Maternal vitamin D deficiency affects approximately 54% of pregnant women globally (1), with prevalence reaching 74–84% in Chinese populations (2, 3). However, these rates remain inadequately characterized in subtropical Asian regions. Vitamin D deficiency is a major global public health challenge associated with adverse maternal and child health outcomes across diverse populations (4, 5). Emerging evidence challenges the traditional focus on deficiency prevention, as maternal fat-soluble vitamins show U-shaped associations with perinatal outcomes (6, 7), where both deficiency and excess pose risks. For vitamin D, optimal birth outcomes are observed at intermediate 25(OH)D levels (6); for vitamin A, both deficiency and excessive intake are linked to adverse fetal outcomes (8); and elevated maternal vitamin E concentrations are associated with macrosomia in Chinese cohorts (7). These findings necessitate a paradigm shift toward population-specific optimization rather than indiscriminate supplementation (9, 10).

While demographic and nutritional determinants have been extensively characterized (5), environmental modifiers, particularly meteorological factors, remain inadequately understood for vitamins A and E. Ambient temperature has emerged as the strongest environmental predictor of vitamin D levels (11, 12), with each 1 °C increase associated with a 3–4% reduction in deficiency risk (11). The mechanistic basis involves UVB-mediated cutaneous synthesis of vitamin D₃, which is highly dependent on solar zenith angle and atmospheric conditions (13, 14). Recent Chinese cohorts confirm marked seasonal variation, with summer 25(OH)D concentrations exceeding winter values by 20–30% (3, 15). However, the influences of humidity, precipitation, and wind speed, especially their interactive effects with temperature, remain largely unexplored. This knowledge gap is particularly evident for vitamins A and E, where meteorological associations have not been systematically investigated despite plausible links through seasonal dietary patterns and temperature-related metabolic changes.

Accumulating evidence reveals nonlinear dose–response relationships, highlighting the need to identify optimal ranges beyond simple deficiency thresholds. For vitamin D, observational studies document U-shaped associations with small-for-gestational-age risk, with nadirs at 25–40 ng/mL (6, 16, 17). Several cohorts report similar patterns for birth weight, where both low (<20 ng/mL) and high (>40 ng/mL) maternal 25(OH)D levels are associated with suboptimal outcomes (18). Vitamin A demonstrates biphasic toxicity: deficiency (<200 ng/mL) correlates with maternal anemia and infections (8), while excessive intake (>10,000 IU/day) increases teratogenic risk fourfold (19). For vitamin E, a large Chinese cohort (n = 8,554) reported that concentrations exceeding 20,000 ng/mL were associated with increased risk of macrosomia (aOR = 1.60) (7). Micronutrient optimization in pregnancy requires context-specific approaches given population heterogeneity in nutritional status and environmental exposures (20, 21). Critically, these nonlinear patterns appear population-specific (6, 7, 18), underscoring the need for regional calibration rather than universal application of Western reference ranges.

Geographic representation in maternal vitamin research remains skewed toward high-latitude temperate (>40°N) and coastal subtropical regions (1, 22), leaving inland subtropical populations underrepresented. Wuhan (30°N, central China) serves as an ideal natural laboratory: its humid subtropical climate features extreme seasonal meteorological variation, allowing examination of environmental influences within a single population (11). Additionally, rapid nutrition transition (23) and increasing maternal obesity prevalence [37.5% in our cohort with BMI > 28 kg/m², consistent with national trends (24)] may result in distinctive vitamin profiles in contemporary Chinese urban populations. Despite these unique characteristics, no previous studies have simultaneously assessed vitamins D, A, and E with comprehensive meteorological integration in inland subtropical China.

