Individual agricultural markets in each district were selected for inclusion on the basis of information obtained from interviews with the district agriculture officer of each district. We created a sample of major agricultural markets that represented potential exposure to aflatoxin among all market maize consumers within the study area. Markets were selected for inclusion based on the following criteria: a) geographic location of the population served by the market, b) having an increased number of maize vendors, c) having a variety of maize vendor types, and d) holding an important position in the maize distribution system for the district.

Large government grain warehouses operated by Kenya’s National Cereals and Produce Board (NCPB) also were included in this study. The NCPB is involved in grain marketing and acts as a strategic grain reserve for food supply functions of the country, including famine relief.

At each market surveyed, vendors were selected to create a sample that included all types of maize vendors represented within each marketplace. Vendor types were store merchant, wholesale maize distributor, small-scale miller, or open market vendor (i.e., migrant vendor who brings products to sell in an open air market). The variability of maize sold determined the number of vendors interviewed at each market. More interviews were conducted at markets with maize from a variety of sources. Maize variability was assessed based on a) the size of each market, b) variety of vendor types present, and c) relationship of the market to major distribution routes within the district. Vendors were systematically selected based on location within the market in order to obtain geographic distribution within the marketplace. Selected maize vendors were administered a face-to-face interview and requested to provide samples of each of their maize products for aflatoxin analysis.

Face-to-face interviews were conducted with maize vendors at the marketplace in Kiswahili, Kikamba, or English. All vendors were administered a standard survey questionnaire. Information was collected on market location, vendor type, vendor trade practices, maize history (as could be recalled by the vendor), and vendor’s assessment of the quality (at the time of purchase) of maize products sampled.

Maize products were sampled from every vendor interviewed. A 1-kg sample was taken from every maize product offered by the vendor. Maize products were dried maize kernels, maize flour (commercial or locally milled), and muthokoi (kernels with the outer hull removed). If the vendor offered the same product from different sources (i.e., maize kernels purchased from local farmers and maize kernels from a distributor), then a 1-kg sample was taken from each.

Most of the samples were collected from 90-kg bags of maize. Multiple samples were taken from different parts of one bag or several bags belonging to one vendor and combined to produce a 1-kg sample for analysis. The maize samples were collected using the respective vendor’s sampling tools (i.e., spikes and scoops). Samples were transported and stored in paper bags. Prepackaged 1- or 2-kg bags of commercial maize flour were also collected for analysis.

The samples were analyzed for total aflatoxin using a slightly modified immunoaffinity method based on Association of Official Analytic Chemists (AOAC) method 991.3 (Trucksess et al. 1991). Briefly, the whole sample was ground to pass a No. 20 sieve, and a 50-g subsample was removed for analysis. Methanol:water (80:20) solvent (100 mL) and 5 g NaCl were added to the 50-g subsample, and the mixture was blended at high speed for 1 min. The mixture was then filtered through a fluted filter paper (Whatman 2V, Whatman plc, Middlesex, UK), and the filtrate was diluted (1:4) with water and refiltered through a glass-fiber filter paper. Two milliliters of the glass-fiber filtrate was placed on an Aflatest P immunoaffinity column (VICAM, Watertown, MA, USA) and allowed to elute at 1–2 drops/sec. The column was washed two times with 5 mL water, and aflatoxin was eluted from the column with 1 mL high performance liquid chromatography (HPLC)-grade methanol. A bromine developer (1 mL) was added to the methanol extract, and the total aflatoxin concentration was read in a precalibrated VICAMSeries-4 fluorometer set at 360 nm excitation and 450 nm emission. Samples containing > 250 ppb were repeated using a 1:49 water:sample dilution ratio. The modified fluorometry method had ≥85% recovery and a 1 ppb limit of detection.

Data were analyzed using SAS computer software (SAS Institute Inc., 2001). We used mixed linear models to investigate the association between the natural log of aflatoxin concentrations in maize samples and questionnaire variables. Nested random effects (i.e., divisions within districts, markets within divisions, and vendors within markets) were added to account for potential correlation among samples. We calculated least-squares means for the fixed effects specified in the models.

We surveyed 65 markets within the four study districts. Within those markets, we interviewed 243 vendors and collected 350 maize products (Table 1). All but two vendors we approached agreed to be interviewed and provide samples.

