Provision of folic acid for reducing arsenic toxicity in arsenic‐exposed children and adults

Abstract

Background

Arsenic is a common environmental toxin. Exposure to arsenic (particularly its inorganic form) through contaminated food and drinking water is an important public health burden worldwide, and is associated with increased risk of neurotoxicity, congenital anomalies, cancer, and adverse neurodevelopment in children. Arsenic is excreted following methylation reactions, which are mediated by folate. Provision of folate through folic acid supplements could facilitate arsenic methylation and excretion, thereby reducing arsenic toxicity.

Objectives

To assess the effects of provision of folic acid (through fortified foods or supplements), alone or in combination with other nutrients, in lessening the burden of arsenic‐related health outcomes and reducing arsenic toxicity in arsenic‐exposed populations.

Search methods

In September 2020, we searched CENTRAL, MEDLINE, Embase, 10 other international databases, nine regional databases, and two trials registers.

Selection criteria

Randomised controlled trials (RCTs) and quasi‐RCTs comparing the provision of folic acid (at any dose or duration), alone or in combination with other nutrients or nutrient supplements, with no intervention, placebo, unfortified food, or the same nutrient or supplements without folic acid, in arsenic‐exposed populations of all ages and genders.

Data collection and analysis

We used standard methodological procedures expected by Cochrane.

Main results

We included two RCTs with 822 adults exposed to arsenic‐contaminated drinking water in Bangladesh. The RCTs compared 400 µg/d (FA400) or 800 µg/d (FA800) folic acid supplements, given for 12 or 24 weeks, with placebo. One RCT, a multi‐armed trial, compared FA400 plus creatine (3 g/d) to creatine alone.

We judged both RCTs at low risk of bias in all domains. Due to differences in co‐intervention, arsenic exposure, and participants' nutritional status, we could not conduct meta‐analyses, and therefore, provide a narrative description of the data.

Neither RCT reported on cancer, all‐cause mortality, neurocognitive function, or congenital anomalies.

Folic acid supplements alone versus placebo

Blood arsenic. In arsenic‐exposed individuals, FA likely reduces blood arsenic concentrations compared to placebo (2 studies, 536 participants; moderate‐certainty evidence).

For folate‐deficient and folate‐replete participants who received arsenic‐removal water filters as a co‐intervention, FA800 reduced blood arsenic levels more than placebo (percentage change (%change) in geometric mean (GM) FA800 −17.8%, 95% confidence intervals (CI) −25.0 to −9.8; placebo GM −9.5%, 95% CI −16.5 to −1.8; 1 study, 406 participants).

In one study with 130 participants with low baseline plasma folate, FA400 reduced total blood arsenic (%change FA400 mean (M) −13.62%, standard error (SE) ± 2.87; placebo M −2.49%, SE ± 3.25), and monomethylarsonic acid (MMA) concentrations (%change FA400 M −22.24%, SE ± 2.86; placebo M −1.24%, SE ± 3.59) more than placebo. Inorganic arsenic (InAs) concentrations reduced in both groups (%change FA400 M −18.54%, SE ± 3.60; placebo M −10.61%, SE ± 3.38). There was little to no change in dimethylarsinic acid (DMA) in either group.

Urinary arsenic. In arsenic‐exposed individuals, FA likely reduces the proportion of total urinary arsenic excreted as InAs (%InAs) and MMA (%MMA) and increases the proportion excreted as DMA (%DMA) to a greater extent than placebo (2 studies, 546 participants; moderate‐certainty evidence), suggesting that FA enhances arsenic methylation.

In a mixed folate‐deficient and folate‐replete population (1 study, 352 participants) receiving arsenic‐removal water filters as a co‐intervention, groups receiving FA had a greater decrease in %InAs (within‐person change FA400 M −0.09%, 95% CI −0.17 to −0.01; FA800 M −0.14%, 95% CI −0.21 to −0.06; placebo M 0.05%, 95% CI 0.00 to 0.10), a greater decrease in %MMA (within‐person change FA400 M −1.80%, 95% CI −2.53 to −1.07; FA800 M −2.60%, 95% CI −3.35 to −1.85; placebo M 0.15%, 95% CI −0.37 to 0.68), and a greater increase in %DMA (within‐person change FA400 M 3.25%, 95% CI 1.81 to 4.68; FA800 M 4.57%, 95% CI 3.20 to 5.95; placebo M −1.17%, 95% CI −2.18 to −0.17), compared to placebo.

In 194 participants with low baseline plasma folate, FA reduced %InAs (%change FA400 M −0.31%, SE ± 0.04; placebo M −0.13%, SE ± 0.04) and %MMA (%change FA400 M −2.6%, SE ± 0.37; placebo M −0.71%, SE ± 0.43), and increased %DMA (%change FA400 M 5.9%, SE ± 0.82; placebo M 2.14%, SE ± 0.71), more than placebo.

Plasma homocysteine: In arsenic‐exposed individuals, FA400 likely reduces homocysteine concentrations to a greater extent than placebo (2 studies, 448 participants; moderate‐certainty evidence), in the mixed folate‐deficient and folate‐replete population receiving arsenic‐removal water filters as a co‐intervention (%change in GM FA400 −23.4%, 95% CI −27.1 to −19.5; placebo −1.3%, 95% CI −5.3 to 3.1; 1 study, 254 participants), and participants with low baseline plasma folate (within‐person change FA400 M −3.06 µmol/L, SE ± 3.51; placebo M −0.05 µmol/L, SE ± 4.31; 1 study, 194 participants).

FA supplements plus other nutrient supplements versus nutrient supplements alone

In arsenic‐exposed individuals who received arsenic‐removal water filters as a co‐intervention, FA400 plus creatine may reduce blood arsenic concentrations more than creatine alone (%change in GM FA400 + creatine −14%, 95% CI −22.2 to −5.0; creatine −7.0%, 95% CI −14.8 to 1.5; 1 study, 204 participants; low‐certainty evidence); may not change urinary arsenic methylation indices (FA400 + creatine: %InAs M 13.2%, SE ± 7.0; %MMA M 10.8, SE ± 4.1; %DMA M 76, SE ± 7.8; creatine: %InAs M 14.8, SE ± 5.5; %MMA M 12.8, SE ± 4.0; %DMA M 72.4, SE ±7.6; 1 study, 190 participants; low‐certainty evidence); and may reduce homocysteine concentrations to a greater extent (%change in GM FA400 + creatinine −21%, 95% CI −25.2 to −16.4; creatine −4.3%, 95% CI −9.0 to 0.7; 1 study, 204 participants; low‐certainty evidence) than creatine alone.

Authors' conclusions

There is moderate‐certainty evidence that FA supplements may benefit blood arsenic concentration, urinary arsenic methylation profiles, and plasma homocysteine concentration versus placebo. There is low‐certainty evidence that FA supplements plus other nutrients may benefit blood arsenic and plasma homocysteine concentrations versus nutrients alone. No studies reported on cancer, all‐cause mortality, neurocognitive function, or congenital anomalies. Given the limited number of RCTs, more studies conducted in diverse settings are needed to assess the effects of FA on arsenic‐related health outcomes and arsenic toxicity in arsenic‐exposed adults and children.

Plain language summary

Providing folic acid to reduce arsenic toxicity in arsenic‐exposed children and adults

What was studied in this review?

Arsenic is a common environmental toxin that affects over 140 million people worldwide. Long‐term arsenic exposure, through the consumption of arsenic‐contaminated drinking water and food, increases the risk of neurotoxicity (damage to the brain or nervous system), skin lesions, birth defects, cancer, and impaired brain development in children. Folic acid may decrease arsenic toxicity by helping to remove arsenic from the body, thus lowering the amount of arsenic in the blood. This review assessed the effect of giving people folic acid through oral supplements, fortified foods, or both, on arsenic‐associated health outcomes and arsenic toxicity in adults and children.

What is the aim of the review?

The aim of this review was to determine whether giving folic acid to arsenic‐exposed children and adults reduced arsenic toxicity.

Key messages

Folic acid supplementation may reduce blood arsenic concentration, and make it easier to get rid of arsenic through the urine, in arsenic‐exposed adults compared to placebo (i.e. dummy pill). This suggests that taking folic acid supplements may reduce arsenic toxicity in arsenic‐exposed adults.

What are the main results of the review?

The review authors found two randomised controlled trials (RCT), with 822 adults, conducted in Bangladesh, which assessed the effect of taking folic acid supplements on the concentration of arsenic and homocysteine (a marker of inflammation and folate deficiency) in plasma, blood, and urine. Both RCTs were funded by government programs. One of the RCTs also assessed the effects of folic acid supplements plus another nutrient supplement called creatine. Neither of the studies reported data on cancer, all‐cause mortality, neurocognitive function, or congenital anomalies.

The results from these studies suggest that taking folic acid supplements, alone or in combination with other nutrients, might reduce blood arsenic and plasma homocysteine concentrations, and potentially improve urinary arsenic methylation profiles (a measure of arsenic toxicity) in adults who had been exposed to arsenic‐contaminated drinking water, compared to placebo.

We judged both RCTs at low risk of bias. We judged the certainty of evidence as moderate for all outcomes included in the comparison of Folic acid supplements alone versus placebo, and low for all outcomes included in the comparison of Folic acid supplements plus other nutrient supplements versus nutrient supplements alone. We downgraded the certainty of the evidence due to the small number of studies. This means that the results may change with further research.

The review highlights the need for more research evaluating the effects of folic acid on arsenic‐related health outcomes and arsenic toxicity in adults and children.

How up‐to‐date is this review?

The review authors searched for studies that had been published up to September 2020.

Background

Description of the condition

Burden of chronic arsenic exposure

Arsenic is a common environmental toxin, and exists in both organic and inorganic forms. In general, organic arsenic is considered to be less harmful than inorganic arsenic; however, depending on the chemical form, some organic compounds are toxic (ATSDR 2016; NTP 2014), and also may undergo degradation, giving rise to bioavailable inorganic arsenic species (Chávez‐Capilla 2016). Acute poisoning by arsenic is rare, but low‐level chronic exposure in humans, through contaminated drinking water, is common throughout the world. The World Health Organization (WHO) estimates that more than 140 million people from more than 70 countries, including Bangladesh, China, India, Chile, Nepal, and areas of the USA, are chronically exposed to arsenic‐contaminated water at levels exceeding the recommended concentration of 10 µg/L (Ahmed 2006; Chiou 2001; Ghosh 2013; Mazumder 2010; Naujokas 2013; Nielsen 2010; Rodíguez‐Lado 2013; Sanders 2012; Smith 2000; Smith 2002; WHO 2008; Yu 2007). The primary source of contaminated drinking water is groundwater exposed to naturally occurring arsenic‐rich geological formations (IARC 2012; Kim 2011). Rice, a staple food for half of the world’s population, can contain high levels of arsenic, which is derived from the soil of paddy fields (Gilbert‐Diamond 2011; Ma 2008; Melkonian 2013; Sohn 2014; Stone 2008; Zavala 2008a; Zavala 2008b). Water contamination remains an issue despite mitigation efforts, and food grown or produced (or both) in arsenic‐rich environments contributes to human arsenic exposure (Banerjee 2013; Carignan 2016; Davis 2012; Karagas 2016; Kippler 2016; Rose 2007). A high amount of arsenic, in both organic and inorganic forms, may also be present in some species of seaweed (Almela 2002; Brandon 2014; Khan 2015Rose 2007), and in fish, shellfish and other types of seafood (Choi 2010; Seo 2016). The average daily intake of total arsenic (including both organic and inorganic forms) from food and beverages in USA and European populations has been estimated to be in the range of 20 to 300 µg/day (DeCastro 2014; EFSA 2009; EFSA 2014; IARC 2012; IPCS 2001; Kurzius‐Spencer 2014; Lynch 2014).

Health consequences associated with arsenic exposure in adults and children

Arsenic exposure affects almost every organ system in the body, including the brain. Chronic exposure to arsenic has been associated with neurotoxicity in adults, as well as increased risk of cancer, diabetes, cardiovascular disease, kidney disease, anemia, and skin disease (ATSDR 2016; Axelson 1980; Cohen 2013; Di Giovanni 2020; Ettinger 2009; Farzan 2013; IARC 1980; IARC 2012; Moody 2018; Moon 2013; Moon 2017; Naujokas 2013; Navas‐Acien 2008; NTP 2014; Prakash 2016; Sidhu 2015; Sung 2015; Tinkelman 2020; Vahidnia 2007; Wang 2020; Xu 2020; Yuan 2018). Arsenic exposure through drinking water has also been linked to excess adult mortality (Argos 2010a; Sohel 2009; Wu 1989; Yuan 2007; Yuan 2010), and adverse pregnancy outcomes, including preterm delivery, stillbirth, spontaneous abortion, and low birth weight (Ahmad 2001; Almberg 2017; Hopenhayn‐Rich 2000; Howe 2020; Huang 2021; Huyck 2007; Kile 2016; Laine 2015; Milton 2005; Rahman 2007; Shih 2017; Wang 2018b; Zhong 2019). There is also limited evidence from human studies suggesting a link between arsenic exposure and risk of neural tube defects (Demir 2019; DeSesso 2001; Kwok 2006; Mazumdar 2015a; Wang 2019; Wu 2011). Exposure to inorganic arsenic during pregnancy is associated with adverse birth outcomes (e.g. low birth weight, preterm birth (Gilbert‐Diamond 2016; Laine 2015)), larger head circumference in infants (Wai 2020), and poorer neurobehavioral performance and development of newborns (Liang 2020; Nyanza 2021; Vahter 2020; Wang 2018a). Arsenic is immunotoxic, and has been associated with weakened immune response to infection during pregnancy and early life, negatively affecting both humoral and cell‐mediated immunity (Ahmed 2020; Attreed 2017; Heaney 2015; Tsai 2021; Welch 2019; Welch 2020). Animal studies have demonstrated that a high dose of arsenic induces embryotoxicity and congenital anomalies, including neural tube defects (Chaineau 1990; Han 2011; Hill 2008; Hill 2009; Morrissey 1983; Wlodarczyk 2001).

Exposure to arsenic in drinking water during early childhood or in utero is reported to increase subsequent mortality in young adults from lung cancer and other lung diseases (Smith 2006). Children have a higher metabolic rate than adults, to support growth and development, which leads to a greater exposure to arsenic, and a greater sensitivity to the adverse effects of arsenic exposure (Bearer 1995). In the first six months of life, children drink seven times more water per kilogram of body weight than adults. Children aged between one and five years consume three to four times more food per kilogram of body weight than adults (Bearer 1995). Because of these metabolically‐driven differences, exposure to arsenic in children younger than three years of age is estimated to be two‐ to three‐fold higher than adults (EFSA 2009; EFSA 2014; Hojsak 2015; Ljung 2011; Meharg 2008; Rintala 2014; Signes‐Pastor 2016; Sirot 2018). Higher exposure to arsenic in drinking water, or higher blood and urinary arsenic concentrations (or both) have been inversely associated with intellectual function and neuropsychological development in children (Dong 2009; Rodríguez‐Barranco 2013; Rodríguez‐Barranco 2016; Tolins 2014; Tsuji 2015; Von Ehrenstein 2007), and adolescents (Manju 2017; Tsai 2003; Wasserman 2018). Observational epidemiological studies support an adverse effect of arsenic on neurodevelopment in children (Parvez 2011; Tolins 2014; Wasserman 2007; Wasserman 2014; Wasserman 2016; Yorifuji 2016).

Biomarkers of arsenic exposure

Arsenic levels measured in blood and urine are reliable biomarkers of arsenic exposure and status (ATSDR 2016; Hall 2006). In general, blood arsenic levels greater than (>) 1 µg/L and urinary arsenic levels > 100 µg/L are considered abnormal (ATSDR 2016), although some studies suggest that health risks of arsenic exposure may be associated with total urinary levels > 50 µg/L (Tseng 2005; Valenzuela 2005; WHO 2001). Because blood and urinary arsenic levels reflect short‐term exposure (i.e. hours to days) to arsenic, these are considered to be biomarkers, which are amenable to modifications driven by interventions (e.g. dietary change (Hall 2006)). Arsenic levels measured in hair and nails reflect more prolonged exposure to arsenic, because arsenic accumulates in these slow‐growing tissues (Hall 2006).

Inorganic arsenic (InAs) can be modified by the addition of one (mono‐) or two (di‐) methyl (CH₃‐) groups, in a process known as methylation, which leads to the organic forms known as monomethyl‐ or dimethyl‐arsenicals. These forms can be present in trivalent (^III) or pentavalent (^V) forms, which are also referred to as monomethylarsonous acid (MMA^III), monomethylarsonic acid (MMA^V), dimethylarsonous acid (DMA^III), and dimethylarsonic acid (DMA^V). The toxicity of arsenic is influenced by the degree of methylation; more highly methylated arsenicals are excreted more readily through the urine, leading to decreased retention, and therefore, decreased biological exposure and toxicity (Buchet 1981). In addition, trivalent metabolites are more toxic than pentavalent metabolites. Therefore, the most toxic arsenicals are InAs^III, MMA^III, and DMA^III, followed by InAs^V, MMA^V, and DMA^V (ATSDR 2016; Styblo 2000). Population groups mainly exposed to arsenic via drinking water, typically excrete 10% to 30% as inorganic arsenicals (InAs^III and InAs^V), 10% to 20% as monomethylated arsenicals (MMA^III and MMA^V), and 60% to 70% as dimethylated arsenicals (DMA^III and DMA^V)(Vahter 2000). Higher levels of monomethyl arsenicals are interpreted as an accumulation of the intermediate product, and thus, the reduced capacity to methylate inorganic arsenic to DMA and an increased toxicity. For example, compared to women, men show higher urinary monomethyl arsenicals, and men also show more frequent skin lesions (Lindberg 2008a). In arsenic‐exposed population studies, participants with skin lesions (versus those without skin lesions) had a higher proportion of monomethyl arsenicals in their urine, which is consistent with a lower arsenic methylation capacity (Li 2011; Valenzuela 2005; Zhang 2014). Arsenic methylation patterns in children differ from those in adults (Concha 1998; Fängström 2009; Skröder Löveborn 2016). In preschool children, urinary total arsenic and monomethyl‐As^V (MMA^V) percentage were shown to be positively associated with the risk of developmental delay (Hsieh 2014). Arsenic is deposited in bone tissue (Adeyemi 2010; Lindgren 1982), thus bone resorption and remodeling rates may modulate urinary arsenic excretion years after arsenic ingestion (Dani 2018).

Several factors affect the extent of arsenic methylation in the body, including dietary factors involved in cellular methylation pathways, e.g. folate, vitamin B12, choline, betaine, and others. Other physiological factors that may affect excretion of arsenic include gut microbiome (Pinyayev 2011; Rubin 2014), gender (Jansen 2016; Lindberg 2008a), pregnancy (Gao 2019; Gardner 2011; Gardner 2012), body mass index (Abuawad 2021; Bommarito 2019; Gribble 2013), and bone remodeling (Dani 2018).

Description of the intervention

Dietary factors that function in cellular methylation pathways can affect methylation of arsenic, leading to more rapid excretion. Among these, folate has been suggested as an important dietary factor to facilitate arsenic methylation and excretion (Carlin 2016). Folate is a general term for the water‐soluble vitamin B9 that is naturally present in foods, which humans are not able to synthesize in vivo, and which therefore, has to be obtained from dietary sources. Folate serves as a carrier for methyl groups required for biochemical reactions within cells, including methylation of arsenic. Thus, dietary interventions to increase folate are potential means to reduce arsenic toxicity, through an increase in arsenic methylation and excretion, and to prevent arsenic‐associated diseases.

Naturally occurring folates exist in many chemical forms, and are unstable, while folic acid is the stable, synthetic, oxidized form, used in supplements and fortified food (Fox 2008). When used as a dietary supplement or fortificant, folic acid is mostly metabolized to the metabolically active, natural forms of folate, including 5‐methyltetrahydrofolate, which is the form found in blood (Pfeiffer 2015; Stover 2004). Blood folate concentration can be reported as red blood cell (RBC) folate or serum/plasma folate. Serum blood folate levels are the earliest indicators of recent exposure, and reflect recent dietary intake (short‐term status). RBC folate is a sensitive indicator of folate status in the preceding 120 days (Bailey 2015). Elevated serum/plasma homocysteine is a sensitive biomarker of folate deficiency; however, elevations in homocysteine are not specific, because homocysteine increases in other B‐vitamin deficiencies, such as B12 deficiency, and can be affected by other factors, such as renal insufficiency and drug treatments (Bailey 2015). Nonetheless, plasma homocysteine is highly responsive to interventions with folate, either alone or in combination with the other methyl donors involved in one‐carbon metabolism, such as betaine, choline, and vitamins B2, B6, and B12.

