Dietary determinants of aflatoxin B₁-lysine adduct in pregnant women consuming a rice-dominated diet in Nepal

Aflatoxin, a toxin produced by Aspergillus fungi, poses health risks in South Asia, including Nepal. Acute exposure can cause severe illness, and chronic low-level exposure is linked to adverse health outcomes, including placental transfer from mother to fetus and impaired child growth. Exposure primarily occurs via consumption of contaminated foods; maize, chilies, spices, oilseeds, and nuts are especially susceptible. Milk can contain aflatoxin M₁ when livestock consume contaminated feed. Populations with limited dietary variety and reliance on susceptible staples (e.g., maize, groundnuts) are at particular risk, while improved dietary diversity might reduce reliance on aflatoxin-prone foods. This study aimed to determine among pregnant Nepali women: (a) whether frequency of consumption of susceptible commodities is associated with aflatoxin exposure, (b) whether greater dietary diversity is associated with lower aflatoxin exposure, and (c) whether aflatoxin exposure varies seasonally.

Prior research documents that aflatoxins contaminate a wide range of foods, particularly maize, groundnuts, chilies, spices, and oilseeds, and that AFM₁ can be present in milk when animals consume contaminated feed. Aflatoxins are hard to detect by consumers and resist thermal inactivation, making household mitigation challenging. Chronic exposure is linked to significant health risks, including fetal exposure and child growth impairment. Populations with monotonous diets, long-term food storage, and reliance on susceptible staples face higher exposure risk. Improved dietary diversity has been proposed as a potential protective factor by reducing dependence on susceptible foods and potentially counteracting toxicity, though evidence remains mixed.

Design and setting: The AflaCohort Birth Cohort Study (2015–2019) in Banke district, Nepal (tropical southern plains). Rolling recruitment enrolled 1675 healthy pregnant women; analysis used pregnancy data from July 2015–August 2016 (data collection paused for 3 months due to nationwide strikes and resumed December 2015). Sample size assumptions: alpha 0.05, power 80%, attrition 20%, design effect 1.5, enabling detection of a −0.207 SD difference in postnatal HAZ per 1-unit increase in log average maternal AFB₁-lys adducts. Eligibility: <30 weeks gestation, age 16–49, singleton pregnancy, resident and planning delivery in study area. Ethical approvals: Nepal Health Research Council (295/2014) and Tufts IRB (11535); informed consent obtained. Data collection: Trained interviewers administered electronic surveys, including a qualitative 7-day and 24-hour food frequency questionnaire (FFQ) with 49 items based on prior assessments in this population. Annual (past year) consumption was also collected. Anthropometry (height, weight, MUAC) measured to nearest 0.1 cm/kg using standard equipment. Biospecimens: Within one week of survey, 3–5 mL venous blood was collected, transported on wet ice to local lab for processing, stored at −80 °C at Patan Academy of Health Sciences, then shipped on dry ice to the University of Georgia for analysis. Biomarker assay: 1650 gestational serum samples analyzed for AFB₁-lysine (AFB₁-lys) adducts, a biomarker of dietary aflatoxin exposure over the prior 2–3 months. Samples were heat-deactivated at 56 °C for 30 min; ~150 µL digested with pronase (pronase:total protein 1:4 w/w) at 37 °C for 3 h. Digests were extracted/purified via Waters MAX SPE cartridge, eluted with 2% formic acid in methanol, vacuum-dried, and reconstituted with 25% methanol in water. Quantification used Agilent 1200 HPLC with fluorescence detection; mobile phases: 20 mM NH4H2PO4 pH 7.2 (A) and methanol (B), gradient separation within 25 min, typical retention ~13 min; Zorbax Eclipse XDB-C18 (5 µm, 4.6×250 mm) with guard column at 25 °C, 1 mL/min; injection 100 µL; excitation/emission 405/470 nm. Calibration with authentic standards generated weekly; QA/QC included one authentic standard per 10 samples and two daily QC samples; average recovery ~90%. AFB₁-lys concentrations adjusted by albumin (UV/Vis spectrophotometry). Values below LOD (0.4 pg/mg albumin) imputed as 0.5×LOD. Dietary diversity: Minimum Dietary Diversity for Women (MDD-W) computed from 24-h FFQ across 10 food groups; dichotomous indicator for MDD ≥5 groups. Covariates: age, education, wealth status, MUAC, hemoglobin, season, and VDC; seasons categorized as autumn, prewinter, winter, spring, summer, monsoon; rainfall during survey comparable to prior years. Wealth index constructed via principal components using housing characteristics, utilities, household size, and assets. Statistical analysis: AFB₁-lys data were non-normal and log-transformed. Data divided into exposure quintiles. Associations assessed using two-sided t-tests/ANOVA and chi-squared tests, OLS regression and quantile regression (QR at 10th, 30th, 50th, 70th, 90th quantiles) to evaluate relationships between frequency of consumption of aflatoxin-prone foods, MDD, season, and AFB₁-lys. Restricted cubic splines and nonparametric smoothing assessed nonlinearity; variance inflation factors checked multicollinearity. Significance p<0.05. Software: Stata 14.2.