Three critical knowledge gaps hinder the optimization of maternal vitamin supplementation in Asian populations. First, environmental determinants beyond latitude, particularly the interactive effects of temperature, humidity, and wind speed, remain uncharacterized for vitamins A and E, despite well-established influences on vitamin D. Second, dose–response relationships may exhibit population-specific thresholds, yet Western-derived reference ranges have not been validated in Chinese cohorts using continuous modeling approaches. Third, most studies focus on single micronutrients, limiting the assessment of potential synergistic or antagonistic interactions among vitamins D, A, and E across multiple perinatal outcomes. However, whether these associations reflect causal effects or late-pregnancy correlates remains uncertain, highlighting the need for prospective studies with serial measurements.

To address these gaps, we conducted a retrospective cohort study of 10,824 third-trimester pregnant women in Wuhan, China (2020–2023), integrating HPLC-measured serum vitamins, daily meteorological data, and comprehensive clinical records. Our objectives were to: (1) characterize the distributions of vitamins D, A, and E and their meteorological determinants, with an emphasis on interactive effects; (2) quantify associations with five perinatal outcomes using sequential adjustment models; and (3) evaluate dose–response patterns through restricted cubic spline regression. By simultaneously assessing three vitamins with environmental integration, this study provides a framework for optimizing maternal micronutrient status in subtropical Asian populations.

Abnormalities in maternal fat-soluble vitamin status during late pregnancy were common in this cohort and were consistently associated with delivery and neonatal outcomes in central China. First, we simultaneously examined three fat-soluble vitamins with comprehensive meteorological integration, revealing temperature as the strongest environmental correlate (3.0% lower vitamin D deficiency risk per 1 °C increase in multivariable models) with significant interaction terms (temperature × humidity and temperature × wind), indicating that the associations between temperature and vitamin status differed according to concurrent humidity and wind conditions. Second, we identified a striking threefold higher odds of macrosomia associated with vitamin A deficiency—a relationship inadequately characterized in previous research—alongside an inverse association between vitamin A insufficiency and low birth weight observed in categorical analyses, which should be interpreted within a broader pattern of non-linear exposure–response relationships rather than as evidence of biological protection. Third, we demonstrated that 41.8% of women exhibited vitamin E excess, which was associated with elevated macrosomia risk, challenging the conventional focus on deficiency prevention alone and revealing a bidirectional risk profile that requires optimization rather than maximization of vitamin status. These findings provide the first systematic characterization of meteorological determinants across multiple vitamins and highlight the urgent need for population-specific evidence to inform nutritional strategies in subtropical Asian populations. Consistent with prior observational literature, vitamin D deficiency was associated with higher odds of macrosomia, highlighting the relevance of maternal vitamin D status to fetal growth outcomes (33).

Our observed vitamin D deficiency prevalence (30.2%) is substantially lower than previously reported rates in northern Chinese populations (2, 3) and closely matches rates documented in other subtropical and tropical regions globally (4, 21), reflecting greater year-round UVB exposure at lower latitudes. This geographic variability in maternal vitamin D status (4) highlights the need for latitude-specific reference ranges and supplementation protocols rather than uniform global guidelines. The 10,824-participant cohort from Wuhan (30°N) provides critical data that fill a gap in the literature for inland subtropical Asian populations.

In addition to geographic variation, vitamin D status was associated with demographic factors. The inverse association between BMI and vitamin D status observed in our cohort (32.6% deficiency in the highest vs. 28.0% in the lowest BMI tertile) aligns with well-established pathophysiological mechanisms (13, 34). Obesity-associated vitamin D deficiency may involve multiple pathways: (1) sequestration of lipophilic 25(OH)D in expanded adipose tissue compartments, reducing bioavailable circulating concentrations; (2) volumetric dilution in larger body pools; and (3) potentially decreased cutaneous synthesis due to reduced outdoor physical activity (13, 34). This relationship underscores the need for BMI-adjusted supplementation protocols, as obese pregnant women require higher doses to achieve equivalent serum 25(OH)D concentrations compared to normal-weight women.

