Rice finding 2207
Source evidence and comparison details for the Rice profile.
Long-grain rice (milling state not stated) · Inorganic arsenic (iAs) · 502 analytical results/samples of rice grain with iAs data (Table 4 sum 1,756 + 7 red rice; text states n = 1,768); 13,608 iAs results overall in food and water.
Selected finding and reference
502 analytical results/samples of rice grain with iAs data (Table 4 sum 1,756 + 7 red rice; text states n = 1,768); 13,608 iAs results overall in food and water. Basis: as reported to EFSA (whole/as-sold food; basis not stated, not dry-weight corrected).
LB: substitution method; LB sets <LOD/<LOQ to zero, UB sets <LOD to LOD and LOD-LOQ to LOQ (Section 2.1.2). This finding is the LB estimate.
HMI baseline position
Low · HMI uncooked-grain iAs reference · 10 sampling families. Mean as reported: 77 µg/kg. HMI median 110; Q1 82.5–84 (bounded); Q3 150 µg/kg. 10 families. Not a safety grade.
Low is below Q1; Moderate includes Q1 through Q3; High is above Q3. Reporting bounds must give the same category throughout.
This finding’s sampling family contributes to the fixed reference.
Sensitivity to weighting
Equal subgroup weight within each family: Low. The position is unchanged under this alternative. This check does not address all sampling, rounding or analytical uncertainty.
Sensitivity to source selection
- Without Alrashdi et al. (2024): Low.
- Without WHO GEMS/Food: Low.
- Without EFSA et al. 2021: Low.
- Without Limmer et al. 2026: Low.
- Without Liu et al. 2022: Low.
- Without Moreda-Pineiro et al. 2020: Low.
- Without Sim et al. 2024: Low.
- Without Tian et al. 2024: Low.
- Without Yan et al. (2025): Low.
- Without Zavala et al. 2008: Low.
These checks show dependence on each sampling family; they are not confidence intervals.
Collection period
2013–2018
No year-by-year sample register is reconstructed from this summary.
Geographic scope
Rice, long-grain
Rice variety or type within one study.
Sampling, analysis and reporting limits
Various laboratory methods reported by data providers; ICP-MS dominant (n=3,521 of food iAs results), also HG-AAS, ICP-HGAES, HPLC-HGAFS, HPLC-ICP-MS.
LB: substitution method; LB sets <LOD/<LOQ to zero, UB sets <LOD to LOD and LOD-LOQ to LOQ (Section 2.1.2). This finding is the LB estimate.
Shared samples and study overlap
Subgroup rows, statistics of the same subgroup and the study total share samples; they are not independent results.
Each table row describes a food/analyte result. Repeated rows do not add independent samples or datasets. No cross-study trend is calculated in this comparison mode.
Reported values, conversion and source locator
| Source statistic | Original result | Displayed unit conversion |
|---|---|---|
| mean | Rice, long-grain N=502; Mean LB = 77 µg/kg (Table 4 prints LB and UB columns: 77 / 87) | none |
Table 4 'Summary statistics (µg/kg) of iAs in different types of rice', p. 15.
No moisture, cooking-yield or species conversion is applied to the original finding above. Any basis conversion used for its HMI comparison is stated separately with the numerical target.
Open the primary publication · HMI source record
Reference and scope
HMI uncooked-grain iAs reference · 10 sampling families. Uncooked grain as analysed (native or unstated moisture), µg/kg. Distribution of subgroup median concentration values.
| Reference (µg/kg) | Q1 | Median | Q3 |
|---|---|---|---|
| Primary HMI reference | 82.5–84 (bounded) | 110 | 150 |
| Alternative weighting | 82.5–84 (bounded) | 110 | 150 |
87 reference units; 10 sampling families. Counts describe evidence units, not independent studies or certainty.
| Contributing family | Reference units | Weight |
|---|---|---|
| Alrashdi et al. (2024) | 6 | 10% |
| WHO GEMS/Food | 54 | 10% |
| EFSA et al. 2021 | 10 | 10% |
| Limmer et al. 2026 | 1 | 10% |
| Liu et al. 2022 | 2 | 10% |
| Moreda-Pineiro et al. 2020 | 2 | 10% |
| Sim et al. 2024 | 1 | 10% |
| Tian et al. 2024 | 1 | 10% |
| Yan et al. (2025) | 7 | 10% |
| Zavala et al. 2008 | 3 | 10% |
Methods
HMI calculates a pooled distribution of subgroup medians across the admitted sampling families. HMI pools subgroup medians from every compatible rice sampling family in the evidence ledger for this analyte, material and basis: published subgroup medians, medians HMI calculates from printed per-sample values and WHO GEMS/Food region-by-year medians (random sampling only). Each family has equal total weight; subgroups have equal weight within a family. Non-detects stay as zero-to-limit intervals, so quartiles carry lower and upper bounds. Means, ranges and maxima never enter as observations. Dry-weight, native-moisture and cooked results are kept in separate references; no moisture conversion is assumed.
Inverse weighted empirical CDF; midpoint at exact cumulative-weight boundaries. At least four reference units and a definitely positive Q1–Q3 spread are required. This is a display eligibility rule, not statistical validation.
Limitations
An evidence-library reference, not a probability sample of the world rice market. Sampling families differ in frame (retail, farm, contaminated sites, monitoring) and period; GEMS/Food regions are reporting regions, not growing origins. Many submissions are non-detects with differing limits, which widens the bounds. A position is descriptive concentration, not a safety or compliance judgment.
Admitted source families
- Alrashdi et al. (2024)
- WHO GEMS/Food
- EFSA et al. 2021
- Limmer et al. 2026
- Liu et al. 2022
- Moreda-Pineiro et al. 2020
- Sim et al. 2024
- Tian et al. 2024
- Yan et al. (2025)
- Zavala et al. 2008
Reference version and interpretation
Frozen evidence selection: 2026-10-22. Method: hmi-pooled-reference-v2.
hmi-pooled-reference-v2:rice-native-iAs:ac23d7340743
The reference is computed by HMI from admitted evidence, not copied from a study baseline. It describes the evidence currently in this scope; it is not a worldwide or current-lot benchmark. Source, year and region filters do not change it. Different HMI scopes are not one common ranking.