Arsenic in Drinking Water and Urinary Tract Cancers: A Systematic Review Update

: Problem : There remains uncertainty around cancer risk at lower levels of arsenic in drinking water. This study updates evidence from our previous review on the relationship between arsenic in drinking water and urinary bladder and kidney cancers (updated search January 2013 to February 2023). Method : Thirty-four studies were retained for review; six met criteria for inclusion in meta-analysis. Risk estimates for bladder and kidney cancer incidence and mortality were analyzed separately using Bayesian multilevel linear models. Results : For bladder cancer incidence, the estimated posterior mean relative risks (RRs) were 1.25 (0.92–1.73), 2.11 (1.18–4.22) and 3.01 (1.31–8.17) at arsenic concentrations of 10, 50 and 150 µ g/L, respectively, with posterior probabilities of 92%, 99% and 100%, respectively, for the RRs to be >1. The corresponding RRs for kidney cancer were 1.37 (1.07–1.77), 1.95 (1.44–2.65) and 2.47 (1.74–3.52), with posterior probabilities of 100%. For bladder cancer, the posterior mean mortality ratios were 1.36 (0.35–6.39), 2.92 (1.24–7.82) and 4.88 (2.83–9.03) with posterior probabilities of 72%, 99% and 100%, respectively. Conclusions : The ﬁndings show increased bladder and kidney cancer risks at lower levels of arsenic in drinking water. Given that many people worldwide are exposed to lower levels of arsenic in drinking water, the public health impacts are substantial.


Introduction
Arsenic (As) is a toxic metalloid that occurs naturally in the Earth's crust.Human exposure to As involves multiple pathways [1][2][3][4][5][6][7], with drinking water being the primary route of exposure for the majority of highly exposed populations [2,7,8].Inorganic As in drinking water is classified as a Group 1 carcinogen by the International Agency for Research on Cancer (IARC), constituting a significant public health issue and impacting millions of people worldwide [1].Various acute and chronic morbidities and malignancies have been observed in populations exposed to high concentrations of inorganic As in drinking water, particularly in arsenic-endemic areas in northern Chile, Taiwan, Argentina and Bangladesh [9][10][11].
Most ingested As is excreted predominantly in urine through the urinary system [12]; thus, urinary tract organs are a key target for As-induced carcinogenesis.Urinary tract cancers, comprising primarily cancers of the urinary bladder and kidney, are a major consequence of long-term exposure to inorganic As.Many epidemiological studies involving highly exposed populations have shown strong associations and dose-response relationships between As in drinking water and bladder cancer and potential associations with kidney cancer [13][14][15][16].The carcinogenic effect of inorganic As is further exacerbated when As exposure is combined with cigarette smoking, which is itself an independent risk factor for urinary tract cancers [17,18].Although the exact mechanism of As-induced carcinogenesis is poorly understood, several hypotheses have been postulated, mostly dependent on the mode of As metabolism in the human body once the metalloid is ingested through drinking water.In humans, As compounds may be associated with several tumorigenesis mechanisms, including modifying epigenetic regulation through inhibiting activities of DNA methyltransferase enzymes and the oxidative methylation of arsenite, direct or indirect damage to DNA, the alteration of cell proliferation and oxidative stress [19][20][21].Arsenic might also exert genotoxic effects through direct and indirect damage to DNA, through the disruption of the DNA repair processes and could interfere with signal transduction pathways and promote immunosuppression [19].Specifically for bladder cancer, As has been observed to interrupt cellular homeostasis through stem cell activators, reprogramming healthy epithelial cells to adopt a malignant phenotype [22].
While the carcinogenic risk at high levels of As in drinking water is well established for urinary tract organs [23][24][25][26][27][28][29][30], there remains uncertainty around the risk at lower levels of exposure, particularly at levels around the current World Health Organization (WHO) maximum acceptable concentration (MAC) of 10 µg/L [31,32].Most epidemiological studies to date that have established strong associations and dose-response relationships typically report on areas of severe exposure where levels of As in drinking water range from 150 to over 1000 ug/L.However, at lower levels of exposure (<150 µg/L), the extent to which health effects may develop remains unclear [31,32].
Previously, we conducted a systematic review and meta-analysis of over 30 years of epidemiological studies on As exposure and bladder and kidney cancer outcomes.The findings showed that exposure to As levels as low as 10 ug/L could double the risk of bladder cancer or, at the very least, increase it by 40%, providing evidence of health effects around the WHO MAC [33].Given the substantial public health consequences associated with exposure to As in drinking water and the potentially large number of people worldwide exposed to lower concentrations of As in drinking water, there may be a need to reassess the provisional WHO advisory limit.Currently, the limit is a provisional guideline related to the capacity of treatment options for individual households [8].
Since our last publication, several studies have been conducted examining associations between As exposure, including lower As levels, and urinary tract cancers.Combining evidence from our previous review with the new studies ensures that the analysis reflects the most current evidence, provides more accurate estimates of the magnitude of the relationship and supports a better understanding the health impacts of As, particularly at lower levels of exposure.Additionally, we updated our analytical approach by applying Bayesian multilevel modeling, which provides more robust results for the association between As exposure and bladder and kidney cancer outcomes.This current work aims, therefore, to summarize and update the evidence from our previous systematic review.Adding the most recent 10 years of data, we present a combined review and meta-analysis to include over 40 years of evidence from multiple studies, examining a continuous range of As exposure from which to better assess and predict cancer risks associated with varying levels of arsenic in drinking water.

