Aflatoxin exposure among lactating women in southern Ethiopia

Abstract In Ethiopia and many other low‐income countries, little is known about the exposure of lactating women to aflatoxin, which is a major health concern to the mother and her nursing infant. We determined the aflatoxin B1 contamination of family foods (AFB1) and urinary aflatoxin M1 (AFM1) of lactating women in Sidama, southern Ethiopia, and compared the levels across agroecological settings (lowland, midland, highland) and two seasons. We conducted two surveys (n = 360) that represented the dry and wet seasons of the locality. AFM1 and AFB1 were determined using enzyme‐linked immunosorbent assay (ELISA). Statistical analysis was made using Mann–Whitney U test and Kruskal–Wallis test. The median (interquartile range) AFB1 was 0.94 (0.63–1.58) ppb. AFB1 was detected in 95.6% of the food samples, and 13.6% exceeded the 2.0 ppb threshold. We observed an increasing trend for aflatoxin exposure from highland to lowland (p < .001), but there was no difference between seasons (p = .743). The median (interquartile range) urinary AFM1 was 214 (undetectable to 2,582) ppt, and AFM1 was detectable in 53.3% of the samples. Urinary AFM1 showed significant difference among agroecological zones (p < .001) but not between seasons (p = .275). A significant but weak correlation was observed between AFB1 and urinary AFM1 (rs = 0.177, p = .001). We concluded that lactating women in Sidama, especially those in the lowland area, have unsafe exposure to aflatoxin.


| BACKG ROU N D
Aflatoxins are a group of highly toxic fungal food contaminates with serious health consequences (Payne & Brown, 1998). Aflatoxin contamination of food is a major risk factor for hepatocellular carcinoma, and it explains 5%-28% of the global burden of hepatic cancer (Liu & Wu, 2010). Exposure to unsafe levels of aflatoxin during early life, including in utero exposure and exposure through breast milk and complementary foods, may lead to growth impairment, immune suppression, and micronutrient deficiencies (Wild, 2017;Williams et al., 2004). From an economic standpoint, market losses secondary to contamination of food and feed with aflatoxins and human capital losses due to aflatoxicosis annually costs the global economy billions of dollars (Wu & Khlangwiset, 2010).
Several types of aflatoxins have been identified so far; however, four types (B 1 , B 2 , G 1 , G 2 ) and two secondary metabolites (M 1 , M 2 ) are significant food contaminants (Williams et al., 2004). Especially, aflatoxin B 1 (AFB 1 ) is regarded as the most potent carcinogen. In humans and animals exposed to unsafe levels of AFB 1 , the toxin metabolizes to aflatoxin M 1 (AFM 1 ), and is secreted into milk and excreted via urine and feces (Marchese et al., 2018). Aflatoxin exposure of humans and animals can be monitored using multiple indices including measurement of contamination of foods, plasma concentration of aflatoxin-albumin adducts, and excretion of hydroxylated metabolites in urine and breast milk. Urinary aflatoxin excretion is considered as a simple index of recent exposure (WHO, 2018).
Studies from low-and middle-income countries suggested that exposure of lactating women to unsafe levels of aflatoxin is common.
Study from western Iran reported that AFM 1 was found in all human breast milk samples tested (Maleki et al., 2015). Similarly, in Nigeria, 95% of women had detectable urinary AFM 1 (Alegbe et al., 2017).
Exposure of infants to AFM 1 via breast milk may cause stunting and growth faltering (Magoha et al., 2014;Sadeghi et al., 2019). A study in Tanzania reported a statistically significant negative relationship between infants' anthropometry and AFM 1 exposure through breastfeeding (Magoha et al., 2014).
Studies suggested that the epidemiology of aflatoxin contamination of the food system is liable seasonal and agroecological variations (Elaridi et al., 2017;Kılıç Altun et al., 2017). Aflatoxin contamination tends frequently to occur in warm and humid settings (Wild, 2017), and various seasonal patterns have also been reported (Elaridi et al., 2017;Kılıç Altun et al., 2017). Accordingly, we analyzed the AFB1 contamination of family foods and urinary excretion of AFM 1 among lactating women in Sidama, southern Ethiopia, and compared the levels across different agroecological settings (lowland, midland, and highland) and seasons (wet and dry).

