Representative aflatoxin B 1 quantification: optimization of sampling intensity in commercial maize lots

Representative sampling is essential for reliable estimation of aflatoxin B1 (AFB1) in maize because contamination can vary considerably within a lot. This study evaluated five sampling intensities (√n, 1.5√n, 2√n, 2.5√n, and 3√n) using 839 commercial maize lots collected from eight states in India during 2024–2025. Each sampling strategy was compared with an operational whole-lot reference composite. Performance was assessed using correct classification rate (CCR), root mean square error (RMSE), mean absolute error (MAE), mean bias, Bland–Altman limits of agreement (LoA), and intraclass correlation coefficient (ICC). The 3√n strategy produced the lowest RMSE (194.9 μg/kg) and the narrowest LoA, whereas 2√n showed the smallest absolute mean bias (2.1 μg/kg) and the highest ICC (0.50; 95% CI: 0.44–0.55). The highest overall CCR was observed for √n (79.9%). However, the differences in absolute errors among the five sampling intensities were not statistically significant (Friedman χ24 = 0.36, p = 0.986), indicating that no sampling intensity consistently outperformed the others across all performance measures. These findings demonstrate that increasing sampling intensity does not necessarily improve AFB1 estimation or classification performance. The results also highlight the importance of evaluating sampling strategies using multiple complementary performance measures rather than relying on a single criterion. The appropriate sampling intensity should therefore be selected according to the intended purpose of testing, while balancing measurement performance and sampling effort. Further validation across different maize lots, contamination levels, geographical regions, seasons, and storage conditions is required before any sampling intensity is considered for wider standardisation or regulatory application.