The relationship between low birth weight and neurological disorders: a prospective cohort study in the UK Biobank

Abstract

Background

Birth weight as a marker of fetal growth has been linked to later health outcomes, but its relationship with adult neurological disorders is less well characterised, and few large studies have evaluated multiple neurological disorders within a unified analytical framework.

Methods

We analyzed 279,842 UK Biobank participants with self-reported birth weight. Birth weight was modeled continuously (per 1-kg increment) and categorically (low < 2.5 kg, reference 2.5–4.0 kg, high ≥ 4.0 kg). Incident multiple sclerosis (MS), Alzheimer’s disease (AD), Parkinson’s disease (PD), epilepsy, migraine and cerebrovascular disease (CVD) were ascertained from hospital and registry records (ICD-10). Logistic regression estimated OR and 95% CI, adjusting for age, sex, education, race, and socioeconomic status, with stratification by sex and age.

Results

Low birth weight was associated with higher risk of AD (OR = 1.31, 95% CI 1.14–1.51), epilepsy (OR = 1.37, 95% CI 1.24–1.51), and CVD (OR = 1.32, 95% CI 1.25–1.40). AD showed a U-shaped pattern, with high birth weight increasing risk (OR = 1.19, 95% CI 1.04–1.36). Each 1-kg increase in birth weight corresponded to 12% lower odds of epilepsy (OR = 0.88, 95% CI 0.84–0.93) and 11% lower odds of CVD (OR = 0.89, 95% CI 0.85–0.94). Findings were consistent across sex and age strata, and no associations were observed for MS, PD, or migraine after full adjustment.

Conclusions

Low birth weight was independently associated with increased risk of AD, epilepsy, and CVD in adulthood, but not with MS, PD, or migraine, after adjustment for measured confounders. These observational findings, which remain susceptible to residual confounding and measurement error, support the Developmental Origins of Health and Disease framework and underscore the enduring influence of early-life growth on neurological health. Future research should incorporate objective birth records, encompass more diverse populations, and explore how birth weight can be integrated into life-course risk prediction models and prenatal preventive strategies.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12883-026-04761-4.

Introduction

With population aging and longer life expectancy, the incidence of neurological disorders has risen substantially. In 2021, approximately 43% of the global population was living with neurological disorders, a 59% increase in cases since 1990 [1]. As leading causes of death and disability, these conditions impose a major public-health burden. Particularly among middle-aged and older adults who are at highest risk of neurodegenerative and cerebrovascular disease (CVD) [2]. Accordingly, the World Health Organization (WHO) places brain health at the center of its 2022–2031 Global Action Plan on Epilepsy and Other Neurological Disorders, emphasizing early screening, prevention, and comprehensive care [3].

Birth weight is a quantifiable marker of fetal growth widely examined in neurodevelopmental and psychiatric research [4, 5]. Experimental work shows that intrauterine growth restriction (IUGR) impairs oligodendrocyte differentiation during neurogenesis [6], more broadly, disrupts neurogenesis and gliogenesis, delays oligodendrocyte maturation and myelination, and alters cerebral vascular development and blood, brain barrier integrity mechanisms that plausibly link suboptimal fetal growth to long-term neurological vulnerability [7–9]. Consistent with these mechanisms, epidemiological studies associate low birth weight with increased risks of cognitive impairment and neurological disorders into adolescence and adulthood, and meta-analytic evidence indicates lower standardized cognitive scores among adults born with low birth weight [9–11].

In addition to low birth weight, high birth weight may also pose long-term health risks through persistent metabolic and vascular mechanisms. Recent population-based studies have consistently demonstrated a U-shaped relationship between birth weight and outcomes including dementia, mild cognitive impairment, and all-cause mortality [12], indicating that both extremes of birth weight are associated with adverse neurological and overall health outcomes. This suggests that deviation from the optimal birth weight range is linked to increased risk [13]. Despite extensive work on brain structure, direct relationships between birth weight and specific adult-onset neurological disorders remain incompletely characterized [14, 15]. Meanwhile, the findings of these studies are inconsistent. Some report a higher risk of dementia, stroke, or epilepsy in individuals with low birth weight, but after adjusting for socioeconomic and perinatal factors find weaker or no associations [16]. In some cases, slightly increased risks appear at the higher end of the birth weight range. These differences reflect variations in populations, outcome definitions, and methods [12, 17], emphasizing the need for large, well-defined cohorts using standardized exposure and outcome criteria.

