Factors Affecting Arsenic Methylation in Arsenic-Exposed Humans: A Systematic Review and Meta-Analysis

Chronic arsenic exposure is a critical public health issue in many countries. The metabolism of arsenic in vivo is complicated because it can be influenced by many factors. In the present meta-analysis, two researchers independently searched electronic databases, including the Cochrane Library, PubMed, Springer, Embase, and China National Knowledge Infrastructure, to analyze factors influencing arsenic methylation. The concentrations of the following arsenic metabolites increase (p< 0.000001) following arsenic exposure: inorganic arsenic (iAs), monomethyl arsenic (MMA), dimethyl arsenic (DMA), and total arsenic. Additionally, the percentages of iAs (standard mean difference (SMD): 1.00; 95% confidence interval (CI): 0.60–1.40; p< 0.00001) and MMA (SMD: 0.49; 95% CI: 0.21–0.77; p = 0.0006) also increase, while the percentage of DMA (SMD: −0.57; 95% CI: −0.80–−0.31; p< 0.0001), primary methylation index (SMD: −0.57; 95% CI: −0.94–−0.20; p = 0.002), and secondary methylation index (SMD: −0.27; 95% CI: −0.46–−0.90; p = 0.004) decrease. Smoking, drinking, and older age can reduce arsenic methylation, and arsenic methylation is more efficient in women than in men. The results of this analysis may provide information regarding the role of arsenic oxidative methylation in the arsenic poisoning process.


Introduction
Arsenic is a toxic metalloid element that is ubiquitous in the environment. The World Health Organization (WHO) and the International Agency for Research on Cancer (IARC) have identified it and its compounds as human carcinogens [1,2]. Arsenic enters an organism via the respiratory tract, alimentary canal, and skin, and it is primarily metabolized in the liver [3]. The metabolic pattern has primarily been regarded as occurring through oxidative methylation [4], which was formerly considered a detoxification pattern [5]. However, monomethylarsonous acid (MMA III ) and dimethylarsonous acid (DMA III ) have recently been shown to be more poisonous than inorganic arsenic (iAs) [6,7]. Therefore, arsenic toxicity is closely related to its metabolism, which in turn is highly dependent on the methylation status and valence states of its metabolites. Recently, research has increasingly focused on the factors influencing arsenic methylation. In addition to the dosage of arsenic exposure, an individual's ethnicity [8], age, sex [9], body mass index (BMI) [10], lifestyle and dietary history [11], and inherited genetic characteristics [12] were also related with the arsenic Based on the JBI criteria, all of the included papers were of high quality, as demonstrated by scores >12 (Table 1). Based on the JBI criteria, all of the included papers were of high quality, as demonstrated by scores >12 (Table 1).

Effect of Arsenic Exposure on Total Arsenic
A total of 12 studies estimated TAs concentration. The pooled analysis showed that the TAs concentration was 3.10-fold higher in the exposed group than in the control group (95% CI, 2.13-4.07; Z = 6.28; p < 0.00001), with significant heterogeneity (p < 0.00001; I 2 = 99%; Figure 2). Int. J. Environ. Res. Public Health 2016, 13,205 1

Effect of Arsenic Exposure on Total Arsenic
A total of 12 studies estimated TAs concentration. The pooled analysis showed that the TAs concentration was 3.10-fold higher in the exposed group than in the control group (95% CI, 2.13-4.07; Z = 6.28; p < 0.00001), with significant heterogeneity (p < 0.00001; I 2 = 99%; Figure 2).

Effect of Arsenic Exposure on Inorganic Arsenic
A total of nine studies estimated iAs concentration. The pooled analysis showed that iAs in the exposed group was 1.07-fold higher than that in the control group (95% CI, 0.61-1. 53; Z = 4.55; p < 0.00001), with significant heterogeneity (p < 0.00001; I 2 = 89%; Figure 3).

Effect of Arsenic Exposure on Monomethyl Arsenic
A total of 10 studies estimated MMA concentration. The pooled analysis showed that the MMA concentration in the exposed group was 1.10-fold higher than that in the control group (95% CI, 0.81-1. 40; Z = 7.34; p < 0.00001), with significant heterogeneity (p < 0.00001; I 2 = 77%; Figure 4).