Folate is critical for supporting rapid fetal growth and development, and thus, is especially important for women who may become pregnant. Failure of the neural tube to close during early embryonic development results in serious congenital anomalies, collectively referred to as neural tube defects (Beaudin 2009; Botto 1999; Greene 2014). A Cochrane Review showed that folic acid supplementation in women planning to become pregnant decreased the incidence and recurrence of neural tube defects in the fetus (De‐Regil 2015). Currently, the US Preventive Services Task Force recommends that all women who are planning, or capable of, pregnancy take a daily supplement containing 400 µg to 800 µg of folic acid, with a goal to reduce the occurrence of neural tube defects (USPSTF 2009). To help achieve these recommended folate intake levels, the US Food and Drug Administration (FDA) mandated fortification of the food supply through enriched cereal grain products (e.g. bread, pasta, rice fortified with 140 μg folic acid per 100 g grain). Population‐wide folic acid fortification has been introduced in 84 countries, and has been effective in reducing the rates of live‐born infants with spina bifida (Atta 2016; Cordero 2015; FFI 2019; Grosse 2016; Rader 2006). The lowest incidence of neural tube defects occurs when the RBC folate concentration is 906 nmol/L or higher (Cordero 2015; Martinez 2018; WHO 2015). The results of a modeled association study indicated that the overall risk of NTDs can be reduced to approximately 6 per 10,000 births when an RBC folate concentration of 1000 nmol/L to 1300 nmol/L is achieved for optimal prevention of folate‐sensitive neural tube defects (Bailey 2015; Crider 2014). Folate and vitamin B12 deficiencies, and high levels of homocysteine in blood (hyperhomocysteinemia) have been reported to occur in populations in Bangladesh and India (FFI 2019; Gamble 2005; Misra 2002). Elevated homocysteine concentration in pregnant women is associated with common pregnancy complications and adverse pregnancy outcomes (Bergen 2012; Dodds 2008; El‐Khairy 2003; Kharb 2016; Kim 2012; Liu 2020; Mishra 2020; Patrick 2004; Shahbazian 2016; Vollset 2000).

Folic acid supplementation in women of child‐bearing age, and population‐wide fortification of staple foods are currently used as public health interventions to reduce the rate of neural tube defects. Because folate detoxifies arsenic through arsenic methylation, this review evaluated and summarized the evidence on whether similar folic acid interventions are efficacious in reducing the public health burden of arsenic‐associated health outcomes in arsenic‐exposed adults and children, including all ages and gender groups.

How the intervention might work

Folate‐mediated one‐carbon metabolism is required for the remethylation of homocysteine to methionine. Methionine is used in the synthesis of S‐adenosylmethionine (SAM) as the methyl donor, a major cellular methyl donor in over 100 methylation reactions, including the methylation of arsenic. One‐carbon metabolism is largely dependent on the availability of folate, which serves as a carrier for one‐carbon (1C) methyl groups essential for the production of SAM. Arsenic excretion in humans involves methylation reactions that use SAM as the methyl donor. The methylation of arsenic is mediated by As(III)‐SAM methyltransferase (AS3MT). AS3MT catalyses the transfer of methyl groups from SAM to inorganic arsenic, producing the monomethyl‐ and dimethyl‐arsenicals (Dheeman 2014; Schlebusch 2015). Several studies link arsenic methylation capacity to the risk for health‐related outcomes of arsenic exposure: a higher percent of MMAs^III+V in urine has been associated with increased risk for bladder, breast, lung, and skin cancer, skin lesions, peripheral vascular disease, hypertension, and atherosclerosis (Chen 2003; Huang 2007; Huang 2008b; Kuo 2017; Li 2013; Lindberg 2008b; López‐Carrillo 2014; López‐Carrillo 2020; Melak 2014; Pu 2007; Steinmaus 2006; Steinmaus 2010; Tseng 2005; Wu 2006), and decreased birthweight (Laine 2015). In a nested case‐control study of As‐induced skin lesions, participants falling into the lowest tertile of the percent of DMAs in the urine were at higher risk for subsequent development of skin lesions, two to seven years later (Niedzwiecki 2018). Given the critical role of folate in methylation reactions, low folate status (through either inadequate intake of folates or genetic variation), affecting folate‐metabolizing enzymes, may impede arsenic methylation and excretion, and thereby, exacerbate arsenic toxicity. The involvement of folate metabolism in arsenic detoxification is supported by animal studies (Spiegelstein 2003). In support, folic acid supplementation has been shown to improve symptoms of chronic arsenic exposure, including arsenical skin lesions (Ghose 2014), and oxidative DNA damage (Guo 2015). In addition, a recent cross‐sectional study suggested that maternal folate status is inversely associated with the distribution of arsenic metabolites in cord serum (Laine 2018). We summarized the findings from observational studies that suggested the association between folate status and arsenic methylation capacity and arsenic‐related health outcomes in Table 3.

1. Summary of observational and intervention studies on folate, arsenic toxicity and arsenic‐related health outcomes.

Study ID	Study design	Participants	Interventions	Outcomes measured	Conclusions
Argos 2009; Argos 2010b	Cross‐ sectional	9833 participants from HEALS cohort in Araihazar, Bangladesh	‐	Daily average vitamin B intakes were estimated by food frequency questionnaire, and urinary arsenic concentrations were measured	Dietary folate, cobalamin, and riboflavin were not associated with creatinine‐adjusted urinary total arsenic concentration
Ashrap 2020	Cross‐sectional	Pregnant women in Puerto Rico	‐	Use of folic acid supplementation was recorded using a questionnaire. Urine was collected at 18 ± 2 weeks, 22 ± 2 weeks, and 26 ± 2 weeks gestation, and blood samples were collected at the first and third visit. Urinary and blood arsenic concentrations were assessed	Use of folic acid supplements was not a significant predictor of arsenic concentrations in urine. The association of folic acid supplements with blood arsenic was not reported
Basu 2011	Nested cross‐sectional	405 people selected from a cross‐sectional survey of 7638 people in an arsenic‐exposed population in West Bengal, India; 192 cases had skin lesions and 213 were controls	‐	19 dietary intake variables were assessed by questionnaire. 16 blood micronutrients and urine percentages of InAs, MMA, and DMA were measured. Associations of urine creatinine and nutritional factors with arsenic metabolites in urine were assessed	Urinary creatinine had the strongest relationship with overall arsenic methylation to DMA. Low serum selenium and low folate were also associated with increased urine MMA percentage
Bommarito 2019	Cross‐sectional	Cohort of 1166 adults living in Chihuahua, Mexico	‐	Nutrient intake was assessed using a food frequency questionnaire. BMI and urine percentages of InAs, MMA, and DMA were also measured	BMI was negatively associated with percentage of urinary InAs and percentage of urinary MMA, and positively associated with percentage of urinary DMA. This association was not influenced by dietary folate, vitamin B12, or vitamin B2 intake
Chen 2005	Case‐control	50 people with arsenic‐induced skin lesions, 35 people without skin lesions from the same village in an arsenic‐endemic area	‐	MTHFR C677T polymorphism and serum folate were assessed and compared between cases and controls	There was no significant difference in MTHFR C677T polymorphism and serum folate concentrations between cases and controls
Chung 2010	Case‐control	150 patients with UC and 300 healthy controls from the Medical Center (National Taiwan University Hospital and Taipei Municipal Wan Fang Hospital)	‐	Urinary arsenic metabolites (total arsenic, percentages of InAs, MMA, and DMA), and plasma folate and homocysteine were assessed; SNPs in 3 genes were genotyped	Participants with UC had higher urinary total arsenic and percentages of InAs and MMA; and lower percentage of DMA, plasma folate and homocysteine levels than controls. Plasma folate concentrations were negatively correlated with homocysteine and positively correlated with percentage of DMA
Chung 2019	Case‐control	178 people with UC and 356 age‐ or sex‐matched controls from the Department of Urology and Family Medicine, China Medical University Hospital, Taichung, Taiwan	‐	Plasma folate, urinary arsenic (total arsenic, percentages of InAs, MMA, and DMA), 8‐OHdG, and global 5‐MedC were assessed	Higher levels of total urinary arsenic, lower global 5‐MedC, and lower plasma folate significantly increased the odds ratios of UC
Desai 2018	Cross‐sectional	328 6 and 7‐year‐old children in Montevideo, Uruguay	‐	Dietary intakes were assessed using the average of two 24‐hour dietary recalls. Outcomes included cognitive performance (general intellectual ability score), and urinary arsenic (total arsenic and MMA)	The associations between dietary folate intake and cognitive performance were stratified by tertiles of folate intake. In the lowest tertile, folate intake was not associated with cognitive performance. In the second tertile, folate intake was associated with higher verbal comprehension, visual‐auditory learning, general intellectual ability, and verbal ability. In the highest tertile, folate intake was associated with higher concept formation, numbers reversed, and cognitive efficiency, and lower visual‐auditory learning. When the association between total urinary arsenic and cognitive performance was stratified by folate intake, there were few consistencies. Higher urinary arsenic was associated with lower concept formation and higher numbers reversed and cognitive efficiency in the lowest tertile of folate. Higher urinary arsenic was associated with higher integration in the second tertile of folate intake and with higher concept formation in the highest tertile of folate intake
Desai 2020a	Cross‐sectional	237 6 to 8‐year‐old children from Montevideo, Uruguay	‐	Serum folate and vitamin B12 levels and urinary arsenic were measured; dietary intakes were assessed using the average of two 24‐hour dietary recalls	No associations were observed between serum folate, serum vitamin B12, or vitamin B12 intake with InAs methylation. Folate intake was inversely associated with urinary percentage of MMA
Desai 2020b	Cross‐sectional	221 6 to 8‐year‐old children from Montevideo, Uruguay	‐	Models assessed the interaction of serum folate and urinary arsenic on broad math and reading scores	There was a significant interaction between serum folate and urinary arsenic on broad reading scores (odds ratio 1.05, 95% CI 1.00 to 1.10)
Desai 2020c	Cross‐sectional	255 6 to 8‐year‐old children from Montevideo, Uruguay	‐	Dietary intakes were assessed using the average of two 24‐hour dietary recalls. Executive function was assessed via three tests: SOC, IED, and SSP	Urinary arsenic was associated with lower scores on executive function tests. When stratified by high and low folate intake, urinary arsenic was associated with lower pre‐executive error score within the IED among children with low folate intake, but not high folate intake; however, the stratified associations were not significantly different from each other
Fisher 2019	Cohort	Pregnant women participating in the MIREC study	‐	Self‐reported folic acid supplement use was collected. Total blood arsenic concentrations were measured at trimester 1 and 3	Use of folic acid supplements was not associated with blood arsenic concentrations
Gamble 2005	Cross‐sectional	300 participants from HEALS in Arailhazar, Bangladesh	‐	Urinary arsenic metabolites (InAs, MMA, DMA) were assessed, and their correlations with plasma folate, homocysteine, and vitamin B12were estimated	Folate, homocysteine, and other factors involved in one‐carbon metabolism influence arsenic methylation. Plasma folate was associated with higher percentage DMA and lower percentage MMA and InAs
Gamboa‐Loira 2018	Cross‐sectional	1027 women exposed to InAs in Northern Mexico	‐	Polymorphisms in five one‐carbon metabolism genes and urinary arsenic (InAs, MMA, DMA) were assessed. Dietary intake was estimated using a food frequency questionnaire. Polymorphisms were assessed as an independent exposure, and as a modifier of nutrient intakes on urinary arsenic	Differences in dietary nutrient intake and genetic variants in one‐carbon metabolism may jointly influence InAs methylation capacity. Individuals with genetic polymorphisms in FOLH1 c.223 and MTR c.2756 had higher percentages of InAs. Polymorphism in FOLH1c.223 modified the association between vitamin B12 and urinary arsenic, but no other significant gene‐nutrient interactions were found
Gao 2019	Cross‐sectional	1613 pregnant women in Bangladesh (4 to 16 weeks' gestation). All participants received and reported taking prenatal supplements with folic acid	‐	Daily dietary folate intake was assessed via a food frequency questionnaire for the previous 12 months. Urinary arsenic metabolites were measured (percentages of InAs, MMA, and DMA). Associations between dietary intake and urinary metabolites at visit 1 were assessed	Folate intake was not associated with any arsenic metabolites (percentages of InAs, MMA, and DMA)
Gardner 2011	Cohort	324 pregnant women exposed to arsenic via drinking water and food in rural Bangladesh, in Matlab, selected from participants of the MINIMat trial where women were randomized to receive either food or iron supplementation at different doses, and all groups received 400 μg folic acid supplementation daily	‐	Total urine arsenic, measured as the sum of InAs, MMA, and DMA, total arsenic in blood, and concentrations of plasma folate and vitamin B12 were measured	Arsenic methylation efficiency improved very early in pregnancy (urinary DMA increased and urinary MMA decreased). Folate and vitamin B12 concentrations were not associated with these changes
Ghose 2014	Open clinical trial	32 people with arsenicosis (who recently switched to arsenic‐free water) and 45 age‐ and sex‐matched arsenic‐affected people	32 people with arsenicosis received 5 mg of folic acid daily; 45 age‐ and sex‐matched arsenic‐affected people received arsenic‐free water	Skin pigmentation, keratosis, chest signs (rales and rhonchi), hepatomegaly, and splenomegaly, flushing of face, solid edema of legs and hands, ascites, and absence of deep reflexes for neuropathy were measured at baseline and at 6 months of treatment	The study found a significant improvement of arsenical skin lesion score in both those with arsenicosis and controls; a significant improvement in systemic disease score from the baseline systemic score in the folic acid treated group and the group treated with arsenic‐free water; and a significantly increased improvement of systematic disease score in the folic acid‐treated group compared to the control group taking arsenic‐free water. Folic acid treatment in arsenicosis may help to reduce clinical symptoms of arsenicosis
Grau‐Pérez 2017	Case‐control	SEARCH for Diabetes in Youth Case‐Control study; USA children and adolescents < 22 years of age, 429 with type 1 diabetes, 85 with type 2 diabetes, and 174 control participants	‐	Arsenic species (InAs, MMA, DMA), and one‐carbon metabolism biomarkers (folate and vitamin B12) were measured in plasma	Percentage of MMA was associated with a lower odds of type 1 diabetes. There was a significant interaction when stratified by folate status: percentage of MMA was associated with a higher odds of type 1 diabetes among individuals with plasma folate above the median, but there was no association among those below the median
Guo 2015	Randomized clinical trial	450 participants for assessment of efficacy of folic acid supplementation on DNA oxidative damage reversal	Participants were randomly assigned to 3 groups: 0.4  mg folic acid/day, 0.8  mg folic acid/day, or placebo (control) for 8 weeks	Urinary 8‐OHdG and creatinine concentration were assessed pre‐ and post‐folic acid supplementation. A multivariate general linear model was applied to assess the individual effects of folic acid supplementation, and the joint effects of folic acid supplementation and hypercholesterolemia on oxidative DNA damage improvement	Folic acid supplementation was independently linked to the reduction of urinary 8‐OHdG/creatinine (a marker of oxidative damage to DNA) in a dose‐related pattern. This effect was stronger in those with hypercholesterolemia
Hall 2007	Cross‐sectional	101 pregnant women who gave birth in Matlab, Bangladesh. All participants were provided with daily iron (60 mg/day) and folic acid (300 μg/day) supplements	‐	Maternal and cord blood pairs concentrations of total arsenic and plasma folate, vitamin B12, and homocysteine levels	Maternal–fetal transport of arsenic readily occurs; compared with maternal urine, maternal blood contained substantially higher proportions of InAs and MMA, but less DMA; maternal homocysteine was a strong predictor of the percentage of DMA in maternal urine, maternal blood, and cord blood. Maternal folate concentrations were associated with lower blood arsenate, but not with any other arsenic metabolites in maternal urine, maternal blood, or cord blood
Hall 2009	Cross‐sectional	165 Bangladeshi 6‐year‐old children	‐	Blood homocysteine, folate, cobalamin, cysteine, total urine arsenic, and arsenic metabolites (InAs, MMA, DMA)	Similar to adults, folate and cysteine facilitate arsenic methylation in children. Higher folate concentrations were associated with lower percentage of urinary InAs, and higher percentage of DMA
Huang 2008a	Case‐control	177 cases of UC and 488 controls in Taiwan; association between UC risk and urinary arsenic and plasma folate	‐	Urinary arsenic (InAs, MMA, DMA) and plasma folate were measured; associations between UC risk, urinary arsenic, and plasma folate were estimated	Higher total arsenic levels, percentage of InAs, percentage of MMA, and primary methylation index were associated with increased risk of UC; higher plasma folate, percentage of DMA, and secondary methylation index were associated with decreased UC risk. A dose‐response relationship was shown between plasma folate levels or methylation indices of arsenic species and UC risk
Hsueh 2020	Case‐control	178 children with developmental delay and 88 without developmental delay in Taiwan	‐	Plasma folate and polymorphisms in MTHFR C677T, MTHFR A1298C, and MTR A2756G were assessed. Total urinary arsenic and arsenic metabolites (percentage of InAs, MMA, DMA) were also assessed	No differences in total urinary arsenic or percentage of arsenic metabolites were found between genotypes of folate metabolism‐related polymorphisms. When stratified by genotypes, total urinary arsenic and arsenic metabolites were not associated with developmental delay, except for a significant interaction between percentage of MMA and MTR A2756G
Kancherla 2017	Case‐control	53 mothers of children with myelomeningocele and 53 controls, from Dhaka Community Hospital in rural Bangladesh	‐	Plasma folate concentrations were assessed and self‐reported folic acid supplement use was recorded	Plasma folate was not different between cases and control, but use of prenatal folic acid supplementation was higher among controls compared to cases. Prenatal folic acid supplement use was associated with a lower odds of myelomeningocele
Kordas 2016	Cross‐sectional	5 to 8 year old children from Montevideo, Uruguay	‐	Urinary arsenic metabolites (total arsenic, InAs, MMA, DMA) were assessed, and dietary intake was estimated from a 24‐hour dietary recall	Higher intake of total dietary folate was associated with lower percentage of MMA and higher percentage of DMA
Kurzius‐Spencer 2017	Cross‐sectional	2420 participants, aged 6 years and older; NHANES, USA	‐	Urinary arsenic metabolites (percentages of InAs, MMA, and DMA), serum and red blood cell folate, serum vitamin B12, and plasma total homocysteine were assessed. Dietary intake was estimated from a 24‐hour recall	Red blood cell folate was associated with a lower percentage of InAs, and dietary vitamin B12 was associated with higher percentage of InAs
Laine 2018	Cross‐sectional	197 pregnant women (BEAR cohort, 188 mom/baby pairs) exposed to arsenic in Mexico	‐	Correlations between folate, homocysteine, vitamin B12 in maternal plasma samples with arsenic (total and percentages of InAs, MMA, DMA) in maternal urine and cord blood were established	Maternal urinary total arsenic was significantly negatively correlated with B12, and positively correlated with homocysteine. Infants born to mothers in the lowest tertile of serum folate had significantly higher mean levels of percentage of MMA in cord serum, and elevated maternal homocysteine was associated with higher cord serum percentage of MMA. Maternal OCM status may influence the distribution of arsenic metabolites in cord serum, which may be indicative of InAs toxicity
Lin 2019	Case‐control	178 children with developmental delay and 88 normal children, from Shin Kong Wu Ho‐Su Memorial Teaching Hospital, Taiwan	‐	Urinary arsenic metabolites (InAs, MMA, DMA) and plasma folate and vitamin B12 were measured	The combination of high plasma folate and high vitamin B12 levels was positively correlated with efficient arsenic methylation capacity (low percentage of MMA, low percentage of InAs, and high percentage of DMA). High percentage of MMA significantly increased, whereas, high percentage of DMA and secondary methylation index decreased the odds ratio of developmental delay in a dose‐dependent manner in both low plasma folate and low vitamin B12 groups. The combination of high plasma folate and high vitamin B12 levels increased arsenic methylation capacity, and indirectly decreased the odds ratio of developmental delay in preschool children
López‐Carrillo 2016	Cross‐sectional	1027 women, at least 20 years old, living in northern Mexico	‐	Intake of 21 micronutrients estimated using a food frequency questionnaire; urinary metabolites of arsenic (percentages of InAs, MMA, DMA) were assessed	Higher intakes of methionine, choline, folate, vitamin B12, zinc, selenium, and vitamin C favor elimination of InAs, mainly by decreasing the percentage of MMA or increasing the percentage of DMA (or both) in urine
Mazumdar 2015a (Ibne Hasan 2015)	Case‐control	57 mothers of children with myelomeningocele and 55 controls, from Dhaka Community Hospital in rural Bangladesh	‐	Maternal arsenic exposure was estimated from drinking water samples taken at the time of enrollment from wells used during the first trimester of pregnancy. Self‐report periconceptional folic acid use and maternal plasma folate status were measured at the time of enrollment	A significant interaction was observed between concentrations of InAs in drinking water and periconceptional folic acid use, on the odds of myelomeningocele. As the concentrations of InAs in drinking water increased from 1 to 25 μg/L, the estimated protective effect of folic acid use on odds of myelomeningocele declined (odds ratio 0.22 versus odds ratio 1.03), and was not protective at higher concentrations of arsenic
Mazumdar 2015b	Case‐control	54 mothers of children with myelomeningocele and 55 controls, from Dhaka Community Hospital in rural Bangladesh	‐	Concentrations of InAs in drinking water was assessed. Maternal plasma folate levels were measured and 14 SNPs in genes involved in folate metabolism were genotyped	Drinking water InAs was associated with increased risk of having a child with myelomeningocele for women with 4 of the 14 studied SNPs in genes involved in folate metabolism; environmental arsenic exposure increases the risk of myelomeningocele by means of interaction with folate metabolic pathways
Mise 2020	Cross‐sectional	104 pregnant women in Japan. Previous publication from this study population reported drinking water arsenic concentrations between 0.01 μg/L to 1.90 μg/L, and arsenic exposure from various food sources (Mise 2019)	‐	Serum folate was assessed, and self‐reported use of folic acid supplements was collected	The women who consumed folic acid supplements more than once per week had significantly higher serum folic acid (serum folate) concentrations than women who consumed no folic acid supplements, and had similar serum folic acid concentrations (serum folate) to women who consumed folic acid supplements less than once per week.
Monroy Torres 2018	Interventional	45 adolescents exposed to arsenic through drinking water	Daily multivitamin supplement for 4 weeks	Urinary excretion of arsenic was compared to baseline. Dietary intake was assessed weekly via 24‐hour recall	Four weeks of supplementation with vitamins and minerals in the adolescent population improved nutritional status, and increased arsenic excretion significantly in the first and second week after intervention. Baseline intake of fiber, vitamin E, and selenium were associated with greater urinary arsenic excretion at week 1 (fiber), at week 2 (vitamin E), and at week 3 (selenium). No associations were reported for folate
Niedzwiecki 2018	Nested case‐control study	Study nested in the HEALS in Bangladesh, in which 876 incident skin lesion cases were individually matched to controls on sex, age, and follow‐up time (703 cases and 705 controls with serum samples were included)	‐	Serum homocysteine, urinary arsenic metabolites (percentages of InAs, MMA, DMA), and 26 SNPs in 13 OCM genes were assessed	Serum homocysteine were independently associated with increased odds, and urinary percentage of DMA with decreased odds of skin lesions. The T allele of MTHFR 677 (i.e. C to T substitution at nucleotide position 677) was associated with hyperhomocysteinemia, higher percentage of MMA, and lower percentage of DMA, but not with skin lesions
Obrycki 2019	Case‐control	55 mothers of children with myelomeningocele and 55 controls, from Dhaka Community Hospital in rural Bangladesh	‐	Maternal dietary intake was assessed using a food frequency questionnaire, and maternal plasma folate concentrations were measured. Maternal arsenic exposure was determined using arsenic concentrations in drinking water and in maternal toenail samples	Folate intake (along with other micronutrients) was higher in cases compared to controls. Plasma folate concentrations did not differ between cases and controls. Arsenic concentrations in drinking water and toenails were higher in controls
Pilsner 2007; Pilsner 2009	Nested case‐control	Study in HEALS in Araihazar, Bangladesh; 274 cases who developed skin lesions 2 years after recruitment, and 274 controls	‐	Plasma folate, total homocysteine, and total cobalamin, and hypomethylated leukocyte DNA were assessed	Folate deficiency (< 9 nmol/L), hyperhomocysteinemia, and low urinary creatinine were each associated with decreased arsenic methylation, and were associated with a higher odds of arsenic‐induced skin lesions
Spratlen 2017	Cross‐ sectional	405 participants from the Strong Heart Study (American Indians in Arizona, Oklahoma, and North and South Dakota)	‐	Urinary arsenic (total arsenic, InAs, MMA, DMA) were assessed. Dietary intake of micronutrients was estimated through food frequency questionnaire	Dietary intake of vitamin B2 and B6, but not folate, were associated with higher percentages of urinary DMA and lower percentages of InAs and MMA. There was a significant interaction between vitamin B6 and folate intake: high folate with low vitamin B6 intake was associated with higher percentage InAs and lower percentage of DMA, while high folate with high vitamin B6 was associated with lower percentage InAs and a non‐significant increase in percentage DMA
Xu 2016	Cross‐sectional	Adults in Chihuahua, Mexico, without diabetes	‐	Micronutrient intake was assessed using a food frequency questionnaire. Concentrations of InAs in drinking water, and concentrations of plasma glucose after an oral glucose tolerance test were analyzed	Among individuals exposed to high concentrations of arsenic in drinking water (> 50 μg/L), elevated folate intake (> 800 DFE) was associated with higher 2‐hour glucose concentrations compared to adequate folate (320 DFE to 800 DFE), while low intake of folate (< 320 DFE) was not associated with a difference in 2‐hour glucose. Among those with adequate folate intake, there was no difference in 2‐hour glucose between those with low versus high arsenic exposure
Zablotska 2008	Cross‐sectional	11,746 individuals recruited from HEALS, 2000‐2002	‐	Skin lesions were identified according to a structured clinical protocol during screening and confirmed with further clinical review. Arsenic concentrations in drinking water and urine samples were measured. Dietary intake of micronutrients was assessed using a food frequency questionnaire	Higher quintiles of folic acid intake (along with other micronutrients) were associated with lower odds of skin lesions. Higher water and urinary arsenic concentrations were associated with higher risk of skin lesions. When stratified by quintiles of micronutrient intake, increasing quintiles of folic acid intake attenuated the association between arsenic exposure and skin lesions
Zhang 2019	Cross‐sectional	USA NHANES (2003‐2012) (N = 11,016)	‐	Serum folate and total urine arsenic (total arsenic, MMA, DMA) were assessed	Higher serum folate was associated with higher percentage of DMA in children, and higher percentage of MMA in adults
Zhu 2018	Cross‐sectional	USA NHANES (2003‐2006) A total of 1161 children (aged 6 to 19 years) and 1938 adults (aged 20 to 85 years)	‐	Serum folate, homocysteine, and vitamin B12, and urine arsenic (total arsenic, MMA, DMA) were assessed	Higher concentrations of serum folate and cobalamin were associated, with higher urinary DMA in both children and adults

