Abstract
Recent evidence suggests that sugar rationing during the first 1000 days of life may shape chronic disease risk in adulthood. Given the established link between mental health and non‑communicable diseases, early sugar restriction could influence later incidence of anxiety and depression. Using UK Biobank data, participants born between October 1, 1951, and March 31, 1956 (n = 46,448) were grouped by duration of exposure to sugar rationing: R270 (n = 9671), R425 (n = 6984), R635 (n = 6622), R817 (n = 7096), R1000 (n = 6856), and never‑rationed groups NR0 (n = 6117) and NR455 (n = 3102). Sweet-foods preference was assessed via questionnaires, while anxiety and depression diagnoses were obtained from linked health records. Survival analyses employed Weibull accelerated failure time models, adjusted for demographic, lifestyle, and health covariates. Structural equation modelling (SEM) examined possible pathways. The R1000 group reported lower preference for sweets (mean score = 5.82 vs. 6.04 in NR455, p = 0.0042) and showed lower hazards of anxiety (HR = 0.67, 95% CI: 0.53–0.84) and depression (HR = 0.71, 95% CI: 0.57–0.87). Associations for anxiety remained robust across sensitivity analyses, including adjustment for later‑life sugar intake, while those for depression weakened and became statistically non‑significant. SEM indicated a direct association between rationing duration and anxiety (γ = –0.144, p = 0.012), but not depression. These findings highlight the potential importance of early developmental environments in shaping mental-health outcomes, with stronger associations observed for anxiety than depression.
Subject terms: Depression, Human behaviour
Introduction
Numerous studies have thoroughly examined the link between quality of diet and human health, revealing that diet influences not only metabolic diseases but also the risk of mental health conditions [1–3]. Both high intake and a greater preference for sweet foods have been associated with increased risks of depression and anxiety [4–9], underscoring the importance of understanding how dietary choices shape mental health. Importantly, food liking often serves as a behavioural proxy for intake, since individuals with stronger preferences for sweet foods tend to consume them in greater quantities [10], which may exacerbate health risks or undermine adherence to dietary interventions.
Depression and anxiety are multifactorial diseases linked to biological, genetic, and environmental factors, with epigenetic mechanisms potentially mediating increased depression risk after adverse life events [11]. Recent findings from the United Kingdom (UK) suggest that sugar rationing during the first 1000 days of life may influence the risk of chronic diseases, particularly diabetes, cardiovascular diseases, and hypertension. Given that depression and anxiety are often associated with other mental conditions and non-communicable diseases [12–18], early-life sugar exposure might also contribute to the incidence of these mental health issues later in life.
This study investigates how early‑life sugar exposure during critical developmental windows influences later preferences for sweet foods, dietary behaviours, and psychiatric outcomes, particularly depression and anxiety, while also examining neurodevelopmental markers such as grey matter volume and inflammatory pathways as potential mediators. Depression and anxiety are shaped by a range of factors, with inflammation and structural brain alterations consistently implicated as key mechanisms [19–25]. Because these pathways can be modulated by nutritional exposures, we examine whether early‑life sugar rationing is associated with long‑term differences in sweet preference, psychiatric risk, inflammatory markers, and grey matter volume. By addressing these questions, this research aims to clarify how early nutrition contributes to mental-health resilience and inform preventive strategies and dietary recommendations for mental health.
Method
Study population
This study used UK Biobank data (application 83988), which recruited >500 000 UK residents aged 37–73 between 2006–2010, with extensive phenotypic and biological measures. Participants born October 1951–June 1954 were identified as exposed to sugar rationing, as previously described [12]. Comparator never‑rationed groups included those born July–December 1954 (post sugar‑rationing) and January–March 1956 (after all food rationing ended). Exclusions were birth/residence outside the UK, multiple births, adoption, Biobank withdrawal, and pregnancy at survey. The final analytic sample comprised 46 448 participants.
Rationed group participants were categorised as follows: R270, born between October 1953 and June 1954, exposed to sugar rationing in‑utero only, for a duration of approximately 270 days (n = 9671); R425, born between April 1953 and September 1953, rationed in‑utero and up to 6 months of age, for approximately 425 days (n = 6984); R635, born between October 1952 and March 1953, rationed in‑utero and up to 12 months of age, for approximately 635 days (n = 6622); R817, born between April 1952 and September 1952, exposed to sugar rationing in‑utero and up to 18 months of age, for approximately 817 days (n = 7096); and R1000, born between October 1951 and March 1952, rationed in‑utero and up to 24 months of age, for approximately 1000 days (n = 6856). These differing in‑utero durations broadly correspond to and are cognisant of early, mid, and late pregnancy trimesters, allowing trimester‑specific exposure to be considered. Two comparison groups were defined as never‑rationed: NR0 (July–December 1954, conceived immediately after sugar rationing ended, n = 6117) and NR455 (January–March 1956, conceived after all food rationing ended, n = 3102) (illustration in Supplementary Figure 1).