Effect of Arsenic Exposure on Inorganic Arsenic
A total of nine studies estimated iAs concentration. The pooled analysis showed that iAs in the exposed group was 1.07-fold higher than that in the control group (95% CI, 0.61-1.53; Z = 4.55; p < 0.00001), with significant heterogeneity (p < 0.00001; I 2 = 89%; Figure 3). A total of 12 studies estimated TAs concentration. The pooled analysis showed that the TAs concentration was 3.10-fold higher in the exposed group than in the control group (95% CI, 2.13-4.07; Z = 6.28; p < 0.00001), with significant heterogeneity (p < 0.00001; I 2 = 99%; Figure 2).

Figure 2.
Effect of arsenic exposure on total arsenic (TAs) concentration in urine. Std., standardized; SD, standard deviation; IV, independent variable; CI, confidence interval.

Effect of Arsenic Exposure on Inorganic Arsenic
A total of nine studies estimated iAs concentration. The pooled analysis showed that iAs in the exposed group was 1.07-fold higher than that in the control group (95% CI, 0.61-1.53; Z = 4.55; p < 0.00001), with significant heterogeneity (p < 0.00001; I 2 = 89%; Figure 3).

Effect of Arsenic Exposure on Monomethyl Arsenic
A total of 10 studies estimated MMA concentration. The pooled analysis showed that the MMA concentration in the exposed group was 1.10-fold higher than that in the control group (95% CI, 0.81-1.40; Z = 7.34; p < 0.00001), with significant heterogeneity (p < 0.00001; I 2 = 77%; Figure 4).

Effect of Arsenic Exposure on Monomethyl Arsenic
A total of 10 studies estimated MMA concentration. The pooled analysis showed that the MMA concentration in the exposed group was 1.10-fold higher than that in the control group (95% CI, 0.81-1.40; Z = 7.34; p < 0.00001), with significant heterogeneity (p < 0.00001; I 2 = 77%; Figure 4).

Effect of Arsenic Exposure on Dimethyl Arsenic
A total of 10 studies estimated DMA concentration. The pooled analysis showed that the DMA concentration in the exposed group was 2.60-fold higher than that in the control group (95% CI, 1.50-3.69; Z = 4.64; p < 0.00001), with significant heterogeneity (p < 0.00001; I 2 = 98%; Figure 5).

Effect of Arsenic Exposure on Inorganic Arsenic Percentage
A total of 10 studies estimated iAs%. The pooled analysis showed that iAs% in the exposed group was 1.00-fold higher than that in the control group (95% CI, 0.60-1.40; Z = 4.86; p < 0.00001), with significant heterogeneity (p < 0.00001; I 2 = 94%; Figure 6).

Effect of Arsenic Exposure on Dimethyl Arsenic
A total of 10 studies estimated DMA concentration. The pooled analysis showed that the DMA concentration in the exposed group was 2.60-fold higher than that in the control group (95% CI, 1.50-3.69; Z = 4.64; p < 0.00001), with significant heterogeneity (p < 0.00001; I 2 = 98%; Figure 5).

Effect of Arsenic Exposure on Dimethyl Arsenic
A total of 10 studies estimated DMA concentration. The pooled analysis showed that the DMA concentration in the exposed group was 2.60-fold higher than that in the control group (95% CI, 1.50-3.69; Z = 4.64; p < 0.00001), with significant heterogeneity (p < 0.00001; I 2 = 98%; Figure 5).

Effect of Arsenic Exposure on Inorganic Arsenic Percentage
A total of 10 studies estimated iAs%. The pooled analysis showed that iAs% in the exposed group was 1.00-fold higher than that in the control group (95% CI, 0.60-1.40; Z = 4.86; p < 0.00001), with significant heterogeneity (p < 0.00001; I 2 = 94%; Figure 6).

