Genetic susceptibility to adverse arsenic-related cardiometabolic outcomes: a systematic review

Search strategy and definitions

Five electronic databases were systematically searched using English search terms: PubMed, Web of Science, Scopus, EMBASE, and TOXLINE via TOXnet. Databases were searched from inception until 24 July 2019, with an updated search on 10 April 2023. Note, TOXLINE was excluded in the updated search as the database was retired in December 2019. Keywords used included indexed MeSH terms as well as free text terms related to ‘genetic polymorphism’, ‘genetic’, ‘arsenic’ and ‘arsenicals’ and were informed by previous publications [1], 16], and a medical information specialist (full search strategy in Supplementary Table 2). No restriction on the language or date was applied. Bibliographies of relevant publications (e.g., reviews, grey literature) and all included studies in the full-text screening were cross-referenced (forward and backward) to identify any additional eligible studies. The GWAS Catalog (https://www.ebi.ac.uk/gwas/) and Google Scholar were searched for additional studies fulfilling the inclusion criteria.

The exposure of interest was single nucleotide polymorphisms (SNPs): a single nucleotide change in the DNA sequence, comprising one of the most abundant types of genetic variation. No specific molecular processes, genes or SNPs were selected as a focus a priori, enabling the synthesis of all available evidence on common genetic variation(s) that may interact with arsenic exposure to impact arsenic-related cardiometabolic outcomes. Other variations in the DNA sequence, including other genetic variants (e.g., indels, microsatellites, repetitive elements), structural variation, and epigenetic variation were excluded. Haplotypes of linked SNPs within a chromosomal region were included.

This study included arsenic-related cardiometabolic diseases such as cardiovascular diseases (CVD, including coronary heart diseases [CHD], cerebrovascular diseases [CeD], peripheral arterial diseases [PAD], deep vein thrombosis [DVT]), diabetes mellitus (including type 1 [T1DM], type 2 [T2DM] and gestational [GDM]), hypertension, obesity, and metabolic syndrome. Subclinical cardiometabolic diseases (carotid atherosclerosis, arterial stiffness, ECG abnormality), and diagnostic markers (blood pressure [systolic, diastolic, pulse], blood glucose levels, carotid intima-media thickness [CIMT], HbA1c% levels) as continuous outcomes were included. Predictive biomarkers (e.g., asymmetric dimethylarginine (ADMA), fatty acid-binding protein 4 (FABP4) for CVDs) were excluded. Standard definitions following commonly used guidelines for the different cardiometabolic outcomes were applied (e.g., ACC/AHA [17] and ESC/ESH [18] guidelines for hypertension, or WHO [19], ADA [20], and NICE [21] guidelines for T2DM). Other diseases, such as skin lesions or cancer, were excluded, as were studies that only assessed the impact of SNPs on arsenic metabolism.

Study selection

Studies that met the inclusion criteria were: i) peer-reviewed quantitative epidemiological studies, ii) studies exploring whether one or more SNPs affects susceptibility to adverse cardiometabolic outcomes related to arsenic (i.e., whether a genetic variant is associated with higher or lower cardiometabolic risk when a person is exposed to arsenic), and iii) studies including arsenic species to which the general population is commonly environmentally exposed (i.e., inorganic arsenic [iAs] exposure or its metabolites [MMA, DMA], see Supplementary Table 3). Studies were excluded if they were studies that i) did not include a quantitative measure of arsenic exposure, ii) focused on forms of arsenic exposure uncommon in the general population such as medical therapy (e.g., melarsoprol treatment for parasitic infections, arsenic trioxide for leukemia), or intentional poisoning (homicide, suicide), ii) were non-human (i.e., in vitro, animal studies), iii) had a secondary study design (e.g., reviews, meta-analyses), or that reported on data previously reported elsewhere, iv) were conference proceedings, case reports or case series, v) lacked an available full-text (after request of full-text from the authors), or vi) had overlapping study populations with another included study within this review (with exception of those studying a different cardiometabolic outcome and partially overlapping study populations). Studies on arsenic-exposed populations that did not include SNP-related arsenic susceptibility or SNP-arsenic interaction analyses were also excluded.