Notably, few population-based studies have systematically examined multiple adult-onset neurological disorders within a single analytical framework. By leveraging the extensive scale and detailed phenotypic data of the UK Biobank, we applied consistent definitions of exposure, standardized covariate adjustments, and uniform sensitivity analyses to evaluate a range of common neurological disorders, thereby improving comparability across distinct outcomes. Furthermore, the study aims to investigate the association between low birth weight (< 2.5 kg) and the risk of adult-onset neurological disorders. It is hypothesized that low birth weight is associated with an elevated risk of such disorders, with a particular focus on Alzheimer’s disease (AD), epilepsy, and CVD. We evaluated non-linear associations across the full birth-weight spectrum, including high birth weight (≥ 4.0 kg), examined sex- and age-specific differences, and assessed whether elevated birth weight confers risk or protection relative to normal birth weight. Therefore, this study aims to establish a robust theoretical foundation for epidemiological and clinical research, while providing valuable insights and evidence-based support for public health strategies designed to mitigate early-life risks.

Methods

Ethics approval and consent to participate

The UK Biobank study received ethical approval from the National Health Services (NHS) National Research Ethics Service (Ref. 21/NW/0157) and was conducted in accordance with the principles of the Declaration of Helsinki. All participants provided written informed consent. Our secondary analysis was conducted under UK Biobank application number [657275], no direct personal identifiers were accessible to the research team, and all data management procedures adhered strictly to the UK Biobank’s information governance framework and data security policies.

Study design and population

Between 2006 and 2010, UK Biobank recruited 502,353 men and women aged 40–69 years across 22 assessment centers in England, Scotland, and Wales to form a prospective cohort for health-outcome research [18]. This study utilized data obtained from the follow-up survey conducted on October 31, 2022 [19].

At the baseline visit, participants completed a touchscreen questionnaire on sociodemographic characteristics (age, sex, race, education), early-life exposures, and lifestyle factors including smoking, alcohol use, and physical activity. A trained nurse conducted a face-to-face interview to confirm key responses and collect detailed personal and family medical histories. Standardized physical exams measured height and weight (for body mass index (BMI)), seated blood pressure (two readings, averaged), grip strength, and lung function via spirometry. Venous blood and spot urine samples were collected for biomarker analysis and long-term storage using UK Biobank protocols. Socioeconomic status was assessed using the Townsend Deprivation Index (TDI), based on national census data linked to residential postcodes. In this analysis, age, sex, race, education, and TDI were included as potential confounders [20, 21]. These were selected a priori because race, education, and area-level deprivation are linked to both birth weight and perinatal outcomes, and are established risk factors for adult neurological disorders, making them plausible confounders in the association between birth weight and neurological disorders [22, 23].

Finally, all baseline data were linked to national hospital episode statistics and death registries to identify incident cases of neurological disorders using linked hospital records [24]. This study examined prevalent neurological disorders. Participants with an established diagnosis at baseline were included, and no restriction to incident cases only was applied. We applied pre-specified plausibility thresholds to birth weight values. Among the 502,353 UK Biobank participants, 222,218 had no recorded birth weight and were therefore excluded. Of the remaining 280,135 participants with available data, those with birth weight values below 0.5 kg or above 6.0 kg (n = 293) were excluded as implausible, resulting in a final analytic sample of 279,842 participants.These plausibility thresholds are widely used in large epidemiological cohorts to minimize data entry errors while maintaining clinically relevant birth weight ranges [25]. The study flowchart is shown in Fig. 1.

Fig. 1.

The study flowchart

Assessment of birth weight

Birth weight was self-reported during the verbal interview/touchscreen assessment in UK Biobank (Data-Field 20022) [26]. In our primary analyses, birth weight was treated as a continuous variable, with increases of 1 kg, to maintain statistical power and reveal linear trends. Acknowledging potential nonlinearity, we also modeled birth weight categorically: low (< 2.5 kg), reference (2.5–4.0 kg), and high (≥ 4.0 kg). These cut points are consistent with the WHO definitions of low and high birth weight [27]. In the categorical analyses, the 2.5–4.0 kg group served as the reference category to facilitate comparisons with existing literature. Participants reported birth weight in kilograms or imperial units, which the UK Biobank converted to kilograms. Validation studies show moderate to good agreement between self-reported and recorded birth weight, though self-reporting may weaken exposure outcome associations and lead to conservative estimates [28].

Assessment of neurological disorders

Hospital inpatient records for all UK Biobank participants have been linked to the NHS Hospital Episode Statistics (HES) database, which provides detailed event-level data on admissions, procedures, and diagnoses coded in ICD-10. Each HES record includes admission and discharge dates, a primary diagnosis field, up to 19 secondary diagnosis fields, and OPCS-4 procedure codes, allowing for precise temporal mapping of incident disease events [24].

