Biomonitoring of nutritional and toxic elements in raw and cooked rice in the Health Effects of Arsenic Longitudinal Study, Bangladesh Authors: Catherine M. Lucey, Tiffany R. Sanchez, Heidi Hannoush, Anirban Basu, Lewis H. Ziska, Muhammed F. Parvez, Habibul Ahsan, Kathrin Schilling Published: 2026 ABSTRACT Rice is a staple food and a major dietary exposure source of both nutrient and toxic elements. Rice studies typically focus on individual trace elements rather than cohesively assess multiple elements and typically measure trace elements in raw grain, disregarding cooking effects. Further, almost no studies link cooked rice profiles to biomarkers of internal exposure. Here, we used 106 matched triplets of raw rice, cooked rice, and urine samples from the Health Effects of Arsenic Longitudinal Study in Araihazar, Bangladesh to 1) demonstrate the effect of cooking on 13 essential and toxic elements in rice, 2) characterize elemental profiles of raw rice, cooked rice, and urine, and 3) identify elements in cooked rice that could influence human elemental exposure. Barium and strontium significantly increased after cooking (mean differences of 241 ng/g and 679 ng/g, respectively), while cesium, selenium, molybdenum, and nickel significantly decreased (mean differences of −12 ng/g, −19 ng/g, −202 ng/g, and −125 ng/g, respectively). Iron, cobalt, molybdenum, and vanadium levels in cooked rice significantly correlated with urinary levels (Spearman’s ρ = 0.28, 0.29, 0.28, and 0.36, respectively). This suggests that rice may be a previously unidentified exposure source for cobalt, vanadium, and molybdenum - three understudied but potentially toxic elements. Overall, we find that elemental concentrations in raw rice may not reliably reflect dietary exposure, as cooking can introduce external inputs and remove endogenous elements. Integrating biomonitoring data, such as urine measurements, is therefore essential for understanding the contribution of rice to total elemental exposure. Introduction Rice is a staple food that plays a central role in global food security. It is critically important to monitor the nutritional quality and safety of rice. During cultivation, rice accumulates a wide variety of trace elements from both soil and irrigation water (Rai et al., 2019; Lu et al., 2019). Concentrations of inorganic elements in rice and agricultural soils, as well as their transfer from soil to rice, have been extensively studied (Meharg et al., 2023). Some trace elements in rice are essential inorganic micronutrients. For example, iron (Fe) (Platel and Srinivasan, 2016), selenium (Se) (Fairweather-Tait et al., 2010), and zinc (Zn) (Platel and Srinivasan, 2016), which populations worldwide are deficient in, are abundant in rice; estimates of Fe, Se, and Zn in raw rice range from 1000 – 5000 ng/g, 100–600 ng/g, and ~10,000 ng/g, respectively (Meharg et al., 2023; Williams et al., 2009; Halder et al., 2020). Rice also contributes smaller amounts of other essential trace elements with lower daily intake requirements, including cobalt (Co), copper (Cu), molybdenum (Mo), and nickel (Ni). Concentrations of Cu are reported as ~1000 ng/g in rice (Meharg et al., 2023), while concentrations of Co, Mo, and Ni are between 100 and 500 ng/g in raw rice (Meharg et al., 2023; Navaretnam et al., 2023; Watanabe et al., 2022). Inorganic toxicants with no known biological role can also bioaccumulate in rice. Arsenic (As) is extensively studied (Jaafar et al., 2021), but rice is also a source of other well-established inorganic toxicants like cadmium (Cd) and lead (Pb) (reviewed in (Rai et al., 2019); Zhao et al., 2024). Cadmium and Pb can be as high as > 1000 ng/g Pb and > 100 ng/g Cd in rice (Sharafi et al., 2019a) despite international food safety guidelines setting maximum permissible concentrations of 50 ng/g Pb and Cd in rice. Additionally, rice can be a source of emerging, potentially toxic environmental contaminants such as vanadium (V) and elements like barium (Ba), strontium (Sr), and cesium (Cs) whose toxicity or biological role are not clearly defined. Strontium and Ba been reported at levels comparable to Cd and Pb, i.e. exceeding 500 ng/g in rice (Lu et al., 2019; Woldetsadik et al., 2024). Typically, trace elements are measured in raw grain (for ex Menon et al., 2022). However, cooked rice, not raw grain, represents the actual dietary exposure, and cooking can change the elemental composition of rice (Mohajer et al., 2024). Most studies comparing raw and cooked rice have focused on already well-understood elements, such as As, Cd, Pb, Zn, and Fe (Mohajer et al., 2024; Menon et al., 2021). There are currently little or no data on the effect of cooking on Ba, Co, Cs, Mo, or Sr even though these elements are common in rice. In laboratory studies where rice is cooked in ultrapure water, cooking generally decreases trace element concentrations by 10–50% depending on the element and cooking method (Mwale et al., 2018; Sharafi et al., 2019b, 2019c). This decrease in toxic elements during cooking typically coincides with a losses of nutrient elements and thus cooking is a trade-off between reduced toxic burden and nutritional value (Sharafi et al., 2019c; Liu et al., 2019). Although experimental studies are useful in describing