| Study design and setting
We collected data in August 2017 (wet season) and the March 2018 (dry season) in Sidama zone, through two independent crosssectional surveys. The actual surveys were carried out in Hawassa Zuria, Dale, and Hula districts, representing the lowland (<1,750 m above sea level (a.s.l)), midland (1,750-2,300 m a.s.l), and highland (>2,300 m a.s.l) agroecological zones of Sidama. Sidama zone covers nearly 10,000 km 2 area and has diverse climatic conditions with altitude ranging from 1,200 m to 2,800 m a.s.l. Depending on the agroecological feature, the mean annual rainfall ranges from 400 to 1,600 mm and average annual temperature varies between 15 and 25°C. In 2017, the zone had nearly four million inhabitants of whom 95% were rural dwellers (Population Census Commission, 2008).
The economy in the area is dependent on rainfed subsistence farming, and major crops grown are maize, enset (Enset ventricosum), and coffee. Enset is the main staple throughout Sidama while maize is so in the lowlands (UNDP, 2002).

| Study participants
All lactating women nursing infants and young children 6-23 months of age were considered eligible for the study. Women nursing infants 0-5 months were excluded because the larger project from which this study originated was designed to evaluate the aflatoxin contamination of complementary foods prepared for children 6-23 months. In the second-round survey, study participants who took part in the first survey were excluded with the concern that information received in the first round may affect the findings of the latter survey.

| Sample size determination and sampling procedure
Priori sample size was determined using G*Power 3.1 program (Faul et al., 2007), assuming that aflatoxin concentration in food and urine will be compared between seasons and among agroecological settings using independent t test and one-way analysis of variance, respectively. Ultimately, assuming 95% confidence level and 80% power, a sample size of 180 subjects per survey round, was deemed adequate to detect a medium effect size of 0.3 (Sullivan & Feinn, 2012).
In both of the seasonal surveys, the study participants were selected using a multistage cluster sampling technique. Initially from the 19 administrative districts of Sidama, Hawassa Zuria, Dale, and Hula districts were selected to represent the aforementioned three agroecological settings. Then from each district, two villages with the required agroecological feature were chosen, and from each village, 30 lactating women were recruited using lottery method.
When a selected woman failed to take part in the study for any reason, we have replaced her with an eligible woman from an adjacent household.

| Measurement and data collection procedures
Socio-demographic information and knowledge and practice of the participants toward mold contamination of foods were assessed using a questionnaire prepared in the local language. The tool was pretested and administered by trained enumerators.

Maternal anthropometric measurements (height and weight)
were taken following standard procedures using calibrated instruments. During the interview visits, 40 ml of morning midstream urine and 100 g of cooked cereals-based foods (mainly maize-based foods and few made up of wheat and teff) and Enset products prepared for household consumption were collected in clean plastic containers and were properly labeled. Food and urine samples then were transported to the Nutrition and Food Science Laboratory of Hawassa University in an icebox and kept frozen at −20°C until analyzed.

| Laboratory analysis
AFB 1 and urinary AFM 1 concentrations were determined using enzyme-linked immunosorbent assay (ELISA) kits (Helica Biosystems Inc., Santa Ana, California, USA). Urine samples were centrifuged at 3,000 rpm for 10 min to remove precipitate, and standards of 0, 150, 400, 800, 1,500, and 4,000 pg/ml were prepared. Samples were diluted 1:20 ratio using distilled water, and the assay protocol provided by the manufacturer was followed to determine urinary AFM 1 (Helica Biosystems, 2020a). Similarly, for AFB 1 analysis, 20 g of the food samples was first ground to fine particle size (50% passed through a 20-mesh screen) and 100 ml of solvent (70% methanol and 30% distilled water) added. After shaking the solution for 5 min, 10 ml of sample was filtered by Whatman qualitative filter paper grade 2 (125 mm diameter) for testing. Ultimately AFB 1 level was determined following standard procedures recommended by Helica Biosystems Inc (Helica Biosystems, 2020b). The limits of detection (LoD) for AFB 1 and AFM 1 tests determined by the manufacturer were 0.2 ppb and 0.15 ppt respectively. Recovery data were 96.4 for AFB 1 and 96.5% for AFM 1 , respectively. For both tests, the coefficient of variation was below 10%.