BEAR: Biomarkers of Exposure to ARsenic; BMI: body mass index; 95% CI: 95% confidence interval; DFE: dietary folate equivalents; DMA: dimethylarsonic acid; DNA: deoxyribonucleic acid; FOLH1: folate hydrolase 1; HEALS: Health Effects of Arsenic Longitudinal Study; IED: intra‐dimensional/extra‐dimensional shift task;InAs: inorganic arsenic;MINIMat: Maternal and Infant Nutrition Interventions in Matlab; MIREC: Maternal‐Infant Research on Environmental Chemicals; MMA: monomethylarsonic acid; MTHFR: methylenetetrahydrofolate reductase; MTR: methionine synthase; N: number; NHANES: National Health and Nutrition Examination Survey; OCM: one‐carbon metabolism; SNP: single nucleotide polymorphism; SOC: Stockings of Cambridge; SSP: Spatial Span;UC: urothelial carcinoma; 5‐MedC: 5‐methyl‐2′‐deoxycytidine;8‐OHdG: 8‐hydroxy‐2′‐deoxyguanosine

Inter‐individual differences, caused by genetic variations, can affect the capacity to methylate arsenic (Gribble 2015; Schlebusch 2015). Genetic variants in folate‐metabolizing enzymes have been shown to affect arsenic metabolism and retention in human and animal studies (Mazumdar 2015b; Schläwicke‐Engstrom 2009; Wlodarczyk 2012; Wlodarczyk 2014). For example, a common genetic variant in the folate‐metabolizing enzyme methylenetetrahydrofolate reductase (MTHFR), MTHFR C677T (i.e. C to T substitution at nucleotide position 677), was associated with lower blood folate concentrations (Bailey 1999; Stover 2011; Tsang 2015), and increased sensitivity to arsenic exposure in a mouse model (Wlodarczyk 2012; Wlodarczyk 2014).

Why it is important to do this review

Inorganic arsenic is a common environmental toxin and an important public health burden. Folate, which plays an essential role in methylation reactions, may lower blood arsenic concentrations in arsenic‐exposed populations, thereby, contributing to the prevention of arsenic‐associated illnesses, including neurotoxicity in children. Hyperhomocysteinemia, with or without folate deficiency, has been reported to occur in populations exposed to arsenic, including children and pregnant women (Hall 2009; Laine 2018; Misra 2002). Elevated total homocysteine concentration, measured in the serum or plasma of pregnant women, has been reported to be associated with common pregnancy complications (pre‐eclampsia), and adverse pregnancy outcomes (prematurity, low birth weight (Bergen 2012; Dodds 2008; El‐Khairy 2003; Kharb 2016; Kim 2012; Liu 2020; Mishra 2020; Patrick 2004; Shahbazian 2016; Vollset 2000)). The provision of folic acid through supplementation or fortification, or both, to reduce neural tube defects may have the added benefit of facilitating arsenic methylation and excretion to decrease arsenic toxicity, particularly in arsenic‐exposed populations with folate deficiency (Dubey 2007; Gamble 2005; NIAT 2006; Pilsner 2009).

Despite the evidence from observational studies and clinical trials, no systematic reviews have been conducted to estimate the effects of folic acid supplementation on arsenic toxicity in children and in adult men and women. This review is important because it was the first Cochrane Review to evaluate and summarize the evidence on folic acid supplementation or fortification, or both, for reducing arsenic toxicity and arsenic‐related health outcomes in children and in adult men and women. The findings of this review will help inform future research and public policy, especially in arsenic‐exposed populations.

Objectives

To assess the effects of the provision of folic acid (through folic acid‐fortified foods or supplements), alone or in combination with other nutrients, in lessening the burden of arsenic‐related health outcomes, and reducing arsenic toxicity in arsenic‐exposed populations.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials (RCTs), with randomization at either the individual or cluster level; and quasi‐RCTs (where allocation of treatment was made, for example, by alternate allocation, date of birth, or alphabetical order).

Types of participants

Arsenic‐exposed populations, of all ages and gender groups (including pregnant women), from any country.

We defined arsenic exposure as exposure to drinking water arsenic levels above the WHO guideline value (i.e. 10 µg arsenic per liter of water (WHO 2008)), or above the national permissible limits, or as exposure to arsenic levels in food (e.g. rice) that may result in measurable blood and urine arsenic exposure, based on the values defined by trial authors.

Types of interventions

Provision of folic acid (through fortified foods or supplements) alone, or in combination with other nutrients.

Interventions may have been given at any dose and for any duration, regardless of frequency of administration. A supplement may have been in a tablet or capsule. Any form of folic acid fortification may have been included. Studies with co‐interventions (e.g. provision of arsenic water filtration systems, education, etc.) were eligible for inclusion, but only if the co‐interventions were the same across study arms.

We planned to assess the following comparisons.

• Folic acid supplements alone versus no intervention or placebo

• Folic acid supplements plus other nutrient supplements versus nutrient supplements alone (exact same formulation of other nutrients in both arms without folic acid)

• Food fortified with folic acid alone versus unfortified food

• Food fortified with folic acid plus other nutrient supplements versus food fortified plus other nutrient supplements alone (exact same formulation of other nutrients but no folic acid)

Types of outcome measures

The following outcomes were prioritized as clinical disease outcomes (primary outcomes), and biological markers or clinical signs (secondary outcomes) of arsenic exposure.

Primary outcomes

• Any type of cancer (as defined by trial authors)

• All‐cause mortality (as defined by trial authors)

• Neurocognitive function, measured only in children (as defined by trial authors; including, but not limited to, the assessments of memory, attention, intelligence, and other cognitive domains)

• Any congenital anomalies, measured only in pregnant women (as defined by trial authors; including, but not limited to, neural tube defects, cleft lip and cleft palate; and detected, for example, during periconceptional and neonatal screenings)

Secondary outcomes

• Blood and urinary arsenic concentration (μg/L), arsenic metabolite proportions and methylation indices (%; as measured by trial author; including, but not limited to, measurements using atomic absorption spectrometry and inductively coupled plasma (ICP)‐mass spectrometry; various forms of arsenic, for example, monomethyl‐ or dimethyl‐arsenic forms)

• Plasma folate concentration (nmol/L; as measured by trial authors; including, but not limited to, measurements using mass spectrometry, competitive protein binding assays, and microbiological assays)

• Plasma homocysteine concentration (μmol/L; as measured by trial authors; including, but not limited to, measurements using high‐performance liquid chromatography, gas chromatography‐mass spectrometry, and immunoassays)

• Skin lesions (as defined by trial authors; including, but not limited to, the assessments of hyper‐ or hypopigmentation, keratosis, and exfoliative dermatitis)

Search methods for identification of studies

We searched the electronic databases listed below in September 2020, for all available years, and without language restrictions.

Electronic searches

We searched the following international and regional sources.

International databases

• Cochrane Central Register of Controlled Trials (CENTRAL; 2020, Issue 10) in the Cochrane Library, which includes the Cochrane Developmental, Psychosocial and Learning Problems Specialised Register (searched 20 October 2020);

• MEDLINE Ovid (1946 to 7 December 2020);

• MEDLINE Ovid In‐Process & Other Non‐Indexed Citations (searched 7 December 2020);

• MEDLINE Ovid Epub Ahead of Print (searched 7 December 2020);

• Embase Ovid (1980 to 1 September 2020);

• Science Citation Index Web of Science (1970 to 3 January 2021);

• Social Sciences Citation Index Web of Science (1970 to 3 January 2021);

• Conference Proceedings Citation Index – Science Web of Science (1990 to 3 January 2021);

• Conference Proceedings Citation Index – Social Science & Humanities Web of Science (1990 to 3 January 2021);

• Cochrane Database of Systematic Reviews (CDSR; 2019, Issue 8), in the Cochrane Library (searched 20 October 2020);

• Database of Abstracts of Reviews of Effects (DARE; 2015, Issue 2. Final Issue), in the Cochrane Library (searched 14 January 2019);

• CINAHL EBSCO (Cumulative Index to Nursing and Allied Health Literature; 2002 to 28 January 2021);

• POPLINE (www.popline.org; searched 14 August 2019);

• ClinicalTrials.gov (clinicaltrials.gov; searched 7 December 2020);

• WHO International Clinical Trials Registry Platform (ICTRP; trialsearch.who.int/; searched 14 January 2021).

Regional databases

• African Index Medicus (AIM; www.globalindexmedicus.net; searched 28 January 2021);

• Index Medicus for the Eastern Mediterranean Region (IMEMR; www.globalindexmedicus.net; searched 28 January 2021);

• PAHO (Pan American Health Library; iris.paho.org/?locale-attribute=es; searched 14 January 2021);

• WHOLIS (WHO Library; dosei.who.int; searched 14 January 2021);

• LILACS (Latin American and Caribbean Health Sciences Literature; lilacs.bvsalud.org/en; searched 07 December 2020);

• Index Medicus for the South‐East Asia Region (IMSEAR; www.globalindexmedicus.net; searched 28 January 2021);

• SciELO (Scientific Electronic Library Online; www.scielo.br; searched 13 January 2021);

• IndMED (Indian Medical Journals; www.india.gov.in/website-indexing-indian-medical-journals-indmed; searched 16 August 2019);

• WPRO (Western Pacific Region Index Medicus; www.wprim.org; searched 07 December 2020).

We report the search strategies for all databases in Appendix 1.

Searching other resources

We scanned the reference lists of the included studies and relevant reviews to identify additional eligible studies. We also contacted relevant study authors to identify any ongoing or unpublished studies.

Data collection and analysis

Selection of studies

Two review authors (SB, EK, or HG) independently screened the titles and abstracts of records retrieved by the search for relevance using Covidence, a screening and data collection tool (Covidence). We resolved any disagreement through discussion, and if required, consulted with a third review author (PAC). We obtained the full texts of potentially relevant reports, which two review authors (SB, EK, HG) independently assessed for eligibility, based on the inclusion criteria (see Criteria for considering studies for this review). We resolved disagreements that occurred at this stage of the eligibility assessment process through discussion with the other review authors, who, when necessary, independently checked the included and excluded studies. We recorded our decisions in a PRISMA diagram (Moher 2009).

Data extraction and management

Two review authors (SB, EK, HG) independently extracted data from the included studies, and recorded them on a data extraction form designed for this review. The data extraction form included the following information.

• General information: title, authors, publication type (e.g. journal article, abstract, book chapter), country of study, funding source of study, year of study, and authors’ conflicts of interest

• Details of study: study aim, design, inclusion and exclusion criteria, unit and method of randomization, study location and duration, sample size, characteristics of participants, procedures for recruiting and selecting participants, method of allocation, and participant attrition

• Intervention and control (comparison group): method used for implementation of intervention, duration, dose and frequency of intervention and control, type of intervention and control, co‐intervention, number of participants allocated to intervention and control groups, and compliance

• Outcomes: any measures of primary and secondary outcomes, time points and method of outcome assessments, and blinding of outcome assessment

Any disagreements between the two review authors (SB and EK) during the data extraction process were resolved through discussion with other review authors (PS, PAC). We entered all extracted data into the latest version of Cochrane Review Manager 5 software, and we checked the data for accuracy (Review Manager 2020). When the information regarding any of the above was insufficient or unclear, we contacted the trial authors to request further details of the study.

Assessment of risk of bias in included studies

Two review authors (SB and EK) independently assessed the risk of bias for each included study, based on the Cochrane RoB 1 tool (Higgins 2011). The review authors assigned one of three ratings (low risk of bias, high risk of bias, or unclear risk of bias) to each of the domains listed below, with justifications for their judgements. Any disagreements were resolved through discussion with other review authors. When study information was insufficient or unclear, we contacted the trial authors to request further details. Domain‐specific criteria for judgements of low, high, or unclear risk of bias are shown in Table 4. More detailed criteria are provided in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

2. Examples of domain‐specific criteria for judgments of low, high, or unclear risk of bias.

Random sequence generation
Low risk of bias	The study used an appropriate randomization method (e.g. computer random number generator, random number table) in the sequence generation process.
High risk of bias	The study used a non‐randomized method (e.g. birth date, case number) in the sequence generation process.
Unclear risk of bias	There was no, or insufficient, information specifying the randomization method. For example, where a study used blocked randomization, but did not specify the process of selecting the blocks, then we classified the study as being at unclear risk of bias.
Allocation concealment
Low risk of bias	The study used an adequate method to conceal the allocation sequence from participants and investigators (e.g. telephone or central randomization, sealed opaque envelopes).
High risk of bias	The study used an inadequate method, such that participants and investigators may have been able to foresee the assignment to intervention groups (e.g. open random allocation, unsealed, or non‐opaque envelopes).
Unclear risk of bias	There was no, or insufficient, information specifying the method of allocation concealment.
Blinding of participants and personnel
Low risk of bias	The study participants and personnel were blinded, and the blinding was unlikely to have been broken; or the lack of blinding was unlikely to have introduced bias.
High risk of bias	There was no, or incomplete, blinding of study participants and personnel, and the lack of blinding was likely to have introduced bias; or blinding was attempted, but it was likely to have been unsuccessful.
Unclear risk of bias	There was no, or insufficient, information on blinding to permit a judgement of high or low risk of bias.
Blinding of outcome assessment
Low risk of bias	The outcome assessors were blinded, and the blinding was unlikely to have been broken; or the lack of blinding was unlikely to have introduced bias.
High risk of bias	There was no, or incomplete, blinding of outcome assessments, and the lack of blinding was likely to have introduced bias; or blinding was attempted, but it was likely to have been unsuccessful.
Unclear risk of bias	There was no, or insufficient, information on blinding of outcome assessments to permit a judgement of high or low risk of bias.
Incomplete outcome data
Low risk of bias	There were no missing outcome data (i.e. all participants randomized to the trial were included in the analysis); or missing data were balanced across groups, and the reasons for missing data were unlikely to have introduced bias.
High risk of bias	There were missing data; missing data were imbalanced across groups; the reasons for missing data were likely to have introduced bias; or ‘as‐treated (per protocol)’ analysis was performed, with substantial differences between the intervention allocated at randomization and the intervention actually received.
Unclear risk of bias	There was no, or insufficient, information to permit a judgement of high or low risk of bias.
Selective reporting
Low risk of bias	It was clear, either by the study protocol or otherwise, that all of the prespecified and expected outcomes were reported.
High risk of bias	Not all prespecified and expected outcomes were reported; reported outcomes were not prespecified (unless the reason for reporting was justified); or outcomes were reported incompletely and could not be used.
Unclear risk of bias	There was no, or insufficient, information to permit a judgement of high or low risk of bias.
Other sources of bias
Low risk of bias	Baseline characteristics (related to outcome measures) or baseline outcome measures were similar across groups; an appropriate adjusted analysis was performed to account for differences in baseline measures across groups; or there were no other sources of bias.
High risk of bias	Baseline characteristics (related to outcome measures) or baseline outcome measures were not similar across groups, and no, or an inappropriate, adjusted analysis was performed; there was a contamination issue whereby the experimental and control interventions were mixed; there was a claim that the study was fraudulent; or there were other sources of bias.
Unclear risk of bias	There was no, or insufficient, information to permit a judgement of high or low risk of bias.