the effects of cooking in idealized conditions, “real-life” studies capture how rice is actually prepared in household settings, including the use of local water sources, cookware, and cooking practices. All of these variables can affect trace element levels in the cooked rice. For example, cooking water contains dissolved elements which rice can absorb through diffusion-driven mass transfer during grain hydration, with incorporation facilitated by interactions with internal grain matrices and starch gelatinization (Shinde et al., 2014). Moreover, cookware can be contaminated with trace elements like Pb, Cd, and Ni (Koo et al., 2020), and cookware has been shown to increase trace element concentrations in rice even when cooking water is uncontaminated (Tarviji et al., 2023). Salt added to rice during cooking may also be a source of additional trace elements (Ismail et al., 2025; Stajic et al., 2025). Therefore, trace element levels in raw rice are an imperfect proxy for dietary exposure. Even when cooked rice is assessed, there is typically no associated biomonitoring. Without biomonitoring, there are no data to assert whether elemental levels in rice meaningfully contribute to elemental exposure in humans; it is entirely possible that elements can be high in rice but not bioaccessible and therefore not relevant for human exposure. Large cohort studies that include urinary element biomonitoring rarely measure elements in food directly, instead relying on self-reported intake to approximate dietary exposure (Gilbert-Diamond et al., 2011). Thus, prior studies have examined either trace element concentrations in rice (raw and sometimes cooked) or associations between reported rice intake and urinary biomarkers but not all three components together. Lastly, studies focus exclusively on either nutrient elements or toxic elements, rather than both, which limits the ability to capture the full elemental profile of rice and potential co-exposures. Here, we address these gaps by (1) quantifying changes in multiple essential and non-essential elements in rice during cooking, (2) identifying inter-elemental dynamics that distinguish raw rice, cooked rice, and urine, and (3) evaluating correlations between cooked rice and urinary metal concentrations in a Bangladeshi cohort. We leveraged matched sample triplets of raw rice, cooked rice, and spot urine collected through the Health Effects of Arsenic Longitudinal Study (HEALS), a large prospective cohort study conducted from 2006 to present in Araihazar, Bangladesh, a roughly 180 km2 upazilla south of Dhaka (Ahsan et al., 2006). In this population, rice composes approximately 75% of the diet (Biswas et al., 2021), reducing confounding from other dietary sources of essential and nonessential elements and emphasizing the importance of rice as a trace element exposure pathway in Bangladesh. Given the high daily intake of rice in Bangladesh, even moderate concentrations can translate into substantial chronic exposure. For example, Cd in Bangladeshi rice has been reported at mean levels of ~37–99 ng/g, among the highest globally (Shahriar et al., 2020), and dietary intake assessments indicate that rice is a major contributor to cadmium exposure in the Bangladeshi population (Al-Rmalli et al., 2012). We evaluated a large 13-element panel that included the essential elements Cu, Ni, Se, Zn, Co, Fe, and Mo and the nonessential elements V, Sr, Ba, Cs, Pb, and Cd. To choose this panel, we prioritized elements that are reported at high levels in raw rice but that have not been assessed in cooked rice, elements that are bioacessible in rice, and elements rarely represented in the rice literature. Further, many of these elements are known geogenic drinking water contaminants in Bangladesh, including in the study site of Araihazar (Supplementary Figure 1). We omitted As and manganese from our panel due to the abundance of existing literature on these two elements in rice (Jaafar et al., 2021). By integrating food chemistry with biomonitoring, this work provides novel insights into dietary exposure pathways and highlights overlooked contaminants in rice-rich diets. Methods HEALS study and sample collection In 2015, during routine follow-up home-visits, trained HEALS investigators collected raw and cooked polished rice samples from the households of a randomly selected subset of HEALS participants to assess the contribution of dietary rice to total metal exposure (Ahsan et al., 2006). Participants were asked to cook some of their household rice according to their usual habit, which in Bangladesh typically involves boiling rice in excess water followed by draining (Melkonian et al., 2013; Rahman et al., 2006). Cooking was performed with the participants' regular utensils and cookware without modification to standard practices, ensuring that rice preparation reflected real-world conditions. The HEALS investigator collected a sample of the cooked rice as well as a sample of raw rice from the same supply/bag of rice. Lastly, the HEALS investigator collected a spot urine sample from the participant who was at home and collected basic biometric data such as age, sex, and weight, as well as betel leaf or cigarette usage (Supplementary Table 1). Spot urine samples were collected in