For this study, we selected ICD-10 codes for neurological disorders of primary interest in neurology practice and research. When determining the outcome, we systematically examine the primary diagnosis field along with up to 19 secondary diagnosis fields, if any of these fields include the predefined ICD-10 codes (multiple sclerosis (MS): G35, AD: G30, Parkinson’s disease (PD): G20, epilepsy: G40, migraine: G43, CVD: I60–I69), the event is classified as having occurred. UK Biobank has also generated algorithmically defined health outcomes by integrating HES, self-reported data, and death registry information. These resources include curated lists of ICD-10 codes and validated phenotype algorithms to enhance case ascertainment and comparability across studies [29].

Statistical analysis

Birth weight was modeled both continuously and categorically. Chi-squared tests were applied to categorical variables, while one-way analysis of variance (ANOVA) was employed for continuous variables. The statistical significance threshold was set at α = 0.05. All statistical analyses were performed using IBM SPSS Statistics (version 27), Stata MP (version 18.0), and R (version 4.4.2). Multivariable logistic regression models were used for each neurological disorder with a complete-case approach, among the 279,842 participants, those missing birth weight or any model covariate were excluded. Table 1 summarizes missing data for key covariates, which led to exclusion of only a small proportion.

Table 1.

Baseline characteristics of the study populations (n = 279,842)

Characteristics	Overall (n = 279,842)
Age (y)	55.23 ± 8.11
Female	170,774 (61.03%)
Race-White	258,822 (92.49%)
TDI	-1.47 (-1.48, -1.46)
Education (college)	95,146 (34.00%)
BMI (kg/m2)	27.31 ± 4.88
Alcohol consumption status
Never	10,107 (3.61%)
Former drinker	9,392 (3.36%)
Current drinker	259,921 (92.88%)
Smoking status
Never	157,712 (56.36%)
Former smoker	93,092 (33.27%)
Current smoker	28,012 (10.01%)
Regular physical activity	188,949 (67.52%)
Regular social activity	121,894 (43.56%)
Hypertension	142,126 (50.79%)
Diabetes	12,166 (4.35%)
Heart disease	15,188 (5.43%)

BMI body mass index, TDI Townsend deprivation index

Missing data: Age = 2, Sex = 2, Race = 875, Education = 4,341, BMI = 997, TDI = 361, Alcohol consumption = 422, Smoking = 1,026, Regular physical activity = 18,975, Regular social contact = 3,429, Hypertension = 2, Diabetes = 2, Heart disease = 2

Because a subset of diagnoses predated the baseline assessment and diagnosis dates were not consistently available across all cases, accurate determination of event timing, including onset and the at-risk population, was not feasible. Consequently, the application of the Cox proportional hazards model was precluded. The logistic regression models were employed to assess disease status at baseline, estimating the odds ratios (OR) and the corresponding 95% confidence intervals (CI) to evaluate the association between birth weight and neurological disorders. In the gender-stratified analysis, the gender variable was excluded from the model specification, while all other covariates were retained. Crude models included birth weight and the outcome only. Multivariable models prespecified adjustment for age, sex (pooled models only), race, education, and TDI. In sex-stratified analyses, the sex term was omitted with other covariates unchanged. Age and sex were included jointly as covariates in the same multivariable model. Stratified analyses were conducted to investigate potential effect modification by age group. Multivariable logistic regression was conducted on complete cases (participants with missing covariate data were excluded). Variance inflation factors (VIFs) indicated no multicollinearity (all VIFs < 2, maximum = 1.03).

Results

Baseline characteristics

A total of 279,842 participants were included and classified into three birth weight groups: low (n = 28,454, 10.17%), middle (n = 213,062, 76.14%), and high (n = 38,326, 13.70%). At baseline, the mean age was 55.23 ± 8.11 years, and females accounted for 61.17% of participants. The mean birth weight was 3.32 ± 0.66 kg. Distributions of age, sex, race, education, and TDI were broadly similar across the three birth-weight categories, with only modest differences in mean age and the proportions of women and more deprived participants (Table 1). Further details are provided in Table 1. In the analytic cohort (N = 279,842), the prevalence of neurological disorders was as follows: AD: 1,684 cases (0.60%), PD: 1,968 cases (0.70%), MS: 1,295 cases (0.46%), epilepsy: 3,667 cases (1.31%), migraine: 4,769 cases (1.70%), and CVD: 11,774 cases (4.21%).