Analysis of sweet food-liking, actual sugar consumption and sugar rationing exposure
To assess relationships between rationing status, sweet food preference, and sugar intake, we analysed data from 15,058 participants who completed the Oxford WebQ (24‑hour dietary recall) and the Food Preference Questionnaire. Average free and total sugar intake was calculated as the mean of all repeated 24‑hour dietary recalls available per participant. Sweet food liking scores (Field ID 20732) were examined alongside specific preferences, including biscuits (Field ID 20615), cake (Field ID 20631), cake icing (Field ID 20632), croissants (Field ID 20648), dark chocolate (Field ID 20652), cheesecake (Field ID 20636), jam (Field ID 20680), marzipan (Field ID 2688), and milk chocolate (Field ID 20691). The selection of these specific items was based on previous work [9]. Group differences across groups were evaluated using independent sample t-test with NR455 as the reference group, and p-values were adjusted for multiple comparisons (Holm method).
Health outcome assessment
Linked health outcome data from UK Biobank (primary care, hospital, and death records; excluding self‑reports) were analysed. Anxiety was defined by ICD‑10 codes F40 and F41 (Field IDs 130904, 130907), and depression by F32 and F33 (Field IDs 130895, 130896). Absolute risk was calculated as the percentage of individuals experiencing anxiety or depression. Two negative control diseases—type 1 diabetes mellitus (ICD‑10 E10, E13, E14) and post‑traumatic stress disorder (ICD‑10 F43)—were included, although they have complex early-life determinants, these determinants are unrelated to sugar intake, metabolic processes, or early dietary exposures. Their inclusion provides a check against spurious associations with conditions lacking plausible mechanistic links to the specific exposure under analysis.
Structural brain imaging using MRI scan data
Volumetric imaging data were obtained from the UK Biobank [26, 27], including 5990 participants with brain MRI, distributed across the rationing groups as follows: R1000 (n = 889), R817 (n = 874), R635 (n = 873), R425 (n = 914), R270 (n = 1258), NR0 (n = 788), and NR455 (n = 394). Dependent variables comprised pre‑processed T1‑weighted structural MRI phenotypes, specifically numerical volume measures for 139 regional grey matter volumes (FAST) generated by UK Biobank imaging pipelines [26]. To account for head size and total brain volume, we additionally used pre‑processed T1‑weighted phenotypes for total white and grey matter, normalised for head size. MRI data were acquired using the UK Biobank standardized protocol on a Siemens Skyra 3 T scanner, as described in Miller et al. [27]. Resting‑state fMRI employed a multi‑band EPI sequence (factor 8, TR = 735 ms, TE = 39 ms, flip angle = 52°, voxel size = 2.4 mm isotropic, 490 volumes, ~6 min 10 s), while task fMRI used the same parameters with 332 volumes ( ~ 4 min 13 s). Structural imaging used a 3D inversion‑recovery sequence (TI = 880 ms, turbo factor 208). Detailed scanning protocols are available in the UK Biobank Brain Imaging Documentation (https://biobank.ctsu.ox.ac.uk/ukb/refer.cgi?id=2367).
Statistical analyses
Survival modelling
Analyses were conducted in R version 4.3.2. Anxiety and depression incidence was modelled using accelerated failure time (AFT) survival regression with a Weibull distribution. For interpretation, model coefficients were converted to an exponential to yield hazard ratios (HRs) and 95% confidence intervals on the proportional hazard scale. This transformation does not violate proportional hazard assumptions. Cases were right‑censored at the last date of available ICD‑10 hospital inpatient data (30 September 2024), ensuring consistent follow‑up. Models adjusted for time‑invariant covariates measured during the UK Biobank baseline assessment: sex, Townsend deprivation index (quintiles), ethnic background (white/other), family history of mental health (yes/no), breastfeeding status (yes/no), birth month, and survey year. Additional covariates included BMI class ( < 18.5, 18.5– < 25, 25– < 30, 30– < 35, 35– < 40, ≥40 kg/m²), alcohol intake frequency (six categories), household size (1, 2–4, ≥5), smoking status (never/previous/current), sleep duration ( < 7, 7– < 9, ≥9 h), and inflammation status (CRP < 3.0 vs ≥3.0 mg/L). These covariates were measured at baseline and treated as fixed values in the models. Descriptive analyses of age at event were also performed. For depression cases, median age at diagnosis was calculated within rationing groups using time‑to‑event data restricted to events. Kernel density estimates of age at event were computed to characterise onset distributions, complementing regression results with measures of central tendency and variability.
Robustness checks
Several robustness checks were implemented. First, a placebo test used the NRIR group (Irish births October 1951–September 1952, n = 154). Ireland’s sugar rationing ended in December 1951, meaning overlap with R1000 involved in‑utero but not postnatal exposure, while overlap with R817 involved no rationing. This provided an external comparator with shared birthdate overlap but differing exposure. Second, missing categorical covariates were imputed using multiple imputation (five imputations, 10 iterations) by chained equations (MICE) via the mice package. Imputation methods matched variable type: logistic regression for binary, polytomous regression for unordered categorical, and proportional odds models for ordered factors. Third, gender interaction analyses were conducted, with stratified models applied where significant modification was detected. Finally, sensitivity analyses incorporated later‑life dietary exposures (average free and total sugar intake) restricted to individuals with complete diet data (n = 21 221), testing whether adult diet influenced associations between rationing exposure and mental health outcomes. In this subsample, participants contributed a median of 1 recall (IQR 1 – 5), with no major differences across rationing groups. Average sugar intake was calculated across available recall data, acknowledging that dietary data were sparse and non-random.