Beyond BMI, age-related and seasonal patterns in vitamin status reflect distinct behavioral mechanisms aside from meteorological factors alone. The higher vitamin D deficiency in younger women aligns with contemporary lifestyle patterns documented across Asian populations, where studies have reported deficiency prevalence exceeding 90% among young women attributed to indoor occupations, sun avoidance behaviors for skin protection, and reduced outdoor activities (35–37). Winter vitamin D nadir (22.6 ng/mL) likely reflects both reduced UVB availability and behavioral thermoregulation through increased clothing coverage and indoor preference during cold periods. The absence of urban–rural differences in our cohort contrasts with studies documenting higher urban deficiency rates (38), suggesting relatively homogeneous lifestyle patterns and sun exposure behaviors throughout our study region, possibly reflecting advanced urbanization where outdoor agricultural work is uncommon and healthcare access is equitable across residential areas.

The strong association between vitamin A deficiency and macrosomia with steep dose–response gradients represents our most clinically significant and novel finding. While excessive vitamin A intake during early pregnancy is associated with congenital malformations (8), the relationship between deficiency and macrosomia remains poorly characterized. Vitamin A has been shown to play critical roles in placental development and nutrient transport in experimental studies (39); deficiency may be associated with altered placental function, though whether this represents a causal mechanism or reflects confounding requires further investigation. Retinoic acid, the active metabolite of vitamin A, has been shown in experimental studies to regulate expression of genes involved in glucose homeostasis and insulin signaling in the placenta, though the relevance of this pathway to the observed associations in our population-based study remains unclear (40). These mechanistic interpretations are speculative, as placental function, glucose metabolism, and dietary intake were not directly measured in this study.

The observed inverse association between vitamin A insufficiency and low birth weight in categorical analyses was compatible with non-linear exposure–response patterns; however, restricted cubic spline analyses showed confidence intervals crossing unity across the exposure range, precluding inference of a protective or biologically optimal concentration. For vitamin D and macrosomia, spline curves were compatible with a U-shaped pattern with a nadir near the mid-range of observed concentrations, but no confidence-interval–supported optimal range could be defined, consistent with prior observational studies reporting non-linear associations between maternal 25(OH)D and fetal growth outcomes (41). For vitamin E, a threshold-like association was observed, with higher concentrations associated with increased odds of macrosomia. This finding is notable given that 41.8% of women in the cohort exceeded 20,000 ng/mL. However, because detailed data on dietary intake and supplement use were unavailable, these associations should be interpreted cautiously and do not permit inference regarding optimal intake levels during pregnancy.

These patterns suggest that maternal vitamin optimization may require balancing risks across multiple perinatal outcomes rather than simply preventing deficiency. However, several critical limitations preclude definitive determination of optimal ranges from our data. First, cross-sectional assessment at a single gestational timepoint cannot establish whether the observed concentrations represent stable maternal status or reflect dynamic changes related to fetal demands, as vitamin status demonstrates substantial longitudinal variation from preconception through pregnancy that cannot be inferred from single-timepoint measurements (42). Second, optimal ranges may differ across perinatal outcomes (e.g., the vitamin A concentration minimizing low birth weight risk may differ from that minimizing macrosomia risk). Third, population heterogeneity in genetics, dietary patterns, and supplement use likely results in individual-specific rather than universal optimal ranges (43). Prospective cohort studies with serial measurements across all trimesters, coupled with randomized dose-finding trials stratified by baseline vitamin status and individual characteristics, are essential to validate these preliminary observations and establish evidence-based supplementation protocols for subtropical Asian populations. This pattern underscores the complexity of maternal nutritional optimization beyond simple deficiency prevention.