Eligibility Criteria
The review protocol was registered on the PROSPERO website (protocol reference #CRD42022381522) [34].Given our previous review included studies published prior to January 2013, the current search results included publications between January 2013 and February 2023.This systematic review included observational epidemiological studies such as ecological, cohort or case-control studies.The following criteria for the inclusion of studies in the review were applied: (1) studies among human participants; (2) arsenic, as the exposure of primary interest, in drinking water, toenail or urine; (3) reported urinary tract cancer incidence, mortality rates and relative effect size estimates (e.g., relative risk, odds ratio or hazard ratio) and their corresponding variability measures; (4) published in the English language and (5) peer reviewed publications.

Electronic Searches
For this review, the original search strategy for the MEDLINE database was updated by including relevant free-text keywords and medical subject heading (MESH) terms [33].In brief, the search strategy was developed based on the key research question concepts in accordance with the PICO format: population-all populations; exposure-various concentration levels of arsenic in drinking water (also arsenic measurements in urine and toenail); outcome-incidence and mortality of urinary tract cancers, including urinary bladder cancer and kidney cancer; and comparator-people not exposed to arsenic or general population.The search included studies that contained keywords or MESH terms such as "arsenic", "bladder cancer", "kidney cancer", "urinary tract cancer", "carcinoma, renal cell", "water", "well water", "water supply", "toenail" and other relevant key terms (Figure S1).
The MEDLINE search strategy was modified and adapted to the specifics of other bibliographic databases considering their search methods, the use of thesauri and search operators.A comprehensive search for relevant published studies was performed in the MEDLINE (Ovid; Table S1), Embase (Ovid; Table S2), Web of Science Core Collection (Table S3), Scopus (Table S4) and Google Scholar (Table S5) bibliographic databases.Adapting the MEDLINE search strategy, additional searches were undertaken for grey literature in the Agricultural & Environmental Science Collection, Europe Pre-print Database and Open Access Theses and Dissertations (Table S6).Lastly, a hand search was performed for relevant studies, and authors were contacted if a full text of a study was not available or important information in a study was missing.

Study Selection
After importing all retrieved information from electronic searches into the Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia), two independent reviewers (AA and BA) performed duplicate screening and retrieved title and abstract records for inclusion in the review, applying the eligibility criteria for studies.After selecting studies based on the title and abstract screening, the full texts of the pre-selected studies were obtained.Similarly, two reviewers (AA and BA) independently screened the full-text studies for inclusion by applying the eligibility criteria.Any disagreements during titles and abstracts or full-text screenings were resolved between two reviewers (AA and BA).If consensus was not reached, a third and a fourth reviewer (NSJ and TJBD) resolved the disagreement.The inclusion and exclusion process of studies in the review is illustrated in the PRISMA flow diagram [35] (Figure 1).

Data Extraction
The Data Extraction and Assessment Template created by the Cochrane Public Health Group was used to guide us in developing and piloting a data extraction form [36]. Using Covidence, two reviewers (AA and BA) independently abstracted data from the selected studies.Any disagreements in the extracted data between the two reviewers (AA and BA) were resolved through discussions.If no consensus was reached, a third and a fourth reviewer (NSJ and TJBD) resolved any disagreement.The following study information was abstracted: study-level characteristics (i.e., study design, location, publication year, funding source, inclusion or exclusion criteria, number of participants, participant recruitment/selection, number of cases, missing data and study follow-up time), exposure characteristics (i.e., individual or group arsenic exposure assessment, arsenic concentrations), outcome definition and measurement (i.e., type of outcome, International Classification of Diseases code, outcome measures-incidence or mortality), absolute and/or relative risk estimates (i.e., mortality rate, standardized mortality ratio (SMR), hazard ratio (HR), odds ratio (OR) and relative risk (RR) and their corresponding confidence intervals, whether confounders or effect modifiers were determined and adjusted in the analysis).If study data were not present, the study authors were contacted requesting missing information.Note that some studies reported on more than one cancer site.* ongoing study, wrong comparison group, not peer-reviewed, not in English, data missing or included in the previous review † Data from the previous review [33].

