|Home | About | Journals | Submit | Contact Us | Français|
This systematic review synthesizes the diverse body of epidemiologic research accrued on inorganic arsenic exposure and respiratory health effects. Twenty-nine articles were identified that examined the relationship between inorganic arsenic exposure and respiratory outcomes (i.e. lung function, symptoms, acute respiratory infections, chronic non-malignant lung diseases, and non-malignant lung disease mortality). There was strong evidence of a general association between arsenic and non-malignant respiratory illness, including consistent evidence on lung function impairment, acute respiratory tract infections, respiratory symptoms, and non-malignant lung disease mortality. Overall, early life exposure (i.e. in utero and/or early-childhood) had a marked effect throughout the lifespan. This review also identified some research gaps, including limited evidence at lower levels of exposure (water arsenic <100 μg/L), mixed evidence of sex differences, and some uncertainty on arsenic and any single non-malignant respiratory disease or pathological process. Common limitations, including potential publication bias; non-comparability of outcome measures across included articles; incomplete exposure histories; and limited confounder control attenuated the cumulative strength of the evidence as it relates to US populations. This systematic review provides a comprehensive assessment of the epidemiologic evidence and should be used to guide future research on arsenic’s detrimental effects on respiratory health.
The World Health Organization (WHO) lists inorganic arsenic (InAs) as one of ten chemicals of major public health concern (World Health Organization, 2010b). Arsenic is a naturally occurring yet life-threatening toxicant to which millions are inadvertently exposed annually (World Health Organization, 2010a). Arsenic contaminates the groundwater (wAs) of many countries, including Bangladesh, Chile, China, India, Mexico, Central Europe, and the United States, at levels which exceed the WHO standard of 10 μg/L (World Health Organization, 2010a).
The International Agency for Research on Cancer classifies InAs and arsenic compounds as group 1 lung carcinogens, meaning there is sufficient evidence to conclude that InAs exposure from inhalation and ingestion causes lung cancer in humans (International Agency for Research on Cancer, 2009). Presently, InAs is the only group 1 lung carcinogen known to be active by both inhalation and ingestion (Smith et al., 2009). Compared to the body of research on InAs and lung cancer, however, the body of research on InAs and non-malignant lung disease is less cohesive and more difficult to characterize.
One of the first studies on InAs exposure and respiratory health dates back to the 1970s in Chile. In 1958 the city of Antofagasta started supplementing its main water supply (wAs levels around 90 μg/L) with river water (wAs levels near 1,000 μg/L) to accommodate the city’s growing population (Smith et al., 2006; Borgoño et al., 1977; Zaldivar, 1980). By 1971 a water treatment plant had been installed and wAs levels were gradually returned to pre-1958 levels. However, during the intervening period before the water was treated, lasting from 1958 to 1970, all residents of Antofagasta, Chile consumed very high levels of InAs (wAs>800 μg/L) in their drinking water (Borgoño et al., 1977). In Borgoño et al. (1977), which occurred several years after peak exposure, children with arsenical skin lesions had a greater prevalence of bronchopulmonary disease history and chronic cough compared to children who did not have arsenical lesions. Although symptoms and functional disabilities associated with non-malignant lung disease generally appear in late adulthood, the Antofagasta study began shedding light on an important result - that developing lungs are particularly vulnerable early in life and that InAs exposure in utero and during childhood can have life-long consequences.
There is only one review article on InAs and non-malignant lung disease, published almost a decade ago (Guha Mazumder, 2007), but no systematic reviews or meta-analyses have summarized the evidence on this body of research. Instead of focusing on a singular aspect of non-malignant lung disease, this systematic review holistically describes the relationship between InAs and non-malignant lung disease, and also groups respiratory health diseases into easily interpretable categories, including lung function, respiratory symptoms, acute respiratory tract infections, chronic non-malignant lung diseases and non-malignant lung disease mortality. We define respiratory health quite broadly but because this outcome is inherently heterogeneous, encompassing numerous phenotypes among people of different ages and various disease trajectories, there may be additional non-malignant lung disease outcomes which we have unintentionally neglected in this review.
Responding to a U.S. congressional mandate, the U.S. National Research Council convened a group of experts in 2012 to evaluate and guide the Environmental Protection Agency’s (EPA) Integrated Risk Information System’s (IRIS) toxicological assessment of InAs (National Research Council, 2014). The National Research Council recommended the EPA conduct systematic reviews on 18 health endpoints of concern, including non-malignant respiratory effects, to support the agency’s assessment. Several published systematic reviews and meta-analyses have already synthesized the epidemiological evidence on the relation between InAs and other health consequences, including lung cancer (Celik et al., 2008), skin lesions and skin cancers (Karagas et al., 2015), cardiovascular disease (Navas-Acien et al., 2005; Moon et al., 2012), hypertension (Abhyankar et al., 2012), adverse pregnancy outcomes (Quansah et al., 2015), chronic kidney disease (Zheng et al., 2014), urinary tract cancers (Saint-Jacques et al., 2014), and type-2 diabetes (Navas-Acien et al., 2006). A systematic review on InAs and respiratory health is greatly needed.
We implemented a systematic approach to identify, evaluate, and synthesize the epidemiologic evidence on this body of research to better understand the effects of InAs exposure on different parameters of non-malignant lung disease in InAs-affected populations. We also examined whether there are timing-, dose- and sex-specific effects.
We implemented an extensive search strategy with guidance from a Reference Librarian at Columbia University Medical Center and in accordance with PRISMA guidelines (Moher et al., 2009):
For bibliographic databases, our search combined comprehensive English terms representing non-malignant respiratory health effects with terms for InAs exposure with the Boolean operator AND (See supplementary material for complete list of search terms). Before conducting a full search, we piloted our search and made slight modifications to meet the specific needs of each database search whenever necessary. We conducted two searches on each bibliographic database. We used both keywords and MeSH terms for PubMed, and similarly for EMBASE, we used keyword and EMTREE/exp terms. Searches were limited to English-language articles published before January 2016. Articles had to meet the following a priori criteria to be eligible for inclusion: 1) contained original human-based research published in a peer-reviewed journal; 2) had a control or referent group; and 3) included an indicator of InAs exposure studied in relation to any one or more of the following outcome categories listed in Table 1.
We downloaded our search results using Endnote and subsequently transferred the references to DistillerSR for title and abstract screening (http://distillercer.com/products/distillersr-systematic-review-software/). After removing exact article duplicates, TS screened all titles and abstracts (if available). JG and MP conducted a blinded independent random check on 20% of the initial search results, with 100% concordance between authors for included and excluded articles. Articles which met the initial screening stages underwent full-text review by all authors.
During the full-text review, we assessed whether multiple publications from the same study population contained duplicate data. We identified multiple publications of the same study (“study” meaning effort to collect primary data and “publication” meaning effort at analyzing the data) by examining author affiliation, study design, cohort name, enrollment criteria, and enrollment dates. When there were multiple publications reporting on the same study we described the relevant publications and noted potential overlap in the results section.
We created and piloted a data extraction form using DistillerSR. The piloted form was then modified to fit our needs. TS extracted information from each article and co-authors spot-checked data extraction.
For publications that examined either exposure or outcome categorically, we presented evidence from the highest category versus the lowest category. We performed additional subgroup analyses on 1) Critical periods of exposure (i.e. in utero and early life exposure); 2) Sex differences; and 3) Low level InAs exposure (wAs<100 μg/L). We specified all subgroup analyses a priori. We deemed a meta-analysis inappropriate due to the heterogeneous nature of the available publications.