Relationship between birth weight and neurological disorders

Figure 2 summarizes associations between birth weight and six neurological disorders (MS, AD, PD, epilepsy, migraine, and CVD). In primary analyses, birth weight was modeled as a continuous variable (per 1-kg increase) to estimate trend associations and maximize statistical power. Figure 2 shows the results for classification and continuous variables adjusted for age and gender. Full evaluation data across variable categories and adjustment models, including unadjusted and adjusted OR stratified by age and gender, are provided in Tables S1-S12 of the supplementary materials to ensure transparency and reproducibility. P values for sex and age interaction terms are reported to facilitate interpretation of potential effect modification (Tables S1–S12).

Fig. 2.

Association of birth weight with neurological disorder. This forest plot illustrates the model performance following adjustment for age and gender

Multiple sclerosis

Across all participants, higher birth weight was associated with lower MS risk in the crude model (OR = 0.91 per 1-kg increase, 95% CI: 0.83–0.98) (Table S1). After adjustment for age and sex, the association weakened and was no longer statistically significant (OR = 0.97, 95% CI: 0.89–1.05) (Fig. 2, Table S1). Further adjustment for education, race, and TDI yielded similar results (OR = 0.97, 95% CI: 0.89–1.06) (Table S1). In sex-stratified analyses, the estimate among females suggested that higher birth weight may be associated with lower MS risk (OR = 0.96, 95% CI: 0.87–1.06), although the 95% CI included 1, among males, there was no statistically significant association (OR = 0.99, 95% CI: 0.84–1.17) (Table S1). Compared with the reference group (2.5–4.0 kg), low birth weight (< 2.5 kg) was associated with higher MS risk (OR = 1.13, 95% CI: 0.94–1.37), though the 95% CI included 1 (Fig. 2, Table S1). A similar pattern was observed in the fully adjusted model (OR = 1.13, 95% CI: 0.93–1.36) (Table S1). Stratified by age, the association of higher birth weight with lower MS risk was significant only among participants younger than 60 years in the crude model (OR = 0.88, 95% CI: 0.80–0.97), but this association attenuated after adjustment and was no longer significant in the fully adjusted model (OR = 0.95, 95% CI: 0.86–1.05) (Table S2). No meaningful association was observed in those aged 60 years or older across all models (Table S2). Overall, the disappearance of the initially protective association after multivariable adjustment indicates that the crude inverse relationship between higher birth weight and MS risk is largely explained by confounding by age, sex, and socioeconomic factors rather than a robust independent effect of birth weight.

Alzheimer’s disease

There was a statistically significant association whereby higher birth weight was related to lower AD risk in the crude model (OR = 0.93 per 1-kg increase, 95% CI: 0.86–0.996) (Table S3), which remained significant after adjustment for age and sex (OR = 0.93, 95% CI: 0.87–0.998) (Fig. 2, Table S3), though it attenuated to non-significance in the fully adjusted model (OR = 0.94, 95% CI: 0.88–1.01) (Table S3). Relative to the reference group (2.5–4.0 kg), low birth weight (< 2.5 kg) was associated with higher AD risk in the age-sex adjusted model (OR = 1.31, 95% CI: 1.14–1.51) (Fig. 2, Table S3), consistent with the fully adjusted estimate (OR = 1.29, 95% CI: 1.11–1.49) (Table S3). Similarly, high birth weight (≥ 4.0 kg) was associated with elevated AD risk in the age-sex adjusted model (OR = 1.19, 95% CI: 1.04–1.36) (Fig. 2, Table S3), with a comparable effect in the fully adjusted model (OR = 1.18, 95% CI: 1.04–1.35) (Table S3). In sex-stratified analyses adjusted for age, the point estimates suggested lower AD risk with higher birth weight, although the associations were not statistically significant, and were somewhat stronger in males (OR = 0.90, 95% CI: 0.81–1.002) than in females (OR = 0.95, 95% CI: 0.87–1.05) (Table S3). Age-stratified results based on the age-sex adjusted model showed that low birth weight was more strongly associated with AD in participants < 60 years (OR = 1.72, 95% CI: 1.16–2.55) than in those ≥ 60 years (OR = 1.26, 95% CI: 1.08–1.47), high birth weight was significantly associated with AD only in the older group (OR = 1.17, 95% CI: 1.02–1.35) (Table S4). In an additional multivariable model with birth weight parameterized as a restricted cubic spline, the adjusted dose response relationship for AD exhibited a U-shaped pattern, characterized by the lowest predicted probability of AD at moderate birth weights and elevated risks at both extremes (Supplementary Figure S1). This non-linear trend aligns consistently with the categorical findings presented in Fig. 2. However, the estimate for the continuous association between birth weight and AD was attenuated to non-significance in the fully adjusted model, suggesting that the overall inverse association is modest and sensitive to confounding; thus, estimates pertaining to mid-range birth weights should be interpreted with caution.