The apparent “protective” association between vitamin A insufficiency and low birth weight warrants cautious interpretation and should not inform clinical practice. This counter-intuitive finding likely reflects methodological artifacts rather than genuine biological benefit. Five alternative explanations merit consideration: First, residual confounding from unmeasured socioeconomic factors, dietary patterns, and supplement use may simultaneously influence vitamin A status and birth weight (44). Second, survival bias could selectively exclude pregnancies with both severe insufficiency and growth restriction, as such fetuses may experience earlier pregnancy loss or preterm delivery (excluded from our ≥37-week cohort). Third, the U-shaped dose–response demonstrated by restricted cubic spline analysis, together with overlapping confidence intervals across the exposure range, indicates that optimal vitamin A status lies within a physiological range rather than insufficiency being beneficial; categorical analyses may therefore reflect statistical artifacts at curve extremes. Fourth, reverse causation cannot be excluded, as maternal vitamin status at 37–40 weeks may be influenced by pre-existing fetal growth patterns through placental adaptations. Finally, chance findings remain possible despite statistical significance, as multiple comparisons across three vitamins, five outcomes, and three exposure categories increase the probability of spurious associations (45). These considerations indicate that vitamin A insufficiency should not be interpreted as protective. Clinical recommendations should continue emphasizing adequate vitamin A intake through balanced diet and appropriate supplementation rather than targeting insufficiency. Future prospective studies with serial measurements, comprehensive dietary assessment, and standardized outcome ascertainment are needed to definitively characterize optimal vitamin A ranges during pregnancy.

In addition to neonatal growth outcomes, delivery-related events also showed consistent associations with maternal vitamin status. Vitamin A deficiency was associated with higher odds of premature rupture of membranes. While experimental studies suggest vitamin A influences epithelial integrity, collagen metabolism, and oxidative balance (46, 47), whether these mechanisms underlie the observed association or reflect residual confounding cannot be determined from our cross-sectional data. Conversely, low to moderate levels of vitamins D and A were associated with lower risks of fetal distress and meconium-stained amniotic fluid, though the mechanisms underlying these associations remain unclear and require investigation in prospective studies (48). Although these associations were modest in magnitude, they suggest that maternal vitamin homeostasis may influence both the structural and functional stability of the delivery process, complementing its well-established role in regulating fetal growth (49).

Our finding that 41.8% exhibited vitamin E excess with associated macrosomia risk aligns with recent Chinese research demonstrating increased odds of large-for-gestational-age with elevated α-tocopherol (7). Mechanisms underlying vitamin E excess-associated macrosomia remain incompletely understood but may involve multiple pathways. At supraphysiological concentrations, α-tocopherol has been shown in experimental studies to exhibit pro-oxidant rather than antioxidant properties (34, 50, 51). However, whether such mechanisms explain the observed association with macrosomia cannot be inferred from the present epidemiological data and may also reflect confounding by unmeasured dietary and supplementation patterns. Experimental studies have demonstrated that oxidative stress can affect placental function and fetal growth through effects on trophoblast invasion, vascular remodeling, and nutrient transfer, though the relevance of this pathway to the observed statistical associations in our population-based study remains speculative (51). Essential micronutrients, including fat-soluble vitamins, are known from experimental studies to function optimally within narrow physiological ranges (40). Whether deviations from these ranges causally contribute to adverse outcomes or simply correlate with other causal factors remains uncertain in observational studies. Additionally, excessive vitamin E has been shown in some studies to potentially affect the absorption and metabolism of other fat-soluble vitamins, particularly vitamin K. However, whether this contributes to the observed associations requires investigation.

The high prevalence of vitamin E excess observed in this cohort may reflect a combination of routine prenatal supplementation, dietary patterns, their interaction, or other unmeasured factors, rather than supplementation alone. In urban China, prenatal multivitamin use is common, and when combined with dietary sources such as vegetable oils, nuts, and fortified foods, total vitamin E intake may exceed physiological requirements. However, because individual-level data on supplement use and detailed dietary intake were unavailable, the relative contribution of these factors cannot be disentangled. Accordingly, the observed association between vitamin E excess and macrosomia should be interpreted with caution, and causal inference cannot be established. Taken together, these findings suggest that both deficient and excessive vitamin E status may be associated with unfavorable pregnancy outcomes, underscoring the importance of avoiding indiscriminate supplementation and the need for future studies incorporating detailed dietary assessment and supplement-use data to inform population-specific nutritional recommendations.