Although no consensus exists on the ideal checklist and scale for assessing methodological quality (Moja et al., 2005), we developed our quality appraisal criteria from existing systematic reviews on InAs exposure (Navas-Acien et al., 2005; Zheng et al., 2014). Quality assessment questions were divided into five categories: exposure assessment, outcome assessment, statistical analysis, data collection, and specific questions for longitudinal studies. Exposure assessment for each study is also described in greater detail in results subsections. All authors independently assessed both the quality and risk of bias for each publication. Together authors resolved any disagreements and discussed any potential risk of bias that may have been overlooked.
The results from our systematic search are diagrammed in Figure 1. Our full search retrieved a total of 3,787 titles and abstracts from which 50 publications met preliminary screening criteria and received full-text review. Our final review included 29 publications from eight countries. Eight publications were from India, nine were from Bangladesh, four were from Chile, three were from the USA, two were from China, and one from Taiwan, Pakistan and Mexico. Articles were published in 21 different journals between 1998 and 2015. Sixteen publications were cross-sectional, eight were longitudinal, four were ecological, and one was case-control.
Exposure also varied widely by timing, duration and level of exposure. Several publications reported more than one relevant endpoint category: nine publications examined lung function, eighteen examined respiratory symptoms, seven examined acute respiratory tract infections, six examined chronic non-malignant lung disease, and five examined non-malignant lung disease mortality.
Nine articles examined the implications from in utero/early life InAs exposure at different lifestages, including endpoints during infancy (Farzan et al., 2015, 2013; Rahman et al., 2011; Raqib et al., 2009), childhood (Recio-Vega et al., 2014; Smith et al., 2013), and adulthood (Dauphine et al., 2011; Smith et al., 2011, 2006). Thirteen publications stratified their results by sex (Parvez et al., 2013; Pesola et al., 2012; Dauphine et al., 2011; Smith et al., 2011; Raqib et al., 2009; Smith et al., 2006; Mazumder et al., 2005; von Ehrenstein et al., 2005; Milton et al., 2001; Mazumder et al., 2000; Tsai et al., 1999; Smith et al., 1998). One publication examined the relationship between arsenic metabolism and lung function (Recio-Vega et al., 2014). Four articles examined the effects of InAs<100 μg/L (Farzan et al., 2015; Feng et al., 2015; Das et al., 2014; Farzan et al., 2013).
Figure 2 shows the quality assessment overview for each publication. Each results subsection applied a more rigorous quality assessment for select publications.
Seventeen publications assessed InAs exposure at the individual level. The remaining twelve assessed InAs exposure ecologically or by arsenical skin lesion status. Most publications had either an incomplete exposure history or did not include exposure from all major sources (water and diet). In general, publications with incomplete exposure histories either had a measure of current urinary arsenic (uAs) or wAs measurements (and/or recent exposure history) but had no record or early life exposure. Alternatively, publications had a measure of pre-natal/early life InAs exposure, but did not have a measurement of later life exposure/exposure contemporaneous with the outcome. Therefore publications using a marker of early-life exposure did not not necessarily include all major exposure sources throughout the life course. Additionally, dietary InAs exposure is a growing concern as a significant contributor to exposure, especially when InAs levels in drinking water are low. Thus publications which incorporated a biomarker of exposure and reported on prior InAs exposure levels, including exposure throughout the life course, had higher quality exposure assessment than publications which only classified InAs exposure according to wAs levels or which used an isolated measurement of uAs.
Respiratory symptom questionnaires and lung function measurements were generally standardized; however, most outcomes on chronic lung disease were self-reported or gleaned through death records and verbal autopsies. Diagnostic criteria for several outcomes, including chronic bronchitis and pneumonia were inconsistent across publications, thus increasing uncertainty in our classification by outcome (respiratory symptom vs. acute respiratory infection vs. chronic respiratory disease) and making comparisons between publications challenging.
All but two publications presented internal comparisons between study groups. Most publications adequately controlled for important potential confounders, including age, sex, and smoking. Participant selection criteria and participation rates were variably reported across publications. Among the eight cohort publications all but three reported follow-up rates and described how those lost to follow-up compared to those remaining in the study. It was often unclear whether the person in charge of collecting information on disease status was blinded to exposure status. Several articles failed to report both adjusted and unadjusted results. These methodological limitations attenuated the cumulative strength of the evidence on InAs and non-malignant lung disease. Overall, this systematic review includes articles of both high quality (robust exposure assessment, adjustment for potential confounders and standardized outcome measures) and low quality (incomplete exposure assessment/history, no adjustment for potential confounders, and non-standardized outcome metrics).
Nine publications examined the relationship between InAs and lung function (Feng et al., 2015; Recio-Vega et al., 2014; Das et al., 2014; Parvez et al., 2013; Smith et al., 2013; Nafees et al., 2011; Dauphine et al., 2011; von Ehrenstein et al., 2005; De et al., 2004). Two articles did not report point estimates for the selected lung function outcomes. De et al. (2004) and Recio-Vega et al. (2014) only reported means, standard deviations and p-values. Table 2 shows the associations on the relationship between InAs and three common spirometric outcomes used to measure lung function: Forced Expiratory Volume in one second (FEV1), Forced Vital Capacity (FVC), and the ratio of FEV1 to FVC (FEV1/FVC).
Spirometry is a commonly used, yet effort-dependent pulmonary function test (PFT) and spirometry test efforts must meet important acceptability criteria. Seven of nine publications used American Thoracic Society (ATS) criteria (Miller et al., 2005)to assess spirometry acceptability (Feng et al., 2015; Das et al., 2014; Recio-Vega et al., 2014; Parvez et al., 2013; Smith et al., 2013; Dauphine et al., 2011; Nafees et al., 2011). Spirometry outcomes were reported in four different ways: raw measured values (mL), predicted values using reference values derived by ethnic group (%predicted), residual values (%predicted - measured), and disease patterns (restrictive vs. obstructive). Although several publications used reference values from a marginally related ethnic group to estimate percent predicted values rather than using raw values and adjusting for sex, age, and height, those authors noted the choice of reference was not critical because the purpose was to compare InAs exposed to unexposed using the same reference values (Feng et al., 2015; Das et al., 2014; Recio-Vega et al., 2014; Dauphine et al., 2011; von Ehrenstein et al., 2005; De et al., 2004). Regardless of how PFT outcomes were reported, the direction of the association between InAs and lung function was consistent; FVC declined with increasing InAs exposure across all nine publications. Six of the seven publications that reported point estimates found a statistically significant decrease in FEV1 and four found a statistically significant decrease in FVC.