| Data management and analyses
We used SPSS 24 for data analysis. Normality of AFB 1 and AFM 1 was visually assessed using histograms and tested via Kolmogorov-Smirnov statistic, and both indices demonstrated right-skewed distributions. Transformations were attempted but failed to normalize the distributions. Accordingly, AFB 1 and AFM 1 were described using median and interquartile range (IQR) and compared among classes of agroecology and seasons using nonparametric tests. Mean AFB 1 and AFM 1 ranks were compared between seasons and across agroecological zones using Mann-Whitney U test and Kruskal-Wallis test, respectively. For Kruskal-Wallis test with global statistical significance, pairwise post hoc tests were performed. For AFB 1 , the proportion of food samples that exceeded the European Union (EU) thresholds of 2.0 parts per billion (ppb) was calculated (European Union, 2006). Correlation between AFB 1 from food samples and urinary AFM 1 was measured using Spearman rank-order coefficient (r s ).

| Ethical considerations
The protocol was reviewed and approved by the Institutional

| Characteristics of the respondents
The basic characteristics of 360 lactating women interviewed for the study are presented in Table 1. The mean (±SD) age of the women was 26.3 (±4.4) years, and about one-third were between 18 and 24 years of age. Nearly all (99.4%) were married, a quarter had no formal education, and 90% identified themselves as farmers or housewives. The median (IQR) monthly household income was 500 (400-800) Ethiopian birr (ETB), and four-fifths earned less than 2,000 ETB per month (at the time of the survey, 1 USD was equivalent to 30 ETB). The median agricultural landholding size was 0.7  Table 2 presents the knowledge and practice of the study participants toward mold contamination of food. Only 5.3% of the respondents were aware of aflatoxin. Most of the households dried their harvests prior to storage, and the popular approach was solar drying by spreading it on bare ground without any protection from soil contamination. In 51% of the households, grains were stored in traditional "Gotera" made up of wood, mud, and straw. Modern or improved stores were not used for a variety of reasons including lack of awareness and unaffordability. Nearly one-third did not do anything to treat or disinfect their stores before stocking grains, and 18.9% reported that their stores were not dry and well ventilated.

| Knowledge and practice toward aflatoxin contamination of food
Only 39.7% of the women reported that they usually sort out moldy grains before storing grains for human consumption. Of them, 51% used the discarded moldy grains for feeding their domestic animals including chickens and milk cows. More than a quarter (28.9%) assumed that providing animals with moldy feeds does not have negative implications to human health. Smaller proportions (3.9%) used moldy cereals for brewing local beverages (Table 2).

| Aflatoxin contamination of food
Of 360 food samples tested (mainly maize-and enset-based foods and few made up of wheat and teff), the median (interquartile range) AFB 1 was 0.94 (0.63-1.58) ppb. AFB 1 was detected in 95.6% of the samples, and in 13.6% of the cases, the level exceeded the upper permissible level of 2.0 ppb set by the EU (European Union, 2006). Table 3 compares the AFB 1 levels between two seasons and across three agroecological settings. We observed no significant difference in AFB 1 mean ranks between wet and dry seasons (p = .743).
However, there was statistically significant difference among the three agroecological zones (p < .001) ( Table 3). When performing post hoc comparisons, we detected significant differences between lowland-midland (p = .014); midland-highland (p < .001); and highland-lowland (p < .001) pairs. Similarly, the proportion of samples that exceeded the upper EU permissible level were 30% and 10% in the lowland and midlands, respectively, while it was only 0.8% in the highland.
However, there was no significant difference between the two seasons (Table 4). Urinary AFM 1 showed statistically significant but weak correlation with dietary AFB 1 (r s = 0.177, p = .001).