Effect of Arsenic Exposure on Inorganic Arsenic Percentage
A total of 10 studies estimated iAs%. The pooled analysis showed that iAs% in the exposed group was 1.00-fold higher than that in the control group (95% CI, 0.60-1.40; Z = 4.86; p < 0.00001), with significant heterogeneity (p < 0.00001; I 2 = 94%; Figure 6).

Effect of Arsenic Exposure on Dimethyl Arsenic
A total of 10 studies estimated DMA concentration. The pooled analysis showed that the DMA concentration in the exposed group was 2.60-fold higher than that in the control group (95% CI, 1.50-3.69; Z = 4.64; p < 0.00001), with significant heterogeneity (p < 0.00001; I 2 = 98%; Figure 5).

Effect of Arsenic Exposure on Inorganic Arsenic Percentage
A total of 10 studies estimated iAs%. The pooled analysis showed that iAs% in the exposed group was 1.00-fold higher than that in the control group (95% CI, 0.60-1.40; Z = 4.86; p < 0.00001), with significant heterogeneity (p < 0.00001; I 2 = 94%; Figure 6).

Effect of Arsenic Exposure on Monomethyl Arsenic Percentage
A total of 12 studies estimated MMA%. The pooled analysis showed that MMA% in the exposed group was 0.49-fold higher than that in the control group (95% CI, 0.21-0.77; Z = 3.42; p = 0.0006), with significant heterogeneity (p < 0.00001; I 2 = 92%; Figure 7).

Effect of Arsenic Exposure on Monomethyl Arsenic Percentage
A total of 12 studies estimated MMA%. The pooled analysis showed that MMA% in the exposed group was 0.49-fold higher than that in the control group (95% CI, 0.21-0.77; Z = 3.42; p = 0.0006), with significant heterogeneity (p < 0.00001; I 2 = 92%; Figure 7).

Effect of Arsenic Exposure on Dimethyl Arsenic Percentage
A total of 12 studies estimated DMA%. The pooled analysis showed that DMA% in the exposed group was 0.55-fold lower than that in the control group (95% CI, 0.80-0.31; Z = 4.41; p < 0.0001), with significant heterogeneity (p < 0.00001; I 2 = 90%; Figure 8).

Effect of Arsenic Exposure on the Primary Methylation Index
A total of nine studies estimated the PMI. The pooled analysis showed that the PMI in the exposed group was 0.57-fold lower than that in the control group (95% CI, 0.94-0.20; Z = 3.04; p = 0.002), with significant heterogeneity (p < 0.00001; I 2 = 94%; Figure 9).

Effect of Arsenic Exposure on Dimethyl Arsenic Percentage
A total of 12 studies estimated DMA%. The pooled analysis showed that DMA% in the exposed group was 0.55-fold lower than that in the control group (95% CI, 0.80-0.31; Z = 4.41; p < 0.0001), with significant heterogeneity (p < 0.00001; I 2 = 90%; Figure 8).

Effect of Arsenic Exposure on Monomethyl Arsenic Percentage
A total of 12 studies estimated MMA%. The pooled analysis showed that MMA% in the exposed group was 0.49-fold higher than that in the control group (95% CI, 0.21-0.77; Z = 3.42; p = 0.0006), with significant heterogeneity (p < 0.00001; I 2 = 92%; Figure 7).

Effect of Arsenic Exposure on Dimethyl Arsenic Percentage
A total of 12 studies estimated DMA%. The pooled analysis showed that DMA% in the exposed group was 0.55-fold lower than that in the control group (95% CI, 0.80-0.31; Z = 4.41; p < 0.0001), with significant heterogeneity (p < 0.00001; I 2 = 90%; Figure 8). A total of nine studies estimated the PMI. The pooled analysis showed that the PMI in the exposed group was 0.57-fold lower than that in the control group (95% CI, 0.94-0.20; Z = 3.04; p = 0.002), with significant heterogeneity (p < 0.00001; I 2 = 94%; Figure 9). A total of nine studies estimated the PMI. The pooled analysis showed that the PMI in the exposed group was 0.57-fold lower than that in the control group (95% CI, 0.94-0.20; Z = 3.04; p = 0.002), with significant heterogeneity (p < 0.00001; I 2 = 94%; Figure 9).