Of the three articles that used ATS criteria and reported raw values adjusted for age, sex, and height, there was a consistent trend between increasing InAs concentration and decreasing FEV1 and FVC. Parvez et al. (2013), a longitudinal post-bronchodilator PFT analysis among 942 Bangladeshi adults, found for every 118.1 μg/L increase in baseline wAs, FEV1 and FVC were lowered by 46.5 ml and 53.1 ml, respectively. Results for uAs were similar. Exposure was measured at least eight years before the outcome was assessed; however, early life exposure history was absent. Nafees et al. (2011), a smaller cross-sectional PFT analysis among 200 Indian adults, found that among participants exposed to wAs≥ 100 μg/L, mean FEV1 and FVC were lowered by 154.3 ml and 221.9 ml, respectively, compared to those exposed to wAs≤ 10 μg/L. Although exposure assessment was based on wAs levels measured at the time of enrollment, participants had been using the same water source for at least one year. Among children, the relationship between InAs and lung function was in the same direction, yet the estimated effect size was greatly attenuated. Smith et al. (2013), a longitudinal PFT analysis among a sample size of 442 Bangladeshi children 7–17 years old found for every 250 μg/L increase in wAs concentration in utero, FEV1 and FVC were lowered by 0.013 ml and 0.007 ml, respectively, although results did not reach statistical significance. Here, exposure assessment was particularly strong, as children’s lifetime exposure history was reconstructed using their mother’s wAs records, where wAs ranged from ≤ 10 μg/L to >500 μg/L.
Whereas FEV1 and FVC values are known to differ by ethnicity, the FEV1/FVC ratio is independent of ethnic group (Quanjer et al., 2012). Five publications reported on InAs and FEV1/FVC (Feng et al., 2015; Parvez et al., 2013; Nafees et al., 2011; von Ehrenstein et al., 2005). von Ehrenstein et al. (2005), a cross-sectional analysis among 281 Indian adults, found a statistically significant trend between presence of arsenical skin lesions and lower FEV1/FVC ratio among men but not among women. The FEV1/FVC ratio was significantly lowered by 0.025 among men with skin lesions compared to men without arsenical skin lesions, adjusting for age, height, and smoking (p=0.03); they used both skin lesions and peak wAs exposure level in the past twenty years in their analysis. Das et al. (2014), Parvez et al. (2013) and Nafees et al. (2011) briefly describe an inverse trend between InAs and the FEV1/FVC ratio, although results did not reach statistical significance. Feng et al. (2015), a cross-sectional analysis among 2360 Chinese adults, found a positive but non statistically significant trend between uAs and FEV1/FVC; however, exposure assessment was based on one measure of uAs among a study population with low levels of exposure (mean uAs=27.6 μg/L).
Few studies were able to capture all major exposure sources from conception through adulthood. Although this is a extremely challenging aspect for any epidemiologic study, three publications had exposure evidence from early life (Smith et al., 2013; Dauphine et al., 2011; Recio-Vega et al., 2014). Early life exposure appeared to detrimentally impact lung function levels in adulthood; however, the relationship between InAs and lung function during childhood and adolescence, periods of continued lung development, was less clear. Dauphine et al. (2011), a pilot study of 97 Chilean adults who were exposed to very high levels of wAs in utero and in early life, found that among adults exposed to wAs>800 μg/L, FEV1 and FVC decreased by 224 ml and 310 ml, respectively, compared to adults who were exposed to wAs ≤ 250 μg/L, although results were borderline significant (FEV1, p=0.06, FVC, p=0.04). Dauphine et al. (2011) had compelling evidence of early life exposure history. Here, exposure history was based on participants’ long-term residence in either Antofagasta or Arica, Chile. Arsenic levels in the drinking water of Arica have always remained around 10 μg/L, whereas Antofagasta’s only source of drinking water averaged 870 μg/L during participants first years of life, but fell to 150 μg/L during later life. Recio-Vega et al. (2014) examined a population of Mexican children aged 6–12 years old who were exposed to wAs between 104–360 μg/L throughout their lives. Here, children’s early life InAs exposure history was assessed by restricting enrollment to participants who had lived in the same community since the child was conceived and used the child’s current uAs level as a proxy for average lifetime exposure. There was a modest decline in %predicted FEV1 and %predicted FVC with increasing quartile of uAs, whereas, %predicted FEV1/FVC remained unchanged. Smith et al. (2013) examined a population of Bangladeshi children aged 7–17 years old exposed to a much wider range of wAs (0.01–1,512 μg/L). There was a slight decrease in FEV1 and FVC, however the results were not statistically significant. Collectively, these three publications implicate early life exposure to InAs as particularly hazardous to lung function.
Three articles presented sex-stratified results (Dauphine et al., 2011; von Ehrenstein et al., 2005; Parvez et al., 2013). In all three, the effect estimate on InAs and lower lung function levels was consistent and statistically significant in men but not women. In Parvez et al. (2013), the effect estimate for FVC was lower among men than among women and the effect estimate for FEV1 was statistically significant in men only. Among women, the direction of association was consistent but not significant. Similarly, in von Ehrenstein et al. (2005) InAs exposure, measured using either wAs or skin lesion status, was significantly associated with lower FEV1 and FVC among men, but not among women. Among women, the direction of association was again consistent but not significant. In Dauphine et al. (2011), early life InAs exposure was borderline significantly associated with lower FVC among men (p=0.06) but not among women. The direction of the association between InAs and FEV1 was consistent among both men and women but was not statistically significant, which may be explained by the small sample size (n=97). Contrarily, Smith et al. (2013) did not find evidence of an interaction by sex (p-value>0.2). Das et al. (2014), a publication on 834 non-smoking adult males in India, also found a statistically significant negative association between wAs and both FEV1 and FVC. Thus, the sex difference findings are largely consistent.
Six articles examined the exposure-response trend of InAs and lung function (Feng et al., 2015; Recio-Vega et al., 2014; Smith et al., 2013; Parvez et al., 2013; Nafees et al., 2011; von Ehrenstein et al., 2005) (Supplementary material). An inverse-trend was generally strong for both FEV1 and FVC. In Nafees et al. (2011) the statistically significant exposure-response relationship between individual-level wAs and decreasing FVC still held at low levels of exposure (wAs<100 μg/L). Dauphine et al. (2011) also found an inverse exposure-response trend, for both FEV1 and FVC. In Recio-Vega et al. (2014) there was a decreasing trend in mean FEV1 and FVC values with increasing quartile of unadjusted uAs, however results did not reach statistical significance. Parvez et al. (2013); Nafees et al. (2011); von Ehrenstein et al. (2005) all found an exposure-response relationship between increasing level of wAs and decreasing levels of FEV1 and FVC across a wide range of wAs exposures. However, Smith et al. (2013) did not find evidence of an exposure-response trend at three levels of wAs (<10 μg/L, 10–499 μg/L and >500 μg/L), nor did Feng et al. (2015) find evidence of a dose-response trend between quartiles of uAs; however, the range of exposure was narrow.
More than half of the included publications in this review examined the relationship between InAs and respiratory symptoms. Articles examined symptoms individually or combined, and reported on anywhere from one to nineteen different symptom-related endpoints. Of the 18 publications on InAs and respiratory symptoms, 11 reported a statistically significant positive association between InAs and at least one respiratory symptom (Das et al., 2014; Bhattacharyya et al., 2014; Paul et al., 2013; Smith et al., 2013; Pesola et al., 2012; Parvez et al., 2010; Ghosh et al., 2007; Mazumder et al., 2005; Guo et al., 2003; Milton et al., 2001; Mazumder et al., 2000). However seven did not find a significant association for any respiratory symptom assessed (Recio-Vega et al., 2014; Farzan et al., 2015, 2013; Amster et al., 2011; Dauphine et al., 2011; Nafees et al., 2011; von Ehrenstein et al., 2005). Table 3 shows the associations of this relationship among the most commonly reported endpoints, chronic cough, shortness of breath and “any respiratory symptom.”
