Heavy metals in Rice

HMI pools compatible evidence across sampling families to calculate fixed rice references. The dry-grain references combine FDA category medians with Carey regional medians for inorganic arsenic and DMA, FDA with Myanmar township medians for total arsenic, and Myanmar with Shi country medians for cadmium (FDA 2013 Myat et al. 2023 Carey et al. 2020 Shi et al. 2020). Native-moisture references combine Saudi and Myanmar grain evidence using the Myanmar study’s measured 6–7% moisture loss (Myat et al. 2023 Alrashdi et al. 2024). Each family receives equal weight; its subgroups share that weight. These references describe subgroup medians in the evidence library, not the median of all rice sold worldwide. Each row’s reported mean or median is compared with that reference; means are never relabelled as medians. Cooked rice remains separate. Low is below Q1, Moderate is Q1–Q3, and High is above Q3; reporting bounds must support one position. The sidebar shows contributing families, conversions, count-weighted and leave-one-family-out sensitivity, and the limitations of single-family references. These are concentration positions, not safety or certainty grades.

Arsenic species and occurrence evidence

Rice occurrence studies cover different chemical species, grain forms and sampling frames (Carey et al. 2020 Shi et al. 2020 FDA 2022). The table brings international surveys, field-grown grain, commercial rice products and FDA cooked-rice composites together by analyte, while retaining each source’s reported statistic and measurement basis. A mean, median and detected range answer different questions; the table labels each explicitly. EFSA’s 2024 organoarsenic assessment adds a separate evidence stream. Its occurrence analysis includes DMA measurements in polished and brown grain, while some processed-food estimates are derived using recipes and an assumption of no processing loss (EFSA 2024). Direct grain observations and model-derived flour, porridge or beverage values must remain distinguishable (EFSA 2024). Organic arsenic is not one interchangeable substance with one toxicological interpretation (EFSA 2024).

Sampling, censoring and shared observations

Sampling and censoring matter as much as the reported concentration. The global cadmium compilation reuses earlier studies and substitutes half the detection limit for low results (Shi et al. 2020). The Henan monitoring summaries also include substituted non-detects (Yan et al. 2025). Their reported medians remain usable source summaries, but they are not medians of detected samples and cannot supply a matched trend against the FDA composites (Shi et al. 2020 Yan et al. 2025). Source rows preserve the counted unit. The Vietnamese field study covers genotypes with paired brown and white grain; the Ethiopian study combines farmers’ grain into area composites (Tran et al. 2024 Desalew et al. 2023). Multiple analyte rows reuse those same materials (Tran et al. 2024 Desalew et al. 2023). Neither more rows nor technical repeats create additional independent studies (Tran et al. 2024 Desalew et al. 2023).

Mercury species and comparison boundaries

Direct mercury measurements remain separate from model outputs. Brombach’s European survey reports both total mercury and methylmercury in a mixture of grain and rice products; its overall range cannot be assigned to a grain-only or baby-food subgroup (C-C et al. 2016). Cui’s spatial model addresses a different question and does not contribute a new measured product distribution to this table (Cui et al. 2022). Original source means, medians, ranges and measurement bases remain visible. Recovered individual measurements and regional summaries supply compatible HMI references; the baseline does not require every study to report a percentile. Unknown moisture bases and unresolved reporting limits remain explicit (Desalew et al. 2023). Neither below-LOQ results nor an absence of detections establishes zero contamination (FDA 2022 Desalew et al. 2023).