| D ISCUSS I ON
The study confirmed that lactating women in Sidama zone, south- detected AFB 1 in 8.8% of the barley, sorghum, teff, and wheat samples collected from different parts of Ethiopia (Ayalew et al., 2006).
Urinary AFM 1 was detected in 53.3% of the lactating women suggesting high levels of recent aflatoxin exposure. Alegbe and his colleagues in Yobe State, Nigeria, reported that 93% of the lactating women excreted AFM 1 in their urine and 82% had AFM 1 in their breast milk (Alegbe et al., 2017). Similarly, urinary AFM 1 was detected in 84% of pregnant women from China (Lei et al., 2013).
Most of the existing studies evaluated the aflatoxin exposure of lactating women using AFM 1 concentration in breast milk. Studies conducted in Nigeria (Anthony et al., 2016) and Tanzania (Fakhri et al., 2019) reported that 90% and 38% of the tested breast milk samples exceeded the EU limit of 0.025 ppb. In Sudan, half of the breast milk samples were highly contaminated with AFM 1 (Elzupir et al., 2012).
We found that the knowledge and the practice of the study participants and their households toward aflatoxin was suboptimal.
Very small proportions were aware of aflatoxin and many harvest and store grains in ways that support proliferation of aflatoxin.
Previous studies in Ethiopia also concluded the same. A study by Beyene and his colleagues in Amhara, Tigray, Oromia, and southern regions of Ethiopia reported that mothers caring for infants and young children had poor knowledge and practice regarding storage and processing of food and the consequences of aflatoxin exposure to humans (Beyene et al., 2016). A study in Wolaita zone of Ethiopia suggested only a very small proportion of farms were aware of aflatoxin and its consequences (Kibret et al., 2019).
Overall, there was an increasing trend for aflatoxin exposure from highland to lowland. This is compatible with the understanding that aflatoxin proliferation is favored by hot climate and the same had been documented by studies conducted in other sub-Saharan Africa countries. A study in Zambia reported that maize and groundnut samples tend to be more frequently contaminated in the lowlands as compared to the other climatic zones (Kachapulula et al., 2017).
In Kenya, contamination of multiple staples including maize was lower in temperate areas than in humid and semiarid zones (Sirma et al., 2016). The agroecological variations that we observed in Sidama can also be partly explained by the fact that maize is an important staple in the lowlands but not in the highland areas (UNDP, 2002). Studies suggest that maize is more frequently contaminated by aflatoxin than other staples (Mahato et al., 2019).
Urinary AFM 1 is a valid biomarker of recent AFB 1 exposure, and a strong correlation had been documented between dietary aflatoxin exposure and urinary excretion (Turner, 2013;Zhu et al., 1987). Several factors may contribute to the weak correlation (r s = 0.177) that we observed between AFB 1 and urinary AFM 1 . First, we only measured the AFB 1 contamination of cereal-based foods, not the total dietary aflatoxin exposures. Further, we did not adjust urinary AFM 1 for creatinine; as a result, AFM 1 variations can partly be explained by interindividual differences in urinary dilution. Parallel to our finding, a study among Tanzania children reported a moderate strength of association between AFB 1 intake and urinary AFM 1 (r = 0.442) (Chen et al., 2018). A study in Brazil among adults observed a modest but significant correlation (r = 0.45) between urinary AFM 1 estimated dietary intake of total aflatoxins (Jager et al., 2016). Studies from Malaysia (Sulaiman et al., 2018) and Brazil (Romero et al., 2010)

| CON CLUS ION
The study confirmed that lactating women in Sidama zone, southern Ethiopia, have unsafe levels of aflatoxin exposure as measured by both AFB 1 and urinary AFM 1 , suggesting possible health concerns both to the women and to their nursing infants. Aflatoxin exposure was higher in the lowland than in the highland, but there were no seasonal differences. The study also showed that the knowledge and practice toward aflatoxin was suboptimal in the area.

ACK N OWLED G M ENTS
This work was supported by Hawassa University-Norwegian Agency for Development Cooperation (NORAD) program. The authors acknowledge Mr. Wondu Wolde-Mariam, CEO of Helica Biosystems, Inc., for freely supplying the ELISA kits and reagents used in this study. The authors are also grateful to Dr Andargachew Gedebo, head of the NORAD program at Hawassa University, for facilitating administrative issues for field data collection, and Eyamo Egata for translating of the data collection tool.

CO N FLI C T O F I NTE R E S T
The authors declare no conflicts of interest. The sponsor had no role in the design, execution, interpretation, or writing of the study.