Effect of Arsenic Exposure on the Secondary Methylation Index
A total of nine studies estimated the SMI. The pooled analysis showed that the SMI in the exposed group was 0.27-fold lower than that in the control group (95% CI, 0.46-0.09; Z = 2.87; p = 0.004), with significant heterogeneity (p = 0.0003; I 2 = 74%; Figure 10).

Subgroup Analyses of the Effects of Arsenic Exposure
In the subgroup analyses, the differences in iAs% and MMA% between the high and low exposure groups were higher in children than in adults (p < 0.00001; Figure 11), while the differences in TAs, DMA, and PMI between the high and low exposure groups were lower in children than in adults (p < 0.0001). Based on sex ratio, groups with >50% men had greater iAs (p = 0.003), iAs% (p = 0.002), and MMA% (p = 0.02) but smaller DMA% (p = 0.01) and PMI (p < 0.00001) than the groups that were ≤50% male. The high-exposure groups higher iAs, iAs%, and MMA% than the low-exposure groups. Furthermore, changes in iAs, MMA, DMA, iAs%, and DMA% were observed in the subjects exposed to drinking water with <50 µg/L arsenic. Significant differences between Asians and Americans were not detected.

Effect of Arsenic Exposure on the Secondary Methylation Index
A total of nine studies estimated the SMI. The pooled analysis showed that the SMI in the exposed group was 0.27-fold lower than that in the control group (95% CI, 0.46-0.09; Z = 2.87; p = 0.004), with significant heterogeneity (p = 0.0003; I 2 = 74%; Figure 10).

Effect of Arsenic Exposure on the Secondary Methylation Index
A total of nine studies estimated the SMI. The pooled analysis showed that the SMI in the exposed group was 0.27-fold lower than that in the control group (95% CI, 0.46-0.09; Z = 2.87; p = 0.004), with significant heterogeneity (p = 0.0003; I 2 = 74%; Figure 10).

Subgroup Analyses of the Effects of Arsenic Exposure
In the subgroup analyses, the differences in iAs% and MMA% between the high and low exposure groups were higher in children than in adults (p < 0.00001; Figure 11), while the differences in TAs, DMA, and PMI between the high and low exposure groups were lower in children than in adults (p < 0.0001). Based on sex ratio, groups with >50% men had greater iAs (p = 0.003), iAs% (p = 0.002), and MMA% (p = 0.02) but smaller DMA% (p = 0.01) and PMI (p < 0.00001) than the groups that were ≤50% male. The high-exposure groups higher iAs, iAs%, and MMA% than the low-exposure groups. Furthermore, changes in iAs, MMA, DMA, iAs%, and DMA% were observed in the subjects exposed to drinking water with <50 µg/L arsenic. Significant differences between Asians and Americans were not detected.

Subgroup Analyses of the Effects of Arsenic Exposure
In the subgroup analyses, the differences in iAs% and MMA% between the high and low exposure groups were higher in children than in adults (p < 0.00001; Figure 11), while the differences in TAs, DMA, and PMI between the high and low exposure groups were lower in children than in adults (p < 0.0001). Based on sex ratio, groups with >50% men had greater iAs (p = 0.003), iAs% (p = 0.002), and MMA% (p = 0.02) but smaller DMA% (p = 0.01) and PMI (p < 0.00001) than the groups that were ď50% male. The high-exposure groups higher iAs, iAs%, and MMA% than the low-exposure groups. Furthermore, changes in iAs, MMA, DMA, iAs%, and DMA% were observed in the subjects exposed to drinking water with <50 µg/L arsenic. Significant differences between Asians and Americans were not detected.