acid-washed 50 mL polypropylene conical tubes that had been rinsed with double deionized water prior to use to minimize background metal contamination. Samples were stored at −20°C, shipped to Columbia University on dry ice, and maintained at −20°C until analysis, consistent with previously described HEALS protocols (Ahsan et al., 2006; George et al., 2013). Rice samples were stored in acid-washed 50 mL tubes and similarly shipped and stored long-term at −20°C. All participants provided verbal informed consent, and all protocols were approved by the Columbia University Institutional Review Board and the Ethical Committee of the Bangladesh Medical Research Council under protocol number AAAA7349. Sample preparation All solutions were prepared using ultrapure nitric acid (HNO3, 65–67%, Optima®, Fisher Scientific) and ultrapure water (≥18.2 MΩ × cm) in new metal-free tubes. All metal-free polypropylene centrifuge tubes (Labcon, Petaluma, CA; MetalFree® certified) were from approved lot numbers that showed no metal contamination during routine testing in the METALab. Raw and cooked rice samples were oven-dried at 85°C to a constant weight, ground into a fine powder with a Retsch MM400 grinder, and stored at room temperature in metal-free tubes. Drying samples ensured grams of raw rice and grams of cooked rice were comparable. To prepare samples for elemental analysis, 0.2 g of each dried and powdered rice sample was weighed to a precision of ± 0.01 mg into acid-cleaned 20 mL microwave PFA vessels. Subsequently, 1.5 mL of ultrapure HNO3 was added to each vessel. Digestion was performed using the MARS 6 Digestion Microwave System (CEM Corp., Matthews, USA). The temperature was gradually increased in 10-minute intervals to 90, 120, 150, and 180°C, then increased to 200°C and held for 20 min. After cooling to room temperature, acid-digested rice samples were transferred into 15 mL metal-free polypropylene centrifuge tubes (Labcon, Petaluma, CA; MetalFree® certified). Then, 150 μL of a 50,000 μg/L gold (Au) stock solution (in 1% HCl) and 150 μL of an internal standard mixture (500 μg/L each of yttrium, rhodium, gallium, and iridium in 2% HNO3 and 1% HCl; Agilent, Santa Clara, CA, USA) were added to each sample. Samples were then diluted to a final volume of 15 mL with ultrapure water (≥18.2 MΩ⋅cm), resulting in a final acid concentration of 10% (v/v) HNO3, a final internal standard concentration of 5 μg/L for each element, and a final Au concentration of 500 μg/L. Gold was included as a stabilizing agent to minimize adsorption of analytes onto the sample introduction system (e.g., tubing and spray chamber). Every batch of microwave digestion included ≥ 4 blanks and ≥ 3 samples of rice flour certified reference materials (CRM) NIST–1568b (National Institute of Standards and Technology, Gaithersburg, MD, USA) and IRMM–804 (Institute for Reference Materials and Measurements, Geel, BE). Seventeen rice samples were digested and analyzed in duplicate (8% of all raw/cooked rice samples). Urine samples were prepared with a “dilute and shoot” method (Schilling et al., 2024). A 0.1 mL aliquot of each study urine sample and of CRMs QM-U-Q1822, QM-U-Q1823, QM-U-Q1824 (Quebec Multi-element External Quality Assessment Scheme, Quebec, CA), ClinChek Level I (RECIPE, Munich, DE), and NIST 2668 Levels 1 and 2 (National Institute of Standards and Technology, Gaithersburg, MD, USA) was diluted 50-fold with 2% HNO3, 0.02% (v/v) Triton X–100 diluent, and 500 μg/L Au (from a 50,000 μg/L Au stock in 1% HCl). An internal standard of 500 μg/L each of yttrium, rhodium, gallium, and iridium in 2% HNO3 and 1% HCl (Agilent, Santa Clara, CA, USA) was added to each sample to achieve a final concentration of 5 μg/L per internal standard element prior to analysis. Blanks were made by replacing the volume of urine with 2% HNO3 but otherwise treated the same. Twenty urine samples were prepared in duplicate (19% of the urine dataset). Total metal analysis with ICP-MS Samples were analyzed for total metals with a PerkinElmer NexION 350S (Waltham, MA, USA) inductively coupled plasma mass spectrometer (ICP-MS) equipped with an Elemental Scientific (ESI) 4DX auto-sampler (Omaha, NE, USA) and fitted with platinum sampler and skimmer cones, PFA-ST nebulizer, and cyclonic quartz spray chamber. The ICP-MS was operated according to standardized instrumental parameters (Schilling et al., 2024). We performed external eight-point calibration with matrix-matched solutions. Urine and rice samples were measured on separate analytical days. Urine samples were bracketed by ≥ 4 method blanks and ≥ 3 samples of each CRM in every analytical run. Rice samples were similarly bracketed by a method blank and CRM (as described in Section 2.2). Rice and urine samples were analyzed for Co, Ni, Zn, Cu, Sr, Mo, Ba, and Pb in no collision cell gas mode, V in ammonia gas mode, and Se and Cd in oxygen gas mode. Cadmium was measured at two isotopes, 113Cd and 111Cd, and the measured concentrations at each isotope were averaged to generate a single Cd measurement. In all ICP-MS runs, method blanks and CRMs were analyzed every 10 samples. The Syngistix software package (v2.5) for the Perkin Elmer NexION 350S was used for data acquisition. All measured values were corrected for minor