Data Analysis
Study inclusion for meta-analyses followed the same criteria as those reported in Saint-Jacques et al. [33].Briefly, studies providing point estimates of As concentrations in drinking water were included, and those reporting annual average As ingestion, cumulative As exposure, duration of artesian water consumption and As levels in urine were excluded.Risk estimates from this current review (6 studies) were combined with those included in our previous review (16 studies; Figure 1), allowing us to examine associations over a broader and more continuous range of concentrations and with a larger number of studies.
Risk estimates for bladder cancer mortality (11 studies) were analyzed separately from risk estimates for bladder cancer incidence (9 studies).Studies reporting risk estimates on incidence for kidney cancer (3 studies) were also analyzed separately.
Bayesian hierarchical/multilevel models were used to estimate bladder and kidney cancer incidence relative risk (RR, OR, HR) and the bladder cancer standardized mortality ratio (SMR) in relation to arsenic exposure.Overall, the Bayesian approach considers both the data and model parameters as random variables, in contrast to frequentist methods.By treating model parameters as unknown random variables, it becomes possible to specify prior distributions for these.The joint prior probability density function for all the model parameters was then combined with the model likelihood function to obtain the joint posterior probability density function.This posterior distribution reflects the updated beliefs about the parameters after incorporating the observed data [37].The approach specified a Gaussian observation submodel for log(Risk) with corresponding standard errors obtained from publications, also known as a random-effects meta-analysis model [38].For bladder cancer, the models assumed random slopes and intercepts (Figures S2 and S3), while for kidney cancer, due to the small number of studies, a random intercept model was used (Figure S4).Posterior mean relative risk estimates, their corresponding 95% credible intervals (CrI) and exceedance probabilities were obtained at 5, 10, 20, 50, 100 and 150 µg/L As concentrations in drinking water.
The analyses were performed using the "brms" R package [39], a front end to the STAN computational platform [40].Non-informative priors were used for model parameters.For each model, four chains of 5000 MCMC (Markov Chain Monte Carlo) samples were run.The first 2500 samples were discarded as the burn-in.Convergence of the MCMC sequences to the posterior distributions was assessed using an R-hat diagnostic statistic (R-hat = 1), effective sample size statistics (>1000) and examining trace plots of simulations.All analyses were performed using R 4.2.2 version [41].

Study Characteristics
The electronic searches generated a total of 5556 records.After removing 2684 duplicate records, 2872 papers were retrieved for title and abstract screening, of which 144 studies were retained for full text review (Figure 1).A total of 110 reports were excluded for a variety of reasons, see Figure 1 for details.In total, 34 original studies met the inclusion criteria: 23 case-control, six ecological and five cohort studies.Of these, 16 studies reported on bladder cancer, nine on kidney cancer and nine on both cancers and other urinary tract cancers.Six studies met the criteria to be included in the meta-analyses: two reported on bladder cancer mortality, one on bladder cancer incidence, two on kidney cancer incidence and one reported incidence for both bladder and kidney cancers.Study regions included: northern Chile; Nova Scotia, Canada; rural USA; Taiwan; Bangladesh and Italy.

Risk of Bias Assessment
The risk of bias of the included studies varied in terms of the method of As exposure measurement, analytical approach, including controlling for potential confounding (particularly for tobacco smoking), handling of missing data and outcome measurement.Thirteen studies reported measures of As levels in well or tap drinking water, six studies reported cumulative As intake and 15 studies measured As concentrations in urine (Tables 1-6).The As concentrations in drinking water ranged from 0.5 µg/L to 3500 µg/L across the studies (Figure 2).Of the ecological studies, one estimated As exposure through linking residential history to previously collected individual-level measurements from wells [42], while the remaining studies relied on As exposure concentrations based on aggregate data.All of the cohort and most case-control studies evaluated As exposure either by directly measuring As levels in urine or in tap/well water or through the individual-level estimation of past As exposure based on residential history or the source of drinking water.Two case-control studies used aggregate data to estimate As exposure levels [43,44].Note(s): 1 All ecological studies assessed As exposure at the group level. 2 ICD = International Classification for Disease for cancer site abstracted, which included bladder and urothelial/transitional cell carcinoma of the bladder.Transitional cell carcinoma of the renal pelvis often share the same etiology as bladder cancer and, as such, were treated as bladder within the meta-analysis as recommended by IARC [9].* Study included in meta-analysis.‡ Confidence intervals were derived from Supplementary Materials.† Urinary tract urothelial carcinoma.§ Bladder cancer.N/A = not reported.Note(s): 1 ICD = International Classification for Disease for cancer site abstracted, which included bladder and urothelial/transitional cell carcinoma of the bladder.Transitional cell carcinoma of the renal pelvis often share the same etiology as bladder cancer and, as such, were treated as bladder within the meta-analysis as recommended by IARC [9].Note(s): * Study included in meta-analysis. 1 ICD International Classification of Disease.‡ Age-adjusted incidence rates per 100,000 person-years; † Renal cell carcinoma (RCC) and transitional cell carcinoma (TCC).Three of the six ecological studies accounted for tobacco smoking, either directly adjusting for smoking [45,49] or controlling for a proxy variable for tobacco smoking, such as social and material deprivation indices [42].Nineteen of 23 case-control studies adjusted for tobacco smoking [44,50,69,72], two of which also controlled for second-hand smoking exposure [51] or education [52].Five cohort studies accounted for tobacco smoking [63][64][65][66]75], one of which controlled for socioeconomic status as a surrogate for smoking [66].