Five articles (Recio-Vega et al., 2014; Smith et al., 2013; Dauphine et al., 2011; Das et al., 2014; Nafees et al., 2011) adapted standardized respiratory symptom questionnaires from the American Thoracic Society (Ferris, 1978), the International Study of Asthma and Allergies in Childhood (Pearce et al., 2007), and the British Medical Research Council (Cotes, 1987).
Five publications combined several respiratory symptoms into one inclusive variable (Farzan et al., 2015; Das et al., 2014; Farzan et al., 2013; Paul et al., 2013; Ghosh et al., 2007). Although these publications included slightly different combinations of symptoms, the positive association between InAs and “any respiratory symptom” was consistent and statistically significant. Paul et al. (2013), a cross-sectional analysis repeated at two time points among Indian adults, found InAs-exposed participants had significantly higher odds of respiratory symptoms (persistent cough, thoracic sounds, throat irritation, loss of breath, hoarseness) compared to unexposed (unadjusted OR=11.45, 95% CI=5.04, 25.97). Here, exposure was primarily assessed by presence or absence of skin lesions and confirmed by wAs and uAs levels among a subset of participants. Ghosh et al. (2007), also examined the same set of respiratory symptoms in a group of 1,114 Indians ranging from 15 to 70 years old. Exposure grouping was assessed by location of residence, and confirmed wAs and uAs level among some participants. Ghosh et al. (2007) found exposed participants had greater odds of respiratory illness compared to unexposed, regardless of presence of skin lesions and after adjusting for age, sex, and smoking. Das et al. (2014), a cross-sectional analysis of 834 adult non-smoking men in India, found the odds of any lower respiratory symptom (dry cough, cough with phlegm, wheeze, chest discomfort, breathlessness on exertion) among InAs exposed males (11–50 μg/L) was 1.23 higher compared to unexposed (<10 μg/L), after adjusting for exhaled CO, indoor air pollution, and agricultural occupation. Exposure assessment was based on residing in either an InAs-endemic village for the past 10 years or InAs-free village for the past 5 years.
Urinary arsenic may be a more important marker of exposure when InAs levels in drinking water are low, as urinary levels reflect all ingestion pathways. Notably, Farzan et al. (2015) and Farzan et al. (2013), two publications on same study population of US infants exposed to InAs in utero, expressed the importance of dietary InAs exposure in their study population and used maternal uAs concentrations during pregnancy to assess in utero exposure. Both publications found a borderline significant positive association between InAs and any respiratory symptom (cough, difficulty breathing, wheeze). Compared to the earlier publication, Farzan et al. (2015) had double the sample size and two additional follow-up periods (8 months and 12 months).
Fourteen publications examined the relationship between InAs and cough (Farzan et al., 2015; Bhattacharyya et al., 2014; Das et al., 2014; Recio-Vega et al., 2014; Smith et al., 2013; Amster et al., 2011; Dauphine et al., 2011; Nafees et al., 2011; Parvez et al., 2010; Mazumder et al., 2005; von Ehrenstein et al., 2005; Guo et al., 2003; Milton et al., 2001; Mazumder et al., 2000). Publications commonly categorized cough by productive/non-productive and severity/frequency. Overall, InAs was significantly positively associated with coughing symptoms in six of the fourteen publications; however, not every publication adjusted for potential confounders, including age, sex and smoking. Of the articles that found a statistically significant positive relationship between InAs exposure and cough, only two sufficiently adjusted for important potential confounders (Parvez et al., 2010; Guo et al., 2003). In Parvez et al. (2010) over 11,000 Bangaldeshi adults were followed over four years. Exposure was measured at the individual level, using both well wAs and uAs. Parvez et al. (2010) found the hazard ratio for incident chronic cough among those with the highest quartile of wAs (>178 μg/L) was significantly greater than those in the lowest quartile of exposure (≤ 10 μg/L) after adjusting for age, sex, BMI, smoking, education, skin lesion status and well-switching status. While exposure history did not extend back to early life, this publication did capture all possible exposure routes, as the association was similarly consistent and significant for uAs. Guo et al. (2003) described a cross-sectional analysis among 620 Chinese adults. Exposure assessment was based on area of residence, with a confirmatory wAs testing on a subset of 25% and 10% of individuals in InAs exposed and unexposed villages, respectively. They reported that participants from the InAs exposed village had 12.8 times the odds of cough compared to those in the unexposed village, after adjusting for sex, age, smoking, and alcohol consumption (95% CI=5.8, 25.9).
Although the four other publications that controlled for potential confounders did not find a statistically significant relationship among InAs and cough (Farzan et al., 2015; Smith et al., 2013; Dauphine et al., 2011; Amster et al., 2011) the direction of the association was consistently positive in all but one publication. Amster et al. (2011) cross-sectionally examined the relationship between InAs and self-reported chronic cough among 2,687 US adult participants in the 2003–2004 and 2005–2005 NHANES cohorts. Curiously, they found participants in the highest quartile of uAs (>17.23 μg/L) had lower odds of chronic cough (defined as having a cough for three consecutive months or more throughout the year) compared to participants in the lowest quartile of uAs (<3.52 μg/L), after adjusting for sex, age, race/ethnicity, education, serum cotinine, serum mercury, urinary creatinine and arsenobetaine (OR=0.49, 95% CI=0.16,1.50); however, they relied on a single measure of uAs as a proxy for past InAs exposure. Urinary arsenic, while a standard marker for recent exposure, is a weak maker of long term exposure. Farzan et al. (2015), also examining a US population, found prenatal uAs exposure was marginally associated with an increased risk of cough (defined as lasting two or more days during in the first year of life) among infants (RR=1.1, 95% CI=1.0, 1.2). Smith et al. (2013), examined three different variables related to cough in the last 12 months: coughing when not having a cold, coughing when having a cold, and dry cough. The odds of coughing when not having a cold among children exposed to higher levels of wAs in utero (≥ 500 μg/L) was 2.23 times that of children exposed to wAs<10 μg/L, after adjusting for age, sex, mother’s education, father’s education and smoking status, and number of rooms in the house (95% CI=0.96, 5.16). Dauphine et al. (2011) examined the prevalence of chronic cough among Chilean adults exposed to very high levels early in life. The odds of chronic cough among those exposed to wAs >800 μg/L before age 10 was 1.3 times the odds among those exposed to wAs<250 μg/L, after adjusting for age, sex, smoking, childhood secondhand smoke, fuel type used in childhood home, occupational air pollution and education. Even though exposure was assessed ecologically, exposure history in this population was very strong (See Section 3.2.2.).