Most (65%) vendors were store merchants, followed by open market vendors (19%), wholesale distributors (10%), and small-scale millers (3%). The most common maize products sold in the market place were maize kernels (69%), followed by muthokoi (18%) and maize flour (12%). Most vendors (89%) reported that their maize products were dry at the time of purchase.

During the study period (June 2004), the maize trade was primarily local. The majority (88%) of maize was locally grown, sold to vendors by local farmers (70%), and bought by local residents (88%). Of the 45 samples representing maize products from outside the local area, 30 (67%) were from Loitokitok and or Busia, and 15 (33%) were from other Kenyan districts.

Aflatoxin levels in market maize indicate widespread aflatoxin contamination. Of the 350 market maize samples collected, 192 (55%) had levels greater than the regulatory limit of 20 ppb. One hundred twenty-one (35%) of the maize samples had aflatoxin levels > 100 ppb (five times the regulatory limit), and 24 (7%) had levels > 1,000 ppb. Aflatoxin levels ranged from 1 ppb (the lower limit of detection) to values as high as 46,400 ppb. Each of the four study districts had a substantial proportion of market maize with aflatoxin levels > 20 ppb (Table 2). Makueni and Kitui districts had the highest proportions of samples, with aflatoxin levels > 20 ppb (65% and 62%, respectively), followed by 51% of maize from Machakos markets and 34% from Thika (Table 2).

Fourteen samples were collected from NCPB warehouses in Makueni, Kitui, and Machakos districts. Of the 14, 8 (57%) had levels > 20 ppb, and 6 (43%) had levels ≥100 ppb. Among NCPB warehouses, samples from the Makueni facility contained the highest levels of aflatoxin, including two that were > 1,000 ppb.

Significant differences were found in the geometric mean (GM) of market maize aflatoxin levels between districts. These differences were consistent with the geographic distribution of aflatoxicosis cases. Makueni and Kitui, the districts with the highest number of aflatoxicosis cases, also had the highest market maize aflatoxin levels. Maize from markets in Makueni had a GM aflatoxin level greater than 2.5 times the upper acceptable regulatory limit [GM = 52.91 ppb; 95% confidence interval (CI), 27.19–103.21 ppb]. Kitui had the second highest GM aflatoxin level, followed by Machakos and Thika (Table 3, Figure 1). When aflatoxin contamination data from Makueni were compared with data from Machakos and Thika, the differences in GM aflatoxin levels were statistically significant (p = 0.0249 and p = 0.0004, respectively).

At the division level, those divisions with one or more aflatoxicosis case-patients had significantly higher aflatoxin levels in market maize than did market maize from divisions with no aflatoxicosis case-patients (GM = 27.70 ppb vs. 6.14 ppb, p = 0.0022).

The aflatoxin GM in locally grown market maize from within the affected area was higher than levels in market maize grown outside the local area. The difference was not, however, statistically significant (GM = 19.84 ppb vs. 9.64 ppb, p = 0.1748). The low aflatoxin level in maize from outside the local area primarily reflects maize from Loitokitok and Busia, major import routes from neighboring African countries. The GM aflatoxin for maize from Loitokitok and Busia was 9.14 ppb (95% CI, 3.32–25.13 ppb).

No significant differences were observed among GM aflatoxin levels of market maize based on whether or not maize was from a store merchant, open market vendor, wholesale distributor, small-scale miller, or other type of maize vendor. Aflatoxin levels did not vary significantly among the types of market maize products (i.e., maize kernels, flour, or muthokoi). No significant differences were seen in aflatoxin levels based on vendor selling practices or whether the maize was wet at the time of purchase.

The dichotomous aflatoxin concentration variable was analyzed to compare the odds of exposure to aflatoxin at levels > 20 ppb by market location, maize history, and vendor type. Significant differences were seen among all four study districts. The odds of exposure to aflatoxin levels > 20 ppb were more than four times higher in samples from Makueni than in samples from Thika [odds ratio (OR) = 4.29; 95% CI, 1.71–10.80]. At the division level, maize samples from markets located in divisions with aflatoxicosis case-patients were three times more likely to have maize aflatoxin levels > 20 ppb compared with samples from markets in divisions not affected by aflatoxicosis (OR = 3.13; 95% CI, 1.69–5.88). Locally grown maize from the affected area was significantly more likely to have aflatoxin levels > 20 ppb compared with maize from other regions of Kenya or imported from other countries (OR = 2.71; 95% CI, 1.12–6.59). Aflatoxin levels by type of vendor did not differ significantly.