Ecological Studies
Three of the six ecological studies reported on bladder cancer incidence, two of which observed a significant increase in risk at lower concentrations of arsenic (Table 1).Saint-Jacques et al. [42] and Krajewski et al. [49] reported an increased risk at As concentrations below the current WHO MAC (10 ug/L).Mendez et al. [45] found an increased risk per unit increase in the logged As concentration.
The ecological studies reporting on bladder cancer mortality were all conducted in the arsenic-endemic Region II area of Chile (Table 1).Of these, Roh et al. [46] found that an increased bladder cancer mortality was associated with an early age at first exposure to arsenic in drinking water, with the highest risk being found in exposures around birth (SMR = 16.0,95% CI: 10.3-23.8).In a separate study, Smith et al. [47] confirmed that an increased mortality risk from bladder cancer persisted in both males and females even up to 40 years after As exposure reduction.Lopez et al. [48] reported increased mortality rates in the arsenic-endemic area than in the rest of Chile (mortality rate ratio = 5.5, 95% CI: 5.2-5.9).

Case-Control Studies
All 14 case-control studies reporting on the relationship between As exposure and bladder cancer focused on incidence (Table 2).Eight were from Taiwan and mostly used urinary As concentrations as the measure of exposure.A higher concentration of urinary As was consistently associated with a higher risk of bladder cancer incidence in Taiwan [50,52,53,59,60,62].Wu et al. [52] reported increased odds of developing bladder cancer at urinary As concentrations ≥ 20.4 µg/g creatinine (OR = 4.68, 95% CI: 3.06-7.14),with the highest risk being found in those who smoked and were exposed to second-hand smoking at urinary As concentration ≥ 15.4 µg/g creatinine (OR = 10.8, 95% CI: 5.16-22.69).One study from Taiwan reporting on drinking water As concentrations observed a significant increase in bladder cancer risk (OR = 2.4, 95% CI: 1.6-3.4) at high As levels (midpoint levels, 745 µg/L vs. 175 µg/L); the effect increased with smoking (OR = 5.7, 95% CI: 3.1-10.3)[43].
In the case-control studies from the USA, Baris et al. [57] and Koutros et al. [61] found no evidence of increased bladder cancer risk at low levels of As in drinking water.However, when the average As exposure was lagged by 40 years, a moderate association with bladder cancer incidence was found in the 'highest' As level (>8.7 µg/L) considered (OR = 1.49, 95% CI: 0.85-2.61).Both current and former smokers had a similar increased risk compared to non-smokers exposed to the same higher level of cumulative As intake (OR = 1.6, 95% CI: 0.9-3.0).Mostafa et al. (2015) found no consistent dose-response association for transitional cell carcinoma of the bladder with increasing As concentrations [56].
The remaining case-control studies reporting on bladder cancer came from the wellknown As-endemic areas in Regions I and II in northern Chile.Ferreccio et al. 2013(a) found an increased odds of bladder cancer incidence with increasing As concentrations in drinking water (OR = 4.44, 95% CI: 2.75-7.15)and a much higher OR in smokers for the highest level of As (>260 µg/L) in drinking water (OR = 15.30,95% CI: 6.75-34.67)[51].Melak et al. [54] also found a similar positive association when considering the percentage of monomethylarsonic acid present in urine.Similarly, Steinmaus et al. [55] observed increasing odds of bladder cancer incidence from exposure to high levels of As in drinking water in relation to age at first exposure, with the highest ORs being found in those exposed in utero or as children (OR = 8.11, 95% CI: 4.31-15.25).

Cohort Studies
Four cohort studies reported on the relationship between As exposure and bladder cancer incidence or mortality (Table 3).Both Yang et al. [63] and Tsai et al. [65] measured bladder cancer incidence in an As-exposed cohort from 18 villages in northeast Taiwan.Yang et al. [63] consistently observed an increased risk of developing bladder cancer at As levels ≥ 100 µg/L in well water; the observed risk in the whole cohort was 8.71 (95% CI: 2.49-30.48),and, separately, in non-smokers it was 4.64 (95% CI: 0.89-24.2) and in smokers it was 16.5 (95% CI: 2.12-128.6).Similarly, Tsai et al. [65] also reported increased risk of bladder cancer incidence with higher levels of cumulative exposure to As in well water (HR = 11.38,95% CI: 1.48-87.74)and urinary As concentrations (HR = 2.78, 95% CI: 0.75-10.39).
Two separate studies from Italy, D'Ippoliti et al. [64] and Nuvolone et al. [66], did not find any excess risk of mortality from bladder cancer with increased cumulative As exposure or As in drinking water, respectively.However, the analyses in both studies were limited by an incomplete adjustment for important individual risk factors such as smoking.