instrumental drift based on internal standard recovery ranging between 91 – 105%. Additionally, the mean method blank was subtracted from each analyzed sample on every analytical day. The limit of detection (LOD) was calculated using the formula: LOD 3.33 × standard deviation from blank measurements (n = 29 for rice and n = 50 for urine). The method detection limit (MDL) was then determined by multiplying the LOD by the dilution factor (75-fold for rice, 50-fold for urine). Measurement accuracy was assessed with percent recovery of certified elements in CRMs, and precision was assessed as the coefficient of variation (CV) between duplicates (20 duplicates for rice and 19 duplicates for urine, representing 9.5% and 18% of all samples, respectively). The MDL, LOD, CRM recovery, and duplicate CV values for rice and urine samples are in Supplementary Tables 2 and 3, respectively. Element concentrations below the MDL were replaced by the MDL divided by the square root of 2. Only one rice sample was below the MDL for Cd, Pb, and Zn and two samples for Fe. For Co and V, 24 and 25 samples were below the MDL (11.8% of 212 total samples). For the urine elemental analyses, one sample was below the MDL for Cu, 8 for Fe, 11 for Ni, and 15 for V (0.9%, 7.5%, 10.4%, and 14.2% of 106 total samples, respectively). Urine specific gravity correction To correct for hydration status, we adjusted urinary element concentrations using urine specific gravity (USG). USG was measured with a hand-held digital refractometer (Atago 4410 PAL–10S, Tokyo, Japan) and values ranged from 1.002 to 1.033 (median, 1.009). Urinary element concentrations were then adjusted using the Levin-Fahey equation with the sample median USG = 1.009 as the USGref value: Cnorm=Cmeasured⋅USGrefUSGmeasured Statistical analysis The final sample size was 106 complete triplets of matched raw rice, cooked rice, and urine samples. Samples were omitted when pairs or triplets were incomplete (i.e., missing a raw rice, cooked rice, or urine sample). One sample triplet was excluded from analysis due to an outlying Co concentration raw rice, and two triplets were removed for outlying Cu concentrations in cooked rice (≥ 6 standard deviations from the mean). To determine whether differences in element content between cooked and raw rice were statistically significant, we used two-sided paired t-tests and a Bonferroni corrected significance threshold of p = 0.0038 to account for multiple comparisons (p = 0.05/13 tests). As data were largely right skewed, we calculated Spearman’s ρ to identify correlations between cooked rice and urinary elements as well as to identify correlations between elements within each sample type (raw rice, cooked rice, and urine element). Like the raw to cooked rice differences, Spearman's correlations were considered significant when p was below the Bonferroni corrected threshold of 0.0038. In addition to the element-by-element analysis, we used correlation matrices and hierarchical clustering to assess groups of elements. Hierarchical clustering was performed on Euclidian distance matrices with Ward’s method, including implementation of Ward’s clustering criterion (hclust function with the ward.D2 method in R). All variables were standardized to zero mean and unit variance prior to hierarchical clustering. All analyses were performed with R Version 4.4.1 in RStudio version 2023.12.1 + 402. Paired t-tests were performed using the rstatix package. Sensitivity analyses We conducted two sensitivity analyses to identify whether measurements below the MDL affected the correlations between cooked rice and urine for V and Co, since there was a relatively high proportion of urine and rice samples below the MDL for these elements. Removing samples below the MDL did not change the strength or magnitude of the correlations. We also conducted a sensitivity analysis for the correlation between urinary and cooked rice Cd stratifying by tobacco use, since smoking and chewing a mix of betel leaf and tobacco are common in the HEALS cohort and both betel leaf usage and cigarette smoking are known to increase urinary Cd (Al-Rmalli et al., 2011). We calculated the Spearman’s correlation coefficient between cooked rice Cd and urinary Cd after stratifying by 1) current betel leaf usage (yes/no) and 2) current cigarette usage (yes/no). In both analyses, ρ was neither significant nor different between groups, and, importantly, was essentially the same value as the non-stratified ρ (data not shown). Results Elemental content of raw rice, cooked rice, and urine In both raw and cooked rice, Zn, Fe, and Cu were by far the most abundant with geometric means (inter-quartile ranges, IQRs) of 9574 (3850), 7948 (6621), and 2872 (1320) ng/g in raw rice and 8949 (3809), 8233 (6681), and 2872 (1202) ng/g in cooked rice, respectively. By contrast, Zn, Sr, and Mo were the most abundant in urine, with means (IQRs) of 303 (381), 141 (229), and 68 (102) μg/L, respectively (Table 1). Degree of variability, as indicated by the IQR, was comparable between cooked and raw rice for each element except that Sr and Ba were more variable in cooked rice and Ni and Mo were less variable in cooked rice relative to raw (Table 1). Urine elemental concentrations varied less than rice concentrations. Effect