Parkinson’s disease

For PD, the crude model indicated a positive association between continuous birth weight and PD risk (OR = 1.06 per 1 kg increase, 95% CI: 0.99–1.13) (Table S5). Compared with the reference group (2.5–4.0 kg), high birth weight (≥ 4.0 kg) was associated with a modest increase in PD risk in the crude model (OR = 1.22, 95% CI: 1.08–1.38) (Table S5), although this association was attenuated and not statistically significant after adjustment for age and sex (OR = 1.03, 95% CI: 0.91–1.16) (Fig. 2, Table S5), with further adjustment for education, race, and TDI yielding nearly identical results (OR = 1.02, 95% CI: 0.90–1.16) (Table S5). Low birth weight (< 2.5 kg) showed no meaningful association with PD in any model (Fig. 2, Table S5). In age-stratified analyses, a significant positive association between continuous birth weight and PD was observed among participants aged ≥ 60 years only in the crude model (OR = 1.08, 95% CI: 1.01–1.17). However, this association disappeared after adjustment for age and sex (OR = 1.01, 95% CI: 0.94–1.09) (Table S6). No statistically significant association was observed among those < 60 years in any model. Given the comparatively smaller number of PD cases and the later typical age at onset, our analyses may still have limited power to detect small effect sizes or associations confined to the oldest age groups.

Epilepsy

Higher birth weight was consistently associated with lower epilepsy risk in the crude model (OR = 0.90 per 1-kg increase, 95% CI: 0.86–0.94) (Table S7), which remained statistically significant after adjustment for age and sex (OR = 0.88, 95% CI: 0.84–0.93) (Fig. 2, Table S7). Further adjustment for education, race, and TDI yielded nearly identical results (OR = 0.89, 95% CI: 0.85–0.94) (Table S7). In the age-sex adjusted model, individuals with low birth weight (< 2.5 kg) had a substantially higher risk of epilepsy compared with the reference group (2.5–4.0 kg) (OR = 1.37, 95% CI: 1.24–1.51), while high birth weight (≥ 4.0 kg) showed no significant association (OR = 1.05, 95% CI: 0.96–1.16) (Fig. 2, Table S7). The association of higher birth weight with lower epilepsy risk was stronger in females (age-sex adjusted OR = 0.87, 95% CI: 0.81–0.93) than in males (OR = 0.90, 95% CI: 0.84–0.97) (Table S7). Age-stratified analyses based on the age-sex adjusted model also indicated a more pronounced association between higher birth weight and lower epilepsy risk among participants aged < 60 years (OR = 0.86, 95% CI: 0.80–0.92) than among those aged ≥ 60 years (OR = 0.92, 95% CI: 0.85–0.98) (Table S8).

Migraine

Higher birth weight was associated with lower migraine risk in the crude model (OR = 0.91 per 1-kg increase, 95% CI: 0.87–0.95) (Table S9), but this association was no longer evident after adjustment for age and sex (OR = 0.99, 95% CI: 0.94–1.03) (Fig. 2, Table S9). Low birth weight (< 2.5 kg) was associated with a modestly increased risk of migraine in the crude model (OR = 1.15, 95% CI: 1.05–1.26), but this association was attenuated and no longer statistically significant after adjustment for age and sex (OR = 1.08, 95% CI: 0.99–1.18) (Fig. 2, Table S9). In contrast, high birth weight (≥ 4.0 kg) remained modestly but significantly associated with increased migraine risk in the age-sex adjusted model (OR = 1.12, 95% CI: 1.02–1.22) (Fig. 2, Table S9), with nearly identical estimates after full adjustment for education, race, and TDI (OR = 1.11, 95% CI: 1.02–1.21) (Table S9). Sex-stratified analyses showed similar effect directions in males and females, with no evidence of interaction. Age-stratified analyses revealed no significant associations in either age group (< 60 or ≥ 60 years) after adjustment (Table S10). Overall, birth weight does not appear to be an independent predictor of migraine, though a small excess risk may persist at the highest birth weights.