Eleven publications examined InAs and a symptom concerning shortness or breath/dyspnea/trouble breathing (Farzan et al., 2015; Bhattacharyya et al., 2014; Das et al., 2014; Recio-Vega et al., 2014; Smith et al., 2013; Pesola et al., 2012; Dauphine et al., 2011; Nafees et al., 2011; Parvez et al., 2010; von Ehrenstein et al., 2005; Mazumder et al., 2000). Overall, the direction of the association was consistently positive for all but one publication (Recio-Vega et al., 2014). Half of the publications used standardized questionnaires and graded scales of dyspnea based on different circumstances in which it arose (Das et al., 2014; Recio-Vega et al., 2014; Smith et al., 2013; Dauphine et al., 2011; von Ehrenstein et al., 2005). Of the five articles that adjusted for potential confounders, including sex, age, and smoking status, four found a significant positive relationship between InAs and dyspnea (Farzan et al., 2015; Smith et al., 2013; Pesola et al., 2012; Dauphine et al., 2011; Parvez et al., 2010). Smith et al. (2013) and Dauphine et al. (2011) found participants exposed to high levels of wAs had greater odds of breathlessness (and at multiple degrees of severity); both publications also had strong exposure assessments. Similarly, Parvez et al. (2010) and Pesola et al. (2012) found significant, positive associations. These two publications reported dyspnea findings from the same study population, the Health Effects of Arsenic Longitudinal Study (HEALS cohort), but Pesola et al. (2012) examined HEALS baseline symptoms, whereas Parvez et al. (2010) excluded those with respiratory symptoms at baseline and examined symptom incidence at HEALS follow-up visits.
Mazumder et al. (2000, 2005) and von Ehrenstein et al. (2005) also reported findings from the same source population of 7,683 participants in a cross-sectional study originally conducted from 1995 to 1996 in 24-South Parganas, West Bengal, India. Here, every member of the household who was present at the time of interview was invited to participate, regardless of age. Mazumder et al. (2000) reported findings from the first iteration of this study, among 6,864 non-smokers. They found age-adjusted prevalence of cough and dyspnea increased with increasing wAs level, particularly among those with arsenical skin lesions; however, findings were not adjusted for other confounders. von Ehrenstein et al. (2005) reported findings from an interview conducted between April 1998 and January 2000. Here, smokers were included, participation was restricted to those 20 years or older, and skin lesion status was used a their primary exposure marker. They found among their 287 participants, odds for respiratory symptoms associated with arsenical skin lesion status were generally elevated in both smokers and non-smokers. The respiratory symptoms reported in Mazumder et al. (2005) came from an interview between 2001 and 2003. They found higher odds of chronic cough (defined as self-reported cough for ≥ 3 months per year for at least two years) among participants with arsenical skin lesions compared to those without lesions; however, chronic cough was a secondary analysis and the finding was not adjusted for important potential confounders, including age, sex, and smoking status.
Three of the four articles which examined in utero exposure found a statistically significant increase in respiratory symptoms among infants (Farzan et al., 2015), children (Smith et al., 2013), and adults (Dauphine et al., 2011). Among 4 month-old infants exposed to InAs in utero (uAs range=0.5–58.3 μg/L), there was an increased risk of acute respiratory symptoms treated with a prescription medication per one natural log unit increase in uAs exposure (RR=1.2, 95% CI=1.0, 1.5) (Farzan et al., 2015). In this US population with low wAs exposure levels, the authors noted the importance of dietary InAs exposure and thus used maternal uAs, which reflects all ingestion pathways, as a proxy for infants’ in utero exposure. Among children, 7–17 years old, those exposed to wAs in utero ranging from 10 μg/L to ≥ 600 μg/L, had increased odds of wheeze, asthma, coughing, and shortness of breath compared to children who were exposed to wAs<10 μg/L in utero and throughout childhood (Smith et al., 2013). Among adults, 32–65 years old, those exposed to wAs levels between 250–800 μg/L in early life, had increased odds in breathlessness when walking at group pace (OR=5.94, 95% CI=1.36, 26.0) compared to adults who were exposed to <250 μg/L in early life (Dauphine et al., 2011).
Five articles either mentioned or presented sex-stratified data on respiratory symptom effects (Pesola et al., 2012; Parvez et al., 2010; von Ehrenstein et al., 2005; Milton et al., 2001; Mazumder et al., 2000). There was some evidence that the odds of cough was greater for females than for males and odds of dyspnea was greater for males; however, evidence was not consistent. Pesola et al. (2012) found the odds ratio for dyspnea, among non-smokers, was greater for males than for females. von Ehrenstein et al. (2005) also found the odds ratio for dyspnea was greater for males than for females but found the odds ratio for cough was greater for females than for males; however, results for either symptom in either sex did not reach statistical significance. Alternatively, Mazumder et al. (2000) found the age-adjusted odds ratio for both dyspnea and cough were greater for females than for males. Milton et al. (2001) found the unadjusted odds ratio for chronic cough was greater for females than for males. Notably, Parvez et al. (2010) mentioned there was not an appreciable difference between sex and thus did not stratify results.
Five articles examined the exposure-response relationship between InAs and respiratory symptoms, with generally strong evidence of a positive trend (Recio-Vega et al., 2014; Smith et al., 2013; Pesola et al., 2012; Parvez et al., 2010; Mazumder et al., 2000). Pesola et al. (2012) found a statistically significant trend between increasing wAs quintile and increasing odds of dyspnea, even at lower levels of wAs exposure (39–91μg/L vs. ≤ 7 μg/L). Parvez et al. (2010) also reported an exposure-response trend between increasing wAs quintiles and increasing risk of chronic cough incidence (wAs range=<7 to >178 μg/L). Interestingly, the highest risk for dyspnea was among those with wAs between 40–90 μg/L. Recio-Vega et al. (2014) presented tabular results between uAs quartiles and several respiratory symptoms, but a modest trend was only seemingly apparent for frequent cough. Mazumder et al. (2000) also presented evidence of a statistically significant exposure-response trend between wAs and both increasing shortness of breath and cough (wAs range=<50 to >800 μg/L); however, the highest prevalence for dyspnea were among those with wAs between 200–499 μg/L. Smith et al. (2013) also examined the exposure-response trends for a subset of respiratory symptoms with an adjusted OR>2.0 in bivariate analysis (in utero wAs range=10 to >600 μg/L). There was a consistent trend between increasing bin of InAs exposure and increasing odds of six respiratory symptoms: wheezing ever, asthma, coughing when not having a cold, shortness of breath when walking fast or climbing, shortness of breath when walking on level ground, wheezing when not having a cold. Interestingly, for most symptoms the highest odds were among those with in utero wAs between 400–599 μg/L.
There were seven publications on InAs and respiratory tract infections (George et al., 2015; Farzan et al., 2015; Recio-Vega et al., 2014; Farzan et al., 2013; Smith et al., 2013; Rahman et al., 2011; Raqib et al., 2009). Overall, there was a modest positive association between InAs exposure and frequency and/or severity of respiratory tract infections (RTI) (Table 5). Articles examined infections combined and individually, including pneumonia and bronchiolitis.
Recio-Vega et al. (2014) reported a statistically significant increase in bronchiolitis among children in the highest quartile of uAs (>181 μg/L) compared to children in the lowest quartiles of uAs (<63 μg/L); however, only a p-value was reported and it was unclear whether results were adjusted for potential confounders. George et al. (2015), a case-control analysis, examined the association between physician-diagnosed WHO-defined pneumonia and uAs levels. There was a positive relationship between uAs level at hospital admission and odds of pneumonia, but only infants with uAs in the third quartile (17–50.9 μg/L) had significantly greater odds of pneumonia than infants in the lowest quartile of uAs (<6 μg/L), after adjusting for urinary creatinine, weight for height, breastfeeding, paternal education, age, and number of people living in the household (OR=2.11, 95% CI=1.10,4.34). Exposure assessment was restricted to uAs either at time of hospitalization or during recovery and did not have a record of past exposure among cases, whereas uAs was measured once at time of enrollment among controls. Smith et al. (2013) also presented some data on self-reported pneumonia. The article reported a statistically significant increase in pneumonia among children exposed to wAs 10–499 μg/L compared to children exposed to <10 μg/L, after adjusting for potential confounders, but the association was not statistically significant at higher exposure levels wAs>500 μg/L (OR=1.43, p-value=0.13).