Effect of Sex on Arsenic Methylation
A total of 12 studies compared the MMA% between men and women. The pooled analysis showed that the MMA% in men was 0.44-fold higher than in women (95% CI, 0.35-0.52; Z = 10.37; p < 0.00001), with significant heterogeneity (p < 0.00001; I 2 = 52%; Figure 12A). A total of 12 studies compared the DMA% between men and women. The pooled analysis showed that the DMA% was 0.33-fold lower in men than in women (95% CI, 0.38-0.28; Z = 12.35; p < 0.00001), with significant heterogeneity (p = 1; I 2 = 0%). A total of six studies compared the SMI between men and women. The pooled analysis showed that the SMI was 0.36-fold lower in men than in women (95% CI, 0.53-0.19; Z = 4.10; p < 0.0001), with significant heterogeneity (p < 0.00001; I 2 = 86%).

Effect of Age on Arsenic Methylation
A total of seven studies compared the iAs% by age. The pooled analysis showed that the iAs% among subjects aged ď50 years was 0.24-fold higher than that among subjects aged >50 years (95% CI, 0.16-0.32; Z = 5.88; p < 0.00001), with significant heterogeneity (p = 0.14; I 2 = 38%; Figure 12D). A total of eight studies compared the MMA% by age. The pooled analysis showed that the MMA% among subjects aged ď50 years was 0.23-fold lower than that among subjects aged >50 years (95% CI, 0.40-0.06; Z = 2.66; p = 0.008), with significant heterogeneity (p < 0.0001; I 2 = 80%); A total of five studies compared the PMI by age. The pooled analysis showed that the PMI for subjects aged ď50 years was 0.22-fold lower than that for participants aged >50 years (95% CI, 0.37-0.07; Z = 2.83; p = 0.005), with significant heterogeneity (p = 0.01; I 2 = 69%).

Effect of Body Mass Index on Arsenic Methylation
A total of four studies estimated the MMA% based on BMI. The pooled analysis showed that the MMA% with a high BMI was 0.18-fold lower than that with a low BMI (95% CI, 0.31-0.04; Z = 2.55; p = 0.01), with significant heterogeneity (p = 0.01; I 2 = 52%; Figure 12E).

Effect of Body Mass Index on Arsenic Methylation
A total of four studies estimated the MMA% based on BMI. The pooled analysis showed that the MMA% with a high BMI was 0.18-fold lower than that with a low BMI (95% CI, 0.31-0.04; Z = 2.55; p = 0.01), with significant heterogeneity (p = 0.01; I 2 = 52%; Figure 12E).

Small-Study Effect Evaluation
Visual inspection of the funnel plot and Egger's test results showed no evidence of significant small-study effects (pA = 0.550, pB = 0.921) (Figure 13).

Small-Study Effect Evaluation
Visual inspection of the funnel plot and Egger's test results showed no evidence of significant small-study effects (p A = 0.550, p B = 0.921) (Figure 13).

Sensitivity Analysis
We conducted a sensitivity analysis for the MMA%. As shown in Figure 14, all of the included studies were distributed evenly from the central line, and none of the studies deviated significantly. Therefore, no individual study influenced the combined results.

Meta-Regression Analysis of Arsenic Exposure Effects
A meta-regression analysis was performed with the MMA% as the Y (outcome) and age (children vs. adults), sex constituent ratio (>50% men vs. ≤50% men), nationality (Asian vs. American), and exposure dose (low exposure (≤50 µg/L) vs. high exposure (>50 µg/L)) as the X variables (factors). Age, as included in this regression model, provided more evidence to illustrate the relationship between age and arsenic methylation ( Table 2).  . Blue-dotted line shows the overall estimated standard mean difference. Evidence for publication bias was not found. SMD, standard mean difference; SE, standard error.

Sensitivity Analysis
We conducted a sensitivity analysis for the MMA%. As shown in Figure 14, all of the included studies were distributed evenly from the central line, and none of the studies deviated significantly. Therefore, no individual study influenced the combined results.  smoking (B). Blue-dotted line shows the overall estimated standard mean difference. Evidence for publication bias was not found. SMD, standard mean difference; SE, standard error.

Sensitivity Analysis
We conducted a sensitivity analysis for the MMA%. As shown in Figure 14, all of the included studies were distributed evenly from the central line, and none of the studies deviated significantly. Therefore, no individual study influenced the combined results.