of cooking on rice elements Cooking substantially altered the elemental profile in an element specific manner. Barium and Sr significantly increased after cooking (mean differences of +679 and +240 ng/g, respectively, Fig. 1A, F). Cesium, Mo, Ni, and Se were significantly and consistently lower in cooked compared to raw rice (Fig. 1B, C, D, and E, respectively, Table 1). Iron, Cd, Co, Cu, Pb, V, and Zn did not differ between raw and cooked rice (Supplementary Figure 2). There was no clear difference in the effect of cooking between elements non-essential to humans (Ba, Pb, Cd, Cs, Sr, V) and those essential to humans (Co, Fe, Zn, Se, Cu, Ni, Mo). Among elements that showed no overall difference, two patterns emerged. First, Cd, Cu, and Zn appeared entirely unaffected by cooking, with nearly identical levels before and after cooking (Supplementary Figure 2 A, C, G). By contrast, Fe, Pb, Co, and V decreased in some samples and increased in others (Supplementary Figure 2 B, D–F). Decreases in Cs, Mo, Ni, and, to a lesser extent, Se and Fe, were inversely related to their respective levels in raw rice (Supplementary Figure 3 D, G, H, J, and F). For Ba and Sr, their increase during cooking showed no relationship to their concentrations in raw rice (Supplementary Figure 3 A and K). Correlations between cooked rice element and urinary element concentrations Urine and cooked rice element concentrations significantly correlated for Co (Spearman’s ρ = 0.29), Fe (Spearman’s ρ = 0.28), Mo (Spearman’s ρ = 0.28), and V (Spearman’s ρ = 0.36, Table 2, Supplementary Figure 4 C, G, and L) but not for the other elements investigated (Table 2, Supplementary Figure 4). The correlation for V was the strongest, and the correlation for Mo had the largest magnitude (Supplementary Figure 4 G and L). The correlations between urinary and cooked rice Co and V were mainly driven by samples where the elements increased during cooking (Supplementary Figure 4 C and L). Elemental profiles of raw rice, cooked rice, and urine Inter-element correlations were weakest in cooked rice, moderate in urine, and strongest in raw rice samples (Fig. 2 A, B, C). In raw rice, the strongest correlations were observed between V and Fe, Pb, and Sr and between Fe and Pb/Sr. Moderate correlations were observed between Ni and Se/Cd and Se and Cu (Fig. 2 A). In cooked rice, the strongest positive correlations were between Ni and Cd/Se, and there was a moderate (ρ = 0.3) negative correlation between Co and Sr (Fig. 2 B). In urine, there were strong (ρ ≈ 0.9) correlations between Ni and both Sr and Co, as well as strong intercorrelations between Se, Cs, Cu, and Pb (Fig. 2 C). Hierarchical clustering patterns differed substantially between raw rice, cooked rice, and urine. We found three main clusters in raw rice Cs-Zn-Mo-Cd, Fe-Pb-V-Ba, and Co-Sr-Se-Ni-Cu (Fig. 2 D) and four in cooked rice: Cd-Cs-Se-Ni, Mo-Cu-Zn-Pb, Ba-Sr, and Fe-V-Co (Fig. 2 E). Strontium and Ba, the only two elements that significantly increased after cooking (Fig. 1), formed a distinct cluster in cooked rice but not in raw rice. Likewise, three of the four elements with variable responses to cooking (some increase, some decrease, e.g., V, Fe, and Co) formed a distinct cluster in cooked rice (Fig. 2 E). There were two major clusters in urine: V-Co-Ba-Fe and Pb-Zn-Se-Cu-Cs-Cd-Mo-Sr-Ni (Fig. 2 F). Of the four elements significantly correlated between urine and cooked rice, three (Co, Fe, and Mo) belonged to the same cluster in urine and three (Fe, V, and Co) belonged to the same cluster in cooked rice (Fig. 2 F). Within a sample type, the hierarchical clustering results and correlation matrices were largely consistent. In urine, Se, Zn, and Pb cluster together and are highly intercorrelated (Fig. 2 C, F). Similarly, Fe, Pb, and V were strongly intercorrelated in raw rice and belong to the same cluster. In cooked rice, where a Cd-Cs association is one of the few substantial positive correlations, Cd and Cs belong to the same leaf (Fig. 2 B, E). Discussion Elemental concentrations in rice and urine relative to regional and global data Elemental concentrations in our rice samples were within reported ranges for most elements (Lu et al., 2019; Navaretnam et al., 2023; Rothenberg et al., 2015). Elemental levels in urine samples were roughly consistent with previously reported results in Bangladeshi adults (Berglund et al., 2011). We observed similar urinary levels of less commonly reported elements like Sr (Zhang et al., 2023) and Mo (Berglund et al., 2011). The urinary metal profile in our subset of HEALS samples closely resembled a larger sampling of HEALS participants, with the exception of much higher (~4x) urinary Ba (Sanchez et al., 2018). This suggests that our sample size, while small, sufficiently represented the HEALS population. Effect of cooking on rice elemental levels Cooking altered concentrations of several elements in rice. Barium and Sr consistently increased, while V, Pb, and Ni increased in some but not all samples. Zinc and Fe remained stable, and Mo and Se consistently decreased. The Ba and Sr increase likely reflects contamination from cooking water because both are common drinking water contaminants in Bangladesh (Haque, 2018; Frisbie et al., 2009; Kato et al., 2013). The