After the duplicates were removed in Endnote, abstracts and titles were blindly double-screened using Rayyan (https://rayyan.ai/). In this stage, all studies on SNPs in relation to arsenic-related disease or arsenic metabolism were included. Ambiguity in the inclusion and exclusion criteria was checked by conflict piloting after the first 500 records were screened, resulting in an inter-rater reliability kappa score of 98.6 %. In the second stage, full-texts were retrieved and double-screened, focusing only on arsenic-related adverse cardiometabolic outcomes as previously defined. Any disagreements between two reviewers were discussed until consensus was reached. To minimise bias, non-English records were included and translated/reviewed by a speaker of the language (i.e., Bengali, Dutch, French, Spanish, German, Serbian). However, few non-English studies were retrieved for full-text screening.

Data extraction

A data extraction tool was developed and pilot-tested on five randomly selected studies, and refined accordingly. Data were extracted by KvD, and verified by MJ. Data extracted included: first author, year of publication, study design, participant characteristics (country, age, sex, ethnicity and/or genetic ancestry), number of participants, years of follow-up (cohort studies), gene(s) and SNP(s) assessed (including their allele/genotype/haplotype frequencies), genotyping method, type of arsenic exposure (e.g., water, urinary or blood arsenic), method of arsenic exposure assessment, adverse cardiometabolic outcome (e.g., CVD, metabolic syndrome), outcome ascertainment method, measures of the combined effects (or interaction) of the SNPs and arsenic exposure on negative cardiometabolic outcomes (e.g., odds ratio [OR] estimates for interaction terms, joint ORs of the combined SNP-arsenic effect, relative excess risk due to the interaction [RERI], attributable proportion due to the interaction [AP], synergy index [SI], p-values for interaction [P_int]), variables adjusted for in the main statistical analyses, and an open field to record additional information. When both p-values and Q-values (p-values adjusted for multiple comparisons) were reported, both were extracted, and results were discussed based on Q-values. When both unadjusted and adjusted measures of effect were reported, both were extracted, but the models that adjusted for the greatest number of confounders were discussed in the information synthesis. This approach was taken even when models adjusted for different variables across studies considering the adjusted effect estimates are more likely to be representative of the “true” effect than crude models.

To create an overview of all genes and SNPs assessed within the studies included review, an overview table was generated listing the corresponding gene, rsID, alleles, and consequence of the variation for each individual SNP (i.e., missense variant, synonymous variant, intron variant, noncoding transcript variant, 5′ UTR variant, 3′ UTR variant, 2KB upstream variant, 500B downstream variant) based on information in the included studies, the dbSNP database (https://www.ncbi.nlm.nih.gov/snp/) and the integrative database GeneCards (https://www.genecards.org/).

Individual study quality assessment

To allow for the incorporation of genetic and environmental components in individual study quality assessment, methodological quality was assessed using study design adapted versions of the quality assessment tool developed by Moon et al. 2017 for studies on environmental arsenic [1], 22] – and the Q-genie tool for the quality assessment of genetic association studies [23]. The tool was pilot-tested on five randomly selected studies, and refined accordingly. No overall score (e.g., good, fair, poor) was given as this can oversimplify important differences in bias, confounding, and overall quality of the different studies. Almost none of the studies explored the same SNP and cardiometabolic outcomes (Table 2, 3), therefore no overall assessment of the certainty of the body of evidence using Grading of Recommendations, Assessments, Development and Evaluation (GRADE) was performed.

Information synthesis

Due to the heterogeneity of included studies [i.e., genes/SNPs assessed, outcome assessed (e.g., CVD, DM, hypertension), and arsenic species assessed], data were descriptively synthesised, and no meta-analyses were performed. Studies were grouped and synthesised by the outcome of interest, and, where possible, further discussed by biochemical processes relevant genes are involved in. Due to the limited number of (heterogenous) studies included, publication bias could not be assessed.