• Random sequence generation (checking for selection bias). We described the method used to generate the allocation sequence, and assessed whether the sequence generation was suitable to minimize selection bias.

• Allocation concealment (checking for selection bias). We described the method used to conceal the allocation sequence, and assessed whether intervention allocation could have been foreseen in advance of, or during, enrolment.

• Blinding of participants and personnel (checking for performance bias). We assessed whether the study participants and personnel (field workers, laboratory technicians, and study investigators) were blinded from knowledge of which intervention a participant received.

• Blinding of outcome assessment (checking for detection bias). We assessed whether the outcome assessors were blinded from knowledge of which intervention a participant received.

• Incomplete outcome data (checking for attrition bias). We assessed the completeness of outcome data by examining attrition and exclusions from the analysis. We examined whether attrition and exclusions were reported, the reasons for attrition or exclusion, and whether missing data were balanced across groups, or were related to outcomes.

• Selective reporting (checking for reporting bias). We assessed whether the included study reported only a subset of outcomes, or only selective data for an outcome.

• Other sources of bias (checking for bias due to problems not covered by the domains above). We assessed other possible sources of bias, if any. For example, we assessed: whether the study had been claimed to be fraudulent; whether the study had contamination bias, which occurs when participants of the control group inadvertently receive the treatment, or are exposed to the intervention; and whether there were any other sources of bias not addressed in the other domains.

Measures of treatment effect

We presented results as reported in included studies. All the reported data in the included studies were continuous. We preferred change data, but where these were not available, we used end point data.

For methods archived for future updates of this review, please refer to Appendix 2 and our protocol (Bae 2017).

Unit of analysis issues

Cluster‐randomized trials

We did not identify any eligible cluster‐randomized trials.

Studies with multiple treatment groups

We presented all intervention groups of a multi‐arm study in the ‘Characteristics of included studies’ table. In the Results section of this review, we only presented data from the relevant intervention groups of a multi‐arm study that met our inclusion criteria (see Criteria for considering studies for this review).

Cross‐over trials

We identified a randomized trial with a cross‐over design, and only used the data from the first period of the trial, or data from the second period for groups continuing to receive the same intervention as during the first period, to avoid any potential risks of a carry‐over effect (i.e. folic acid supplementation during the first period would affect outcomes among the groups that switched to placebo during the second period).

Dealing with missing data

For each included study, we determined whether summary data for an outcome (e.g. standard deviations) were missing, and whether prespecified outcome data were missing. We also reported levels of attrition for each included study in the Characteristics of included studies, with reasons for dropouts. In trials where data were missing due to participant dropout, we determined whether the reasons for dropout were well documented and unrelated to study arms.

Assessment of heterogeneity

We assessed clinical and methodological heterogeneity among studies by examining the variability in study design, participants, intervention, outcomes, and risk of bias. We found significant heterogeneity between included studies, primarily across study participants (i.e. baseline arsenic exposure and nutritional status) and co‐interventions (i.e. provision of arsenic‐removal water filter); therefore, we did not combine the included studies in a meta‐analysis, and as a result, were not able to assess statistical heterogeneity.

Assessment of reporting biases

We attempted to minimize reporting biases by comprehensively searching for eligible studies (including ongoing and unpublished studies) using multiple sources and databases. We assessed selective outcome reporting bias within each included study, as described in the Assessment of risk of bias in included studies section.

Data synthesis

Due to the small number of included studies and the heterogeneity across study participants (i.e. differences in baseline arsenic exposure and nutritional status) and co‐interventions (i.e. use of arsenic‐removal water filters), we were not able to conduct a meta‐analysis. A folic acid intervention will likely have different effects for arsenic‐exposed populations based on baseline folate status. In addition, a lower baseline exposure to arsenic (or co‐intervention designed to lower arsenic exposure) may alter the effect of folic acid intervention on arsenic toxicity‐related outcomes. Therefore, we provided a description of the results from each included study. In order to avoid introducing bias into the narrative synthesis, we attempted to report the results of each included RCT without inappropriate emphasis on the findings of any one particular study. Our narrative summary includes the following descriptions:

• study design and participants with any details available about baseline arsenic exposure and nutritional status;

• types of interventions and their implementation;

• outcomes measured in the included studies and study findings;

• any adverse outcomes or potential harms found if available in the included studies; and

• any pertinent contextual details available for the included studies.

Table 3 describes observational and interventional studies that did not meet our inclusion criteria, but provided data on arsenic toxicity (including arsenic species in blood and urine, and oxidative DNA damage), and arsenic‐related health outcomes (including skin lesions, diabetes, neural tube defects, cancer, cognitive performance, developmental delay) in relation to folate. We also described observational studies that assessed the association between the use of folic acid supplementation and folate status in arsenic‐exposed populations.

Subgroup analysis and investigation of heterogeneity

We did not perform our preplanned subgroup analyses due to insufficient data.

Sensitivity analysis

We did not perform our preplanned sensitivity analyses due to insufficient data.

Summary of findings and assessment of the certainty of the evidence

We used GRADEpro GDT to create the summary of findings tables (GRADEpro GDT). We created summary of findings tables for 'Folic acid supplements alone versus placebo' and 'Folic acid supplements plus other nutrient supplements versus other nutrient supplements alone'. The summary of findings tables include a narrative description of:

• any type of cancer

• all‐cause mortality

• neurocognitive function

• any congenital anomalies

• blood arsenic concentration (μg/L)

• urinary arsenic concentration (μg/L) and urinary arsenic metabolite proportions (%)

• plasma homocysteine concentration (μmol/L)

None of the studies measured cancer, all‐cause mortality, neurocognitive function, or congenital anomalies. We used either change data or end of follow‐up data for blood or urinary arsenic or plasma homocysteine concentration, as reported in the study reports.

Two review authors (SB and EK) independently assessed the certainty of evidence for each outcome, using the GRADE approach (Balshem 2011). Disagreements were resolved through discussion with other review authors. We graded evidence as one of four levels of certainty (high, moderate, low, or very low), depending on the presence of the following five factors: 1) within‐study risk of bias (e.g. limitations in study design and implementation, such as a lack of allocation concealment or blinding, and a large loss to follow‐up); 2) indirectness of evidence; 3) unexplained heterogeneity or inconsistency of results; 4) imprecision of results (e.g. few participants, few events); and 5) high probability of publication bias.

Results

Description of studies

Results of the search

We identified 12,930 records from the electronic searches (see Search methods for identification of studies), and imported them into Covidence (Covidence). After removing 4584 duplicates, we screened the titles and abstracts of 8346 records. A total of 8246 records were clearly irrelevant, leaving 100 records for which we retrieved the full text (see Figure 1). For this review, we included two randomised controlled trials (RCT), in eight records, that met our inclusion criteria (FACT 2015; NIAT 2006). We excluded 26 records, and identified a further 20 duplicate records. Two RCTs are ongoing or unpublished (NCT02235948; NCT03384862). Details of all studies are provided in the Characteristics of included studies tables, Characteristics of excluded studies tables, and Characteristics of ongoing studies tables.

1.

PRISMA study selection flowchart

non‐RCT: non‐randomised controlled trial

Although the inclusion of observational studies was beyond the scope of this review, we provided a summary of 41 observational studies (44 records) on folate, arsenic toxicity and arsenic‐related health outcomes in Table 3.

Included studies

We identified eight records, including seven full‐text articles and one conference abstract, for two RCTs that met our inclusion criteria (FACT 2015; NIAT 2006). These trials reported the effects of folic acid supplementation, alone or in combination with other nutrients, in arsenic‐exposed men and women. We did not identify any RCTs examining the provision of folic acid through fortified foods, or on the effects of the provision of providing folic acid to arsenic‐exposed children.

Study design and participants

The two included studies were randomized, double‐blind, controlled trials of supplementation with folic acid. The participants in both included RCTs were recruited from the Health Effects of Arsenic Longitudinal Study (HEALS), a prospective cohort study of 11,746 married adults, aged 20 to 65 years old, who were drinking from their current well for at least three years and living within a 25 km² region in Araihazar, Bangladesh. We contacted the trial authors of both studies, and obtained further information, indicating that there was no overlap between the participants of the two included RCTs.

FACT 2015 had a cross‐over design; we included the data from the first period of the study (up to 12 weeks), and from the second period of the study (12 to 24 weeks) for groups continuing to receive the same intervention as the first period.

FACT 2015 randomly selected 622 participants from the HEALS cohort, which comprised both folate‐deficient and folate‐sufficient individuals who had been drinking well water with arsenic concentrations higher than 50 µg/L for at least one year. NIAT 2006 enrolled 200 participants, randomly selected from the 550 participants who were previously found to have low plasma folate concentrations (9 nmol/L or lower) in a cross‐sectional study nested within the HEALS.

The baseline mean plasma folate concentrations in the intervention groups in FACT 2015 ranged from 15.4 nmol/L to 17.9 nmol/L; those in NIAT 2006 ranged from 7.8 nmol/L to 8.3 nmol/L. The participants in both trials had been exposed to a wide range of water arsenic concentrations. NIAT 2006 reported that 81% of the study participants used drinking water with arsenic concentrations greater than 10 µg/L, and 62% of the participants used drinking water with arsenic concentrations greater than 50 µg/L. The baseline mean water arsenic concentrations in the intervention groups in FACT 2015 ranged from 120.4 µg/L to 146.6 µg/L; those in NIAT 2006 ranged from 104 µg/L to 106 µg/L. Baseline nutritional variables (e.g. plasma folate, vitamin B12, homocysteine), and arsenic exposure variables (e.g. water arsenic, urinary, and blood arsenic) did not differ among the intervention groups in either trial.

Interventions and comparisons

We aimed to assess the following two comparisons:

• folic acid supplements alone versus placebo; and

• folic acid supplements plus other nutrient supplements versus nutrient supplements alone.

FACT 2015 was a 24‐week study that included 1) a placebo group; 2) a group that received folic acid supplements (either at 400 µg/d or 800 µg/d); 3) a group that received folic acid supplements (400 µg/d) in combination with creatine supplements (3 g/d); and 4) a group that received creatine supplements (3 g/d) alone. The study authors compared the folic acid supplementation groups to the placebo group, and compared the folic acid plus creatine supplementation group to the placebo group. Participants received arsenic removal water filters at the beginning of the trial to provide low arsenic water (< 10 µg/L), and were instructed to use filtered water for drinking and cooking. Participants received and retained two bottles of pills (one bottle containing pills of folic acid or placebo; the second bottle containing pills of creatine or placebo). Village health workers either witnessed or enquired about compliance daily. All bottles were retained, and pills were counted at the end of the first period of the cross‐over (i.e. at week 12). The percentage of pills taken by each participant ranged from 79.1% to 100%.

NIAT 2006 compared the provision of folic acid supplements (400 µg/d) with placebo, given to the participants for 12 weeks of the study period. Each study participant was assigned one bottle containing 100 tablets of folic acid or placebo. Field staff retained the bottles of folic acid or placebo and visited each participant's home daily to observe the participant taking the tablet.

No RCTs assessed the effects of food fortified with folic acid, alone or in combination with other nutrients, versus unfortified food or food fortified with the exact same formulation of other nutrients.

Outcomes

Neither RCT provided data on the review's primary outcomes of interest (i.e. any type of cancer, all‐cause mortality, neurocognitive function, and congenital anomalies), or skin lesions. The included RCTs reported data on total arsenic concentrations in blood and urine (normalized to urinary creatinine), and concentrations of blood arsenic metabolites (i.e. inorganic arsenic (InAs), monomethylarsonic acid (MMA), and dimethylarsonous acid (DMA)), and the proportion of total urinary arsenic excreted as inorganic arsenic (percentage of InAs), MMA (percentage of MMA), and DMA (percentage of DMA). Both included RCTs also reported plasma concentrations of folate and homocysteine; FACT 2015 reported plasma folate concentrations in a figure, so we contacted the trial author and obtained the actual numerical values. FACT 2015 did not report the direct comparison of the folic acid (400 μg/d) plus creatine (3 g/d) supplementation group to the creatine supplementation (3 g/d) group for the outcomes in the review, except for urinary arsenic metabolites.

Funding sources

The included studies were both funded by government programs.

Excluded studies

Although the inclusion of observational studies was beyond the scope of this review, there is substantial evidence of the association between folate status and arsenic‐related health outcomes from observational studies. Table 3 provides a brief description of the available evidence from observational studies that were not eligible for this review, but provided data on folate and arsenic toxicity or arsenic‐related health outcomes.

We excluded 26 studies after full‐text screening: 19 due to ineligible study design and that did not contribute data to Table 3; 4 due to ineligible participants; and 3 due to ineligible interventions. Descriptions of excluded studies, with the reasons for exclusion, are set out in the Characteristics of excluded studies tables.

Ongoing studies

We identified two ongoing studies. Descriptions of ongoing studies are provided in the Characteristics of ongoing studies tables. We categorized a study as ongoing if the primary outcome paper has not yet been published but available information indicates that the study meets review eligibility criteria.

NCT02235948 is a randomized, double‐blind, placebo‐controlled trial, aimed at determining whether folic acid supplementation is effective on arsenic lowering in a chronic, low‐level arsenic‐exposed adult population. Outcome measures include changes of arsenic metabolites between baseline and week eight. The estimated study completion date was December 2016. The ClinicalTrials.gov record states the outcome of oxidative DNA damage was published; however, the primary outcome (changes of arsenic metabolites) paper has not yet been published. We emailed the study authors, but did not receive additional information on other published or unpublished outcomes from this trial.

NCT03384862 is a randomized, placebo‐controlled trial with children aged 8 to 10 years old in Bangladesh, aimed at evaluating whether folic acid and vitamin B12 supplementation can increase arsenic methylation, lower blood arsenic, and improve arsenic‐related reductions in cognitive function. Outcome measures include change in arsenic methylation patterns in blood and urine, change in arsenic level in blood, change in monomethyl arsenic level in blood, and change in cognitive function test score, measured up to 12 weeks.

Risk of bias in included studies

See Figure 2 and Figure 3 for an overall assessment of the risk of bias for included studies. Details of included studies are provided in Characteristics of included studies.

2.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies

3.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study

Allocation

Sequence generation

Both trials reported the studies as randomized, but the methods used to generate the random sequence were not described. We contacted the trial author and obtained further information, which indicated that in both studies, a computer‐generated random sequence was provided by a statistician. Thus, we rated these studies as being at low risk for this domain.

Allocation concealment

In FACT 2015, a pharmacist in the field (Bangladesh) distributed identical pill bottles, labeled with a barcode, to field staff sequentially as they enrolled participants, and bottles were distributed in the order of the random assignment list generated by a statistician. Thus, we rated this study as being at low risk of bias for this domain. NIAT 2006 did not provide a description of the methods used to conceal allocation, so we contacted the trial author and obtained further information that indicated that all pill bottles were identical, and labeled with the study participant's ID numbers with a barcode label. Thus, we assessed this trial as being at low risk of bias for this domain.

Blinding

Blinding of participants and personnel

Both trials reported the studies as double‐blind, thus we assessed the studies as being at low risk of bias for blinding of participants and personnel. In both studies, all study participants, field workers, laboratory technicians, and study investigators, except for two data management specialists, were blinded for the entire duration of the study.

Blinding of outcome assessment

In both trials, the outcomes were assessed by laboratory technicians blinded to the study arms. Thus, we assessed both studies as being at low risk of detection bias.

Incomplete outcome data

In FACT 2015, 12 participants out of 622 enrolled were dropped over the course of the study: 6 were dropped for adverse events, including abdominal cramps, hypertension, severe vertigo and bilateral hydronephrosis; 3 due to pregnancy; 1 due to missing baseline blood sample; and 2 for unexplained reasons. All analyses were performed on an intention‐to‐treat basis. Since the attrition rate (1.9%) was low, and the reasons for withdrawal seem to be unrelated to either treatment, we assessed this trial as being at low risk of bias for this domain.

Out of the 200 study participants enrolled in NIAT 2006, six dropped out because they were unavailable to meet with the field staff to receive the folic acid or placebo tablet daily. Of these six participants, three had been assigned to the folic acid group and three to the placebo group. Because the number of dropouts was balanced across groups, and the reasons for dropping out seemed to be unrelated to either treatment, we assessed this trial as being at low risk of bias for this domain.

Selective reporting

We had access to a clinical trial registration for FACT 2015, and assessed this trial as being at low risk of outcome reporting bias based on the prespecified outcomes provided in the registry.

For NIAT 2006, the outcome measures were reported as described in the methods section, thus we rated this trial as being at low risk of bias for this domain.

Other potential sources of bias

In FACT 2015, household‐level arsenic removal water filters were given to all participants at baseline, and the participants were instructed to exclusively use filtered water. However, no detailed information was provided as to whether the compliance of filter usage was balanced across groups. Due to the uncertainty of the filter compliance, we assessed this trial as being at unclear risk of bias for this domain.

In NIAT 2006, groups appeared balanced at baseline, and other sources of bias were not apparent. Therefore, we assessed this trial as being at low risk of bias for other potential biases.

Effects of interventions

See: Table 1; Table 2

Summary of findings 1. Folic acid supplements alone versus placebo for reducing arsenic toxicity in arsenic‐exposed adults.