As noted in section 3.3.1, Farzan et al. (2015) and Farzan et al. (2013) examined the same study population in the USA. Both analyses found in utero InAs exposure was associated with a higher risk of infection during the first year of life. Farzan et al. (2013) found an increased risk of lower RTI (LRTI) treated with a prescription medication (defined by RSV, pertussis, bronchitis, bronchiolitis, pneumonia) among 4 month old infants exposed in utero.
Articles by Rahman et al. (2011) and Raqib et al. (2009) were both from the same study region in Bangladesh. Whereas Raqib et al. (2009) had a much smaller study population (N=140), Rahman et al. (2011) studied 115,850 person-years of observation time from 1,552 participants. Raqib et al. (2009) found an increase in acute respiratory infection (defined by cough, fever or shaking chills, and fast breathing, difficulty breathing or chest retractions) with increasing in utero uAs level among infants 6–12 months old. Rahman et al. (2011) similarly found an increased relative risk of LRTI (defined by cough and/or difficulty breathing combined with rapid respiration) with increasing maternal uAs.
Four publications used maternal uAs as a marker of gestational exposure but did not have a measure of the child’s post-natal exposure (Farzan et al., 2015, 2013; Rahman et al., 2011; Raqib et al., 2009). Farzan et al. (2015) found that associations with InAs exposure during pregnancy were strong for infections at 4 months of age and weaker for infections at 8 or 12 months of age. Whereas Rahman et al. (2011) found that associations with InAs exposure during pregnancy were stronger for LRTI at 6–11 months of age than for LRTI at 0–5 months. Smith et al. (2013) reported greater odds of pneumonia only among older children 7–17 years old who were prenatally exposed (wAs 10–499 μg/L) compared with never exposed.
Raqib et al. (2009) found the correlation between in utero InAs and acute respiratory infections among boys (r=0.72, p=0.0001) was significantly greater than the correlation among girls (r=0.38, p=0.32).
Three publications examined RTIs at lower levels of exposure and report modest findings. Rahman et al. (2011) found a statistically significant trend between increasing quintile of mothers’ uAs and both increasing risk of LRTI and severe LRTI. Urinary arsenic exposure levels in Rahman et al. (2011) ranged from 6–977 μg/L, and even at the lowest quintiles of in utero exposure (39–64 μg/L vs. <39 μg/L) the association remained significant (RR=1.28, p=0.03). Arsenic exposure levels in Farzan et al. (2015) and Farzan et al. (2013) were much lower (mean uAs=3.7 μg/L, wAs range=0.01–67.5 μg/L), and had smaller effect estimates (RR=1.1, 95% CI=0.9, 1.4) of uAs and total number of ARTI. The interquartile range in uAs levels in George et al. (2015) ranged from 4.5–41 μg/L among controls, and from 8.0–64 μg/L among cases. George et al. (2015) found odds of pneumonia were greatest among the third quartile of uAs. Although Smith et al. (2013) did not formally test the exposure-response relationship between prenatal wAs exposure and pneumonia, the article did present some results from multiple logistic regression at two levels of wAs (10–499 μg/L and ≥ 500 μg/L) but a trend was not apparent.
Six publications examined the relationship between InAs and chronic non-malignant lung disease (Table 5). Few articles in this category examined the same disease outcome, and thus the relationship between InAs and specific chronic non-malignant lung diseases in adults was unclear. The specific disease outcomes presented here include pulmonary artery dilatation, bronchiectasis, emphysema, and chronic bronchitis.
Four publications examined the relationship between InAs and chronic bronchitis (Amster et al., 2011; Dauphine et al., 2011; Milton et al., 2001; von Ehrenstein et al., 2005). Milton et al. (2001), a cross-sectional analysis among 218 non-smoking Indian adults, found participants with arsenical skin lesions had greater risk of chronic bronchitis (defined as sputum production on most days for at least three consecutive months for more than two successive years with the presence of rhonchi and/or crepitation) compared to those without skin lesions, after adjusting for sex but not age (OR=3.0, 95% CI=1.6, 5.3). Exposure assessment in this analysis was multi-faceted. Exposed participants came from one of three InAs affected villages and had physician-confirmed arsenicosis, whereas unexposed participants were randomly recruited from a village without InAs-contaminated wells and had no evidence of arsenicosis. von Ehrenstein et al. (2005), a cross-sectional analysis among a total of 287 Indian adults, found that among smoking males, those with skin lesions had greater age-adjusted odds of chronic bronchitis (defined as cough and phlegm for at least 12 months) compared to those without skin lesions, however results did not reach statistical significance (OR=2.0, 95% CI=0.7, 5.8). Dauphine et al. (2011), a cross-sectional analysis among 97 Chilean adults (see section 3.2.2.), also found a modest association (OR=1.02, 95% CI=0.09, 11.6), but only three total cases of chronic bronchitis were reported. Alternatively, Amster et al. (2011), a cross-sectional analysis among 5,365 US adults, found those in the highest quartile of uAs (>17.23 μg/L) had lower odds of self-reported chronic bronchitis compared to those in the lowest quartile (<3.52 μg/L); however, results did not reach statistical significance even after adjusting for sex, age, race, education, serum cotinine, arsenobetaine and urinary creatinine (OR=0.77, 95% CI=0.24, 2.51). The direction of the association in Amster et al. (2011) is inconsistent with the other three publications on chronic bronchitis. Despite its large sample size, it is plausible that the discrepant results from Amster et al. (2011) may be related to incomplete history of prior As exposure. Although uAs is a good biomarker of recent exposure, it does not adequately reflect long-term InAs exposure. Further, the range of exposure in Amster et al. (2011) was also narrow (mean exposed uAs=172.3 μg/L, mean unexposed uAs=3.52 μg/L).
Amster et al. (2011) also reported greater odds of self-reported emphysema among those with the highest quartile of uAs compared to those in the lowest quartile, although results did not reach statistical significance (OR=1.29, 95% CI=0.17, 9.82). Mazumder et al. (2005) found a statistically significant increase in bronchiectasis (determined by high-resolution computed tomography) among those with skin lesions. One of the major limitations of this publication was that only participants with self-reported cough were given a referral for bronchiectasis testing and of those who were referred, those who underwent CT-testing differed by exposure group. Overall, 25% of exposed (27/108) and 7% of unexposed (11/150) underwent CT-test. Bhattacharyya et al. (2014) found a statistically significant increase in pulmonary artery dilatation (also determined by high-resolution computed tomography) among those with skin lesions; however, findings were not adjusted for potential confounders, including sex and smoking status.
Dauphine et al. (2011) was the only publication in this category with exposure assessment in early life; however there were not enough cases of chronic bronchitis to adjust for potential confounders.
