichever occurred first. All-cause mortality All-cause mortality was obtained from registry data (field 40000), with survival time calculated from baseline assessment (field 53) until death or censoring on March 25, 2023. Prevalent diabetes Participants with prevalent diabetes at baseline were excluded in a sensitivity analysis. Prevalent diabetes was defined according to the UK Biobank baseline self-report and medication data, including doctor-diagnosed diabetes, self-reported diabetes, and use of antidiabetic medication at baseline, following the UK Biobank diabetes adjudication framework described by Eastwood et al.46. Aging biomarkers We assessed biological aging using multiple established composite measures of ageing spanning clinical, molecular, and multi-omics domains. Together these biomarkers capture diverse aspects of the aging process across physiological, epigenetic, proteomic, metabolomic, and genomic layers. Detailed computational formulas and procedures are provided in the eMethods. Specifically, Phenotypic age (PhenoAge) and Klemera–Doubal method (KDM) biological age were derived from standard clinical biomarkers following published algorithms47,48. DNA methylation PhenoAge was estimated using 513 CpG sites via the elastic-net model49, while proteomic and metabolomic aging clocks were constructed from large-scale plasma proteomics and NMR-based metabolomics profiling, respectively50–53. Relative telomere length (field 22191) was measured by qPCR and provided by UK Biobank54. Organ-specific biological age Plasma proteomic profiling was conducted using the UK Biobank Olink® Explore 3072 platform on blood samples collected during participants’ baseline assessments (2006–2010), as part of the UK Biobank Pharma Proteomics Project (UKB-PPP), encompassing over 54,000 individuals55. Organ-specific biological age was computed using previously published linear models trained in the UK Biobank by Oh et al.25. These models were derived using LASSO regression to predict chronological age based on organ-enriched protein panels identified via GTEx transcriptomic profiling. Each of the 13 organ systems (e.g., brain, heart, liver, kidney, intestine, pancreas, muscle, adipose, artery, immune) was modeled independently, with protein coefficients and intercepts publicly available25. Non-organ-enriched proteins were exclusively used to train the organismal aging model, whereas the conventional model was developed using the complete proteomic profile. The predicted age for each organ was calculated as the weighted sum of z-scored protein levels and an organ-specific intercept. The detailed formulas and procedures for organ-specific biological age and organ age gap are provided in the eMethods. Differential expression and pathway enrichment analysis The analytical methods and in-house quality control measures for proteomic assay have been published elsewhere55. To identify differentially expressed plasma proteins between the early-life sugar-rationed and non-rationed groups, we performed differential expression analysis using proteomic data representing 2,911 proteins. For each protein, the log₂ fold change (log₂FC) and nominal p value were calculated. A volcano plot was generated using the EnhancedVolcano R package (v1.18.0)56, with significance thresholds set at FDR < 0.05 and |log₂FC | > 0.25 to highlight proteins exhibiting both statistically significant and biologically meaningful expression changes. To investigate the potential biological implications of these protein-level differences, we conducted a pathway enrichment analysis using a curated set of KEGG pathways57. Normalized enrichment scores (NES) were calculated to reflect the relative pathway activity between the two groups, and statistical significance was assessed based on a false discovery rate (FDR) threshold of 0.05. Statistical analysis Detailed statistical methods are provided in the eMethods. Continuous variables were summarized as means ± SD and categorical variables as frequencies (%). Missing data (Supplementary Table 4) were imputed using multiple chained equations across 20 datasets58. Group differences were assessed with χ² tests for categorical variables and Mann–Whitney U tests for continuous variables. Cox proportional hazards models estimated hazard ratios (HRs) with 95% confidence intervals (CIs), assessing the association between sugar rationing and the incidence of hallmark-related diseases. Proportional hazards assumptions were verified using Schoenfeld residuals59, with no violations detected. Associations of lifestyle and early-life sugar rationing with organ-specific biological aging were assessed using linear regression models. Covariate adjustment was guided by a directed acyclic graph (Supplementary Fig. 9): Model 1 included age, sex, and ethnicity; Model 2 additionally accounted for month and place of birth, real food prices, parental disease history (CVD, diabetes, hypertension), maternal smoking, breastfeeding, and survey year. All covariates were determined at birth to reduce endogeneity concerns. To address potential biases from time trends or improved disease detection, we compared outcomes across 6-month birth cohorts after sugar rationing ended. We hypothesized that, after adjusting for covariates, disease risk among adults born post-December 1954 (unexposed group) would be comparable, with HRs close to 1. We applied a counterfactual-based mediation analysis to estimate the proportion of the association between early-life sugar rationing and all-cause mortality mediated through hallmark-related diseases. Details are shown in the eMethods. All statistical analyses were performed using R version 4.0.2 (R Core Team, Vienna, Austria), with a significance level set at p < 0.05 (two-tailed). To control for inflated false-positive rates due to multiple testing, we applied a Bonferroni correction across 83 disease outcomes, yielding a significance threshold of p ≤ 0.0006. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Reporting Summary Transparent Peer Review file Source data Source Data