Folic acid supplements alone versus placebo for reducing arsenic toxicity in arsenic‐exposed adults
Patient or population: adult men and women who had been exposed to arsenic‐containing drinking water Setting: households in Arailhazar, Bangladesh Intervention: folic acid supplements (400 µg/d or 800 µg/d for 12 weeks) Comparison: placebo
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with placebo	Risk with folic acid supplements (400 µg/d or 800 µg/d)
Any type of cancer	‐	‐	‐	(0 RCTs)	‐	No studies reported on any type of cancer
All‐cause mortality	‐	‐	‐	(0 RCTs)	‐	No studies reported on all‐cause mortality
Neurocognitive function	‐	‐	‐	(0 RCTs)	‐	No studies reported on any type of cancer
Any congenital anomalies	‐	‐	‐	(0 RCTs)	‐	No studies reported on any congenital anomalies
Blood arsenic Measured by change in total blood arsenic, InAs, MMA, and DMA from baseline to week 12	Two studies reported that folic acid supplementation (400 µg/d or 800 µg/d) decreased total blood arsenic concentrations to a greater extent than the placebo group, which suggests the beneficial effects of folic acid supplementation on reducing body burden of arsenic. One study (406 participants) reported that the GM of blood arsenic concentrations decreased by 17.8% (95% CI −25.0 to −9.8) in the folic acid (800 µg/d) supplementation group and by 9.5% (95% CI −16.5 to −1.8) in the placebo group. One study (130 participants) reported that within‐person total blood arsenic decreased more in the 400 µg/d folic acid supplementation group (M 13.62%, SE ± 2.87) than in the placebo group (M 2.49%, SE ± 3.25). Blood InAs concentration decreased by 18.54% (SE ± 3.60) in the folic acid (400 µg/d) supplementation group and by 10.61% (SE ± 3.38) in the placebo group (P = 0.075). Blood MMA concentration decreased more in the folic acid (400 µg/d) supplementation group (M 22.24%, SE ± 2.86) than in the placebo group (M 1.24%, SE ± 3.59; P < 0.001). There was little to no change in blood DMA concentration in either the folic acid (400 µg/d) or placebo groups.		‐	536 (2 RCTs)	⊕⊕⊕⊝ Moderatea	‐
Urinary arsenic Measured by change in percentage (%) of InAs, MMA, and DMA from baseline to week 12	Two studies reported a greater percentage increase in % of DMA and decrease in % of MMA and % of InAs in the folic acid groups than in the placebo groups, suggesting that folic acid supplementation enhances arsenic methylation. One study (352 participants) found that the mean decreases in % of InAs and % of MMA, and the mean increase in % of DMA were greater in the folic acid (400 µg/d or 800 µg/d) groups than in the placebo group (P < 0.05). The mean within‐person change in % of InAs was −0.09% (95% CI −0.17 to −0.01) for the 400 µg/d folic acid group, −0.14% (95% CI −0.21 to −0.06) for the 800 µg/d folic acid group, and 0.05% (95% CI 0.00 to 0.10) for the placebo group. The mean within‐person change in % of MMA was −1.80% (95% CI −2.53 to −1.07) for the 400 µg/d folic acid group, −2.60% (95% CI −3.35 to −1.85) for the 800 µg/d folic acid group, and 0.15%, (95% CI −0.37 to 0.68) for the placebo group. The mean within‐person change in % of DMA was 3.25% (95% CI 1.81 to 4.68) in the 400 µg/d folic acid group, 4.57% (95% CI 3.20 to 5.95) in the 800 µg/d folic acid group, and −1.17% (95% CI −2.18 to −0.17) in the placebo group. One study (194 participants) reported that the mean percentage decrease in % of InAs in the folic acid (400 µg/d) group (M −0.31%, SE ± 0.04) were greater than those in the placebo group (M −0.13%, SE ± 0.04; P < 0.001). The mean percentage decreases in % of MMA in the folic acid (400 µg/d) group (M −2.6%, SE ± 0.37) were greater than those in the placebo group (M −0.71%, SE ± 0.43; P < 0.001). The mean percentage change in % of DMA increased to a greater extent in the folic acid (400 µg/d) group (M 5.9%, SE ± 0.82) than in the placebo group (M 2.14%, SE ± 0.71; P < 0.001).		‐	546 (2 RCTs)	⊕⊕⊕⊝ Moderatea	‐
Plasma homocysteine Measured by change in plasma homocysteine concentrations from baseline to week 12	Two studies reported that supplementation of folic acid (400 μg/d) decreased plasma total homocysteine concentrations, whereas no changes of plasma total homocysteine were observed in the placebo group. In one study (254 participants), homocysteine decreased more in the folic acid group (GM −23.4%, 95% CI −27.1 to −19.5) than in the placebo group (GM −1.3%, 95% CI −5.3 to 3.1). In one study (194 participants), within‐person homocysteine decreased more in the folic acid group (M −3.06 µmol/L, SE ± 3.51), than in the placebo group (M −0.05 µmol/L, SE ± 4.31).		‐	448 (2 RCTs)	⊕⊕⊕⊝ Moderatea	‐
*The risk in the intervention group (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval;DMA: dimethyl arsinic acid; GM: geometric mean; InAs: inorganic arsenic; M: mean; MMA: monomethylarsonic acid; RCT: randomised controlled trial; SE: standard error
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect

^aDowngraded by one level due to imprecision; two studies with relatively small sample sizes

Summary of findings 2. Folic acid supplements plus other nutrient supplements versus nutrient supplements alone for reducing arsenic toxicity in arsenic‐exposed adults.

Folic acid supplements plus other nutrient supplements versus nutrient supplements alone for reducing arsenic toxicity in arsenic‐exposed adultsa
Patient or population: adult men and women who had been drinking from a household well with water arsenic concentrations greater than 50 μg/L for at least 1 year Setting: households in Arailhazar, Bangladesh Intervention: supplementation of folic acid (400 µg/d) in combination with creatine (3 g/d) for 12 weeks Comparison: creatine (3 g/d) alone
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with creatine (3 g/d) alone	Risk with supplementation of folic acid (400 µg/d) plus creatine (3 g/d)
Any type of cancer	‐	‐	‐	(0 RCTs)	‐	No studies reported on any type of cancer
All‐cause mortality	‐	‐	‐	(0 RCTs)	‐	No studies reported on all‐cause mortality
Neurocognitive function	‐	‐	‐	(0 RCTs)	‐	No studies reported on any type of cancer
Any congenital anomalies	‐	‐	‐	(0 RCTs)	‐	No studies reported on any congenital anomalies
Blood arsenic Measured by change in GM of blood arsenic concentrations from baseline to week 12	The GM of blood arsenic concentration decreased 14% (95% CI −22.2 to −5.0) in the folic acid plus creatine supplementation group and by 7% (95% CI −14.8 to 1.5) in the creatine supplementation group.		‐	204 (1 RCT)	⊕⊕⊝⊝ Lowb	‐
Urinary arsenic Measured by percentage (%) of InAs, MMA, and DMA from baseline to week 12	The mean % of InAs was 13.2 (SE ± 7.0) in the folic acid plus creatine supplementation group and 14.8 (SE ± 5.5) in the creatine supplementation group. The mean % of MMA was 10.8 (SE ± 4.1) in the folic acid plus creatine supplementation group and 12.8 (SE ± 4.0) in the creatine supplementation group. The mean % of DMA was 76 (SE ± 7.8) in the folic acid plus creatine supplementation group and 72.4 (SE ± 7.6) in the creatine supplementation group.		‐	190 (1 RCT)	⊕⊕⊝⊝ Lowb	‐
Plasma homocysteine Measured by change in GM of plasma homocysteine concentrations from baseline to week 12	The GM of plasma total homocysteine concentration decreased by 21% (95% CI −25.2 to −16.4) in the folic acid plus creatine supplementation group and by 4.3% (95% CI −9.0 to 0.7) in the creatine supplementation group.		‐	204 (1 RCT)	⊕⊕⊝⊝ Lowb	‐
*The risk in the intervention group (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; DMA: dimethyl arsinic acid; GM: geometric mean; InAs: inorganic arsenic; MMA: monomethylarsonic acid; RCT: randomised controlled trial; SE: standard error.
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect

^aOne RCT compared folic acid plus creatine supplementation to creatine supplementation (FACT 2015). The study reported results for folic acid plus creatine supplementation versus placebo, and creatine supplementation versus placebo, but not folic acid plus creatine supplementation versus creatine supplementation for the outcomes in this study, except for urinary arsenic metabolites. For this review, we reported the data from folic acid plus creatine supplementation versus creatine supplementation.
^bDowngraded by two levels due to imprecision: only one study with small sample size

Because there were only two included RCTs, which were heterogeneous across study participants (i.e. differences in baseline arsenic exposure and nutritional status), co‐interventions (i.e. use of arsenic‐removal water filters), and outcome measures, we were unable to conduct a meta‐analysis (FACT 2015; NIAT 2006). A folic acid intervention will likely have different effects for arsenic‐exposed populations, based on baseline folate status. A lower baseline exposure to arsenic (or a co‐intervention designed to lower arsenic exposure) may alter the effect of folic acid intervention on arsenic toxicity‐related outcomes. Therefore, we described the results from each RCT separately.

We presented Folic acid supplements alone versus placebo in Table 1, and Folic acid supplements plus other nutrient supplements versus nutrient supplements alone in Table 2.

Folic acid supplements alone versus placebo

Both studies compared folic acid supplements of 400 μg/d versus placebo. FACT 2015 also compared folic acid supplements of 800 μg/d versus placebo.

Primary outcomes

Neither study reported on any of the primary outcomes of interest: any type of cancer; all‐cause mortality; neurocognitive function; or any congential anomalies.

Secondary outcomes

Blood arsenic

Both trials reported on total blood arsenic concentrations.

The FACT 2015 study assessed blood arsenic concentrations in two folic acid supplementation groups (400 µg/d (FA400) and 800 µg/d (FA800)), compared to a placebo group. All study participants, from all three groups, received arsenic‐removal filters. Blood arsenic concentration was lower in the FA800 group and was not different in the FA400 group, compared to the placebo group at any follow‐up time point. At week 12, the FA800 group had a GM of 7.19 μg/L (95% CI 6.56 to 7.87; 150 participants), and the FA400 group had a geometric mean (GM) of 8.40 μg/L (95% confidence interval (CI) 7.55 to 9.36; 153 participants), versus the placebo group (GM 7.62 μg/L, 95% CI 6.72 to 8.64; 101 participants).

Compared to the placebo group, the FA800 group showed a greater decline in blood arsenic concentrations from baseline to the end of weeks 1, 6, and 12. By week 12, GM blood arsenic decreased by 3.7% (95% CI −10.7 to 3.8) in the FA400 group, 17.8% (95% CI −25.0 to −9.8) in the FA800 group, and 9.5% (95% CI −16.5 to −1.8) in the placebo group.

The FACT 2015 study also reported blood arsenic concentrations during the second cross‐over phase, comparing those in the folic acid supplementation groups who continued supplementation (FA400/FA and FA800/FA) with the placebo group at week 24. At week 24, the FA800/FA group had a GM of 7.69 μg/L (95% CI 6.60 to 7.96; 76 participants), and the FA400/FA group had a GM of 8.52 μg/L (95% CI 7.30 to 9.95; 77 participants), versus the placebo group (GM 8.12 μg/L, 95% CI 7.23 to 9.13; 102 participants). Compared to the placebo group, blood arsenic decreased more from week 13 to week 24 in the FA800/FA group (mean difference (MD) −0.12 μg/L, 95% CI −0.24 to −0.00; P = 0.04). The study authors reported that the difference between blood arsenic concentrations for the FA800/FA group and the FA800 group that switched to placebo in the second phase was not significant; therefore, the study authors combined these groups in secondary analyses. At the end of 24 weeks, the within‐person decrease in blood arsenic in the two FA800 groups was greater (GM −12.3%, 95% CI −19.6 to −4.3) than in the placebo group (GM −2.3%, 95% CI −8.6 to 4.4; P = 0.05).

NIAT 2006 reported that folic acid supplementation (400 µg/d) decreased the mean (M) total blood arsenic concentrations from baseline (M 9.86 μg/L, standard error (SE) ± 0.62) to week 12 levels (M 8.20 μg/L, SE ± 0.50; P < 0.001; 68 participants). Arsenic concentrations did not change in the placebo group from baseline (M 9.59 μg/L, SE ± 0.63) to week 12 levels (M 9.14 μg/L, SE ± 0.61; P = 0.10; 62 participants). Mean within‐person total blood arsenic levels decreased in the folic acid supplementation group (M −13.62%, SE ± 2.87), and in the placebo group (M −2.49%, SE ± 3.25; P = 0.02).

NIAT 2006 also reported the percentage of within‐person change in blood InAs, MMA, and DMA concentrations in the folic acid (N = 68) and placebo (N = 62) groups. The percentage of change was defined as the difference between baseline and 12‐week blood arsenic, expressed as a percentage of the baseline measure.

Blood InAs concentration decreased in the folic acid group (M −18.54%, SE ± 3.60), and the placebo group (M −10.61%, SE ± 3.38). The change in absolute values of blood InAs concentration were reported for the folic acid group (M −0.58 μg/L, standard deviation (SD) ± 0.91), and the placebo group (M −0.32 μg/L, SD ± 0.73).

Blood MMA concentration decreased more in the folic acid group (M −22.24%, SE ± 2.86), than in the placebo group (M −1.24%, SE ± 3.59; P < 0.001). Changes in absolute values of blood MMA concentration were reported for the folic acid group (M −1.08 μg/L, SD ± 1.46), and in the placebo group (M 0.19 μg/L, SD ± 1.19; P = 0.53).

Blood DMA concentration did not change from baseline to week 12 in either the folic acid group (M 3.24 μg/L, SE ± 0.17 to 3.24 μg/L, SE ± 0.19), or the placebo group (M 3.17 μg/L, SE ± 0.19 to M 3.24 μg/L, SE ± 0.20; P = 0.81), which is likely due to the short circulating half‐life of DMA, and rapid excretion of DMA in urine, as noted by the study authors.

We graded the certainty of the evidence for blood arsenic as moderate (2 studies, 536 participants; downgraded once for imprecision).

Urinary arsenic

The FACT 2015 study reported the proportion of arsenic metabolites (percentage of InAs, percentage of MMA, and percentage of DMA) in urine at baseline (week 0), and after 1 week, 6 weeks, and 12 weeks of folic acid supplementation among 352 participants (FA400 group N = 133; FA800 group N = 129; placebo N = 90). Within‐person changes in the proportion of arsenic metabolites over the first 6 weeks were similar to those over 12 weeks; therefore, they used a parsimonious model, dividing results into three categories, (week 0, week 1, and weeks 6 plus 12).

The mean decreases in the percentages of InAs and MMA, and the mean increase in the percentage of DMA were greater in the FA400 and FA800 groups than in the placebo group at weeks 6 plus 12 (P < 0.05).

At weeks 6 plus 12, the percentage of InAs was reported in the FA400 group (M 13.6%, SE ± 5.6; 125 participants), the FA800 group (M 12.4%, SE ± 4.7; 122 participants), and the placebo group (M 15.5%, SE ± 5.5; 84 participants). The mean within‐person changes from baseline to weeks 6 plus 12 were higher in the FA800 group (M −0.14%, 95% CI −0.21 to −0.06; P < 0.001), and the FA400 group (M −00.09%, 95% CI −0.17 to −0.01; P < 0.05), than in the placebo group (M 0.05%, 95% CI 0.00 to 0.10; P > 0.05).

The mean percentage of MMA in the FA400 group was 11.2% (SE ± 4.1; 125 participants); in the FA800 group, it was 10.7% (SE ± 4.0; 122 participants); and in the placebo group, it was 13.4% (SE ± 3.6; 84 participants). Within‐person changes decreased from baseline to week 12 in the FA800 group (M −2.60%, 95% CI −3.35 to −1.85; P < 0.001), and in the FA400 group (M −1.80%, 95% CI −2.53 to −1.07; P < 0.001), but not in the placebo group (M 0.15%, 95% CI −0.37 to 0.68; P > 0.05).

The mean percentage of DMA in the FA800 group was 76.9% (SE ± 6.3; 122 participants); in the FA400 group, it was 75.3% (SE ± 7.0; 125 participants); and in the placebo group, it was 71.1% (SE ± 7.5; 84 participants). From baseline to week 12, the within‐person changes increased in the FA800 group (M 4.57%, 95% CI 3.20 to 5.95; P < 0.001), and in the FA400 group (3.25%, 95% CI 1.81 to 4.68; P < 0.001), and decreased in the placebo group (M −1.17%, 95% CI −2.18 to −0.17; P > 0.05).

Data were provided for changes in urinary arsenic metabolites at week 24 for the FA800/FA and FA400/FA groups, but not for the placebo group. InAs decreased in the FA800/FA group (M −0.12%, 95% CI −0.19 to −0.04), and in the FA400/FA group (−0.09%, 95% CI −0.17 to −0.02). MMA decreased in the FA800/FA group (M −1.27%, 95% CI −2.01 to −0.54), and in the FA400/FA group (M −1.63%, 95% CI −2.36 to −0.90). DMA increased in the FA800/FA group (M 2.66%, 95% CI 1.31 to 4.01), and in the FA400/FA group (M 2.81%, 95% CI 1.54 to 4.08).

NIAT 2006 reported the percentage of within‐person changes (from baseline to week 1, or from baseline to week 12) in the proportions of total urinary arsenic excreted as InAs, MMA, and DMA, which were assessed in the 194 participants who completed the trial (folic acid group N = 96; placebo group N = 98).

In unadjusted analyses, the proportion of total urinary arsenic excreted as InAs decreased to a greater extent in the folic acid group (M −0.31%, SE ± 0.04) than in the placebo group (M −0.13%, SE ± 0.04; P < 0.001).

At week 12, the proportion of total urinary arsenic excreted as MMA decreased to a greater extent in the folic acid group (M −2.6%, SE ± 0.37) than in the placebo group (M −0.71%, SE ± 0.43; P < 0.001). The proportion of total urinary arsenic excreted as DMA increased to a greater extent in the folic acid group (M 5.9%, SE ± 0.82) than in the placebo group (M 2.14%, SE ± 0.71; P < 0.001).

NIAT 2006 also reported percentage changes (from baseline to week 1, or from baseline to week 12) in the ratios of MMA to InAs (MMA: InAs; known as the primary methylation index (PMI)), and DMA to MMA (DMA: MMA; known as the secondary methylation index (SMI)), as secondary outcome variables.

From baseline to week one, the mean percentage change in the PMI (adjusted for covariates) in the folic acid group was 0.04% (SE ± 0.03; P > 0.05); in the placebo group, it was 0.08% (SE ± 0.04; P < 0.05). From baseline to week 12, the mean percentage change in the PMI in the folic acid group was 0.11% (SE ± 0.05; P < 0.05); in the placebo group, it was 0.13% (SE ± 0.05; P < 0.001). The study authors noted that the effect on the PMI was likely so small because the percentages of both InAs and MMA decreased in the folic acid supplementation group.

From baseline to week one, the mean percentage change in the SMI (adjusted for covariates) in the folic acid group was 1.25% (SE ± 0.18; P < 0.001); in the placebo group, it was 0.04% (SE ± 0.26; P > 0.05). From baseline to week 12, the mean percentage change in SMI in the folic acid group was 2.88% (SE ± 0.43; P < 0.001); in the placebo group, it was 0.40% (SE ± 0.32; P > 0.05).

We graded the certainty of the evidence for urinary arsenic as moderate (2 studies, 546 participants; downgraded once for imprecision).

Plasma folate concentration

Plasma folate was measured to show evidence of compliance with folic acid supplementation.

FACT 2015 assessed plasma folate concentrations in two folic acid supplementation groups (FA400 and FA800), and in the placebo group. Since geometric means (GM; nmol/L) and 95% CIs of plasma folate levels were provided in a figure, we requested and obtained the actual numerical values from the trial authors.

Plasma folate concentrations increased in both the FA800 (from baseline GM 14.52 nmol/L, 95% CI 13.19 to 15.98; to 12‐week GM 56.84 nmol/L, 95% CI 49.47 to 65.32; 121 participants), and the FA400 groups (from baseline GM 13.51 nmol/L, 95% CI 12.26 to 14.90; to 12‐week GM 44.04 nmol/L, 95% CI 39.77 to 48.77; 125 participants), to a greater extent than in the placebo group (from baseline GM 13.63 nmol/L, 95% CI 12.21 to 15.22; to 12‐week GM 17.7 nmol/L, 95% CI 15.42 to 20.32; 84 participants).

Plasma folate concentrations did not change during the second phase of FACT 2015 in the FA800/FA group (12‐week GM 54.57 nmol/L, 95% CI 44.78 to 66.50; 24‐week GM 48.56 nmol/L, 95% 41.21 to 57.21); the FA400/FA group (12‐week GM 42.82 nmol/L, 95% CI 36.55 to 50.16; 24‐week GM 41.23 nmol/L, 95% CI 36.20 to 46.95); or in the placebo group (12‐week GM 17.70 nmol/L, 95% CI 15.40 to 20.35; 24‐week GM 16.04 nmol/L, 95% CI 14.07 to 18.29); participant numbers not reported.

NIAT 2006 reported plasma folate concentrations before and after 12 weeks of folic acid (400 μg/d) or placebo supplementation. The mean plasma folate concentration changes were higher in the folic acid group (baseline M 8.28 nmol/L, SE ± 4.74; to 12‐week M 61.57 nmol/L, SE ± 27.67) than in the placebo group (baseline M 7.81 nmol/L, SE ± 2.96; to 12‐week M 8.86 nmol/L, SE ± 9.14). The within‐person changes in the folic acid group were higher (M 53.29 nmol/L, SE ± 27.84; P < 0.001; 96 participants) than in the placebo group (M 1.05 nmol/L, SE ± 9.32, P > 0.05; 98 participants).

Plasma homocysteine concentration

Plasma homocysteine was measured to show evidence of compliance with folic acid supplementation.

FACT 2015 reported means and geometric means of plasma homocysteine at baseline and week 12, and the percentage of change in the geometric mean from baseline to week 12.

Plasma homocysteine concentrations decreased in both the FA800 (from baseline M 14.0 μmol/L, SD ± 10.9; to 12‐week M 9.6 nmol/L, SD ± 4.1; 122 participants) and the FA400 groups (from baseline M 13.8 μmol/L, SD ± 9.2; to 12‐week M 9.8 μmol/L, SD ± 3.3; 125 participants), to a greater extent than in the placebo group (from baseline M 13.1 μmol/L, SD ± 5.7; to 12‐week M 13.0 μmol/L, SD ± 5.9; 84 participants; Treatment group differences P < 0.05).