One-way ANCOVA and post hoc tests
To compare brain MRI traits across seven rationing groups, one‑way ANCOVA was performed. Levene’s test assessed homogeneity of variances. Post hoc two‑sample t‑tests compared each rationing group with the never-rationed (NR0) group. Covariates included age, sex, BMI, Townsend deprivation index, and brain volume normalised for head size. False discovery rate (FDR) corrections were applied across the full ACOVA set, and only post hoc comparison that remained significant after FDR correction are reported.
Structural equation modelling (SEM)
Structural equation modelling (SEM) was employed to explore the relationships between sugar rationing duration, inflammation, mental health outcomes (anxiety and depression), and three latent constructs: preference for sweetness, sugar consumption, and brain MRI traits. The analysis was restricted to participants from the rationed groups (R270, R425, R635, R817, R1000) and one never‑rationed group (NR0) with complete data across all variables (n = 2960). Given the outcomes were lifetime binary diagnoses in a complete-case subset, the SEM provides an exploratory cross-sectional summary of associations rather than causal modelling of time-to-event risk. The latent variable for sweetness preference was derived from individuals’ liking scores for a range of sweet foods, including biscuits, cake, chocolate, jam, marzipan, and related items. Sugar consumption was represented as a latent construct based on daily intake of free sugars and total sugars. Brain MRI traits were defined by combining grey matter volume (GMV) measurements from regions that demonstrated significant differences between rationed and never‑rationed groups in post hoc analyses. Confirmatory factor analysis was used to estimate these latent variables prior to inclusion in the SEM.
The SEM specified direct regression paths whereby anxiety and depression were predicted by rationing group, inflammation status (measured by C‑reactive protein, CRP), and brain MRI traits. Covariances were specified to account for plausible shared variance, reflecting hypothesised correlations not fully explained by direct paths and consistent with prior reports of correlated dietary behaviours, neurodevelopmental traits, and inflammatory markers. Sugar intake was modelled as a function of sweetness preference, while inflammation was predicted by sugar intake. In addition, covariances were incorporated to account for correlations not explicitly captured by regression paths. These included covariances between sweetness preference and rationing group, sugar intake and rationing group, sweetness preference and brain MRI traits, and brain MRI traits and inflammation. Model fit was evaluated using the Root Mean Square Error of Approximation (RMSEA) and the Standardised Root Mean Square Residual (SRMR), providing indices of overall adequacy of the specified model. The analyses were performed using the ‘lavaan 0.6–19’ package in R [28]. The analysis code used for this analysis available at https://github.com/hanavratilova/sr_ukb.
Results
Among the eligible 46 448 UK Biobank participants that met the criteria for this study, 37 229 participants were grouped into rationed groups and 9219 were grouped into the never-rationed groups. The average follow-up time was 15.72 years, during which 2776 (5.98%) participants experienced anxiety events and 3147 (6.78%) experienced depression events. The characteristics of each group are presented in Supplementary Table 1.
Early‑life sugar rationing relates weakly to adult sweet food preferences but not to sugar consumption
The correlation between sweet food liking, sugar intake, and degree of exposure to sugar (via rationing status) during gestation and infancy was examined to assess whether early-life sugar rationing influences later preferences. Adult sugar consumption patterns were consistent across groups, with mean daily free sugar intake ranging from 58.1 to 59.2 g/day and total sugar intake from 122.4 to 123.5 g/day (Fig. 1a, b). No statistically significant differences in sugar consumption were observed. Distributions stratified by mental health conditions are shown in Supplementary Table 2. Average liking scores for sweet foods were similar across groups (5.82–6.04 on a 9‑point scale; Fig. 1c), with the only significant difference between R1000 (mean = 5.82) and NR455 (mean = 6.04; adjusted p-value = 0.025). Although the difference is modest, it infers a subtle pattern that warrants further study in the context of early-life nutrition and later food preferences.
Fig. 1. Relationship between sugar rationing status, sweet liking, and actual sugar intake.

Distribution of (a) free sugar intake, (b) total sugar intake, and (c) sweet food liking scores across the seven rationing groups. (d) Relationship between average sweet food liking score and average free sugar intake. (e) Correlation plot of sugar intakes and liking for sweet foods across the seven rationing groups. Square size and colour correspond to the correlation value. Raw p‑values are displayed in the figure for visualization; Holm‑adjusted p‑values are reported in the text.
Analyses revealed a positive association between sweet liking and sugar intake (Fig. 1d). In contrast, rationing status showed no meaningful association with sugar intake or sweet preferences, with coefficients near zero (Fig. 1e). Correlations with free and total sugar intake ranged between −0.005 and +0.008, and associations with specific sweet foods were similarly weak. While not statistically significant, inspection suggests subtle differences between rationed groups (R1000, R817) and never‑rationed groups (NR0, NR455), hinting at possible long‑term dietary influences.