Cerebrovascular disease

Higher birth weight was associated with lower CVD risk in the crude model (OR = 0.93 per 1-kg increase, 95% CI: 0.90–0.95) (Table S11), and this association persisted after adjustment for age and sex (OR = 0.89, 95% CI: 0.86–0.91) (Fig. 2, Table S11), with a similar estimate after full adjustment for education, race, and TDI (OR = 0.90, 95% CI: 0.88–0.93) (Table S11). Compared with the reference group (2.5–4.0 kg), low birth weight (< 2.5 kg) was associated with a higher CVD risk in the age-sex adjusted model (OR = 1.32, 95% CI: 1.25–1.40) (Fig. 2, Table S11), which remained elevated after full adjustment (OR = 1.27, 95% CI: 1.20–1.34) (Table S11). In contrast, high birth weight (≥ 4.0 kg) was not associated with CVD risk in either the age-sex adjusted (OR = 1.003, 95% CI: 0.95–1.06) or fully adjusted models (OR = 0.99, 95% CI: 0.93–1.04) (Table S11). In sex-stratified analyses, the association of higher birth weight with lower CVD risk was stronger in females (OR = 0.84, 95% CI: 0.81–0.87) than in males (OR = 0.94, 95% CI: 0.90–0.97) (Table S11). In age-stratified analyses, the association was more pronounced among participants < 60 years (OR = 0.85, 95% CI: 0.81–0.89) and remained statistically significant in those ≥ 60 years (OR = 0.91, 95% CI: 0.88–0.94) (Table S12). High birth weight was not independently associated with CVD risk in either age group.

Discussion

In this large, population-based prospective cohort, birth weight was differentially associated with the risk of six neurological disorders of clinical and research interest in adulthood. Compared with the reference range (2.5–4.0 kg), low birth weight (< 2.5 kg) was associated with moderately higher odds of AD, epilepsy, and CVD, corresponding to about 30–40% increases in risk in age- and sex-adjusted models. For instance, an age- and sex-adjusted OR of approximately 1.3 for CVD corresponds to an estimated absolute increase of 1 to 2 additional CVD cases per 100 individuals with low birth weight, compared to those with normal birth weight, assuming a baseline CVD prevalence of approximately 4%. In contrast, no statistically significant associations were observed for MS or PD after adjustment for age and sex. For migraine, low birth weight showed only a modest, non-significant risk increase after adjustment, suggesting substantial confounding by demographic factors. The lack of significant associations between low birth weight and MS or PD is consistent with previous studies [30, 31]. Where previous cohorts have reported weaker or null associations for stroke or dementia, CI generally overlap our estimates, and differences in age structure, outcome definitions, and covariate adjustment may account for much of the apparent heterogeneity [32].The increased risk of AD is consistent with UK Biobank neuroimaging sub-studies linking both low and high birth weights to reduced brain reserve and higher dementia incidence [15]. Although the categorical associations were robust, the inverse linear trend for continuous birth weight attenuated after full adjustment, so very small deviations in birth weight should not be over-interpreted. Similarly, the strong association with epilepsy supports a developmental origin of seizure susceptibility, as observed in cohorts where very low birth weight conferred substantially elevated risk [33]. For migraine, low birth weight (< 2.5 kg) showed a modest, non-significant risk increase after adjusting for age and sex (OR = 1.08, 95% CI: 0.99–1.18), suggesting confounding by these factors. In contrast, high birth weight (≥ 4.0 kg) was linked to a small but significant increase in migraine risk, which remained after full adjustment for education, race, and TDI (fully adjusted OR = 1.11, 95% CI: 1.02–1.21, Table S9).Given the small effect sizes, weakened associations after age and sex adjustment, and the fact that hospital records likely capture only severe or treatment-resistant migraine cases, we consider the migraine findings largely null, possibly influenced by residual confounding and selection bias. High birth weight was not associated with MS, PD, epilepsy, or CVD in multivariable models (Tables S3, S5, S7, S9, and S11). Our findings extend this vulnerability into mid and late adulthood, reinforcing low birth weight as a persistent marker of neurological risk.

These patterns align with the Developmental Origins of Health and Disease (DOHaD) framework, which posits that adverse intrauterine conditions reflected by low birth weight may program long-term alterations in brain structure and function [34]. In particular, fetal undernutrition, hypoxia, and maternal stress can trigger adaptive changes in endocrine and metabolic systems and induce persistent epigenetic modifications, thereby shaping neural circuit formation, vascular development, and cardiometabolic set-points across the life course [35, 36]. Growing evidence indicates that such early-life exposures exert lifelong influences on neurological health [37, 38]. In addition, our findings and previous work on cardiometabolic and multimorbidity outcomes are consistent with a “dual-risk” pattern in which both fetal undergrowth and overgrowth may be adverse, with modestly elevated risks observed at the upper tail of the birth-weight distribution in some cohorts [39].Importantly, further adjustment for education, race, and socioeconomic status only modestly attenuated the effect estimates, which remained statistically significant for AD, epilepsy, and CVD, supporting the robustness of the primary associations. The stratified analysis indicated that associations were more pronounced among women and participants younger than 60 years, suggesting effect modification by sex and age. Experimental and epidemiological DOHaD research indicates that adverse intrauterine environments can program sex-specific cardiometabolic and vascular trajectories, with women born small for gestational age showing increased risks of hypertension, coronary disease, and stroke later in life [12, 40] These programmed effects may interact with women’s higher lifetime risk of dementia and CVD, particularly around the menopausal transition when declining estrogen levels reduce neuroprotective and vasoprotective influences. In addition, sex differences in cardiovascular risk profiles and healthcare-seeking behavior may amplify early-life vulnerabilities or increase the likelihood that neurological disorders such as migraine are diagnosed and recorded in women [41]. The stronger associations in individuals aged < 60 years may reflect an age-dependent expression of early-life programming, consistent with DOHaD evidence that the impact of low birth weight on cardiometabolic and cognitive outcomes tends to be most evident in midlife and attenuates in older age because of competing risks and selective survival [42, 43].