lack of an association between raw rice concentrations and post-cooking increases in Ba and Sr also supports an external source. Ba and Sr have similar chemical properties to each other and behave chemically similarly to calcium (Ca). Calcium in rice increases when raw grains are soaked in a Ca-enriched solution (Rahman et al., 2019), supporting our hypothesis that Ba and Sr in cooked rice originate from cooking water. The substantial Ca concentrations observed in local groundwater of up to 64,000 μg/L in shallow wells further supports the plausibility of uptake processes of divalent cations during rice cooking (Dhar et al., 2008). The lack of correlation between cooked rice and urinary Ba and Sr in our data agrees with evidence that Sr in drinking water does not correlate with urinary Sr, but contradicts data in that same study showing an association between drinking water and urinary Ba (Kato et al., 2013). For samples where V increased, V could also come from cooking water. Although data on V as a drinking water contaminant are sparse (Parvin et al., 2022), V and As co-occur in groundwater due to redox sensitivity, and it is likely that V is present in regions like Bangladesh where As contamination is high (Ahmed et al., 2010). Reliable and comprehensive well water data for the studied households were not available, as water samples were improperly preserved for the studied elements (e.g., lack of acidification and uncontrolled storage conditions) and rice/urine were analyzed approximately six years post-collection, potentially altering elemental concentrations. Previous studies have reported substantial decreases in Pb during cooking (Ortiz et al., 2021). However, Pb levels increased in roughly half of our rice samples, potentially from Pb leaching from aluminum pots during cooking. Aluminum pots, which are widely used in Bangladesh, are contaminated with Pb (Weidenhamer et al., 2017), unlike stainless steel (Ortiz et al., 2021), and Pb and Cd from these pots can leach into the cooked food (Koo et al., 2020; Binkhorst et al., 2025). However, cookware type was not directly assessed in this study, and this explanation remains speculative. Consistent with our findings, Menon et al. showed that cooking did not change Zn and Fe in white rice, even across a number of different cooking methods, and showed the same decrease in Mo after cooking (Menon et al., 2024). Molybdenum is likely lost during pre-cooking rinsing of the rice rather than during cooking itself (Shahriar et al., 2022). However, Menon et al. found that Se did not decrease during cooking, in contrast to our results (Menon et al., 2024). Menon et al. additionally reported that Cu stayed the same only when rice was cooked in excessive water – but not with a parboil method – suggesting that Cu might be particularly sensitive to cooking method (Menon et al., 2024) and that HEALS households likely used excessive water to cook rice. Inter-elemental dynamics in raw rice, cooked rice, and urine Our urine hierarchical cluster analysis agreed with a larger HEALS urine analysis that also showed clustering between Sr/Ni and Cs/Se (Sanchez et al., 2018). Selenium, Zn, and Pb likely cluster together and intercorrelate because all three are divalent metals regulated by binding to metallothionein (Krężel and Maret, 2021). The urine cluster and correlation patterns differed from both raw and cooked rice, except for co-clustering of Pb and Zn both cooked rice and urine, consistent with the weak to moderate correlations between cooked rice and urinary elements. Further, groupings of elements in raw rice did not predict groupings of elements in cooked rice. This suggests that the factors determining abundance within the rice grain (storage, degree of uptake from soil, binding environment, etc.) are distinct from the factors determining the effect of cooking. For instance, Ba and Sr, the two elements that increased during cooking, did not cluster together at all in raw rice but formed a distinct leaf in cooked rice. It was surprising that Fe and Cd neither clustered together nor substantially correlated in raw rice because Cd uptake is partly governed by activation of Fe transporters in low Fe environments (Nakanishi et al., 2006). We expected Zn and Cd to co-cluster because Cd mimics Zn in plants such that Zn and Cd uptake co-occur (Nan et al., 2002). The inclusion of Cs in the Cs-Zn-Cd cluster in both raw and cooked rice could be because Cs is taken up by potassium (K) channels in plants (Rai and Kawabata, 2020), and Zn uptake in rice is enriched when K-based fertilizers are used (Shahid et al., 2016). In this way, Cs, Zn, and Cd (by extension of its proposed Zn mimicry), could all be responsive to K in the growing environment. The strongest similarity between raw and cooked rice is the identical leaf of Ni and Se. This cluster suggests that Ni and Se are similarly prevalent in the rice grain and that cooking does not affect their relative abundances to one another. In raw rice, this cluster includes Cu, suggesting that Cu may also have similar uptake behavior to Se and Ni in rice, but different responses to cooking. In rice, Se is primarily incorporated into the amino acids cysteine (Cys) or methionine (Met) (Terry et al., 2000; Sors et al., 2005). Given