• Biomarker prevalence and levels: 94% of pregnant women had detectable AFB₁-lys adducts (range 0.4–147 pg/mg albumin); mean 3.2 ± 8.3 pg/mg; geometric mean 1.37 pg/mg (95% CI: 1.3–1.4).
• Diet patterns (24 h): 100% consumed rice; 62% pulses; 12% nuts/seeds; dairy 43% vs meat 25%; fruits 33%, eggs 12%, dark green leafy vegetables 28%, vitamin A-rich sources 20%. Only 39% met MDD (≥5 of 10 groups).
• Bivariate associations: Older age and higher MUAC were negatively associated with AFB₁-lys; hemoglobin and winter season were positively associated.
• Weekly consumption: 31% consumed groundnuts, 3% maize, 2% both; mean frequencies: maize 2.4 ± 2.2, groundnuts 2.8 ± 2.2 (occasions/week). AFB₁-lys higher among maize and groundnut consumers vs non-consumers (7.9 vs 2.4 pg/mg; p<0.0001).
• Socioeconomic patterns: Weekly maize consumption did not vary by wealth quintile and was positively associated with maternal education (p<0.01). Weekly groundnut consumption was positively associated with wealth (p<0.01) but not education; both maize (p<0.05) and groundnut (p<0.001) consumption frequencies were higher among women recruited in winter.
• Annual consumption: 83% maize, 97% groundnut, 100% chilies, 93% milk. Annual milk consumption was positively associated with AFB₁-lys (p<0.05); annual maize or groundnut consumption showed no association; limited variability prevented testing chilies.
• Regression results: In adjusted OLS, weekly groundnut consumption (β=0.730, p<0.001) and winter season (β=2.339, p<0.001) predicted higher AFB₁-lys. In QR, each additional weekly maize occasion associated with higher AFB₁-lys at the 30th (0.094 pg/mg, p<0.05), 50th (0.112 pg/mg, p<0.05), and 70th (0.109 pg/mg, p<0.05) quantiles, but not at 10th or 90th. Each additional weekly groundnut occasion associated with higher AFB₁-lys at the 30th (0.058 pg/mg, p<0.001), 50th (0.085 pg/mg, p<0.01), 70th (0.133 pg/mg, p<0.001), and 90th (0.133 pg/mg, p<0.001) quantiles, but not at 10th. Restricted cubic splines suggested no threshold effect (consistent with linear relationship).
• Milk and DD: Milk consumption in the past year was associated with higher AFB₁-lys at the 10th (0.63, p<0.01) and 50th (0.23, p<0.05) quantiles. Dietary diversity was not associated with exposure in OLS; in QR, DD was positively associated only at the 10th quantile (0.064, p<0.05).
• Seasonality: Winter recruitment was positively associated with AFB₁-lys at all quantiles (range 0.313–1.101 pg/mg, p<0.001).

The study demonstrates widespread aflatoxin exposure among pregnant women in a rice-dominant diet context, with exposure levels strongly linked to the frequency of consuming known high-risk foods—groundnuts and maize—and to seasonality, particularly higher levels in winter. These findings directly address the research questions by identifying specific dietary drivers of exposure and showing that general dietary diversity did not confer protection against aflatoxin exposure. The associations with milk intake suggest indirect exposure via AFM₁ in dairy, likely from contaminated feed, highlighting the need to consider animal-source food pathways. Despite rice being a staple, it was not associated with aflatoxin biomarker levels in this analysis, although it may harbor low levels and should not be dismissed. Compared with some prior studies, overall exposure levels were lower, but still prevalent, underscoring a significant public health concern. The seasonal pattern suggests post-harvest handling/storage and environmental conditions may elevate contamination risk in winter. Policy and program implications include prioritizing interventions targeting groundnut and maize safety, improving feed quality to reduce AFM₁ in milk, and implementing measures along the value chain and at household level; dietary diversity alone is insufficient as a mitigation strategy in this context.

This study identifies groundnut and maize consumption and winter seasonality as key determinants of maternal aflatoxin exposure in Nepal, with additional indications of exposure via milk. Dietary diversity was not protective against exposure. These findings support integrated aflatoxin mitigation strategies, including regulatory standards, improved agricultural and post-harvest practices, feed quality control to reduce AFM₁ in dairy, and food safety interventions throughout the value chain and at household level. Future research should assess contamination and exposure contributions from commonly used spices and other ubiquitous foods (e.g., black pepper, nutmeg, cumin, coriander, garlic) and dairy products, and further investigate rice’s role, as well as evaluate intervention effectiveness across seasons.

• AFM₁ in milk was not directly measured; milk associations are inferential.
• Low variability in annual chili consumption limited ability to test its association with exposure.
• Data collection interruptions due to nationwide strikes may have influenced seasonal distribution of measurements.
• Single-district setting may limit generalizability to other regions.
• Dietary data based on 7-day and 24-hour FFQs and annual recall are subject to recall/reporting bias.
• Values below LOD were imputed as half-LOD, which may affect estimates at the low end of exposure.