Plasma homocysteine concentration decreased from baseline to week 12 in the FA400 group by 23.4% (95% CI −27.1 to −19.5; P < 0.001; 153 participants), but not in the placebo group (−1.3%, 95% CI −5.3 to 3.1; P > 0.05; 101 participants).

Homocysteine concentrations were not reported for the second phase of the FACT 2015 study.

NIAT 2006 reported plasma total homocysteine concentrations before and after 12 weeks of folic acid (400 μg/d) or placebo supplementation.

Mean plasma homocysteine concentration decreased in the folic acid group at the group level (baseline M 11.14 μmol/L, SE ± 4.03; 12‐week M 8.08 μmol/L; SE ± 2.29), and at the within‐person level (M −3.06 μmol/L, SE ± 3.51; P < 0.001; 96 participants). There was very little change in the placebo group at the group level (baseline M 11.41 μmol/L, SE ± 8.21; 12‐week M 11.37 μmol/L, SE ± 9.29), or at the within‐person level (M −0.05 μmol/L, SE ± 4.31; P > 0.05; 98 participants).

We graded the certainty of the evidence for plasma homocysteine as moderate (2 studies; 448 participants; downgraded once for imprecision).

Skin lesions

Neither of the included RCTs assessed skin lesions.

Folic acid supplements plus other nutrient supplements versus nutrient supplements alone

FACT 2015 studied this comparison. They compared folic acid (400 μg/d) plus creatine (3 g/d) to a placebo group, and compared creatine supplementation (3 g/d) to the placebo group. Articles reporting on the FACT 2015 study did not report the direct comparison of the folic acid plus creatine supplementation group to the creatine supplementation group for the outcomes in this review, except for urinary arsenic metabolites.

This review aimed to compare folic acid supplements plus other nutrient supplements versus nutrients supplements alone (Types of interventions). Therefore, for the purpose of this review, we report the data from the folic acid plus creatine supplementation group and the creatine only supplementation group.

Primary outcomes

FACT 2015 did not report on any of the primary outcomes of interest: any type of cancer; all‐cause mortality; neurocognitive function; or any congential anomalies.

Secondary outcomes

Blood arsenic

FACT 2015 reported geometric means (GM) of blood arsenic concentrations and the percent change in blood arsenic from baseline to each follow‐up time point (i.e. week 1, week 6 and week 12). Blood arsenic concentration decreased by 14% (95% CI −22.2 to −5.0; 103 participants) in the folic acid plus creatine group from baseline (GM 9.26 μg/L, 95% CI 8.38 to 10.24) to week 12 (GM 7.96 μg/L, 95% CI 7.04 to 9.00). Blood arsenic concentration decreased by 7% (95% CI −14.8 to 1.5; 101 participants) in the creatine only group from baseline (GM 8.73 μg/L, 95% 7.72 to 9.87) to week 12 (GM 8.12 μg/L, 95% CI 7.20 to 9.15).

We graded the certainty of the evidence for blood arsenic as low (1 study, 204 participants; downgraded twice for imprecision).

Urinary arsenic

FACT 2015 reported the proportion of arsenic metabolites (percentage of InAsc, percentage of MMA, and percentage of DMA) in urine at baseline and after supplementation at week 1, week 6 and week 12. At week 12, the mean percentage of InAs was 13.2% (SE ±7.0; 88 participants) in the folic acid plus creatine supplementation group and 14.8% (SE ± 5.5; 88 participants) in the creatine supplementation group. The mean percentage of MMA was 10.8% (SE ± 4.1; 88 participants) in the folic acid plus creatine supplementation group and 12.8% (SE ± 4.0; 88 participants) in the creatine supplementation group. The mean percentage of DMA was 76% (SE ± 7.8; 88 participants) in the folic acid plus creatine supplementation group and 72.4% (SE ± 7.6; 88 participants) in the creatine supplementation group. The study authors stated that changes in all urinary arsenic metabolites were not significantly different between the folic acid plus creatine group and the creatine group.

We graded the certainty of the evidence for urinary arsenic as low (1 study, 290 participants; downgraded twice for imprecision).

Plasma folate concentration

FACT 2015 reported the mean concentrations of plasma folate at baseline and week 12. In the folic acid plus creatine group, the plasma folate concentration was 15.3 nmol/L (SE ± 8.8; 96 participants) at baseline and 81.1 nmol/L (SE ± 144.1; 88 participants) at week 12. In the creatine group, the plasma folate concentration was 16.0 nmol/L (SE ± 7.6; 93 participants) at baseline and 41.5 nmol/L (SE ±101.1; 88 participants) at week 12. The study authors stated that the changes in the creatine group were similar to the placebo group, and the changes in the folic acid plus creatine groups were similar to the FA400 group.

Plasma homocysteine concentration

FACT 2015 reported the geometric mean of plasma homocysteine at baseline and week 12 and the percent change in the geometric mean from baseline to week 12. Plasma total homocysteine concentration decreased by 21% (95% CI −25.2 to −16.4) in the folic acid plus creatine supplementation group from baseline (GM 11.8 μmol/L, 95% CI 10.9 to 12.7) to week 12 (GM 9.3 μmol/L, 95% CI 8.8 to 9.8; P < 0.001; 103 participants); and by 4.3% (95% CI −9.0 to 0.7) in the creatine supplementation group from baseline (GM 11.4 μmol/L, 95% CI 10.5 to 12.3) to week 12 (GM 10.9 μmol/L, 95% CI 10.1 to 11.7; P = 0.09; 101 participants).

We graded the certainty of the evidence for plasma homocysteine as low (1 study, 204 participants; downgraded twice for imprecision).

Skin lesions

FACT 2015 did not assess skin lesions.

Discussion

Summary of main results

We included two double‐blind, placebo‐controlled randomised controlled trials (RCT) on folic acid supplementation in 822 adults who had been exposed to arsenic‐contaminated drinking water in Bangladesh. We also identified two ongoing RCTs in children (NCT03384862; NCT02235948). We did not identify any RCTs that assessed effects of providing folic acid through fortified food.

Neither of the included RCTs measured any of our primary outcomes: cancer, all‐cause mortality, neurocognitive function, and congenital anomalies. Due to clinical heterogeneity between the two included RCTs (i.e. differences in baseline arsenic exposure and nutritional status of the study participants, and co‐intervention), we could not combine the study results in a meta‐analysis; therefore, we provided a narrative description.

Folic acid supplements alone versus placebo

The available evidence from the two RCTs indicated that 12 weeks of folic acid supplementation might lower blood arsenic concentrations. In study participants with low baseline plasma folate status, folic acid supplementation of 400 µg/d was shown to reduce total blood arsenic and MMA concentrations to a greater extent than placebo. In the mixed folate‐deficient and ‐replete study population who received arsenic‐removal water filters as the co‐intervention, 800 µg/d folic acid supplements (but not 400 µg/d folic acid supplements) reduced blood arsenic levels to a greater extent than placebo. Folic acid supplementation (400 µg/d or 800 µg/d) to study participants in both studies increased the proportion of total urinary arsenic excreted as DMA to a greater extent than placebo, suggesting that folic acid supplementation in this population enhances arsenic methylation, thereby facilitating urinary excretion of arsenic. Results from both included studies also showed that supplementation of folic acid (400 µg/d or 800 µg/d) decreased plasma homocysteine concentration more than placebo in arsenic‐exposed adults.

Folic acid supplements plus other nutrient supplements versus nutrient supplements alone

In one study, 400 µg/d of folic acid supplementation plus creatine reduced blood arsenic and homocysteine concentrations to a greater extent than creatine alone, but did not change the proportion of total urinary arsenic metabolites.

Taken together, the available evidence from two RCTs demonstrated consistent effects of folic acid on blood arsenic and urinary arsenic methylation profiles, suggesting beneficial effects of folic acid supplementation on reducing arsenic toxicity in arsenic‐exposed adults.

Overall completeness and applicability of evidence

This review includes data from two RCTs that took place in rural Bangladesh. Both studies were conducted in adults who had been exposed to arsenic‐contaminated drinking water and who were either folate‐deficient or both folate‐deficient and folate‐sufficient; no studies were conducted in children. The studies assessed the effects of folic acid supplements at a limited range of doses (400 µg/d and 800 µg/d doses). Only one study included co‐interventions to reduce arsenic exposure from drinking water (e.g. arsenic‐removal water filters) and neither study assessed the effect of folic acid in the presence of other mitigation strategies. Although the available evidence shows that folic acid supplementation lowered blood arsenic concentrations, we cannot rule out the possibility that effects of folic acid supplements may be different in other contexts, populations (e.g. children), and with different doses and co‐interventions.

This review only included RCTs, neither of which assessed the effects of folic acid on our prespecified primary outcomes of cancer, all‐cause mortality, neurocognitive function, and congenital anomalies. However, there is substantial evidence of the associations between folate and arsenic toxicity, or arsenic‐related health outcomes from observational studies (e.g. cohort study, case‐control study) and non‐randomised intervention studies. Although the inclusion of observational studies was beyond the scope of this review, our literature search identified 41 relevant observational studies. Of these, 23 were cross‐sectional, 13 were case‐control, 2 were cohort, and 3 were interventional studies. The studies evaluated potential associations between folate and other one‐carbon metabolites, arsenic toxicity, or arsenic‐related health outcomes. Twenty‐eight out of 41 studies reported favorable associations between folate and arsenic metabolites, or arsenic‐related health outcomes; two reported favorable associations between folic acid supplements and folate status in an arsenic‐exposed population; nine studies did not find any associations between dietary factors and urinary arsenic excretion patterns; and two studies reported unfavorable associations. For a description of these studies, which we plan to use for future updates of this review, see Table 3.

Given the limited number of RCTs identified, more studies are needed to assess the effects of provision of folic acid on arsenic‐related health outcomes and arsenic toxicity in arsenic‐exposed adults and children.

Quality of the evidence

Both included RCTs had a low risk of selection bias, performance bias, detection bias, attrition bias, and reporting bias, suggesting that both studies were well designed and conducted.

We assessed the certainty of the evidence across studies using the GRADE approach for outcomes measured in the RCTs: blood and urinary arsenic, and homocysteine concentrations.

Folic acid supplements alone versus placebo

We assessed the certainty of the evidence for blood and urinary arsenic outcomes and plasma homocysteine as moderate. We downgraded the evidence by one level for the imprecision of results, because there were only two included studies, with relatively small sample sizes.

Folic acid supplements plus other nutrient supplements versus nutrient supplements alone

We assessed the certainty of the evidence for blood and urinary arsenic outcomes and plasma homocysteine as low for this comparison. We downgraded the evidence by two levels for imprecision of results, due to the small sample size (N < 204) from only one RCT.

Potential biases in the review process

In order to minimize the possibility of introducing bias in the review process, two review authors independently assessed eligibility for inclusion of the trials, carried out data extraction, assessed risk of bias for each included study, and assessed the certainty of the evidence for each outcome. Although the process of assessing risk of bias may involve subjective judgements, we provided the rationale to support our judgement for each risk of bias domain. Any disagreements were discussed, and the third review author was involved when necessary. To minimise reporting bias, we attempted to create the search strategy as inclusively as possible, and conducted comprehensive searches for eligible studies from a wide range of international and regional databases. Although we planned to generate funnel plots to assess the possibility of publication bias, we were unable to do this, because we only included two studies.

Agreements and disagreements with other studies or reviews

We are unaware of similar systematic reviews on this topic that include randomised controlled trials (RCT) only. Three nonsystematic literature reviews have examined the relationship between nutritional status or intake (e.g. folate) and arsenic toxicity and arsenic‐related health risk from animal studies, case reports, and observational studies (Bozack 2018; Hall 2012; Kile 2008). Two of these reviews also described the evidence from the RCTs included in this review, reporting results on urinary arsenic metabolites (Bozack 2018; Hall 2012) and concentrations of arsenic and arsenic metabolites in blood (Bozack 2018); they did not report on any other outcome included in this review (i.e. plasma homocysteine and plasma folate), nor did they evaluate the risk of bias either. All three reviews concluded that higher folate status or intake is associated with improved arsenic methylation (e.g. lower percentage of MMA and higher percentage of DMA in urine), which is similar to the results from the studies included in this review. They also found that higher folate status or intake is beneficial for arsenic‐related health outcomes, which is similar to the observational studies described in Table 3.

Meta‐analyses of RCTs conducted in arsenic‐unexposed populations have demonstrated that folic acid supplementation decreased plasma homocysteine (Duffy 2014; HLTC 1998; HLTC 2005), which is consistent with the results from this review in arsenic‐exposed populations.

Authors' conclusions

Implications for practice.

Moderate‐certainty evidence from two included randomised controlled trials (RCTs) found that folic acid supplementation of 400 µg/d or 800 µg/d likely reduces blood arsenic concentrations, facilitates arsenic methylation, and reduces plasma homocysteine concentrations in adults exposed to arsenic‐contaminated water, regardless of their baseline folate status. The studies also found low‐certainty evidence that folic acid supplementation plus other nutrients likely reduces blood arsenic concentrations and plasma homocysteine concentrations compared to nutrient supplementation alone. This suggests that folic acid supplementation will likely improve arsenic toxicity in adults who had been exposed to arsenic‐contaminated drinking water.

Neither study measured the effects of providing folic acid (through folic acid fortified foods or supplements), alone or in combination with other nutrients, on arsenic‐related health outcomes (i.e. cancer, mortality, neurocognitive function, congenital disorders, and skin lesions) in arsenic‐exposed adult populations. Neither of the RCTs assessed whether folic acid interventions would be efficacious in reducing arsenic toxicity and arsenic‐induced outcomes in arsenic‐exposed children.

Implications for research.

A broad understanding of the role of folic acid supplementation across varying arsenic‐exposed populations — such as by country, age, and gender groups — is lacking. Given the limited number of RCTs identified, more studies conducted in diverse settings with large populations are needed to assess the effects of provision of folic acid on arsenic‐related health outcomes and arsenic toxicity in arsenic‐exposed adults and children. Both short‐term RCTs of biological markers and long‐term observational studies of chronic disease outcomes have the potential to contribute needed information to guide public health decision‐making. Where feasible, RCTs that measure arsenic‐related health outcomes and arsenic toxicity (e.g. concentrations of total and methylated arsenic metabolites), are needed. Due to the uncertainty of the effect of folic acid interventions on improving arsenic‐related health outcomes, there is ethical equipoise for future RCTs. However, a longer‐term trial assessing health outcomes (i.e. cancer, all‐cause mortality, neurocognitive function, congenital anomalies) may have other ethical considerations, such as including populations that are folate‐deficient or including women who are, or may become pregnant, or providing co‐interventions to lower water arsenic concentrations. There are few observational studies that investigate the associations of folic acid and outcomes in arsenic‐exposed pregnant women, yet this is an important area of study, given that exposure to inorganic arsenic during pregnancy is associated with adverse birth outcomes and congenital anomalies. Research that assesses a dose‐response effect of folic acid supplements on arsenic toxicity and arsenic‐induced illnesses, and potential side effects of this intervention is also lacking. Meta‐analyses of programmatic or observational evidence will inform the potential effectiveness of folic acid interventions, and future systematic reviews of RCTs on this topic will provide useful information on efficacy and dosage of folic acid interventions in arsenic‐exposed populations.

History

Protocol first published: Issue 5, 2017

Appendices

Appendix 1. Search strategies

Cochrane Central Register of Controlled Trials (CENTRAL), in the Cochrane Library

#1 MeSH descriptor: [Metals, Heavy] this term only
#2 MeSH descriptor: [Heavy Metal Poisoning, Nervous System] this term only
#3 (metalloid* or metal poison* or heavy metal*):ti,ab,kw (Word variations have been searched)
#4 MeSH descriptor: [Arsenic] this term only
#5 MeSH descriptor: [Arsenic Poisoning] this term only
#6 MeSH descriptor: [Arsenicals] explode all trees
#7 arsen*:ti,ab,kw (Word variations have been searched)
#8 (arsenic* or arsenite* or arsenate*):ti,ab,kw (Word variations have been searched)
#9 (monomethylarsonic or mono‐methylarsonic):ti,ab,kw (Word variations have been searched)
#10 (monomethylarsonous or mono‐methylarsonous):ti,ab,kw (Word variations have been searched)
#11 (dimethylarsinous or di‐methylarsinous):ti,ab,kw (Word variations have been searched)
#12 (trimethylarsine or tri‐methylarsine):ti,ab,kw (Word variations have been searched)
#13 #1 or #2 or #3 or #4 or #5 or #6 or #7 or #8 or #9 or #10 or #11 or #12
#14 MeSH descriptor: [Folic Acid] explode all trees
#15 MeSH descriptor: [Folic Acid Deficiency] this term only
#16 folate*:ti,ab,kw (Word variations have been searched)
#17 folic:ti,ab,kw (Word variations have been searched)
#18 folinic*:ti,ab,kw (Word variations have been searched)
#19 folacin*:ti,ab,kw (Word variations have been searched)
#20 vitamin B9:ti,ab,kw (Word variations have been searched)
#21 methyltetrahydrofolate:ti,ab,kw (Word variations have been searched)
#22 methylTHF:ti,ab,kw (Word variations have been searched)
#23 metafolin:ti,ab,kw (Word variations have been searched)
#24 leucovorin:ti,ab,kw (Word variations have been searched)
#25 MeSH descriptor: [Micronutrients] this term only
#26 MeSH descriptor: [Food, Fortified] this term only
#27 MeSH descriptor: [Functional Food] this term only
#28 (food* N3 (fortif* or functional* or supplement*)):ti,ab,kw (Word variations have been searched)
#29 (micro‐nutrient* or micronutrient*):ti,ab,kw (Word variations have been searched)
#30 #14 or #15 or #16 or #17 or #18 or #19 or #20 or #21 or #22 or #23 or #24 or #25 or #26 or #27 or #28 or #29
#31 #13 and #30

MEDLINE, MEDLINE In‐Process & Other Non‐Indexed Citations, and MEDLINE Epub Ahead of Print Ovid

1 Metals, Heavy/
2 Heavy Metal Poisoning, Nervous System/
3 (metalloid$ or metal poison$ or heavy metal$).tw,kw.
4 Arsenic/
5 Arsenic Poisoning/
6 exp Arsenicals/
7 arsen$.mp.
8 (arsenic$ or arsenite$ or arsenate$).mp.
9 (monomethylarsonic or mono‐methylarsonic).mp.
10 (monomethylarsonous or mono‐methylarsonous).mp.
11 (dimethylarsinous or di‐methylarsinous).mp.
12 (trimethylarsine or tri‐methylarsine).mp.
13 or/1‐12
14 exp Folic Acid/
15 Folic Acid Deficiency/
16 folate$.mp.
17 folic.mp.
18 folinic$.mp.
19 folacin$.mp.
20 vitamin B9.mp.
21 5‐methyltetrahydrofolate.mp.
22 5‐methylTHF.mp.
23 metafolin.mp.
24 leucovorin.mp.
25 Micronutrients/
26 Food, Fortified/
27 Functional Food/
28 (food$ adj3 (fortif$ or functional$ or supplement$1)).tw,kw.
29 (micro‐nutrient$ or micronutrient$).tw,kw.
30 or/14‐29
31 13 and 30

Embase Ovid

1 Heavy metal/
2 Heavy Metal Poisoning/
3 (metalloid$ or metal poison$ or heavy metal$).tw,kw.
4 Arsenic/
5 Arsenic Poisoning/
6 exp organoarsenic derivative/
7 arsen$.mp.
8 (arsenic$ or arsenite$ or arsenate$).mp.
9 (monomethylarsonic or mono‐methylarsonic).mp.
10 (monomethylarsonous or mono‐methylarsonous).mp.
11 (dimethylarsinous or di‐methylarsinous).mp.
12 (trimethylarsine or tri‐methylarsine).mp.
13 or/1‐12
14 exp Folic Acid/
15 Folic Acid Deficiency/
16 folate$.mp.
17 folic.mp.
18 folinic$.mp.
19 folacin$.mp.
20 vitamin B9.mp.
21 5‐methyltetrahydrofolate.mp.
22 5‐methylTHF.mp.
23 metafolin.mp.
24 leucovorin.mp.
25 trace element/
26 Fortified food/
27 Functional Food/
28 (food$ adj3 (fortif$ or functional$ or supplement$1)).tw,kw.
29 (micro‐nutrient$ or micronutrient$).tw,kw.
30 or/14‐29
31 13 and 30
32 limit 31 to embase