Four publications stratified their results by sex (Bhattacharyya et al., 2014; Mazumder et al., 2005; von Ehrenstein et al., 2005; Milton et al., 2001). In general, associations among males seemed to be stronger than associations among females. Bhattacharyya et al. (2014) found the association between InAs exposure and pulmonary artery dilation was stronger in men (OR=7.53, 95% CI=2.15, 25.82) than in women (OR=4.49, 95% CI=1.11–23.53), results were adjusted for age but not smoking. Mazumder et al. (2005) also found the association between InAs exposure and bronchiectasis was stronger in men (OR=13.0, 95% CI=2.6, 62) than in women (OR=6.1, 95% CI=0.6, 62), after adjusting for age, smoking, and self-reported tuberculosis history. von Ehrenstein et al. (2005) reported findings among smoking males and non-smoking females, but not among non-smoking males. Whereas the direction of association between InAs and chronic bronchitis was again positive among smoking men (OR=2.0, 95% CI=0.7, 5.8), there was a negative association among non-smoking females (OR=0.3, 95% CI=0.03, 2.7), although neither finding reached statistical significance. Alternatively Milton et al. (2001) found the association between InAs exposure and chronic bronchitis among non-smokers was much stronger in women (OR=10.3, 95% CI=2.4, 43.1) than in men (OR=1.6, 95% CI=0.8, 3.1), although results were not adjusted for age.
None of the six article in this category examined dosimetry.
Five publications examined InAs and non-malignant lung disease mortality (Argos et al., 2014; Smith et al., 2011, 2006; Tsai et al., 1999; Smith et al., 1998) (Table 6). Publications studied several relevant causes of death, including bronchiectasis, COPD, bronchitis, emphysema, pulmonary tuberculosis, and non-malignant lung disease (a combination of primary pulmonary hypertension and diseases of the respiratory system).
Overall, there was strong evidence between InAs exposure and increased non-malignant lung disease mortality. However, there was also a moderate risk of outcome misclassification, as all publications were based on death certificates or verbal autopsy, which may be subjectively decided. Tsai et al. (1999), an ecological analysis in Taiwan, found significantly increased bronchitis mortality rates between 1971 and 1994, however it is possible that the findings may have been influenced by exposure misclassification, as there was greater variability of InAs in the exposure area. Argos et al. (2014), a longitudinal analysis of over 26,000 adults with over 220,000 person-years of follow-up time, also found greater hazards of non-malignant lung disease mortality among participants with the highest tertile of creatinine-adjusted uAs (≥ 332 μg/g Cr) compared to those with the lowest tertile (<132.5 μg/g Cr), after adjusting for important potential confounders, including sex, age, BMI, education, and smoking (HR=1.75, 95% CI=1.15, 2.66). Results were similar for wAs. Although Argos et al. (2014) did not capture early life InAs exposure, results from exposed populations in Antofagasta, Chile, with known lifetime exposure, show comparable findings. The 1998, 2006 and 2011 publications of Smith et al. studied the same region in Chile. Arsenic exposure was particularly well documented in this time period for Antofagasta, Chile, so there was little risk of exposure misclassification in the three Chilean ecological publications. Smith et al. 1998 and 2006 reported variations on one outcome of interest to this review, COPD. While Smith et al. (1998) found no overall increase in deaths from COPD, (SMR=1.0, 95% CI=0.8, 1.1), the analysis covered a shorter time period, 1989–1993 and included mortality at all ages. Smith et al. (2006) subsequently examined mortality from 1989–2000, focused on subjects between the ages 30–49, and subdivided COPD deaths into two categories, bronchiectasis and “other COPD”. They found significantly increased SMRs for COPD (SMR=7.6, 95% CI=3, 15.6) and bronchiectasis (SMR=46.2, 95% CI=21.1, 87.7), especially among those exposed in utero during peak InAs exposure. Interestingly, Smith et al. (2011) further explored the time trend and interplay between the respiratory and immune systems by estimating mortality rate ratios of pulmonary tuberculosis between two regions in Chile. This analysis showed tuberculosis mortality rate ratios started to increase 10 years after high InAs exposure.
Three articles examined whether high InAs exposure in utero and during childhood was associated with increased mortality in adulthood (Smith et al., 2006, 2011, 1998). All three publications came from Chile. In these articles, mortality was most pronounced when exposure occurred in utero and at very high levels. People exposed early in life died of bronchiectasis, COPD, and pulmonary tuberculosis at a younger age and at greater numbers compared to those exposed to InAs only in adulthood or later in childhood. Among these three ecological investigations, when exposure was not restricted by age, as in Smith et al. (1998) and Tsai et al. (1999), the relationship between InAs and increased respiratory mortality was attenuated. Upon further examination of Smith et al. (1998), however, the highest rates of COPD mortality were among adults aged 30–39 year olds who were exposed to high levels of InAs in utero and during childhood.
Four articles stratified their findings by sex (Smith et al., 2011; Tsai et al., 1999; Smith et al., 1998, 2006). Evidence of sex-specific effects was heterogeneous. In Smith et al. (2011), men had higher rates of pulmonary tuberculosis mortality than women. However, women had higher SMRs for chronic bronchitis, bronchiectasis and “other COPD” compared to men in articles by Smith et al. (2006); Tsai et al. (1999). Interestingly, in Smith et al. (1998), the SMR for COPD was null among men, but InAs exposed women had statistically significantly lower SMR for COPD.
Effect estimates were generally higher in the publications from Chile than in the publications in Taiwan or Bangladesh. Of note, exposure in Chile (mean wAs=870 μg/L) was much higher than in Bangladesh (mean wAs ≈ 34.5 μg/L). Argos et al. (2014) presented evidence on the exposure-response relationship. This publication found a statistically significant trend between increasing tertile of uAs and increasing magnitude of effect on non-malignant lung disease mortality.
This systematic review provides an extensive introduction to the field and should be used to guide future research on InAs and respiratory health. Our review identifies 29 publications that examine how InAs exposure maybe associated with different facets of respiratory health, including consistent evidence on lung function impairment, acute respiratory tract infections, respiratory symptoms, and non-malignant lung disease mortality particularly at high levels of exposure. Our review also uncovers some common shortcomings in this body of evidence, including incomplete exposure histories, poorly defined outcomes and limited confounder control.
While acknowledging the importance of our findings, there are some limitations worth noting. As with any systematic review, our findings may have been affected by publication bias. However, we have searched two complimentary databases and have performed an intensive gray literature search. The term non-malignant lung disease is loosely used to describe the non-cancerous effects of InAs on respiratory system. We have attempted to organize outcomes by clinically meaningful categories but this approach could have unintentionally complicated our interpretation of the relationship between InAs and lung disease. Our adapted quality assessment questionnaire is somewhat problematic. A publication may appear to be of higher or lower quality because the binary quality assessment was unable to capture nuanced details or authors failed to report needed details. Thus, interpreting how well a study did on our quality assessment should be done with caution. Several articles have used inappropriate reference values to predicted lung function estimates. In lieu of presenting the evidence on lung function using set cut-points to define obstructive and restrictive lung disease, we have chosen to present lung function as a continuous measure. While this is less clinically meaningful, we believe the former classification would have been misleading.
Although the findings from this systematic review show strong evidence of a general association between InAs and non-malignant respiratory illnesses, there is somewhat more uncertainty of a relationship between InAs and any single non-malignant respiratory outcome or pathological process. Arsenic could potentially contribute to different non-malignant lung diseases through several mechanisms, including altered wound repair, increased inflammation, and immune system dysregulation. Exactly how and which of these pathways contribute to non-malignant lung disease is still poorly understood and should be further investigated.