Meta-Regression Analysis of Arsenic Exposure Effects
A meta-regression analysis was performed with the MMA% as the Y (outcome) and age (children vs. adults), sex constituent ratio (>50% men vs. ≤50% men), nationality (Asian vs. American), and exposure dose (low exposure (≤50 µg/L) vs. high exposure (>50 µg/L)) as the X variables (factors). Age, as included in this regression model, provided more evidence to illustrate the relationship between age and arsenic methylation ( Table 2). Meta-analysis estimates, given named study is omitted Figure 14. Sensitivity analysis for the percentage of monomethyl arsenic, which was affected by arsenic exposure. CI, confidence interval.

Meta-Regression Analysis of Arsenic Exposure Effects
A meta-regression analysis was performed with the MMA% as the Y (outcome) and age (children vs. adults), sex constituent ratio (>50% men vs. ď50% men), nationality (Asian vs. American), and exposure dose (low exposure (ď50 µg/L) vs. high exposure (>50 µg/L)) as the X variables (factors). Age, as included in this regression model, provided more evidence to illustrate the relationship between age and arsenic methylation ( Table 2).

Discussion
The present results, based on a meta-analysis of the current best evidence, indicate that the dose of arsenic exposure is negatively related with arsenic methylation capacity and women have a better methylation capacity than men. Smoking and drinking, and potentially age and BMI, are also associated with poorer methylation capacity.
Recent studies suggest that exposure to arsenic may increase the risk of certain diseases, including arsenic-induced skin lesions [40], peripheral artery disease [41], hypertension [42,43], cardiovascular disease [44], diabetes [45], skin cancer and urothelial cancer [46], and reproductive system damage [47]. The primary mode of arsenic metabolism is generally considered to be methylation ( Figure 15). However, the toxicities and targets of iAs III , iAs V , MMA III , MMA V , DMA III , and DMA V that are produced in the metabolic process are not consistent.