Association between early-life exposure to sugar and risk for anxiety and depression
The absolute risk of anxiety and depression was calculated across groups to assess likelihood of onset. Anxiety risk ranged from 5.45% to 6.29%, lowest in R817 (5.45%) and R1000 (5.79%), and highest in NR0 (6.29%) and NR455 (6.28%). Depression risk ranged from 6.00% to 7.23% across groups, with R817 showing the lowest prevalence (6.00%), R1000 at 6.69%, and NR0 the highest (7.23%) (Supplementary Table 3). To examine differences in anxiety and depression risk, Weibull accelerated failure time models were expressed as proportional hazards, with NR455 as the reference group (Fig. 2). In fully adjusted models (Model 3), longer durations of early‑life sugar rationing were associated with significantly reduced hazards of anxiety: R1000 (HR = 0.67, 95% CI 0.53–0.84, p = 0.0015), R817 (HR = 0.70, 95% CI 0.52–0.94, p = 0.045), and R635 (HR = 0.73, 95% CI 0.58–0.91, p = 0.020). Shorter‑exposed groups (R425, R270) and NR0 did not differ from NR455. A similar gradient was observed for depression, with reduced hazards in R1000 (HR = 0.71, 95% CI 0.57–0.87, p = 0.0026), R817 (HR = 0.73, 95% CI 0.56–0.96, p = 0.036), and R635 (HR = 0.76, 95% CI 0.62–0.95, p = 0.025), while shorter‑exposed groups showed no differences. Negative control diseases showed no associations in the fully adjusted model (Model 3): for type 1 diabetes, R1000 HR 1.09 (95% CI: 0.91–1.47, p > 0.05), R817 HR 1.25 (95% CI: 0.86–1.82, p > 0.05), and R635 HR 1.18 (95% CI: 0.88–1.59, p > 0.05); for PTSD, R1000 HR 0.89 (95% CI: 0.66–1.22, p > 0.05), R817 HR 0.77 (95% CI: 0.51–1.15, p > 0.05), and R635 HR 0.79 (95% CI: 0.57–1.09, p > 0.05). These results are illustrated in Supplementary Figure 2. Collectively, prolonged rationing (R1000, R817, R635) was associated with later onset of anxiety and depression, whereas shorter exposure or none conferred no benefit.
Fig. 2. Hazard Ratio for anxiety and depression.

a Hazard ratios for anxiety across rationing groups and b Hazard ratios for depression across rationing groups. Model 1 = unadjusted model; Model 2 = adjusted with sociodemographic factor (sex, BMI, Townsend quintiles) and season (month of birth, survey year); Model 3 = adjusted with sociodemographic, season, health factor (family history of mental health, inflammation, breastfed as child), and lifestyle (smoking status, household size, alcohol intake, sleep duration). P-value adjusted using Benjamini-Hochberg correction.
Several robustness checks were conducted. In the placebo comparison group born in Ireland (NRIR), no significant associations were observed with anxiety or depression, indicating that the observed protective association in UK cohorts were not explained by birth timing but were specific to rationing exposure (Supplementary Table 4). Sensitivity analyses addressing missing data confirmed the main findings, with associations suggesting lower risk observed for R1000, R817, and R635 (Supplementary Table 5). Interaction analyses showed stronger association in women, with significant gender modification for R1000, R425, and NR0; however, the direction of association was consistent across sexes. Adjustment for later‑life sugar intake revealed that the associations for anxiety remained significant, notably for R1000 (HR = 0.63, 95% CI: 0.44–0.89, p = 0.011), whereas associations for depression were attenuated, suggesting adult diet may partly mediate depression risk (Supplementary Table 6). Together, these checks confirm that the associations are not driven by birth timing, missing data, or gender imbalance. The attenuation of significance for depression risk after sugar-intake adjustment implies dietary behaviour plays a role, while persistent associations for anxiety suggest that early‑life nutrition may have long‑lasting influences independent that remain after adjustment of measured adult sugar intake in the available sub-sample, though this adjustment is limited by measurement error and subsample size.
Differences in median age of onset highlight long‑term impacts (Fig. 3). For anxiety, onset was latest in R1000 (61.3 years), R425 (61.3), and R817 (62.1), compared to NR455 (55.6). For depression, ages of onset were consistently earlier than those for anxiety, with R1000 (59.1) and R817 (58.4) occurring later than NR455 (54.6). Across groups, depression was diagnosed at younger ages than anxiety overall, but rationed cohorts showed delayed onset, suggesting reduced vulnerability. The higher median ages of onset observed here, as compared to those observed in wider populations, reflects the UK Biobank recruitment of participants aged 39 – 69 years old and the availability of linked health records primarily from 1997 onwards, which likely under-captures earlier diagnoses. Mental health conditions were also historically under-diagnosed, as fewer individuals sought care in young adulthood. The small peak visible at ages 20 – 30 suggests that the true age of onset may be younger than the reported median, but this is not fully recorded in the health records available to us.
Fig. 3. Density of age at event for Anxiety and Depression.

Density of age at event for (a) Anxiety and (b) Depression across Rationing Status. The x‑axis represents age at diagnosis in years, while the y‑axis shows the probability density, indicating the relative concentration of diagnoses at different ages. Each curve corresponds to a rationed group, with peaks marking the ages where diagnoses were most common.
Differences in neuroimaging phenotypes across rationing status groups
We examined brain imaging data alongside rationing exposure groupings. In one-way ANCOVA analyses, 80 of the 139 brain regions of grey matter volume were significantly different among the different rationing status groups (Supplementary Table 7). Post-hoc analysis further revealed significant differences in 11 out of the 80 brain regions between most of the rationing groups and NR0. These regions included the brain stem, occipital fusiform gyrus, and regions of cerebellum (crus I, crus II, VI, VIIIa, VIIIb bilaterally; Fig. 4). Most of these regions were cerebellar or brainstem structures, highlighting a predominance of subcortical involvement. These 11 regions were next summarised into a latent Brain MRI Trait for SEM analysis; this construct is exploratory in nature and serves to generate hypotheses about potential neurodevelopmental pathways.
Fig. 4. Heatmap showing significant differences in brain regions between rationing groups and the never-rationed group (NR0).