It is noteworthy that distinct dysregulation mechanisms may partially account for the heterogeneity observed in epidemiological associations across neurodevelopmental and neurodegenerative diseases [44]. In the context of AD, lower birth weight has been associated with reduced total brain volume across the lifespan, as well as compromised structural integrity of the hippocampus and white matter tracts. These alterations may contribute to diminished brain reserve capacity and potentially lower the threshold for the clinical manifestation of neurodegenerative disorders. Regarding epilepsy, fetal growth restriction is frequently concomitant with perinatal hypoxia-ischemia and neonatal seizures, which can result in persistent cortical and subcortical injuries [45, 46]. Such neurological sequelae are linked to an elevated risk of developing epilepsy during childhood. In CVD, both fetal growth restriction and low birth weight have been associated with long-term vascular and metabolic dysregulations, including sustained endothelial dysfunction, increased arterial stiffness, and insulin resistance [47, 48]. At the upper end of the distribution, high birth weight or macrosomia often reflects intrauterine overnutrition driven by maternal obesity and gestational diabetes. Maternal hyperglycemia and the resulting fetal hyperinsulinemia promote excessive somatic and adipose growth, chronic low-grade inflammation, and insulin resistance, and have been linked to altered neonatal brain microstructure and neuroinflammatory signaling. These physiological changes may serve as plausible biological pathways linking early-life growth impairment to an increased susceptibility to stroke and other cerebrovascular events in later life [49, 50]. Collectively, these findings indicate that neurodegenerative diseases, such as AD and, to a certain extent, epilepsy, are more closely associated with early-life influences on brain development, neuronal reserve capacity, and network excitability. In contrast, neurovascular disorders, including CVD, are predominantly linked to vascular and cardiometabolic programming pathways, specifically endothelial dysfunction, arterial stiffness, and insulin resistance. While there is partial overlap in underlying cardiometabolic risk factors, the primary pathophysiological mechanisms appear to differ between neurodegenerative and neurovascular diseases [51–53].

Our observational findings support the potential of targeted prenatal interventions to reduce neurological risks, from a public health perspective. The associations between low birth weight and AD, epilepsy, and CVD add to evidence highlighting the need for improved maternal and perinatal care [54]. In particular, modifiable maternal factors such as diet quality, smoking, and glycaemic control are central to optimising fetal growth. Ensuring adequate maternal nutrition and micronutrient intake, promoting healthy gestational weight gain, and providing equitable access to antenatal care can reduce the prevalence of fetal growth restriction and low birth weight [55]. Parallel efforts to prevent and manage gestational diabetes, and to achieve complete avoidance of active and passive smoking during pregnancy, may help to limit both fetal undergrowth and overgrowth and thereby reduce long-term neurological and cardiometabolic burden [56–58]. Integrating birth-weight information with other indicators of fetal wellbeing (such as fetal growth trajectories and placental function) could help identify infants at elevated long-term neurological risk who may benefit from closer developmental follow-up and aggressive management of modifiable risk factors across the life course. Notably, the association of higher birth weight with lower risk of AD, epilepsy, and CVD extended across much of the reference range (2.5–4.0 kg), indicating that risk is not confined to conventional “low birth weight” thresholds [9]. Taken together with evidence that very high birth weight is linked to later cardiometabolic risk [12, 13]. This pattern, in light of existing DOHaD and perinatal research, suggests prevention strategies should focus on optimizing fetal growth within a healthy range preventing both undergrowth and overgrowth rather than merely normalizing birth weight.