Beyond individual vitamin-outcome associations, understanding the environmental determinants of maternal vitamin status is critical for designing effective interventions. Our systematic meteorological analysis identifies temperature as the primary environmental correlate, with each 1 °C increase associated with a 3.0% lower risk of vitamin D deficiency and a 2.4% lower risk of vitamin A deficiency. Previous studies in subtropical China have shown similar temperature-vitamin D correlations (11), but our findings extend these observations by: (1) demonstrating temperature effects on vitamin A; (2) identifying significant meteorological interactions―temperature × humidity and temperature × wind speed―indicating significant effect modification, such that the associations between temperature and vitamin status differed under concurrent atmospheric conditions; and (3) revealing that vitamin E excess peaks in autumn (46.6%) compared to summer and winter (38.4–38.8%), suggesting season-specific dietary or metabolic patterns.

Temperature emerged as the predominant environmental correlate, and the observed non-linear relationships and interaction terms are compatible with combined biological and behavioral influences. The strong temperature-vitamin D association observed in our study is consistent with multiple plausible mechanisms, including enhanced cutaneous vitamin D3 synthesis through increased UV-B exposure (13, 14), behavioral changes such as reduced clothing and increased outdoor activity, and altered dietary patterns during warmer periods (12, 52). Dietary patterns in meat-eaters versus vegetarians significantly influence vitamin D status (14), and similar dietary shifts may occur seasonally. Importantly, because meteorological variables were standardized, interaction coefficients reflect relative effect modification rather than absolute changes in vitamin concentrations. These interaction effects suggest heterogeneity in associations across climatic contexts, rather than uniform temperature effects. The observed temperature-humidity interaction may operate through both physical and behavioral mechanisms, though the relative contributions of each cannot be disentangled in this observational study. Behaviorally, thermal discomfort from combined heat and humidity may discourage outdoor activity, with studies documenting paradoxical seasonal patterns where vitamin D levels are lower in summer than winter in hot, humid climates (37, 53).

The association between temperature and vitamin D status exhibited threshold-dependent patterns in previous large cohort studies, 25(OH)D concentration changes were minimal below 20 °C, showed maximum variation between 20–25 °C when outdoor activity was most comfortable, and then declined above 25 °C as thermal discomfort increased (11). Combined climatic conditions consistently outperform single-factor models in predicting maternal vitamin D status across subtropical populations, but the relative importance of specific factors varies by climate zone. In a population-based study of 1,502 pregnant women in Taiwan, the determinant hierarchy reversed between climate zones: southern tropical regions showed sunlight factors exceeding dietary factors, while northern subtropical regions exhibited dietary factors exceeding sunlight factors (12). Multiple studies across diverse geographic settings have confirmed that vitamin D deficiency relates to combined environmental factors rather than any single predictor (54).

These findings suggest that integrated climatic condition assessment may help identify periods of elevated deficiency risk and inform geographically-tailored intervention strategies. Importantly, seasonal variation in vitamins A and E may also be strongly influenced by season-dependent changes in food availability and dietary composition (e.g., intake of liver/egg/dairy and carotenoid-rich fruits and vegetables for vitamin A, and vegetable oils, nuts, and seeds for vitamin E), rather than meteorological factors alone (55, 56). Because detailed dietary and supplement data were unavailable, we cannot disentangle direct meteorological effects from seasonal dietary changes, and the observed seasonality should be interpreted as reflecting a combination of environmental and behavioral factors. From a public health perspective, these temperature-by-humidity, temperature-by-season, and climate zone-by-predictor interactions demonstrate that vitamin supplementation strategies should consider local meteorological patterns and population-specific seasonal variations rather than relying solely on calendar-based seasonal categorizations (20). Enhanced nutritional surveillance is particularly warranted during winter months and in regions characterized by low ambient temperatures combined with high humidity, where vitamin deficiency risk is substantially elevated.