Ecological Studies
Four ecological studies reported on kidney cancer, with half of these reporting on incidence (Table 4).Saint-Jacques et al. [42] observed increased kidney cancer incidence in Nova Scotia, Canada, at As levels ≥ 5 µg/L in drinking water for males and both sexes combined.Similarly, Krajewski et al. [49] observed a strong association between the incidence of kidney cancer with increasing quartiles of cumulative As exposure in the USA.
The two ecological studies on kidney cancer mortality were from the arsenic-endemic Region II area of Chile.Unlike for bladder cancer in the same study, Roh et al. [46] did not find a significant association between kidney cancer mortality and age at first exposure to high levels of As in drinking water.Instead, those exposed very early in life (at birth) had the lowest risk for kidney cancer (SMR = 1.3, 95% CI: 0.6-2.5).However, Smith et al. [47] found an elevated risk of mortality from kidney cancer that persisted in both males and females even up to 40 years after As exposure reduction.

Case-Control Studies
All nine case-control studies reporting on the relationship between As exposure and kidney cancer focused on incidence, and these were mostly from Taiwan (n = 7) (Table 5).Six studies [69-74] used urinary As concentrations as the metric of exposure and observed a consistent dose-response relationship between the incidence of kidney cancer and the concentration of urinary As.Similarly, Chen et al. [44] found a positive relationship between the incidence of kidney cancer and an increasing arseniasis grade, reporting rates of 3.92, 4.71, 5.31 and 8.35 per 100,000 person-years for arseniasis grades 0, 1, 2 and 3, respectively.

Cohort Studies
Three cohort studies reported on the relationship between As exposure and kidney cancer incidence or mortality (Table 6).Tsai et al. [65] explored kidney cancer incidence in an As-exposed cohort from 18 villages in northeast Taiwan and found an increased risk at higher levels of As in well drinking water (HR = 1.66, 95% CI: 0.53-5.15)and cumulative exposure to As in well water (HR = 2.37, 95% CI: 0.72-7.82).D'Ippoliti et al. [64] reported a similar magnitude of effect for kidney cancer mortality in both males (HR = 1.93, 95% CI: 0.63-5.96)and females (HR = 3.03, 95% CI: 0.37-25.22)using a cumulative As exposure metric.However, using urinary As concentrations as the metric of exposure, García-Esquinas et al. [75] did not find any excess mortality from kidney cancer; rather, the hazard ratios decreased with increasing urinary As levels (HR = 0.44, 95% CI: 0.14-1.40).

Meta-Analyses
Twenty-two studies were considered in the meta-analyses (16 from the previous review).Among these, nine studies reported on bladder cancer incidence, 11 on bladder cancer mortality and three on kidney cancer incidence.All studies used As in drinking water as the metric of exposure.Similar to the previous review, we found that exposure to increasing levels of As in drinking water was associated with an increased risk of bladder cancer incidence and bladder cancer mortality (Figures 3-5).For bladder cancer, the posterior mean relative risks estimated at 10, 50 and 150 ug/L were 1.25 (95% CrI: 0.92-1.73),2.11 (95% CrI: 1.18-4.22)and 3.01, (95% CrI: 1.31-8.17),respectively (Table 7).Exceedance probabilities for the estimated risks to be above 1 were consistently above 90%, suggesting a significant excess risk from exposure to arsenic in drinking water (Table 7; Figure 5A).The corresponding effect sizes for bladder cancer mortality were 1.36 (95% CrI: 0.35-6.39),2.92 (95% CrI: 1.24-7.82)and 4.88 (95% CrI: 2.83-9.03).The probability for the risk of dying began to increase at arsenic exposure levels between 5 and 10 ug/L and became significantly higher thereafter (Figure 5B).For kidney cancer, the posterior mean relative risks estimated at 10, 50 and 150 ug/L were 1.37 (95% CrI: 1.07-1.77),1.95 (95% CrI: 1.44-2.65)and 2.47 (95% CrI: 1.74-3.52),respectively.There was a 91% probability for the risk to exceed 1 at levels of arsenic exposure as low as 5 ug/L (Figure 5C).Overall, exceedance probabilities above 80%, which are generally considered to be consistent with a significant excess risk within a Bayesian framework [76,77], were observed at arsenic levels of 4 ug/L, 8 ug/L and 13 ug/L for kidney and bladder cancer incidence and bladder cancer mortality, respectively (Figure 5).