Discussion
The present results, based on a meta-analysis of the current best evidence, indicate that the dose of arsenic exposure is negatively related with arsenic methylation capacity and women have a better methylation capacity than men. Smoking and drinking, and potentially age and BMI, are also associated with poorer methylation capacity.
Recent studies suggest that exposure to arsenic may increase the risk of certain diseases, including arsenic-induced skin lesions [40], peripheral artery disease [41], hypertension [42,43], cardiovascular disease [44], diabetes [45], skin cancer and urothelial cancer [46], and reproductive system damage [47]. The primary mode of arsenic metabolism is generally considered to be methylation ( Figure 15). However, the toxicities and targets of iAs III , iAs V , MMA III , MMA V , DMA III , and DMA V that are produced in the metabolic process are not consistent. Although the average urinary values for iAs%, MMA%, and DMA% among arsenic-exposed people are 10%-30%, 10%-20%, and 60%-70% respectively, there are differences between individuals. Epidemiological studies suggest that high MMA%, low DMA%, and low SMI are associated with the incidence of arsenic-related diseases [33,34,48,49]. In the present study, arsenic exposure led to high levels of iAs% and MMA% and low levels of DMA%, FMR, and SMR, indicating inefficient methylation. The changes in the subgroup exposed to drinking water with arsenic >100 µg/L were more significant. Those results suggest that the arsenic methylation capacity declines with increased doses of exposure, implying that highly toxic iAs III and MMA III at the cellular Although the average urinary values for iAs%, MMA%, and DMA% among arsenic-exposed people are 10%-30%, 10%-20%, and 60%-70% respectively, there are differences between individuals. Epidemiological studies suggest that high MMA%, low DMA%, and low SMI are associated with the incidence of arsenic-related diseases [33,34,48,49]. In the present study, arsenic exposure led to high levels of iAs% and MMA% and low levels of DMA%, FMR, and SMR, indicating inefficient methylation. The changes in the subgroup exposed to drinking water with arsenic >100 µg/L were more significant. Those results suggest that the arsenic methylation capacity declines with increased doses of exposure, implying that highly toxic iAs III and MMA III at the cellular or blood level may lead to arsenic-related injuries. Continued arsenic exposure may cause insufficient metabolic-associated factors or enzymes such as S-adenosylmethionine and glutathione (Figure 15, 1 and 2). Additionally, a considerable amount of glutathione is consumed by oxidative stress. It is worth noting that inefficient methylation has also been found in a low-exposure subgroup (<50 µg/L); that is, chronic low-exposure might also be deleterious for health.
The subgroup analysis showed that children had a lower methylation capacity than adults, which may be due to metabolism-related organs and enzymes during growth in children. In contrast, Chowdury et al. [50] suggested that children might have better methylation capacity, but the study lacked sufficient statistical data. Therefore, further studies are required to confirm the differences between children and adults.
Although no significant differences were detected between Asians and Americans, remarkable differences in the distribution of urinary arsenic species were detected among Chinese, Chileans, and Mexicans; however, the sample sizes of Chinese and Chilean participants were too small [8]. Another study reported ethnic differences in arsenic methylation capacity [16]; indigenous people in Chile had better methylation than individuals of European ethnic origin. Smith et al. [51] suggested that indigenous peoples have less severe clinical manifestations and the ethnic differences were derived from the different genetic locus of arsenic-metabolizing enzymes.
Women had better methylation capacities than men in not only the subgroup analyses of arsenic exposure but also the meta-analysis of sex effects. One possible mechanism is that betaine formed from choline oxidation can donate its methyl group to homocysteine to form methionine. Then, choline can be derived from phosphatidylcholine, which is up-regulated by estrogen [52]. Thus, estrogen may indirectly promote arsenic metabolism (Figure 15, 3). Better methylation capacity and the effect of estrogen in were present for women of childbearing age and not adolescents [53]. In addition, men might be exposed to more factors that inhibit arsenic methylation, such as drinking and smoking.
Smoking was associated with low methylation capacity in the present study, supporting a previous study, in which inefficient methylation and cardiovascular disease were related with older age and smoking [54]. Both smoking and arsenic exposure can stimulate cells to release free radicals and consume antioxidants, resulting in oxidative stress and tissue damage [55]. In addition, some chemicals in cigarettes can influence the enzymes involved in the methylation processes, especially those involved in the second methylation phase. Moreover, smoking itself could be a pathway for arsenic exposure if the cigarettes contain trace amounts of arsenic. Alcohol consumption could affect the methylation processes, because alcohol results in liver damage, which is the primary organ associated with arsenic metabolism. Because smoking and drinking are more common behaviors in men than in women, sex, smoking status, and drinking status might confound each other.
In the meta-analysis of the effect of age on arsenic methylation, MMA% and PMI increased with age; SMI was somewhat, but not significantly, higher. Similarly, Huang et al. reported that MMA% increased, while both DMA% and SMI decreased [39]. Higher MMA% and lower DMA% and SMI are potentially risk factors for arsenicosis. However, there is little evidence that either iAs% or PMI is related with arsenicosis [56]. Therefore, older people might have poorer methylation capacity and are susceptible to arsenic damage. Many organs are senescent, which may block the methylation process, and older age might also be associated with longer arsenic exposure. The present study also indicated that higher BMIs were associated with a lower MMA%, suggesting that BMI might be positively related with methylation ability. Further studies regarding the relationship between the key molecular factors and arsenic methylation are needed.

Conclusions
In summary, the methylation of arsenic is influenced by a variety of factors. Arsenic exposure, smoking, drinking, and older age can reduce the capacity of arsenic methylation. Furthermore, arsenic methylation is more efficient in women than in men. The present study was limited by the obvious heterogeneity in the data. Although the heterogeneity was diminished in the subgroup analyses, it remained high. This heterogeneity might have been related with the variation in a number of characteristics among the studies, such as ethnicity, exposure duration, nutrition and dietary factors, and lifestyle. However, there was no evidence of significant publication bias, and the sensitivity of the articles was satisfactory. Thus, the method was suitable to analyze the relationship between the factors and arsenic methylation, resulting in reliable conclusions.