Science Citation Index, Social Sciences Citation Index, Conference Proceedings Citation Index ‐ Science, Conference Proceedings Citation Index ‐ Social Science & Humanities via Web of Science

#17 #16 AND #8
DocType=All document types; Language=All languages;

#16 #15 OR #14 OR #13 OR #12 OR #11 OR #10 OR #9
DocType=All document types; Language=All languages;

#15 TOPIC: ((food* near/3 (fortif* or functional* or supplement*)))
DocType=All document types; Language=All languages;

#14 TOPIC: (leucovorin)
DocType=All document types; Language=All languages;

#13 TOPIC: (metafolin)
DocType=All document types; Language=All languages;

#12 TOPIC: (5‐methylTHF)
DocType=All document types; Language=All languages;

#11 TOPIC: (5‐methyltetrahydrofolate)
DocType=All document types; Language=All languages;

#10 TOPIC: (vitamin B9)
DocType=All document types; Language=All languages;

#9 TOPIC: (folate* or folic or folinic* or folacin*)
DocType=All document types; Language=All languages;

#8 #7 OR #6 OR #5 OR #4 OR #3 OR #2 OR #1
DocType=All document types; Language=All languages;

#7 TOPIC: ((trimethylarsine or tri‐methylarsine))
DocType=All document types; Language=All languages;

#6 TOPIC: ((dimethylarsinous or di‐methylarsinous))
DocType=All document types; Language=All languages;

#5 TOPIC: ((monomethylarsonous or mono‐methylarsonous))
DocType=All document types; Language=All languages;

#4 TOPIC: ((monomethylarsonic or mono‐methylarsonic))
DocType=All document types; Language=All languages;

#3 TOPIC: ((arsenic* or arsenite* or arsenate*))
DocType=All document types; Language=All languages;

#2 TOPIC: (arsen*)
DocType=All document types; Language=All languages;

#1 TOPIC: ((metalloid* or metal poison* or heavy metal*))
DocType=All document types; Language=All languages;

Cochrane Database of Systematic Reviews (CDSR) and Database of Abstracts of Reviews of Effects (DARE), in the Cochrane Library

#1 MeSH descriptor: [Metals, Heavy] this term only
#2 MeSH descriptor: [Heavy Metal Poisoning, Nervous System] this term only
#3 (metalloid* or metal poison* or heavy metal*):ti,ab,kw (Word variations have been searched)
#4 MeSH descriptor: [Arsenic] this term only
#5 MeSH descriptor: [Arsenic Poisoning] this term only
#6 MeSH descriptor: [Arsenicals] explode all trees
#7 arsen*:ti,ab,kw (Word variations have been searched)
#8 (arsenic* or arsenite* or arsenate*):ti,ab,kw (Word variations have been searched)
#9 (monomethylarsonic or mono‐methylarsonic):ti,ab,kw (Word variations have been searched)
#10 (monomethylarsonous or mono‐methylarsonous):ti,ab,kw (Word variations have been searched)
#11 (dimethylarsinous or di‐methylarsinous):ti,ab,kw (Word variations have been searched)
#12 (trimethylarsine or tri‐methylarsine):ti,ab,kw (Word variations have been searched)
#13 #1 or #2 or #3 or #4 or #5 or #6 or #7 or #8 or #9 or #10 or #11 or #12
#14 MeSH descriptor: [Folic Acid] explode all trees
#15 MeSH descriptor: [Folic Acid Deficiency] this term only
#16 folate*:ti,ab,kw (Word variations have been searched)
#17 folic:ti,ab,kw (Word variations have been searched)
#18 folinic*:ti,ab,kw (Word variations have been searched)
#19 folacin*:ti,ab,kw (Word variations have been searched)
#20 vitamin B9:ti,ab,kw (Word variations have been searched)
#21 5‐methyltetrahydrofolate:ti,ab,kw (Word variations have been searched)
#22 5‐methylTHF:ti,ab,kw (Word variations have been searched)
#23 metafolin:ti,ab,kw (Word variations have been searched)
#24 leucovorin:ti,ab,kw (Word variations have been searched)
#25 MeSH descriptor: [Micronutrients] this term only
#26 MeSH descriptor: [Food, Fortified] this term only
#27 MeSH descriptor: [Functional Food] this term only
#28 (food* N3 (fortif* or functional* or supplement*)):ti,ab,kw (Word variations have been searched)
#29 (micro‐nutrient* or micronutrient*):ti,ab,kw (Word variations have been searched)
#30 #14 or #15 or #16 or #17 or #18 or #19 or #20 or #21 or #22 or #23 or #24 or
#25 or #26 or #27 or #28 or #29
#31 #13 and #30

CINAHL EBSCO

S30 S12 AND S29
S29 S13 OR S14 OR S15 OR S16 OR S17 OR S18 OR S19 OR S20 OR S21 OR S22 OR S23 OR S24 OR S25 OR S26 OR S27 OR S28
S28 (micro‐nutrient* or micronutrient*)
S27 (food* N/3 (fortif* or functional* or supplement*))
S26 (MH "Functional Food")
S25 (MH "Food, Fortified")
S24 (MH "Micronutrients")
S23 leucovorin
S22 metafolin
S21 5‐methylTHF
S20 5‐methyltetrahydrofolate
S19 vitamin B9
S18 folacin*
S17 folinic*
S16 folic
S15 folate*
S14 (MH "Folic Acid Deficiency")
S13 (MH "Folic Acid+")
S12 S1 OR S2 OR S3 OR S4 OR S5 OR S6 OR S7 OR S8 OR S9 OR S10 OR S11
S11 (trimethylarsine or tri‐methylarsine)
S10 (dimethylarsinous or di‐methylarsinous)
S9 (monomethylarsonous or mono‐methylarsonous)
S8 (monomethylarsonic or mono‐methylarsonic)
S7 (arsenic* or arsenite* or arsenate*)
S6 arsen*
S5 (MH "Arsenicals+")
S4 (MH "Arsenic Poisoning")
S3 (MH "Arsenic")
S2 (metalloid* or metal poison* or heavy metal*)
S1 (MH "Metals, Heavy")

POPLINE

(arsenic or metal or poison) AND (folic or folate or "vitamin B9")

ClinicalTrials.gov

1 ArsenicAND (Folate OR folic OR folinic OR folacin OR “vitamin B9” OR 5‐methyltetrahydrofolate OR 5‐methylTHF OR metafolin OR leucovorin OR “fortified food” OR “functional food” OR “food supplement” OR micro‐nutrient OR micronutrient)

2 Arsenicals AND (Folate OR folic OR folinic OR folacin OR “vitamin B9” OR 5‐methyltetrahydrofolate OR 5‐methylTHF OR metafolin OR leucovorin OR “fortified food” OR “functional food” OR “food supplement” OR micro‐nutrient OR micronutrient)

3 arsenite AND (Folate OR folic OR folinic OR folacin OR “vitamin B9” OR 5‐methyltetrahydrofolate OR 5‐methylTHF OR metafolin OR leucovorin OR “fortified food” OR “functional food” OR “food supplement” OR micro‐nutrient OR micronutrient)

4 arsenate AND (Folate OR folic OR folinic OR folacin OR “vitamin B9” OR 5‐methyltetrahydrofolate OR 5‐methylTHF OR metafolin OR leucovorin OR “fortified food” OR “functional food” OR “food supplement” OR micro‐nutrient OR micronutrient)

5 “heavy metal”AND (Folate OR folic OR folinic OR folacin OR “vitamin B9” OR 5‐methyltetrahydrofolate OR 5‐methylTHF OR metafolin OR leucovorin OR “fortified food” OR “functional food” OR “food supplement” OR micro‐nutrient OR micronutrient)

6 metalloidAND (Folate OR folic OR folinic OR folacin OR “vitamin B9” OR 5‐methyltetrahydrofolate OR 5‐methylTHF OR metafolin OR leucovorin OR “fortified food” OR “functional food” OR “food supplement” OR micro‐nutrient OR micronutrient)

7 “metal poison” AND (Folate OR folic OR folinic OR folacin OR “vitamin B9” OR 5‐methyltetrahydrofolate OR 5‐methylTHF OR metafolin OR leucovorin OR “fortified food” OR “functional food” OR “food supplement” OR micro‐nutrient OR micronutrient)

8 monomethylarsonic AND (Folate OR folic OR folinic OR folacin OR “vitamin B9” OR 5‐methyltetrahydrofolate OR 5‐methylTHF OR metafolin OR leucovorin OR “fortified food” OR “functional food” OR “food supplement” OR micro‐nutrient OR micronutrient)

9 mono‐methylarsonicAND (Folate OR folic OR folinic OR folacin OR “vitamin B9” OR 5‐methyltetrahydrofolate OR 5‐methylTHF OR metafolin OR leucovorin OR “fortified food” OR “functional food” OR “food supplement” OR micro‐nutrient OR micronutrient)

10 monomethylarsonous AND (Folate OR folic OR folinic OR folacin OR “vitamin B9” OR 5‐methyltetrahydrofolate OR 5‐methylTHF OR metafolin OR leucovorin OR “fortified food” OR “functional food” OR “food supplement” OR micro‐nutrient OR micronutrient)

11 mono‐methylarsonous AND (Folate OR folic OR folinic OR folacin OR “vitamin B9” OR 5‐methyltetrahydrofolate OR 5‐methylTHF OR metafolin OR leucovorin OR “fortified food” OR “functional food” OR “food supplement” OR micro‐nutrient OR micronutrient)

12 dimethylarsinous AND (Folate OR folic OR folinic OR folacin OR “vitamin B9” OR 5‐methyltetrahydrofolate OR 5‐methylTHF OR metafolin OR leucovorin OR “fortified food” OR “functional food” OR “food supplement” OR micro‐nutrient OR micronutrient)

13 di‐methylarsinous AND (Folate OR folic OR folinic OR folacin OR “vitamin B9” OR 5‐methyltetrahydrofolate OR 5‐methylTHF OR metafolin OR leucovorin OR “fortified food” OR “functional food” OR “food supplement” OR micro‐nutrient OR micronutrient)

14 trimethylarsine AND (Folate OR folic OR folinic OR folacin OR “vitamin B9” OR 5‐methyltetrahydrofolate OR 5‐methylTHF OR metafolin OR leucovorin OR “fortified food” OR “functional food” OR “food supplement” OR micro‐nutrient OR micronutrient)

15 tri‐methylarsineAND (Folate OR folic OR folinic OR folacin OR “vitamin B9” OR 5‐methyltetrahydrofolate OR 5‐methylTHF OR metafolin OR leucovorin OR “fortified food” OR “functional food” OR “food supplement” OR micro‐nutrient OR micronutrient)

WHO International Clinical Trials Registry Platform (ICTRP; apps.who.int/trialsearch)

arsenic AND folic

African Index Medicus (AIM), Index Medicus for the Eastern Mediterranean Region (IMEMR), PAHO (Pan American Health Library), WHOLIS (WHO Library) and LILACS (Latin American and Caribbean Health Sciences Literature)

(metalloid$ or metal poison$ or heavy metal$ or arsen$ or arsenic$ or arsenite$ or arsenate$ or monomethylarsonic or mono‐methylarsonic or monomethylarsonous or mono‐methylarsonous or dimethylarsinous or di‐methylarsinous or trimethylarsine or tri‐methylarsine) and (folate$ or folic or folinic$ or folacin$ or vitamin B9 or 5‐methyltetrahydrofolate or 5‐methylTHF or metafolin or leucovorin or micro‐nutrient$ or micronutrient$ or functional food$ or fortified food$)

IMSEAR (Index Medicus for the South‐East Asia Region)

(metalloid$ or metal poison$ or heavy metal$ or arsen$ or arsenic$ or arsenite$ or arsenate$ or monomethylarsonic or mono‐methylarsonic or monomethylarsonous or mono‐methylarsonous or dimethylarsinous or di‐methylarsinous or trimethylarsine or tri‐methylarsine) and (folate$ or folic or folinic$ or folacin$ or vitamin B9 or 5‐methyltetrahydrofolate or 5‐methylTHF or metafolin or leucovorin or micro‐nutrient$ or micronutrient$ or functional food$ or fortified food$)

SciELO (Scientific Electronic Library Online)

(arsenic or metal or poison) AND (folic or folate or "vitamin B9")

IndMED (Indian Medical Journals)

(arsenic or metal or poison) AND (folic or folate or "vitamin B9")

WPRO (Western Pacific Region Index Medicus)

arsenic AND folic
arsenic AND folate
metal AND folic
metal AND folate

Appendix 2. Unused methods archived for future updates of this review

Measures of treatment effect	Dichotomous outcomes Had there been RCTs that reported dichotomous outcomes, we would have calculated the risk ratio and presented it with 95% confidence intervals.
	Continuous outcomes Had studies used different measurement methods for the same continuous outcomes, we would have calculated the standardised mean difference and presented it with 95% confidence intervals.
Unit of analysis issues	Cluster‐RCTs For each cluster‐RCT, we would have examined whether clustering was accounted for in the analysis. If not, we would have attempted to calculate effective sample sizes by using the ICC. If we could not calculate it, we would have contacted the trial authors for further detail, or obtained external estimates of the ICC from similar studies. We would have conducted a sensitivity analysis to examine the impact of variation in the ICC (see Sensitivity analysis).
	Studies with multiple treatment groups For trials with more than two intervention groups, we would have assessed which intervention group(s) of a multi‐arm study was (were) relevant to the review, and only included data from the directly relevant group(s). If a single study had more than one relevant intervention group, or more than one control group, or both, we would have created a single pairwise comparison, where appropriate, by combining all relevant intervention groups into a single intervention group, and all control groups into a single control group. For dichotomous outcomes, we would have summed both the sample sizes and the number of people with events across groups. For continuous outcomes, we would have combined means and standard deviations. In subgroup analyses, where the control group was shared by multiple intervention arms, we would have divided the control group over the number of subgroup categories, and included each pairwise comparison separately. For dichotomous outcomes, we would have divided the number of events and total participants. For continuous outcomes, we would have divided the number of total participants only; the means and standard deviations would have remained unchanged.
Dealing with missing data	Had summary data for an outcome (e.g. standard deviations) been missing, we would have based calculations on other reported measurements, if possible. When prespecified outcome data were missing, we would have contacted the trial authors to request them. Where missing data were not supplied, we would have reported the available data alone, without data imputation. For all relevant outcomes, we would have attempted to conduct an intention‐to‐treat (ITT) analysis, as far as possible, by including the participants randomised to each group, irrespective of whether they actually received the allocated intervention. When data were missing due to participant dropout, and if the reasons for dropout were well documented and unrelated to study arms, we would have conducted the analysis by including participants who completed the trial. Had it been possible, we would have conducted a sensitivity analysis to examine the impact of studies with missing data in the overall assessment of intervention effect.
Assessment of heterogeneity	Had we identified more eligible trials in this review, we would have assessed statistical heterogeneity by using the Chi² and I² statistics included in the forest plots. We would have considered a P value < 0.10 in the Chi² test as evidence of heterogeneity of intervention effects. We also would have used the I² statistic, which indicates the percentage of the variability due to heterogeneity rather than sampling error or chance, based on the thresholds listed below. 0% to 40%: might not be important 30% to 60%: may represent moderate heterogeneity 50% to 90%: may represent substantial heterogeneity 75% to 100%: considerable heterogeneity In addition, as an estimate of the between‐study variability, we would have reported Tau² from the random‐effects meta‐analysis. If there had been heterogeneity among studies, we would have explored the potential reasons for heterogeneity by conducting prespecified subgroup analyses, and we would have been cautious in the interpretation of those results with high levels of unexplained heterogeneity.
Assessment of reporting biases	Had we identified more than 10 studies reporting the same outcome of interest in this review, we would have generated funnel plots, and examined asymmetry that indicates the possibility of publication bias. We also would have considered reasons for asymmetry other than publication bias, such as differences in methodological quality among studies (Page 2021). For example, smaller trials tend to use less rigorous methodological approaches than larger trials, which may result in spuriously larger intervention effects. Had there been evidence of publication bias and small study effects, we would have taken it into account in the overall assessment and interpretation of intervention effects.
Data synthesis	We intended to conduct a meta‐analysis to yield an overall (pooled) estimate of the intervention effect when more than one study could be appropriately combined (e.g. studies examining the same intervention and outcomes, with comparable methods and approaches, in similar populations). For example, we would have assessed the appropriateness of combining studies by considering whether they: 1) used comparable measurement tools or scales for outcome assessment; and 2) examined the same intervention and outcomes in similar populations (e.g. age (adults versus children) and reproductive status (pregnancy versus non‐pregnancy)), and considered the comparability of timing of the outcome measurement (i.e. scale comparable, timing comparable). To examine heterogeneity among studies, we would have conducted subgroup analyses, based on prespecified factors (e.g. folic acid alone versus folic acid plus other nutrients). We would have performed a random‐effects analysis, since we anticipated natural heterogeneity among studies in terms of study populations, comparisons, and interventions (e.g. doses and durations of intervention). We would have pooled the outcome data using the inverse variance method. We would have attempted to conduct an ITT analysis by including all participants randomized to each group, where appropriate.
Subgroup analysis and investigation of heterogeneity	Had there been more eligible trials to include in this review, we would have conducted the following subgroup analyses to investigate heterogeneity among studies. Baseline folate status: deficient (defined as RBC folate < 340 nmol/L or plasma/serum folate < 10 nmol/L), versus non‐deficient, versus mixed, unknown, or unreported MTHFR C677T genotype: CC versus CT, versus TT, versus unknown or unreported, where there is sufficient sample size; or CC versus CT/TT, versus unknown or unreported Sex: male versus female, versus mixed, unknown, or unreported Pregnancy: yes versus no, versus mixed, unknown, or unreported Daily dose: ≤ 400 μg/d folic acid versus > 400 μg/d folic acid Nutrient composition: folic acid alone versus folic acid plus other nutrients Duration: < three months versus three months or more
Sensitivity analysis	Had there been more eligible trials to include in this review, we would have conducted sensitivity analyses to investigate the effects listed below. Effect of excluding studies at high or unclear risk of bias in both the random sequence generation and allocation concealment domains, and in one more of the other domains Effect of excluding cluster‐randomised trials, where appropriate adjustments for clustering were not made Effect of different ICC values of cluster‐randomised trials Effect of excluding studies with high levels (> 20%) of missing outcome data

^{CC: CC genotype, two copies of C base
CT: CT genotype, one copy of C base and one copy of T base
ICC: intra‐cluster correlation coefficient
MTHFR: Methylenetetrahydrofolate reductase
RBC: red blood cell
TT: TT genotype, two copies of T base}

Characteristics of studies

Characteristics of included studies [ordered by study ID]

FACT 2015.