Summary of Findings
This work updated evidence from our previous systematic review on the relationship between long-term exposure to As in drinking water and urinary bladder and kidney cancers.An additional 34 studies published between January 2013 and February 2023 were found.Among these, 16 studies from the USA, Chile and Taiwan reported statistically significant associations between As exposure and the risk of developing bladder cancer or dying from the disease (seven studies assessed As concentrations in drinking water, seven studies measured As levels in urine and two studies evaluated cumulative As exposure).Eleven studies, conducted in the USA, Bangladesh, Chile and Taiwan, reported a significant relationship between As exposure and the risk of developing kidney cancer or dying from the disease (four studies measured As levels in drinking water, six measured As levels in urine and one measured cumulative As exposure).
Although controlling for confounding is challenging in ecological studies, half of the included studies accounted for tobacco smoking, either directly or indirectly (e.g., using a proxy variable) [42,45,49].All but four of the case-control studies and all cohort studies accounted for tobacco smoking behavior.Smoking is an independent risk factor for urinary tract cancers, and not adjusting for smoking may overestimate the magnitude of the association between As exposure and urinary tract cancers [90].Indeed, studies that did not account for smoking [44,50,69] had a stronger dose-response relationship than studies that controlled for smoking.In addition, over 70% of the studies did not report information concerning missing data or excluded substantial proportions of the study participants due to missingness.Without properly testing the 'missing completely at random' assumption, results based on a complete-case analysis should be interpreted with caution as these effect size estimates might be biased.
As exposure levels in most of the ecological studies and some of the case-control studies were determined based on geographic or group-level measurements, and their results might have been prone to nondifferential misclassification bias.Risk estimates from the studies evaluating As exposure in well or tap drinking water were measured within a limited range of As levels.Interestingly, these estimates varied considerably across the studies, within geographic areas or even in regions with the same As concentrations.These differences could be explained by variability in exposure (e.g., duration of exposure and As species), exposure measurement methods (e.g., historical data, longitudinal assessment, and measurement technique), individual factors (e.g., genetic susceptibility, lifestyle, and behavioral factors) and study methodological approaches (e.g., study design, participant sampling and data analysis).