Cardiometabolic outcomes

Studies reported on different cardiometabolic outcomes, including: hypertension or blood pressure (SBP, DBP, PP) [24], 28], 31], 36], metabolic syndrome [25], obesity [39], T2DM or insulin resistance [32], 35], 38], 40], 41], GDM [37], carotid atherosclerosis or CIMT [26], 27], 30], 33], CVD [34], CHD [34], stroke [34] or ECG abnormalities consistent with MI or ischemia, conduction defect, arrhythmia, arterial enlargement or ventricular hypertrophy or prolonged ventricular repolarization (Table 1) [29].

Hypertension and blood pressure (i.e., systolic, diastolic, pulse pressure) were assessed using a mercury or automated sphygmomanometer [24], 28], 31]. Hypertension was defined as SBP≥130 mmHg and/or DBP≥80 mmHg [24], or as SBP ≥140 mmHG and/or DBP ≥90 mmHg [28]. The studies on T2DM defined DM as fasting blood glucose (FBG) ≥126 mg/dL [38], 41], HbA1c ≥6.5 % [32], 35], 38], self-reported physician diagnosis or T2DM treatment [38], and 2-h post glucose ≥200 mg/dl [40], with homeostasis model assessment of insulin resistance (HOMA2-IR) computed using fasting glucose and insulin values [41]. GDM was defined at 24–28 weeks of pregnancy as FBG ≥5.1 mmol/L, or 1 h plasma glucose (PG) ≥10 mmol/L and 2hPG ≥8.5 mmol/L after 75 g oral glucose tolerance test (OGTT) [37]. Carotid atherosclerosis was defined based on the intima-media thickness (>1 mm), plaque score (≥1), and maximum level of stenosis of the ECCA (≥50 %) [26], 27], 30]. CVD, CHD and stroke diagnoses were based on a validated verbal autopsy procedure [34], and obesity was defined as those with body mass index (BMI) values of ≥30 kg/m² Finally, metabolic syndrome cases were those with three or more of the following identified FPG ≥110 mg/dL, triglycerides ≥150 mg/dL, HDL ≤40 mg/dL for men or ≤50 mg/dL for women, SBP ≥130 mmHg, DBP ≥85 mmHg, and waist girth ≥90 cm for men ≥80 cm for women [34]. Studies focused predominantly on non-fatal outcomes, with only one of the studies focusing on fatal events (Table 1) [34].

Environmental arsenic exposure

Environmental arsenic exposure was measured in water [26], [27], [28], [29], [30], [31], [32], [33], [34], [40], spot morning [24], 41] or spot [25], [33], [34], [35], [36], [37], [38] urine (adjusted [24], 25], 33], 34], 38], 41] or not adjusted for creatinine [35], [36], [37]), blood [31], 36], and blood plasma [39]. Most studies performed elemental analyses, with some performing speciation analyses (Table 1) [25], [33], [34], [35], [37], 38], 41]. In the SNP-arsenic-cardiometabolic disease analyses, most studies focused on total arsenic (tAs) or inorganic arsenic (iAs) [24], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [[38], [39], [40], [41] with only two studies including analyses exploring the interactions between SNP and arsenic metabolism markers: Primary methylation index [PMI] (defined by MMA/iAs) [25], Secondary Methylation Index (SMI) (defined by DMA/MMA) [25], iAs% [37], MMA% [25], 37], or DMA% [25], 37]. Studies reported predominantly on point arsenic exposure [24], [25], [26], [27], [[30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41] or cumulative arsenic exposure (CAE) [26], 28], 29], and for two studies it was assumed that multiple arsenic measurements were performed [28], 29].

Chemical analyses were performed using inductively couple plasma mass spectrometry (ICP-MS) [32], 36], 38], 39], high-performance liquid chromatography coupled to ICP-MS (HPLC-ICP-MS) [25], 33], 34], 37], 41], high-resolution ICP-MS (HR-ICP-MS) [31], anion exchange liquid chromatography coupled to ICP-MS (AEC-ICP-MS) [38], hydride generation atomic absorption spectroscopy AAS (HG-AAS) [26], 27], 30], 32], 40], graphite furnace AAS (GF-AAS) [24], 31], 33], 34], liquid chromatography-atomic fluorescence spectrometry (LC-AFS) [35] or results from previous reports (without detail on the chemical analyses performed) [28], 29].