The numbers inside each cell represent the adjusted p-values.
Complex interplay of sugar rationing and mental health
To examine the relationships between sugar rationing exposure, liking of sweet foods, sugar intake, inflammation levels (as determined by CRP measurement), brain MRI traits, and mental health outcomes, we performed a SEM analysis. In this complete-case subset, outcomes were modelled as lifetime binary diagnoses: accordingly, the SEM provides an exploratory view of associations among exposures, mediators, and outcomes. The model shows a good fit to the data, with an RMSEA of 0.057 and SRMR of 0.059 (Fig. 5). The estimates for the latent variables (sweet foods liking, sugar intake, and brain MRI traits) show significant factor loadings, indicating that the observed variables are good indicators of the latent constructs (Supplementary Table 8 and Supplementary Figure 2). The direct effect of rationing status on anxiety was significant and negative (γ = –0.144, p = 0.012), indicating shorter sugar rationing increased adult anxiety risk. In contrast, depression showed no significant direct effect (γ = –0.062, p = 0.241), and indirect pathways via inflammation or brain MRI traits were unsupported.
Fig. 5. Directional associations among food preference, sugar intake, mental health outcomes, inflammation (CRP level) and brain MRI trait.

Oval shape represents a latent variable. The sugar rationing variable represents the duration of days exposed to sugar rationing. The inflammation level is indicated by the CRP measurement. Anxiety and Depression is captured as a binary outcome, indicating either presence or absence. The beta (β) coefficients represent paths between endogenous variables. The gamma (γ) coefficients represent paths from exogenous variables to endogenous variables. The phi (φ) value represents variance or covariance of the exogenous variable. The magnitude of the coefficient indicates the strength of the relationship, while the sign (positive or negative) indicates the direction. Red arrows represent negative correlation; Green arrows represent positive correlation. Dashed Arrow: Non-significant or weaker relationships. Solid Arrow: Significant or strong relationships. One-Direction Arrow: Causal relationship from one variable to another. Two-Direction Arrow: Correlation or bidirectional relationship.
The relationship between sweet taste liking and sugar intake is suggestive of a positive association between preference and intake (φ = 0.18, p < 0.001). Additionally, the covariance between sweet taste liking and brain MRI traits is significant (γ = 0.06, p = 0.003), indicating a modest but meaningful link between taste preference and structural brain traits. Nonetheless, grey matter volume differed between prolonged rationing and never‑rationed groups, suggesting possible neurodevelopmental influences beyond dietary preference.
Discussion
This study explored the long-term effects of early-life sugar exposure on later-life sugar preferences, brain structure, and mental health. Evidence suggests that sugar rationing during the first 1000 days of life was associated with differences in sweet food preferences, psychiatric/mental-health outcomes, and neurodevelopmental traits, underscoring the critical role of early nutrition in shaping health and behaviour. Participants exposed to the longest period of rationing (R1000) reported lower preference for sweet foods and significantly lower hazards of anxiety in adulthood; these protective associations persisted after adjustment of adult sugar intake in the subsample of dietary data, though this adjustment is limited. An association with depression was also observed but attenuated after adjustment, suggesting mediation by later diet.
Although other foods were rationed post-war, sugar rationing ended earlier and saw rapid uptake once reintroduced [12]. NR0 therefore represents a mixed exposure group, in which sugar was available but other foods remained restricted. NR455 (conceived after all rationing ended) was chosen as the reference group, as it provides a fully post-rationing baseline. Hazard ratios for NR0 relative to NR455 were null, supporting NR455 as a better-controlled comparator for evaluating sugar rationing duration. While NR455 helps distinguish sugar availability from broader rationing influences, we acknowledge that rationing duration may also capture wider early-life conditions (e.g., healthcare, infections, socioeconomic shifts).
The observed associations in R1000, R817, and R635 highlight sugar’s potential role in shaping preferences and mental health. Birth timing effects were further examined using an Irish comparison group (NRIR), where no significant associations were found; however, the small sample size and wide confidence intervals mean these null findings provide only weak supportive evidence rather than a decisive control. Sensitivity analyses adjusting for later sugar intake showed reduced effect sizes, suggesting partial mediation. Consistent with prior evidence of bidirectional links between sugar and mental health [7, 8, 29, 30], individuals with anxiety generally consumed more sugar. Because only a subset of participants contributed to dietary data, and selection into the diet subsample could bias attenuating patterns, particularly for depression, these results should be interpreted cautiously. Adult behaviours used as covariates such as BMI, smoking, alcohol intake, and sleep were measured only at baseline and treated as fixed values. As these behaviours may change across adulthood and could affect the causal pathway from early-life rationing exposures to later mental health outcomes, adjustment for them risks partial overcontrol, and our estimates should be interpreted as conservative associations.
While these patterns align with our hypothesis, residual confounding from unmeasured socioeconomic, environmental, or diagnostic factors cannot be fully excluded. In particular, maternal stress during pregnancy may have influenced offspring development independently of food rationing. This stress was potentially heightened in UK mothers due to more direct wartime exposure compared to Irish mothers. Although healthcare systems varied across birth cohorts, the stronger associations in older cohorts with longer rationing duration suggest that such variation is unlikely to explain the observed gradients. Diagnoses were coded using standardised ICD-10 code, which reduces variation. However, changes in detection and service use over time may still have influenced recording, particularly before 1997 when linked records become more reliable. This is reflected in the faint peak at ages 20–30, earlier than the median age of onset for both anxiety and depression. Findings should therefore be interpreted as associations, while underscoring the potential importance of sugar exposure in influencing later-life mental health risk.