Quantitative Synthesis of the Association between As Exposure and Bladder and Kidney Cancers
The association between As exposure in drinking water and bladder and kidney cancer outcomes were quantitatively assessed over a broad and continuous range of As concentrations.After controlling for study differences, we observed positive non-linear relationships of increased risks of developing bladder or kidney cancers or dying from bladder cancer with increasing As exposure levels; similar associations were reported in our previous review [33].The results from the Bayesian random-effects meta-analysis models suggested that exposure to 10 µg/L of As in drinking water was associated on average with a 25% excess bladder cancer risk (posterior mean RR = 1.25, 95% CrI: 0.92-1.73).Exposure to As levels of 50 and 150 µg/L resulted on average in a doubling (posterior mean RR = 2.11; 95% CrI: 1.18-4.22)and tripling of the risk (3.01, 95% CrI: 1.31-8.17),respectively.The risk estimates for bladder cancer obtained from bootstrap resamplings in our previous review were of a similar magnitude (predicted RR = 1.4,95% CI: 0.35-4.0;2.3, 95% CI: 0.59-6.4;and 3.1, 95% CI: 0.80-8.9)for As levels at 10, 50 and 150 µg/L, respectively) [33].Additionally, exposure to 10 µg/L of As in drinking water was associated on average with a 36% excess bladder cancer mortality (posterior mean RR = 1.36, 95% CrI: 0.35-6.39)and with a tripling (posterior mean RR = 2.92, 95% CrI: 1.24-7.82)and quintupling (posterior mean RR = 4.88, 95% CrI: 2.83-9.03) of risk at As levels of 50 and 150 µg/L, respectively.The mortality estimates from bootstrap resamplings reported in our previous review were moderate (predicted SMR = 1.0, 95% CI: 0.15-38; 1.7, 95% CI:0.49-40; and 2.2, 95% CI:0.54-41) for As levels at 10, 50 and 150 µg/L, respectively) in comparison to the current findings.The updated findings suggest a stronger relationship between lower concentrations of As in drinking water and bladder cancer mortality.Our reviews, which now summarize 40 years of epidemiological data, support evidence of an increased cancer risk around the current WHO MAC guideline limit (10 µg/L).
The findings reported here were in contrast with those from a recent systematic review, which reported no association between exposure to low-level As in drinking water (<150 µg/L) and bladder cancer risk [91].However, it is important to note that the inconsistent findings were from a review of studies using different inclusion and exclusion criteria (e.g., excluding ecological studies and studies reporting at least two exposure categories), which might have resulted in reduced statistical power.In addition, the other review combined various As exposure measurements (e.g., cumulated, average lifetime exposure and yearly mean exposure) and diverse exposure categories, which could potentially result in the misclassification of exposure and likely trend the estimated risk towards the null.Another systematic review evaluated the relationship between As exposure in drinking water and the risk of bladder cancer by applying a meta-regression analysis [15].Similar to our findings, they reported a non-linear dose-response relationship in support of an increased risk of bladder cancer with increasing As exposure levels.However, in their subgroup analyses, the authors noted that although the dose-response relationship was sustained at As levels < 100 µg/L, the association was not statistically significant.
In this study, we also modeled the risk of developing kidney cancer at various As levels in drinking water.The modeled estimates suggested that exposure to 10 µg/L of As in drinking water was associated on average with a 37% excess kidney cancer risk (posterior mean RR = 1.37, 95% CrI: 1.07-1.77),with the risk doubling (posterior mean RR = 1.95, 95% CrI: 1.44-2.65)and almost tripling (posterior mean RR = 2.47, 95% CrI: 1.74-3.52)at As levels of 50 and 150 µg/L, respectively.These results were aligned with our previous review findings, where we reported a dose-response relationship between increased kidney cancer mortality and increasing As concentrations in drinking water [33].The results were also consistent with those of a recent systematic review that reported a significant increase in kidney cancer incidence and mortality in people exposed to drinking water contaminated with an As level > 100 µg/L [92].Nonetheless, although we found an association at 10 µg/L of As in drinking water, studies reporting on lower concentrations of As exposure are lacking.Additional studies are required to confirm the excess risk of developing and dying from kidney cancer observed in our review and in the previous review.
Findings from this updated review support our previously published results [33], indicating that individuals exposed to As concentrations of 50 and 150 µg/L have two and three times the risk of developing bladder cancer, respectively.The updated evidence suggests that people exposed to As in drinking water at a concentration of 10 µg/L have a 25% increased bladder cancer risk; this is slightly lower than what has been previously reported.Also, in the current meta-analysis, we found a stronger association between As concentrations in drinking water and bladder cancer mortality compared to the previous review.For As concentrations of 50 and 150 µg/L, the risk increased three-fold and five-fold, respectively.Interestingly, our analysis also revealed that exposure to an As concentration of 10 µg/L in drinking water resulted in a 36% excess bladder cancer mortality, a finding that was not observed in our previous review.Kidney cancer incidence risk was not assessed in the previous review as none of the identified studies met the inclusion criteria for the meta-analysis.However, this updated review suggests that people exposed to 10 µg/L of As in drinking water have, on average, a 37% excess risk of kidney cancer.Notably, the risk substantially increased, with it doubling and nearly tripling at As levels 50 and 150 µg/L, respectively.
This systematic review and meta-analysis provides evidence of an excess risk of bladder and kidney cancer at lower concentrations of As in drinking water, particularly around the WHO provisional MAC guideline limit of 10 µg/L.Recent modelling of As concentrations in groundwater suggests that 94 to 220 million people might be exposed to elevated levels of As (>10 µg/L) in drinking water [93].Given that this large number of people may be exposed to these lower levels worldwide [94], the public health consequences of arsenic in drinking water are substantial.
Arsenic mitigation in drinking water is an important public health priority.Various mitigation measures, such as accessing alternative water sources and utilizing centralized and non-centralized drinking water treatment technologies, have proven to be effective in reducing exposure to arsenic [95,96].The successful implementation of mitigation strategies often necessitates substantial behavioral change and community acceptance alongside the enforcement of public policies.Although there is no evidence indicating that lower socioeconomic status groups disproportionately reside in arsenic-affected areas, disparities in exposure arise from differences in rates of protective behaviors and psychological factors favoring such behaviors [97][98][99].Measures such as the universal screening of private well water quality and state laws mandating arsenic testing during real estate transactions can partially address the socioeconomic disparity in water testing [100,101].
Moreover, the implementation of effective mitigation strategies poses financial challenges as it requires investments in the installation, operation and maintenance of treatment systems.This can be particularly daunting for communities, regions or countries with limited resources [102].Therefore, it is crucial to invest in the research and development of cost-effective technologies, capacity building, policy reforms and community engagement to achieve sustainable and long-term solutions for reducing arsenic concentrations in drinking water.

Strengths and Weakness of the Review
This review has several strengths.First, it drew from multiple bibliographic databases to find the related published and grey literature on the topic.The search strategy followed established robust criteria, applying the PRESS Peer Review of Electronic Search Strategies Guideline and the PRISMA-S checklist [103,104].A broad sensitive search strategy was adopted by not restricting the review to certain study designs; rather, we included all epidemiological designs-ecological, case-control and cohort studies-allowing all relevant findings to be qualitatively and quantitatively synthesized.Second, similar to our previous review, this update quantified the risks of developing or dying from urinary tract cancers over a wide range of As exposures, including lower levels (<150 µg/L), for which there is a clear knowledge gap.Third, the meta-analysis was performed using a Bayesian framework, which is known to be a robust analytical approach that accounts for the uncertainty around the heterogeneity of variance [105,106].Fourth, conducting the meta-analysis on studies included in the current (6 studies) and previous (16 studies) reviews allowed for an inference over a broader and more continuous range of As concentrations and an increased statistical power.Finally, the independent analyses of incidence and mortality outcomes likely minimized biases related to exposure misclassification and ascertainment.Mortality data are prone to misclassification (e.g., inconsistent methods used to determine cause of the death) relative to incidence data.This review has some limitations.First, publication bias may be present in the review, since we only included studies published in peer-reviewed scientific journals, which may favorably publish studies with significant results.However, such a bias should be minimal, as almost a third of the included studies reported non-significant findings, and more than half of the excluded studies found statistically significant associations.Second, we piloted the ROBINS-E tool [107] that was specifically developed to assess the risk of bias in non-randomized studies.However, the tool did not account for specific methodological and reporting characteristics of As-related exposure studies (e.g., various types of exposure matrix, challenges with measuring and defining or categorizing As exposure).As such, we evaluated studies based on simple core methodological aspects (e.g., whether a study adjusted for possible confounding factors, whether As exposure was measured at the individual or group level, whether missing data were present and how these were handled).Third, the results from the meta-analysis performed on kidney cancer risk should be interpreted with caution given that the estimates were modeled based only on three studies [106].Lastly, due to methodological heterogeneity (e.g., study designs and analytical approaches) among the studies with missing data and among the studies without missing data (or with near-complete data), we were not able to distinguish the direction and magnitude of bias related to missing data.