Overall, we leveraged the ultra-large scale of the UK Biobank cohort to investigate the associations between birth weight and a broad spectrum of adult neurological disorders within a unified analytical framework, thereby generating comparative risk estimates that are seldom attainable in smaller studies [30, 31, 33]. Second, we applied both continuous and categorical modeling of birth weight and utilized flexible statistical methods to examine potential non-linear relationships, which enabled the detection of U-shaped association patterns for certain outcomes [59]. Third, we conducted stratified analyses by sex and age to assess effect modification and observed stronger associations among females and individuals under 60 years of age, offering novel insights into the populations and life stages in which early growth may be most relevant to neurological vulnerability. Meanwhile, this study has several limitations. First, birth weight was self-reported retrospectively in mid- to late adulthood, which may introduce recall bias and measurement error. Validation studies suggest that self-reported birth weight shows only moderate agreement with official records [60], and misclassification of birth-weight categories is likely. Importantly, such error is probably non-differential with respect to neurological disorders in this cohort, which would bias associations toward the null and render our estimates conservative [28].Second, we were unable to adjust for gestational age or to distinguish small- and large-for-gestational-age infants from those appropriate for gestational age, and key maternal health variables (such as maternal BMI, smoking, hypertension, and gestational diabetes) were not available. The use of plausibility thresholds for birth weight (0.5–6.0 kg) may have under-represented extreme growth phenotypes, slightly limiting generalizability to individuals at the tails of the birth-weight distribution. Furthermore, neurological disorders were obtained from hospital episode statistics and linked registries. Although these sources are widely used and generally reliable, they are subject to diagnostic miscoding and may fail to capture milder or outpatient-managed cases, particularly for migraine and epilepsy. Similar issues may also affect CVD, where transient or less severe vascular events might be under-recorded. Any such misclassification is likely to be non-differential with respect to birth weight and would therefore tend to attenuate the true associations. These limitations may result in reduced etiological specificity and should be considered when interpreting the stronger associations observed for epilepsy and CVD. These limitations may result in residual confounding and reduced etiological specificity, future linkage with birth registration and maternity records may help mitigate this issue. Moreover, survival and selection biases may be present. Individuals with low birth weight who survived to reach eligibility for enrollment in the UK Biobank at age ≥ 60 years may constitute a relatively healthier subgroup compared to their counterparts in the source population, potentially exhibiting more favorable risk profiles. This survival effect could lead to attenuation of observed associations among older participants and result in age-stratified estimates being biased toward the null [42, 43]. Additionally, the observational design precludes strong causal inference, and residual confounding by unmeasured factors (for example, genetic susceptibility or unrecorded early-life environmental exposures) remains possible. Longitudinal analyses and genetic approaches such as Mendelian randomization will be required to more robustly assess whether the observed associations between birth weight and adult neurological disorders are causal. Finally, the UK Biobank cohort is predominantly of European ancestry and healthier than the general population, limiting generalizability and precluding detailed examination of effect modification by race [61, 62].

Conclusions

Low birth weight was associated with modestly increased odds of AD, epilepsy, and CVD in adulthood, but not with MS, PD, or migraine, after adjustment for measured covariates. These observational findings, which remain vulnerable to residual confounding and selection bias, support the view that early-life growth contributes to later-life neurological vulnerability and suggest that perinatal indicators could be considered alongside traditional vascular and metabolic factors in life-course risk assessment. Future studies in more diverse populations with prospectively recorded birth and gestational data, richer maternal health information, and causal-inference approaches will be essential to clarify mechanisms and to determine whether modifying early-life exposures can reduce the burden of neurological disorders.

Abbreviations

AD

Alzheimer’s disease

ANOVA

analysis of variance

BMI

body mass index

CI

confidence intervals

CVD

cerebrovascular diseases

DOHaD

Developmental Origins of Health and Disease

HES

Hospital Episode Statistics

MS

multiple sclerosis

NHS

National Health Services

WHO

World Health Organization

OR

odds ratios

PD

Parkinson’s disease

TDI

Townsend deprivation index

IUGR

intrauterine growth restriction

VIFs

Variance inflation factors

Authors’ contributions

MXY and LHY: Conceptualization, Methodology and Writing-Original Draft; MXY: Project administration and Funding Acquisition; LHY and MXY: Writing-review and editing; LHY, CY, MDH and ZY, JFY: Data Curation & Validation; GXJ, WMJ, and XJ: Software; MXY, XY, WWH and FXY: Supervision.

Funding

Chongqing Talents : Exceptional Young Talents Project (cstc2021ycjh-bgzxm0008).

Data availability

The data for this research are available on the UK Biobank under application number 657275 (www.ukbiobank.ac.uk/).

Declarations

Ethics approval and consent to participate

The UK Biobank study received ethical approval from the National Health Services National Research Ethics Service (Ref. 21/NW/0157) and was conducted in accordance with the principles of the Declaration of Helsinki. All participants provided written informed consent. Our secondary analysis was conducted under UK Biobank application number [657275], in compliance with data access and research use policies.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.