Study strengths include large sample size, standardized HPLC vitamin measurements with rigorous quality control, daily meteorological monitoring on individual blood collection dates, comprehensive outcome ascertainment, and sequential regression models. However, several limitations should be considered. First, the cross-sectional assessment of vitamin concentrations in late pregnancy precludes causal inference and raises the possibility of reverse causation, particularly for fetal growth outcomes such as macrosomia. Second, single measurements at 37–40 weeks do not capture temporal variability across pregnancy. In addition, meteorological exposures were assessed using daily values on the date of blood collection, which may not fully capture the biologically relevant exposure window for vitamin D status that reflects cumulative environmental conditions over preceding weeks, potentially leading to exposure misclassification and attenuation of observed associations. Third, and representing a primary limitation, information on dietary intake and supplement use was unavailable, which may result in substantial residual confounding. Dietary patterns and supplementation practices likely vary systematically with socioeconomic and health-seeking behaviors that independently influence perinatal outcomes, and the absence of these data particularly precludes definitive attribution of the observed vitamin E excess or its associations with pregnancy outcomes. Additionally, data on gestational diabetes and hypertensive disorders were not available, further limiting our ability to adjust for important pregnancy complications. Fourth, although vitamin status categories were defined using widely accepted guideline-based cut-offs, these thresholds have not been specifically validated for late-pregnancy women in subtropical Chinese populations, and population-specific optimal ranges may differ. Finally, the hospital-based, single-center urban sample may limit generalizability to rural or other regional populations with different environmental exposures, dietary patterns, and healthcare access. Future studies using prospective cohort designs with serial vitamin measurements across pregnancy, together with biologically meaningful averaging or lag-based assessment of meteorological exposures (e.g., over several weeks prior to blood collection), are needed to better characterize cumulative environmental influences and identify critical intervention windows. Such approaches, potentially complemented by targeted interventional and mechanistic studies, would help clarify causal pathways and inform population-specific nutritional recommendations.

These observations are hypothesis-generating and should not be interpreted as clinical recommendations; nevertheless, several potential clinical implications warrant further investigation: (1) The 30.2% deficiency prevalence and its association with macrosomia suggests that the potential value of vitamin D screening during early pregnancy deserves evaluation in randomized trials (57, 58). Recent randomized controlled trials in diverse populations have demonstrated variable efficacy of vitamin D supplementation during pregnancy, with genetic polymorphisms in vitamin D receptor genes potentially modulating individual response to supplementation, underscoring the need for personalized approaches. (2) The strong association between vitamin A deficiency and macrosomia, as well as potential optimal effects at intermediate concentrations, suggest that vitamin A status assessment may warrant consideration, though optimal intervention strategies require investigation in randomized trials. (3) Monitoring strategies for vitamin E excess, including dietary counseling and reassessment of routine multivitamin supplementation protocols, warrant further investigation in prospective studies, particularly for formulations with high vitamin E content. (4) Our findings suggest that the integration of meteorological data into nutritional risk stratification frameworks warrants further evaluation as a hypothesis-generating approach, particularly in winter months and in regions characterized by low temperatures and high humidity.

In conclusion, abnormalities in maternal vitamins D, A, and E during late pregnancy were associated with a range of adverse delivery and neonatal outcomes. Ambient temperature emerged as a key meteorological factor associated with maternal vitamin status, with significant interactions with other environmental variables. Vitamin A deficiency showed the strongest observed association with macrosomia (approximately threefold higher odds), while excess vitamin E was associated with a 60% higher odds of macrosomia. The complex, non-linear dose–response relationships observed in this study suggest that achieving appropriate maternal vitamin status requires more than preventing deficiency alone and should consider population-specific distributions. These findings underscore the need for population-specific nutritional surveillance and prospective trials to inform evidence-based supplementation strategies in subtropical Asian populations.