Conclusions
The present review provides consistent evidence of an association between moderate and higher As concentrations (≥50 µg/L) in drinking water with bladder and kidney cancer incidence and bladder cancer mortality.In addition, it shows an increased bladder and kidney cancer risk at lower levels of As around the current WHO provisional MAC of 10 µg/L.People exposed to 10 µg/L of As in drinking water may be at a 25% or 36% increased risk of developing or dying from bladder cancer, respectively, and a 37% increased kidney cancer risk.Moreover, the pooled analysis showed that people drinking water contaminated with As at concentrations as low as 5 µg/L could be at an excess risk of developing kidney cancer.However, as these results were derived from a small number of studies, it would be advisable to confirm the pattern once additional studies reporting on kidney cancer incidence become available.Additional research is necessary to further assess the relationship between urinary tract cancers and lower concentrations of As in drinking water, including exposure close to, or below, the current regulatory limits.In addition, given the challenges associated with the current As detection methods, future research should focus on development of cost-effective and reliable solutions that are applicable in real-world setting.This is particularly important given the fact that several millions of people worldwide are regularly drinking water containing naturally occurring As around the current WHO 10 µg/L MAC.The public health consequences of exposure to As in drinking water are both substantial and widespread.We suggest that the current policies and recommendations regulating As concentrations in drinking water should be revised, particularly through lowering the current WHO provisional MAC guideline limit.

Figure 1 .
Figure 1.PRISMA flowchart illustrating the study selection process for the review and meta-analysis.Note that some studies reported on more than one cancer site.* ongoing study, wrong comparison group, not peer-reviewed, not in English, data missing or included in the previous review † Data from the previous review[33].
Note(s): * Study included in meta-analysis.1 ICD International Classification of Disease.

Figure 5 .
Figure 5. Exceedance probability profile for estimated risk as a function of varying levels of arsenic in drinking water.Results are shown for: (A) bladder cancer incident relative risk (RR); (B) standardized bladder cancer mortality ratio (SMR); and (C) kidney cancer incident relative risk (RR).

Table 1 .
Summary of results from ecological studies reporting on arsenic (As) exposure and the risk of bladder cancer.

Table 2 .
Summary of results from case-control studies reporting on arsenic exposure and the risk of bladder cancer.
(Period of As water samples collection not reported.Participants used artesian well water for > 10 years when recruited.The interview included history of well-water consumption, residential history, lifestyle factors.As water concentration in Peimen, Hsuechia, Putai and Ichu townships ranged between 350-1140 µg /L (BFD-endemic area), in surrounding areas was <350 µg /L.
(Residential history was collected; each city/town was linked to a water As measurement for that city/town so that an As concentration could be assigned to each year of each subject's life.)
≥300(3535 wells were sampled (1998-1999) in 61/64 districts in Bangladesh, 27% of hand-pumped tube wells contained >50 µg/L of As.The mean As concentration data was linked to where participants lived during a biopsy.)

Table 2 .
Cont.Study included in meta-analysis.§ Upper urinary tract urothelial carcinoma; £ Transitional cell carcinoma; + Urinary tract urothelial carcinoma ; † Former smokers; ‡ Current smokers.ICD codes were not reported for any of the studies.

Table 3 .
Summary of results from cohort studies reporting on arsenic (As) exposure and the risk of bladder cancer.

Table 4 .
Summary of results from ecological studies reporting on arsenic (As) exposure and the risk of kidney cancer.

Table 4 .
Cont.Study included in meta-analysis.1Allecological studies assessed As exposure at the group level. 2 ICD = International Classification of Disease.

Table 5 .
Summary of results from case-control studies reporting on arsenic (As) exposure and the risk of kidney cancer.

Table 6 .
Summary of results from cohort studies reporting on arsenic (As) exposure and the risk of kidney cancer.

Table 7 .
Modeled mean risk, credible intervals and associated exceedance probabilities for bladder and kidney cancer outcomes estimated at varying levels of arsenic in drinking water.