similar Cu, Ni, and Se behavior in our data, we hypothesize that Ni and Cu are also binding to Cys and Met in Se-containing enzymes. Cys-Cu and Met-Cu bonds have been found in silico (Krishna Deepak et al., 2017) and in vitro (Rubino et al., 2010), and Ni-Se-Cys bonds have been found in bacteria (Eidsness et al., 1989) but not yet in rice. The strong negative correlation between Cs, Ni, and Se in raw rice and the magnitude of loss during cooking supports that Cu and Ni are in comparable binding environments. Cobalt, iron, molybdenum, and vanadium correlate in cooked rice and urine The observed correlations between cooked rice and urinary element levels suggest that cooked rice may be a previously unrecognized exposure route for Co, Mo, and V and a dietary source of Fe. It is plausible that rice is a relevant source of these elements because Hattori et al. found that rice had the highest Mo among numerous foods (Hattori et al., 2004), much V exposure is known to be through food (ATSDR, 2012), and dietary Co exposure is dominated by rice due to volume eaten (González et al., 2026). For V, the correlation appears driven by samples where V increased, so rice may only be a dietary V source when cooking introduces additional V. The correlation in Fe between cooked rice and urine is encouraging because Fe deficient anemia is a severe public health issue in South Asia (Kassebaum et al., 2014); our data suggest that efforts to biofortify rice with additional Fe (Lopez de Romaña et al., 2023) may be effective in reducing anemia risk. Besides Fe, the elements significantly correlated in cooked rice and urine are all poorly studied in the context of human exposure and toxicity. Molybdenum and Co are essential micronutrients that are components of a human oxidase cofactor (Mendel and Kruse, 2012) and vitamin B12, respectively (Agency for Toxic Substances and Disease Registry, 2024), but excessive Mo and Co intake could be detrimental. Molybdenum supplementation may have beneficial effects in cases of iron deficient anemia (Grech, 2021), but there is also emerging evidence of an association between Mo exposure and reduced male fertility (Meeker et al., 2008). Chronic Co exposure is not well studied; however, chronic inhalation can have negative pulmonary effects, and Co is potentially carcinogenic (Agency for Toxic Substances and Disease Registry, 2024). Like Mo, V is thought to be beneficial at low doses but toxic at high ones (Ścibior et al., 2020). Whether V should be considered an essential element is actively debated, but there is increased concern about V as an emerging environmental contaminant because industrial V use is on the rise (Watt et al., 2018). Of course, rice is not the sole component of a meal, and it is possible that observed correlations are confounded by trace elements in other ingredients. Distinguishing between rice and other sources of these elements is a necessary next step in understanding exposure risk. For example, the correlation between rice Mo and urinary Mo could be affected by legume consumption, since legumes are a staple food high in bioaccessible Mo (Mauer et al., 2024). However, other ingredients like leafy greens are equally likely to contribute trace elements such as Cd (Al-Rmalli et al., 2012) where we did not observe a rice to urine correlation. Furthermore, HEALS participants eat many fewer grams of beans, vegetables, and meats than they do of rice ( >10x fewer) (Chen et al., 2004). Taken together, the three correlations in cooked rice and urine Co, Mo, and V suggest that current biomonitoring efforts have been insufficient in identifying potentially toxic components of a rice-enriched diet. Confirmatory work assessing Co, Mo, and V exposure in other dietary components and drinking water is needed to clarify where these elements come from. Further, these data reveal a need for additional toxicological assessment of Co, Ba, Mo, V because the health consequences of dietary exposure to these elements are not clear. Elements not correlated to urinary levels Barium, Cs, Ni, Se, and Sr in rice significantly changed during cooking but did not significantly correlate with their urinary levels. The large decrease in Cs during cooking suggests that risk assessments of Cs in raw rice may be overestimating Cs exposure risk. Loss of Se during cooking is particularly important given the high prevalence of Se deficiency in Bangladesh and widespread efforts to biofortify or retain Se in food to mitigate deficiencies (D'Amato et al., 2020). The loss of Se during cooking suggests that Se fortification of rice may need to “overshoot” to ensure that sufficient Se is retained in cooked rice. However, biofortification of Se in rice is a delicate balance, as Se is toxic at high doses (Fairweather-Tait et al., 2011). We only assessed total Se, but the bioavailability of Se depends on its species (Fairweather-Tait et al., 2010). Therefore, it is possible that total Se decreased during cooking but bioavailable methylated species were retained. Cooking did not change levels of Pb, Co, Cd, Zn, Fe, or Cu. No change in Cd and Pb is concerning because both are toxic at very low doses as well as bioaccesible in rice (Ortiz et al., 2021; Howard et al., 2024; Zhuang et al., 2020). An average Bangladeshi diet