It is important to situate our findings within the broader historical and contemporary context of sugar consumption and its established links to mental health. Historical data show rising UK sugar consumption after rationing ended, peaking in the early 2000s. Despite declines in sugar‑sweetened beverage intake, current levels of free sugar intake among UK adults remain above recommended thresholds. According to the National Diet and Nutrition Survey (2019–2023), only 9% of children and 19% of adults adhere to recommended levels, with mean intakes in most age groups—particularly adolescents and working‑age adults—hovering around double the recommended threshold [31]. This continuity underscores the relevance of contrasting rationing with today’s excess, highlighting how early restriction may confer resilience against anxiety.
Analyses of potential pathways involving inflammation, brain morphology, and rationing duration were conducted. No significant association emerged between rationing duration and sugar intake, suggesting that adult inflammation status does not account for the association. However, our analysis was limited to CRP and did not include other inflammatory markers such as interleukin‑6 (IL‑6), which others have found more reliably associated with depression [19, 32]. Neurodevelopmental mechanisms were therefore considered. Prior evidence showed that food preferences, brain morphology, and mental health are interconnected [33]. Early‑life rationing may have induced lasting neurodevelopmental changes, influencing dietary preferences and psychiatric resilience. Observed differences in grey matter volume support the view that nutritional deprivation alters synaptic connectivity in hypothalamic circuits [34].
Although our study specifically examined sugar rationing, prior evidence from famine and malnutrition cohorts demonstrates that early-life nutritional restriction increases risk of affective disorders and alters brain development [35–38]. Mechanistic studies further show that metabolic imbalance can affect cerebellar tissue [39] and that sugar-rich diets reduce serotonin neuron density and availability [40]. Such early-life nutritional disruptions appear to exert disproportionate long-term effects on brain function and mental health [41].
Our findings should be interpreted in consideration of possible limitations. DNA methylation data are not yet available in UK Biobank, preventing assessment of epigenetic mechanisms underlying sugar rationing effects. Although recent evidence suggests prenatal sugar exposure may interact with genetic predisposition to influence later outcomes [13], we did not test this possibility. SEM enabled evaluation of causal pathways, but the observational design limits causal inference and residual confounding cannot be excluded. In particular, personality traits, which are known to be associated with both dietary behaviours and psychiatric outcomes [42], were not measured and could possibly confound the observed associations. Finally, as analyses focused on UK births during rationing, generalisability may be limited, since other countries differed in timing, duration, and policy context.
Overall, our finding show that limiting sugar exposure during the first 1000 days was associated with enduring differences on sweet preference and psychiatric outcomes. Rationing was associated with reduced sweet liking and lower risk of anxiety, while depression appeared more closely linked to inflammatory pathways. The observed associations likely reflect neurodevelopmental influences during critical windows rather than altered taste alone. Despite increased sugar consumption after rationing, early nutritional environments appears to have lasting relevance for mental health. Future research should incorporate epigenetic and gene–environment analyses, expanded inflammatory profiling, and neuroimaging to clarify mechanisms by which early nutrition shapes psychiatric resilience.
Supplementary information
Acknowledgements
The study utilised data from UK Biobank, see http://www.ukbiobank.ac.uk. We thank UK Biobank volunteers for their involvement.
Author contributions
HFN contributed to the data analysis, interpretation of the results, and drafting the article. ADW contributed to the interpretation of the results and critically revised the manuscript for important intellectual content. NG contributed to the conception and design of the study, interpretation of the results, and critically revised the manuscript for important intellectual content. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted. NG is guarantor of the paper.
Funding
UK Biobank access was funded by the School of Biosciences, University of Surrey, with additional support from Beasiswa Pendidikan Indonesia grant and the Indonesian Endowment Fund for Education awarded to HFN.
Data availability
The UK Biobank database is accessible to researchers upon successful application through the UK Biobank website. Our study utilised the UK Biobank under research application number 83988. The code used for this study was tailored to the UK Biobank data and is no use as a standalone without access to the UK Biobank.
Competing interests
The authors declare no competing interests.
Ethical approval
The UK Biobank has ethical approval from the North West Multi-center Research Ethics Committee (REC reference number: 16/NW/0274). Written informed consent was obtained for all UK Biobank study participants. Ethical and governance approval was also obtained from the Research Integrity and Governance Office, University of Surrey (EGA reference: FHMS 22-23 273 EGA). All methods were performed in accordance with the relevant guidelines and regulations.
Footnotes
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
The online version contains supplementary material available at https://doi.org/10.1038/s41398-026-04376-w.