Study characteristics
Methods	Design: randomised, double‐blind, placebo‐controlled trial with cross‐over design Unit of randomization: individual Date of study: not reported; participants recruited into the Health Effects of Arsenic Longitudinal Study between October 2000 and May 2002
Participants	Location/setting: households in Arailhazar, Bangladesh Sample size: 622 participants Number of withdrawals/dropouts: 12 participants Sex: men and women Mean age: 39 years old in the folic acid intervention groups and 38 years old in placebo group Inclusion criteria: participants were recruited from the Health Effects of Arsenic Longitudinal Study and were randomly selected among those who had been drinking from a household well with well water arsenic concentration greater than 50 μg/L for at least 1 year. Exclusion criteria: pregnant women, individuals taking nutritional supplements, individuals with protein in their urine, and individuals with known renal disease, diabetes, or gastrointestinal or other health problems
Interventions	Intervention 1 (N = 153): 400 µg folic acid/day Intervention 2 (N = 151): 800 µg folic acid/day Intervention 3 (N = 101): 3 mg creatine/day Intervention 4 (N = 103): 3 mg creatine plus 400 µg folic acid/day Control (N = 102): placebo Administration: after 12 weeks of intervention (first phase), folic acid intervention groups 1 (400 µg folic acid) and 2 (800 µg folic acid) were randomly divided, with half continuing to receive their assigned supplements, and the other half receiving placebo throughout the remaining study period, until week 24. The participants in intervention groups 3 (3 mg creatine) and 4 (3 mg creatine plus 400 µg folic acid) were switched to receive placebo after week 12.
Outcomes	Outcomes: Outcomes included total blood arsenic, urinary arsenic metabolites, total urinary arsenic, plasma folate, and total homocysteine. Total blood arsenic was measured using a PerkinElmer Elan DRC II ICP‐MS. Urinary arsenic metabolites were measured by HPLC and ICP‐MS. Total urinary arsenic was measured by graphite furnace atomic absorption spectrophotometry. Plasma folate was measured by radioimmunoassay, and total homocysteine was measured by HPLC. Timing of outcome assessment: Total blood arsenic was measured at baseline, and after 1, 6, 12, 13, 18, and 24 weeks of intervention. Urinary arsenic metabolites were measured at baseline, and after 1, 6, 12, 13, 18, and 24 weeks of intervention. Plasma folate was measured at baseline, and after 12 and 24 weeks of intervention. Plasma homocysteine was measured at baseline, and after the 12‐week intervention.
Notes	Funding source: R01CA133595, P42ES10349, P30ES09089, and R00ES018890 grants from National Institutes of Health (USA) Conflicts of interest: none Comment(s) Only data from the first phase (i.e. up to week 12), or from the second phase of the study (12 to 24 weeks) for groups continuing to receive the same intervention as the first phase were included in this review, to avoid any potential risks of a carry‐over effect. Study participants comprised of both folate‐deficient and folate‐sufficient individuals. All participants received arsenic‐removal water filters at baseline, and were instructed to use filtered water.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Comment: randomization was conducted separately for men and women, and in blocks. Within each block, the order of treatments was randomly permutated by a statistician using a computer‐based software program that generates a random sequence.
Allocation concealment (selection bias)	Low risk	Comment: the study indicated that a pharmacist in the study field distributed barcode‐labeled pill bottles to field staff sequentially as they enrolled participants; bottles were distributed in the order of the random assignment list generated elsewhere.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Comment: the study indicated that except for 2 data management specialists who assigned letters to treatments, all study investigators, fieldwork teams, village health workers, and study participants were blinded to interventions for the entire duration of the study, through the use of blind‐labeled pill bottles and participant identification barcodes on all pill bottles and biological samples
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: laboratory technicians were blinded for the entire duration of the study
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: of the 622 participants enrolled, 12 participants were dropped. Reasons for dropout were explained and did not appear to relate to supplementation.
Selective reporting (reporting bias)	Low risk	Comment: the study's prespecified outcome measures were reported as per registered protocol in ClinicalTrials.gov (clinicaltrials.gov/ct2/show/NCT01050556?id=NCT01050556&draw=2&rank=1)
Other bias	Unclear risk	Comment: there is uncertainty of arsenic‐removal water filter compliance. No information was provided on whether the compliance of filter usage was balanced across groups.

NIAT 2006.

Study characteristics
Methods	Design: randomised, double‐blind, placebo‐controlled trial with 2 arms Unit of randomization: individual Date of study: not reported; participants recruited into the Health Effects of Arsenic Longitudinal Study between October 2000 and May 2002
Participants	Location/setting: households in Arailhazar, Bangladesh Sample size: 200 participants Number of withdrawals/dropouts: 6 participants Sex: men and women Mean age: 40 years old in folic acid intervention group and 38 years old in placebo group Inclusion criteria: participants in this study were randomly selected from the 550 participants who fell into the lowest tertile for plasma folate (lower than 9 nmol/L) in the cross‐sectional study of 1650 participants enrolled from the Health Effects of Arsenic Longitudinal Study. Exclusion criteria: pregnancy, cobalamin deficiency (lower than 185 pmol/L), or taking vitamin supplements
Interventions	Intervention (N = 96): 400 µg folic acid per day Control (N = 98): placebo Administration: the intervention lasted 12 weeks. One bottle containing 100 tablets of folic acid or placebo was assigned to each participant. The field staff retained the bottles of folic acid or placebo, and returned to each participant’s home daily to observe the participant taking the folic acid or placebo tablet.
Outcomes	Outcomes: blood arsenic (total, inorganic, MMA, DMA), urinary arsenic metabolites, plasma folate, and total homocysteine. Proportions of total urinary arsenic excreted as inorganic arsenic, indicated as % inorganic arsenic; DMA, indicated as % DMA; and MMA, indicated as % MMA. Urinary arsenic metabolites were measured by HPLC and ICP‐MS. Data were presented as percentage of within‐person changes from baseline. In a subset (N = 130) of the 194 participants, blood concentrations of total arsenic, inorganic arsenic, MMAs, and DMAs were measured by HPLC and ICP‐MS. Plasma folate was measured by radioimmunoassay and total homocysteine was measured HPLC. Timing of outcome assessment: Urinary arsenic metabolites were measured before and after 1 week and 12 weeks of intervention. Blood arsenic metabolites were measured before and after the 12‐week intervention. Plasma folate and total homocysteine were measured before and after 12 weeks of intervention.
Notes	Funding source: RO1 ES011601, 5P30ES09089, and 1 P42ES10349 grants from the National Institutes of Health (USA) Conflicts of interest: none Comment(s): none
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Comment: the trial is reported as randomised, but no explanation was provided for the method used to generate random sequence. We contacted the trial author, and obtained further information, indicating that a computer‐generated random sequence was provided by a statistician.
Allocation concealment (selection bias)	Low risk	Comment: there is no report on allocation concealment, or the method used to conceal the allocation sequence. We contacted the trial author, and obtained further information indicating that all pill bottles were identical, and labeled with the participant’s ID numbers with a barcode label.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Comment: trial is reported as double‐blinded to the interventions.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Comment: laboratory technicians were blinded for the entire duration of the study.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: of the 200 participants enrolled, 6 participants were dropped because they were unable to meet with the field staff to receive the interventions on a daily basis; 3 were randomly assigned to the folic acid group and 3 to the placebo group. The reasons for dropout were unlikely to introduce bias.
Selective reporting (reporting bias)	Low risk	Comment: the study's outcome measures were reported as described in the methods section.
Other bias	Low risk	Comment: the intervention groups appeared similar in terms of baseline characteristics. The study appears to be free of other sources of bias.

DMA: dimethyl arsinic acid; DRC: dynamic reaction cell; HPLC: high‐performance liquid chromatography; ICP‐MS: inductively coupled mass spectrometry; ID: identifier; MMA: monomethylarsonic acid

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Becerra‐Romero 2012	This record is a conference abstract, describing a cross‐sectional study that assessed the associations between micronutrient intake (using a food frequency questionnaire) and urinary arsenic metabolites. The abstract reported associations for some nutrients (i.e. vitamin B12, vitamin C, copper, zinc, iron, selenium, sodium, calcium), but did not report data on dietary folate or folic acid intake. This record’s study design is outside the scope of this review, and because the study did not provide data on folate exposure, we did not include it in Table 3.
Bhattacharjee 2014	This study was conducted in mice, to assess the protective effects of folic acid and vitamin B12 supplementation (alone or in combination) on arsenic‐induced cardiotoxicity. This review only included studies conducted in humans; therefore, this study is outside the scope of this review.
Bocca 2020	In this study, concentrations of arsenic (along with other heavy metals) were measured in samples from 53 pregnant women in Spain (blood samples at trimester 1 and at delivery; cord blood; and urine at each trimester). Nutritional status was not assessed in this study; therefore, this study was outside the scope of this review.
Chen 2009	This is a review article that describes findings from the Health Effects of Arsenic Longitudinal Study (HEALS) in Bangladesh. This type of article does not provide primary data and is outside the scope of this review.
Desai 2020d	This cross‐sectional study was conducted among 290 children (aged 6 to 8 years) in Uruguay, with low level arsenic exposure, to assess the association between B‐vitamin intake and urinary arsenic metabolites. Study authors already assessed the association between folate intake and urinary arsenic methylation in a previous publication; therefore, the micronutrients assessed in this article included vitamin B12, vitamin B6, thiamin, riboflavin, niacin. The exposures in this article are outside the scope of this review.
Dooley 2007	This record describes findings from NIAT 2006 (Gamble 2007). This type of article does not provide primary data and is outside the scope of this review.
Dye 2008	This article describes findings from the NIAT 2006 (Gamble 2007). This type of article does not provide primary data and is outside the scope of this review.
Freeman 2009	This is an editorial article. This type of article does not provide primary data and is outside the scope of this review.
Gamble 2007	This record is a short review article. This type of article does not provide primary data and is outside the scope of this review.
Hall 2012	This is a review article. This type of article does not provide primary data and is outside the scope of this review.
Li 2017	This is an editorial article. This type of article does not provide primary data and is outside the scope of this review.
Liu 2004	Individuals suffering from arsenic poisoning due to coal underwent assessment for nutritional (including serum folic acid, vitamin B12) and inflammatory biomarkers. This study did not assess arsenic in drinking water or food. Therefore, the population of this study is outside the scope of this review.
Mazumdar 2017	This article is a protocol for a case‐control study in Bangladesh, which aims to assess the association between arsenic exposure and neural tube defects. This protocol describes a study design outside the scope of this review. This type of article did not provide primary data for inclusion in Table 3.
Milton 2010	This is a case‐control study among women in Bangladesh, which assessed arsenic concentrations in water among malnourished women (BMI < 18.5) and controls (BMI 18.5 to 24.9). The study design is outside the scope of this review. The study did not assess folate as an exposure and is therefore not included in Table 3.
Moitra 2018	Arsenicosis patients (N = 37), arsenic exposed volunteers (N = 20) and healthy volunteers (N = 20) of an arsenic endemic area were examined for the pattern of fungi in palmar arsenical keratosis. The supplementation of iron and folic acid to the person for 12 weeks affected the incidence of skin fungi. This study is a non‐randomised intervention study, thus the study design is outside the scope of this review. The type of intervention prevents assessment of the independent effects of folic acid; therefore this study was not included in Table 3.
Mungan 2013	Coal miners in Turkey (N = 81) were recruited into this cross‐sectional study to assess associations between serum concentrations of homocysteine, vitamin B12, cystatin C, and folate. Mining can lead to arsenic exposure; however, this study did not assess arsenic in drinking water or food. Therefore, the population of this study is outside the scope of this review.
NCT01360723	This study protocol describes an observational study (case‐control) assessing differences in DNA methylation and levels of plasma folate, homocysteine, S‐adenosylhomocysteine, and S‐adenosylmethionine between individuals with and without arsenic‐related urothelial carcinoma. This study design is outside the scope of this review, and this article does not provide primary data for inclusion in Table 3.
NCT01749982	This is a study protocol that describes a randomised clinical trial to assess the effects of choline and betaine supplementation for 8 weeks on urinary arsenic metabolites. The intervention of this study protocol is outside the scope of this review.
NCT02908581	This is a study protocol for a non‐randomized trial in Bangladesh that will provide iron and folic acid supplementation or placebo for 12 weeks to people with arsenicosis. The primary outcome is changes in the type of fungus on the skin, and the secondary outcome is the score of palmar arsenical keratosis. The intervention prevents assessment of the independent effects of folic acid on study outcomes. This type of study design and intervention is outside the scope of this review.
NCT03930017	This is a study protocol among pregnant women and their newborns in Bangladesh. The study aims to assess the association between arsenic exposure and one‐carbon micronutrients during pregnancy, on maternal and infant systemic immune function and respiratory illness. The primary and secondary outcomes listed in this protocol are outside the scope of this review.
Potera 2015	This is an editorial for the FACT 2015 (Peters 2015). This type of article does not provide primary data and is outside the scope of this review.
Sarmah 2016	This is a review article. This type of article does not provide primary data and is outside the scope of this review.
Schmidt 2019	This is a review article. This type of article does not provide primary data and is outside the scope of this review.
Tauheed 2017	This study was conducted among a subset of participants from a larger case‐control study that is included in Table 3 of this review (Mazumdar 2015a). The full case‐control study was conducted in Bangladesh, among mothers who gave birth to infants with and without neural tube defects. This article focused on plasma concentrations of histone 3, so the study authors excluded participants with samples that did not pass quality control criteria (n = 12 excluded among cases, n = 15 excluded among controls). Maternal age, infant age, and infant sex were not different between included and excluded participants, except that included mothers were more likely to report using folic acid supplements during pregnancy. Maternal plasma folate concentrations and reported folic acid supplementation did not differ between cases (N = 45) and controls (N = 40), although these differences were significant in the full study. This study does not provide additional primary data within the scope of this review.
Watson 2020	This study was conducted within the National Health and Nutrition Examination Survey (NHANES) in the USA, and reported on levels of heavy metal concentrations in blood and arsenic metabolites in urine. Neither folate status nor intake were assessed; therefore, this study is outside the scope of this review.
Yin 2009	This case‐control study was conducted among women with (N = 30) and without (N = 30) an anembryonic pregnancy in Shanxi province in China. Serum concentrations of folate and homocysteine, as well as urinary concentration of 55 elements (including arsenic) were assessed. Concentrations of arsenic were not reported. Women with anembryonic pregnancies had lower folate concentrations, higher homocysteine concentrations, and a higher odds of low folate status (< 7.93 nmol/L) and elevated homocysteine (≥ 14.64 umol/L), compared to women with normal pregnancies. Exposure to arsenic is unknown in this specific study population. The population is outside the scope of this review.

BMI: body mass index; DNA: deoxyribonucleic acid; FACT: Folic acid and Creatine Trial; NIAT: Nutritional Influences of Arsenic Toxicity

Characteristics of ongoing studies [ordered by study ID]

NCT02235948.

Study name	Public title: Effects of folic acid supplementation on arsenic lowering Scientific title: Efficacy and safety of folic acid supplementation lowering arsenic in a chronic, low‐level exposed arsenic population: a randomised, double‐blind, placebo‐controlled clinical trial
Methods	Design: randomised, placebo‐controlled trial
Participants	Location: China Sample size: 450 (estimated) Sex: all Age: 18 years and older Inclusion criteria Men or women older than 18 years of age, and chronically exposed to arsenic (arsenic concentration of the drinking water > 10 µg/L) Population who had no folic acid supplementation in the two weeks before the study Women of childbearing age agreed to use a reliable contraception method during the study Everyone volunteered to participate and signed informed consent Exclusion criteria Pregnant or breastfeeding women Allergic to folic acid Having clearly defined allergic history Reported long‐term use of folic acid and other vitamin B Having obvious signs or laboratory abnormalities that could affect the efficacy of folic acid Unsuitable to participate in the study based on the judgment of the investigators Do not agree to cancel medications that may affect serum folate concentration during the study period Participants who plan to become pregnant during the study, or move out of the area within the study period
Interventions	Intervention: folic acid supplementation of 800 µg/d Control: placebo
Outcomes	Primary outcome: change of urinary arsenic metabolites between baseline and week 8, measured by HPLC Timing of outcome assessment: baseline and week 8
Starting date	July 2014
Contact information	Xiao Xiao, Wenzhou Medical University
Notes	Study start date: July 2014 Study end date: estimated study completion date is listed as December 2016 on ClinicalTrials.gov. Sponsors and collaborators: Wenzhou Medical University Comment: this study is registered at ClinicalTials.gov as NCT02235948. The trial record for NCT02235948 shows that the study has published on the outcome of oxidative DNA damage (Guo 2015).

NCT03384862.

Study name	Public title: Nutrition, arsenic and cognitive function in children Scientific title: same as public title
Methods	Design: randomized, placebo‐controlled trial
Participants	Location: Bangladesh Sample size: 239 (actual) Sex: all Age: 8 years to 10 years Inclusion criteria: age‐eligible children of parents enrolled in the Health Effects of Arsenic Longitudinal Study cohort Exclusion criteria Siblings Twins Children who are not attending school Children with known physical disability or known chronic illness
Interventions	Intervention: 5 μg of vitamin B12 plus 400 μg of folic acid Control: placebo
Outcomes	Primary outcomes Change in arsenic methylation Change in blood arsenic Change in blood monomethyl arsenic Other outcome(s): change in WASI‐II cognitive function test score Timing of outcome assessment: up to 12 weeks
Starting date	January 2018
Contact information	Mary Gamble, Columbia University
Notes	Study start date: January 2018 Study end date: estimated study completion date is listed as December 2019 at ClinicalTrials.gov. Sponsors and collaborators: Columbia University and National Institute of Environmental Health Sciences (NIEHS) Comment: registered at ClinicalTrials.gov as NCT03384862.

DNA: deoxyribonucleic acid; HPLC: high‐performance liquid chromatography; WASI‐II: Wechsler Abbreviated Scale of Intelligence

Differences between protocol and review

Differences between the protocol and the review are described in Appendix 2.

The corresponding author changed from JPP‐R to PAC for the review.

Given the small number of randomised controlled trials included in the review and lack of results on our primary outcomes, we included observational and non‐randomised intervention studies in Table 3, as part of a narrative assessment of the evidence, to provide useful data for interpreting the potential benefits of folic acid supplementation for arsenic‐related health outcomes. We rescreened records from searches to identify observational and non‐randomised intervention studies that assessed associations between folate status or folic acid supplemention and any health outcome related to arsenic toxicity.

In addition to blood and urinary concentration of total arsenic (Bae 2017), the secondary outcomes of this review also included concentration of arsenic metabolites in blood and urine, as well as their proportions and methylation indices, because these are used as a standard measure of arsenic methylation capacity, which is linked to arsenic toxicity.

We did not include the secondary outcome of plasma folate concentration in the summary of findings tables (Bae 2017). Homocysteine represents a functional biomarker of folate status, and elevated homocysteine concentrations have well‐documented implications for adverse health effects. Homocysteine represents a better biomarker for arsenic toxicity, while plasma folate was used as biomarker of participant adherence in one study. Therefore, we deemed homocysteine to be a more relevant health outcome to include in the summary of findings tables, compared to plasma folate.

The POPLINE database ceased on 1 September 2019, so was not searched after this date. Database of Abstracts of Effects (DARE) has had no new content since 2015, so the search was not updated after 2019. The IndMed database was discontinued, so a search was not updated after 16 August 2019.

Data from the second period of a cross‐over trial were included only for groups who continued with the same treatment as the first period of the trial, to ensure no influence of cross‐over effects impacted the results.

Contributions of authors

Sajin Bae and Elena Kamynina conceived and designed the review, prepared the protocol, conducted database searches, screened eligible studies, extracted data, conducted the risk of bias assessment, assessed the certainty of evidence using the GRADE approach, analyzed and interpreted the data, and wrote the review. Adetutu F Farinola participated in the preparation of the search strategy and an early draft of the review protocol. Heather M Guetterman conducted database searches, screened eligible studies, extracted data, analysed and interpreted the data, and wrote the review. Juan Pablo Peña‐Rosas contributed to the production of the GRADE evidence profiles and the generation of the summary of findings tables. Patricia A Cassano analysed and interpreted the data. Sajin Bae, Heather M Guetterman, Patrick J Stover, and Patricia A Cassano co‐ordinated the review. Patrick J Stover and Patricia A Cassano conceived and designed the review. Patrick J Stover, Patricia A Cassano, Robert J Berry, and Juan Pablo Peña‐Rosas provided feedback on the contents and the methods. Marie Caudill and all authors provided overall feedback on the review. Juan Pablo Peña‐Rosas had overall responsibility for the review protocol, and Patrick J Stover is the corresponding review author and the guarantor of the review.

Sources of support

Internal sources

• Evidence and Programme Guidance Unit, Department of Nutrition for Health and Development, World Health Organization (WHO), Switzerland

  Dr Juan Pablo Peña‐Rosas is a full‐time member of staff at the WHO

External sources

• Bill & Melinda Gates Foundation, USA

  WHO acknowledges financial support from the Bill & Melinda Gates Foundation for the development of systematic reviews of the evidence on the effects of nutrition interventions.

• National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) of the National Institutes of Health (NIH), USA

  Sajin Bae and Heather M Guetterman would like to acknowledge the support from the NIDDK of the NIH, under award number: T32‐DK007158 (Nutrition Training grant). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIDDK or the NIH.

• Nutrition International (formerly Micronutrient Initiative), Canada

  WHO acknowledges the financial support of Nutrition International (formerly Micronutrient Initiative) for completion of systematic reviews of the evidence on nutrition actions.

• Centers for Disease Control and Prevention (CDC), USA

  WHO acknowledges financial and technical support from the National Center on Birth Defects and Developmental Disabilities for the completion of systematic reviews of nutrition interventions.

Declarations of interest

Sajin Bae: none known

Elena Kamynina: none known

Adetutu F Farinola: none known

Heather M Guetterman: none known

Marie Caudill: none known

Patrick J Stover: none known

Patricia A Cassano: none known

Robert Berry was an employee of the Centers for Disease Control and Prevention (CDC) while working on this review. The CDC partially funded this review, but had no control over the design or conduct of the review. The author alone is responsible for the views expressed in this publication; the views do not necessarily represent the official position, decisions, policy, or views of the CDC.

These authors contributed equally to this work.

These authors contributed equally to this work.

These authors contributed equally to this work.

These authors contributed equally to this work as senior authors.

These authors contributed equally to this work as senior authors.

New