This review found strong evidence of an association between increasing InAs and decrements in FVC and FEV1. Some in vitro evidence indicates InAs exposure may alter the expression of an important protease in lung function, matrix metalloproteinase (MMP-9), and alter ATP-dependent Ca2+ signaling (Sherwood et al., 2011; Olsen et al., 2008). Both in turn could affect the signaling pathways for cell migration and alter the airway epithelial barrier by restricting proper wound repair and compromising innate airway defense mechanisms.
Alternatively, there is some evidence that reduced lung function is associated with increased risk for future cardiovascular morbidity and mortality (Sin et al., 2005; Yoon et al., 2014), another health system strongly associated with InAs exposure and increased inflammation (Abhyankar et al., 2012; Navas-Acien et al., 2005). This is also supported by some recent epidemiologic evidence of lung inflammation biomarkers and lung function (Olivas-Calderon et al., 2015; Parvez et al., 2008). Although evidence in this review on InAs and lung function in children was inconsistent, a reduction in FVC in young adulthood has been associated with diastolic dysfunction in middle age (Cuttica et al., 2015). Two publications in this review hint at InAs and cardiopulmonary effects: Argos et al. (2014) in their linkage of exposure to pulmonary hypertension mortality; and Bhattacharyya et al. (2014) in their examination of pulmonary arterial dilatation. Together, these findings add to the complex yet understudied interplay between the cardiovascular and pulmonary systems in relation to InAs exposure.
Chronic InAs exposure may also have biphasic effects on the immune system. In an in vivo model of influenza infection, among mice having first been exposed to 100 μg/L wAs, Kozul et al. (2009) observed that InAs exposed mice first display a compromised immune response to respiratory infections, but at later time points, these mice display an excessive cellular inflammatory response (Kozul et al., 2009). The current review found some evidence of positive relationship between in utero InAs exposure and acute respiratory tract infections among infants and compelling evidence on bronchiectasis among adults. Additionally, respiratory symptoms like chronic productive cough and reoccurring RTI may be early signs of non-cystic-fibrosis bronchiectasis (Pasteur et al., 2010).
The heterogeneity of the chronic non-malignant lung diseases examined is indicative of a potential area for future research. The publications on chronic non-malignant lung disease have some limitations, including insufficient confounder control and disease prevalence rather than incidence. While we believe there is compelling evidence on bronchiectasis, it is based on only two publications (Mazumder et al., 2005; Smith et al., 2006). In Bangladesh, Mazumder et al. (2005) found the odds of CT-diagnosed bronchiectasis among participants with skin lesions was 10 times that of participants without skin lesions, after adjusting for age, sex, smoking and self-reported tuberculosis history. The size of this effect estimate is compelling given the results are based on 21 total cases of bronchiectasis. There is also strong evidence on bronchiectasis mortality. In Chile, Smith et al. (2006) reports the SMR for bronchiectasis for men and women exposed to very high InAs levels in utero and early childhood was more than 46 times that of the rest of the country, although results were based on nine total cases of Bronchiectasis.
Research elucidating the pathophysiology of a single non-malignant lung disease as it relates to InAs exposure, like bronchiectasis, based on a careful clinical diagnosis of disease and a strong history and range of InAs exposure, would help corroborate these findings and further our understanding of the relationship between InAs and a single non-malignant respiratory outcome or pathological process.
There is growing evidence that InAs exposure in utero and during early life can have detrimental health consequences throughout the lifespan. This review finds adverse respiratory consequences can be detected even in the first months of life. There is some evidence that in utero InAs exposure increased both the frequency and severity of RTI among infants (Rahman et al., 2011; Farzan et al., 2015). Yet less evidence on the potential effects of early life arsenic exposure among adolescents. There is also strong evidence that mortality later in life is most pronounced when exposure occurs in utero and at very high levels. People exposed early in life seemingly die at a younger age and at greater numbers compared to those exposed to InAs only in adulthood or later in childhood (Smith et al., 2011, 2006). However, less is known about the mechanisms that contribute to latent disease development (Bailey et al., 2016).
Publications on other health effects related to InAs exposure indicate that men and women seem to be susceptible to different types of effects (Vahter et al., 2007). In general, effect estimates for InAs and lung function, respiratory symptoms, infection, and non-malignant lung disease morbidity seem to be stronger in men than in women; however, effect estimates for InAs and non-malignant lung disease mortality seem to be stronger in women than in men. Sex-differences in arsenic metabolism may also lend biological plausibility to the aforementioned differences. Inorganic arsenic is biomethylated in the body and forms monomethylarsonic acid (MMA) and dimethylarsinic acid (DMA). Methylated arsenicals are excreted more rapidly in urine than InAs; however, the trivalent forms of the methylated metabolites, MMAIII and DMAIII, are thought to be the most toxic species (Thomas et al., 2007). Several studies indicate that women have better methylation efficiency, as suggested by higher percentage of DMA and less MMA, compared to men (Lindberg et al., 2008; Gamble et al., 2005), perhaps contributing to the increased morbidity in men.
Although we had intended to perform a subgroup analysis on low-level wAs exposure (wAs <100 μg/L), wAs levels in most publications was >100 μg/L. Thus low-level effects are still poorly understood. Additionally, publications used different measures of exposure and incorporated different exposure contrasts, thus making it difficult to study whether similar exposure levels yielded similar effect estimates.
Of note, six of the included publications used the presence of arsenic-induced skin lesions as the primary exposure marker. While their presence may confirm prior InAs exposure, their absence does not exclude exposure. Arsenic-induced skin lesions are early biomarkers of effect (Haque et al., 2003). The presence of skin lesions also represents susceptibility to a certain degree, as not all people chronically exposed to InAs develop lesions and those with lesions appear to be susceptible to other adverse health outcomes, including lung and bladder cancer (Cuzick et al., 1992, 1984). Indeed, three articles in this review also find a greater magnitude of effect (on lung function, respiratory symptoms and non-malignant lung disease mortality) among people with skin lesions compared to InAs exposed people without skin lesions (von Ehrenstein et al., 2005; Ghosh et al., 2007; Argos et al., 2014).
In this systematic review we have summarized the epidemiologic literature on different ways in which InAs is associated with respiratory health to better inform stakeholders on the existing evidence in this field and to identify future avenues of research. In short, associations between InAs and respiratory health have been noted throughout the lifespan: in infancy, there is growing evidence that in utero InAs exposure is associated with increased frequency and severity of respiratory tract infections; in childhood, evidence of respiratory symptoms also begin to appear; and in adulthood, there is consistent evidence that InAs exposure is associated with deficits in lung function (particularly FVC) and increased reports of coughing and breathing problems. Limitations including potential publication bias; non-comparability of outcome measures across publications; incomplete exposure histories; potential effect modification by smoking; insufficient adjustment for indoor/outdoor air pollution and other important confounders attenuated the cumulative strength of the evidence on InAs and non-malignant lung disease as it relates to US populations. Greater attention should be paid both to examining whether sex differences exist and further elucidating the dose-response relationship between InAs and respiratory health, particularly at wAs levels <100 μg/L. Early life InAs exposure appears to have lasting effects on the respiratory system. This extensive review should be used to guide future research in order to define relevant mechanisms involved and implement effective public health measures.
This work was supported by the National Institute of Environmental Health Sciences [P42 ES010349]. We thank Lindsay M. Greenawalt, MLIS for assistance with designing PubMed and Embase literature searches.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.