References
- 1.Jacka FN, Pasco JA, Mykletun A, Williams LJ, Hodge AM, O’Reilly SL, et al. Association of western and traditional diets with depression and anxiety in women. Am J Psychiatry. 2010;167:305–11. [DOI] [PubMed] [Google Scholar]
- 2.Phillips CM, Shivappa N, Hébert JR, Perry IJ. Dietary inflammatory index and mental health: A cross-sectional analysis of the relationship with depressive symptoms, anxiety and well-being in adults. Clinical Nutrition. 2018;37:1485–91. [DOI] [PubMed] [Google Scholar]
- 3.Adan RAH, van der Beek EM, Buitelaar JK, Cryan JF, Hebebrand J, Higgs S, et al. Nutritional psychiatry: Towards improving mental health by what you eat. Eur Neuropsychopharmacol. 2019;29:1321–32. [DOI] [PubMed] [Google Scholar]
- 4.Xie X, Li Y, Zhang Y, Lin X, Huang M, Fu H, et al. Associations of diet quality and daily free sugar intake with depressive and anxiety symptoms among Chinese adolescents. J Affect Disord. 2024;350:550–8. [DOI] [PubMed] [Google Scholar]
- 5.Fujiwara A, Murakami K, Suga H, Sasaki S. Cross-sectional associations of intakes of starch and sugars with depressive symptoms in young and middle-aged Japanese women: three-generation study of women on diets and health. Nutrients. 2022;14:2400. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Zhang L, Sun H, Liu Z, Yang J, Liu Y. Association between dietary sugar intake and depression in US adults: a cross-sectional study using data from the National Health and Nutrition Examination Survey 2011–2018. BMC Psychiatry. 2024;24:110. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Knüppel A, Shipley MJ, Llewellyn CH, Brunner EJ. Sugar intake from sweet food and beverages, common mental disorder and depression: prospective findings from the Whitehall II study. Sci Rep. 2017;7:6287. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Kose J, Cheung A, Fezeu LK, Péneau S, Debras C, Touvier M, et al. A comparison of sugar intake between individuals with high and low trait anxiety: results from the NutriNet-Santé study. Nutrients. 2021;13:1526. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Navratilova HF, Whetton AD, Geifman N. Artificial intelligence driven definition of food preference endotypes in UK Biobank volunteers is associated with distinctive health outcomes and blood based metabolomic and proteomic profiles. J Transl Med. 2024;22:881. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Hwang L-D, Lin C, Gharahkhani P, Cuellar-Partida G, Ong J-S, An J, et al. New insight into human sweet taste: a genome-wide association study of the perception and intake of sweet substances. Am J Clin Nutr. 2019;109:1724–37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Penner-Goeke S, Binder EB. Epigenetics and depression. Dialogues Clin Neurosci. 2019;21:397–405. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Gracner T, Boone C, Gertler PJ Exposure to sugar rationing in the first 1000 days of life protected against chronic disease. Science (1979) 2024; published online Oct 31. 10.1126/science.adn5421. [DOI] [PMC free article] [PubMed]
- 13.van den Berg GJ, von Hinke S, Wang RAH Prenatal sugar exposure shapes late-life human capital and health. PNAS Nexus 2025;4. 10.1093/pnasnexus/pgaf301. [DOI] [PMC free article] [PubMed]
- 14.Zheng J, Zhou Z, Huang J, Tu Q, Wu H, Yang Q, et al. Exposure to sugar rationing in first 1000 days after conception and long term cardiovascular outcomes: natural experiment study. BMJ. 2025;391:e083890. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Zheng J, Zhou Z, Huang J, Tu Q, Wu H, Yang Q, et al. Sugar rationing in the first 1000 days after conception and long-term risk of metabolic dysfunction-associated steatotic liver disease and major adverse liver outcomes: a natural experiment study. Clin Gastroenterol Hepatol 2025; published online Oct. 10.1016/j.cgh.2025.10.015. [DOI] [PubMed]
- 16.Katon W, Lin EHB, Kroenke K. The association of depression and anxiety with medical symptom burden in patients with chronic medical illness. Gen Hosp Psychiatry. 2007;29:147–55. [DOI] [PubMed] [Google Scholar]
- 17.Niles AN, Dour HJ, Stanton AL, Roy-Byrne PP, Stein MB, Sullivan G, et al. Anxiety and depressive symptoms and medical illness among adults with anxiety disorders. J Psychosom Res. 2015;78:109–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Abbas U, Hussain N, Tanveer M, Laghari RN, Ahmed I, Rajper AB. Frequency and predictors of depression and anxiety in chronic illnesses: A multi disease study across non-communicable and communicable diseases. PLoS One. 2025;20:e0323126. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Ye Z, Kappelmann N, Moser S, Davey Smith G, Burgess S, Jones PB, et al. Role of inflammation in depression and anxiety: Tests for disorder specificity, linearity and potential causality of association in the UK Biobank. EClinicalMedicine. 2021;38:100992. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Toker S, Shirom A, Shapira I, Berliner S, Melamed S. The association between burnout, depression, anxiety, and inflammation biomarkers: C-Reactive protein and fibrinogen in men and women. J Occup Health Psychol. 2005;10:344–62. [DOI] [PubMed] [Google Scholar]