of ~350 g of rice per day (Yunus et al., 2019) would contribute substantial Pb and Cd: 3.6 μg Pb and 13.7 μg Cd at the average cooked rice Pb and Cd levels we observed. Despite this, we did not observe a correlation between urinary Cd or Pb and cooked rice. While urine is a reliable biomarker for Cd exposure, for Pb, it is possible that there may be a correlation between rice and biological matrices that are more representative of long-term exposure, such as blood (Martinez-Morata et al., 2023). The retention of Zn after cooking should be a benefit because Zn deficiencies are common in South Asia (Kassebaum et al., 2014). However, the near-zero correlations between rice Zn and urinary Zn in our data suggest Zn from cooked rice contribute very little to urinary Zn. Our results are particularly interesting given widespread biofortification efforts to increase Zn in rice, since our data suggest that rice Zn may not meaningfully contribute to internal Zn dose. Zinc in rice likely has low bioavailability because Zn is predominantly bound to the nonbioavailable “anti-nutrient” phytate (Shen et al., 2023). Although phytate is more abundant in brown rice, polished (white) rice has enough phytate to complex Fe2+, Zn2+, Ca2+, Mg2+, Mn2+, and Cu2+ and reduce their bioavailability (Schlemmer et al., 2009). Interestingly, we see a moderate association between cooked rice Cu and urinary Cu and a significant correlation for cooked rice Fe and urinary Fe. Though both Cu and Fe can bind phytate, like Zn (Schlemmer et al., 2009), the effect of phytate binding on bioavailability might be more severe for Zn than Cu or Fe. These data suggest that rice may be a better substrate for Fe or Cu fortification rather than Zn. Limitations The main study limitations are a lack of matched contemporaneous well water data, participant-specific cooking methods data, and non-rice dietary data. With participant-level data on cooking method and cooking utensils, we could have assessed whether different techniques affected elemental profiles. For elements with variable changes during cooking (Fe, Pb, Co, V), cooking method may determine whether concentrations increase or decrease. At the same time, we captured real-life variation in personal cooking preference which makes our findings more generalizable. A lack of well water data or dietary data beyond rice meant we were unable to assess the relative contributions of rice, water, or other foods to total exposure. With well water and more comprehensive dietary data, we could have clarified which matrix contributes the most to the urinary element profile. Further, we did not directly assess bioavailability of each element in the rice and instead used the correlation between cooked rice and urine as a surrogate measure for assessing bioavailability. Direct bioaccessibility measurements would better indicate whether the bioacessible fraction of each element, rather than the total amount of each element, changed during cooking. Given novel associations between rice and urinary Co, Mo, and V, bioaccessibility studies for these three elements are important for future research. Conclusions Here we demonstrate that cooking changes the elemental profile of rice in an element-specific manner, highlighting the importance of analyzing cooked food, rather than only raw food, when assessing elemental exposure profiles. Despite a strong correlation between cooked rice and urine, there is no correlation between raw rice and urine for Fe, V, Co, or Mo (data not shown), suggesting that only cooked rice accurately reflects the exposure potential for these elements. We also showed that cooked rice may be a novel dietary source of the understudied elements Co, Mo, and V, and identified new relationships between trace elements in rice. Given the critical role of rice in feeding the world’s population, it is essential to continuously monitor the elemental profile of rice to accurately assess both the health benefits and health risks of a rice-based diet. The research presented here is directly in line with the UN’s 2030 agenda aiming to enhance food security and nutrition by linking environmental health with human well-being. First, we demonstrate that elements that are not typically assessed in rice can be significant exposures in humans, and the bioaccessibility of these elements should be further studied. Second, we can anticipate that the elemental profiles of rice will change over time, perhaps substantially, and current research will need to be continuously updated as the rice market evolves. The elemental composition of rice could be positively affected by biofortification efforts, or negatively affected as industrial use of elements like Co and V becomes more common. As industry evolves, so will the composition of industry-related contamination of water and soil, and thus changes to the prevalence of toxic elements in rice and risk for human exposure. Lastly, elemental concentrations in rice are known to change under the pressures of climate change (Wang et al., 2025), including levels of Mo in rice (Wang et al., 2020). As the global climate continues to change and worsen, the role of rice in dietary exposures and the necessity of monitoring its elemental quality will change as well. Supplementary Material