- 21.Milaneschi Y, Kappelmann N, Ye Z, Lamers F, Moser S, Jones PB, et al. Association of inflammation with depression and anxiety: evidence for symptom-specificity and potential causality from UK Biobank and NESDA cohorts. Mol Psychiatry. 2021;26:7393–402. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Han K-M, Ham B-J. How inflammation affects the brain in depression: a review of functional and structural MRI studies. J Clin Neurol. 2021;17:503. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Stone EA, Lin Y, Quartermain D. A final common pathway for depression? Progress toward a general conceptual framework. Neurosci Biobehav Rev. 2008;32:508–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Moon C-M, Kim G-W, Jeong G-W. Whole-brain gray matter volume abnormalities in patients with generalized anxiety disorder. Neuroreport. 2014;25:184–9. [DOI] [PubMed] [Google Scholar]
- 25.Schmaal L, Veltman DJ, van Erp TGM, Sämann PG, Frodl T, Jahanshad N, et al. Subcortical brain alterations in major depressive disorder: findings from the ENIGMA Major Depressive Disorder working group. Mol Psychiatry. 2016;21:806–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Alfaro-Almagro F, Jenkinson M, Bangerter NK, Andersson JLR, Griffanti L, Douaud G, et al. Image processing and Quality Control for the first 10,000 brain imaging datasets from UK Biobank. Neuroimage. 2018;166:400–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Miller KL, Alfaro-Almagro F, Bangerter NK, Thomas DL, Yacoub E, Xu J, et al. Multimodal population brain imaging in the UK Biobank prospective epidemiological study. Nat Neurosci. 2016;19:1523–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Rosseel Y lavaan: An R Package for Structural Equation Modeling. J Stat Softw 2012;48. 10.18637/jss.v048.i02. [DOI]
- 29.Sanchez-Villegas A, Zazpe I, Santiago S, Perez-Cornago A, Martinez-Gonzalez MA, Lahortiga-Ramos F. Added sugars and sugar-sweetened beverage consumption, dietary carbohydrate index and depression risk in the Seguimiento Universidad de Navarra (SUN) Project. Br J Nutr. 2018;119:211–21. [DOI] [PubMed] [Google Scholar]
- 30.Jacques A, Chaaya N, Beecher K, Ali SA, Belmer A, Bartlett S. The impact of sugar consumption on stress driven, emotional and addictive behaviors. Neurosci Biobehav Rev. 2019;103:178–99. [DOI] [PubMed] [Google Scholar]
- 31.Office for Health Improvement and Disparities. National Diet and Nutrition Survey 2019 to 2023: report. 2025 https://www.gov.uk/government/statistics/national-diet-and-nutrition-survey-2019-to-2023 (accessed Sept 11, 2025).
- 32.Lamers F, Milaneschi Y, Smit JH, Schoevers RA, Wittenberg G, Penninx BWJH. Longitudinal association between depression and inflammatory markers: results from the netherlands study of depression and anxiety. Biol Psychiatry. 2019;85:829–37. [DOI] [PubMed] [Google Scholar]
- 33.Zhang R, Zhang B, Shen C, Sahakian BJ, Li Z, Zhang W, et al. Associations of dietary patterns with brain health from behavioral, neuroimaging, biochemical and genetic analyses. Nature Mental Health. 2024;2:535–52. [Google Scholar]
- 34.Lewis S. Feast or famine. Nat Rev Neurosci. 2016;17:466. 10.1038/nrn.2016.90. [DOI] [PubMed] [Google Scholar]
- 35.Brown AS, van Os J, Driessens C, Hoek HW, Susser ES. Further evidence of relation between prenatal famine and major affective disorder. Am J Psychiatry. 2000;157:190–5. 10.1176/appi.ajp.157.2.190. [DOI] [PubMed] [Google Scholar]
- 36.Hulshoff Pol HE, Hoek HW, Susser ES, Brown AS, Dingemans A, Schnack HG, et al. Prenatal exposure to famine and brain morphology in schizophrenia. Am J Psychiatry. 2000;157:1170–2. 10.1176/appi.ajp.157.7.1170. [DOI] [PubMed] [Google Scholar]
- 37.Li C, Zhu N, Zeng L, Dang S, Zhou J, Yan H. Effect of prenatal and postnatal malnutrition on intellectual functioning in early school-aged children in rural western China. Medicine (Baltimore). 2016;95:e4161. 10.1097/MD.0000000000004161. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Razzaq FA, Calzada-Reyes A, Tang Q, Guo Y, Rabinowitz AG, Bosch-Bayard J, et al. Spectral quantitative and semi-quantitative EEG provide complementary information on the life-long effects of early childhood malnutrition on cognitive decline. Front Neurosci. 2023;17:1149102. 10.3389/fnins.2023.1149102. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Thau-Zuchman O, Pallier PN, Savelkoul PJM, Kuipers AAM, Verkuyl JM, Michael-Titus AT. High phenylalanine concentrations induce demyelination and microglial activation in mouse cerebellar organotypic slices. Front Neurosci. 2022;16:926023. 10.3389/fnins.2022.926023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Dunn GA, Thompson JR, Mitchell AJ, Papadakis S, Selby M, Fair D, et al. Perinatal Western-style diet alters serotonergic neurons in the macaque raphe nuclei. Front Neurosci. 2023;16:1067479. 10.3389/fnins.2022.1067479. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Beuriat PA, Cristofori I, Gordon B, Grafman J. The shifting role of the cerebellum in executive, emotional and social processing across the lifespan. Behav Brain Funct. 2022;18:6. 10.1186/s12993-022-00193-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Kang W, Tiego J, Hellyer PJ, Trender W, Grant JE, Chamberlain SR, et al. Validation of an abbreviated Big Five personality inventory at large population scale: Psychometric structure and associations with common psychiatric and neurological disorders. Compr Psychiatry. 2024;134:152514. 10.1016/j.comppsych.2024.152514. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The UK Biobank database is accessible to researchers upon successful application through the UK Biobank website. Our study utilised the UK Biobank under research application number 83988. The code used for this study was tailored to the UK Biobank data and is no use as a standalone without access to the UK Biobank.
