Electronic waste exposure and DNA damage: a systematic review and meta-analysis

Protocol and registration

A review protocol was developed and registered with the International prospective register of systematic reviews (PROSPERO) with registration number CRD42020201149, and it is available from https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=201149.

Eligibility criteria

This review focused on observational studies on human populations exposed to e-waste recycling. We included studies that were original peer-reviewed publications, assessed e-waste exposure and biomarkers of DNA damage, and involved human populations, including women and children. We excluded studies that were not original (e.g., reviews, conference proceedings, letters to the editor, and abstracts) and did not report biomarkers of DNA damage in human populations.

Information sources and search strategy

Articles published in English from January 2000, investigating the associations between e-waste exposure and biomarkers of DNA damage were retrieved through the following three major databases: MEDLINE (Academic Search Complete, CINAHL Complete, Education Research Complete, GreenFILE, Health Source: Nursing/Academic Edition, Library, Information Science & Technology Abstracts), ProQuest, and Scopus. The search terms used included the following keywords: (“electronic waste” OR “e-waste” OR “WEEE”) AND (“DNA damage” OR “chromosomal aberration” OR “DNA strand breaks” OR “micronucl*” OR “Sister chromatid exchanges” OR “oxidative DNA damage” OR “genotox*” OR “oxidative stress”). Reference lists of selected articles were hand-searched for relevant publications that were not captured by the electronic search. The search strategy and results of the various databases are presented in (Supplementary Table 1).

Study selection

The study selection was conducted in 2 phases. In phase 1, two reviewers independently screened titles and abstracts of publications retrieved from the electronic databases and hand searches. Publications that did not meet the inclusion criteria were excluded. After phase 1 screening, 32 publications advanced to phase 2 (see supplementary material), full-text screening. In phase 2, the same reviewers independently examined the full-text publications for inclusion. Any discrepancies between reviewers were resolved by a third reviewer. Studies that were excluded at this stage are presented in Supplementary Table 2 with reasons. Finally, a total of 20 publications met the inclusion criteria.

Diagrammatic representation of study selection

The flowchart representing the process of study selection is presented in Figure 1. In the initial search of three electronic databases, a total of 822 articles were retrieved. Duplicates of 106 were identified and removed, with a total of 717 articles making it to the title and abstract screening stage. After the title and abstract screening, a total of 685 articles were excluded, allowing 32 articles to advance to the full-text screening stage.

Figure 1:

PRISMA flow chart illustrating the process of selecting the studies included in the review.

Out of the 32 full-text articles screened, 12 were excluded either because they were not original articles (5), did not meet the inclusion criteria (3), were retracted articles (1) or shared the same population and outcome with another study already included (3) (Supplementary Table 2). A total of 20 publications were included in this review, of which seven studies were within the occupational setting, and the rest were ecological studies. Seven of the 20 studies included in this systematic review were included in the meta-analysis.

Data extraction

The type of data extracted from each of the selected studies was both qualitative and quantitative. The qualitative data include author and year of publication, details of study design (exposure setting and population), the country where the study was conducted, and methods details (samples type, targeted chemicals, outcome measures). The quantitative data extracted included the sample size for each study, and the main findings expressed as means and standard deviations if applicable (Table 2). Data extraction was performed by two reviewers. One reviewer extracted the data, and the second reviewer compared the extracted data with the original report.

Risk of bias (quality) assessment

The risk of bias (methodological quality) of each included study was assessed using the modified version of the Newcastle-Ottawa Scale (NOS) for cross-sectional studies developed by Elyasi et al. [18] to ensure that the conclusions and findings of the reviews were based on the best available evidence. The tool was adjusted to include exposure assessment and a comparison group.

The score for each cross-sectional study was calculated based on four categories: group selection (three items), comparability (two items), exposure measurement (one item), and outcome measurement (one item). The items in the first two categories, ’group selection’ and ’comparability’, were awarded a maximum of two stars and one star for items in the remaining categories. The NOS score ranged from 0 (lowest grade) to 10 (highest grade). Studies that scored above the median were considered high quality [19], that is, >5 in this review. Two reviewers carried out the risk of bias assessment. Any disagreements were addressed by discussion between the two reviewers or by the intervention of a third independent reviewer.

Statistical analysis

The effect size was calculated using standardized mean differences (SMD) since the studies were conducted in different settings (ecological or occupational). Heterogeneity was determined using Cochran’s χ² test and quantified using the I² test. The null hypothesis for heterogeneity was that all studies share a common mean difference for MN frequency. The I² describes the percentage of differences across studies attributed to heterogeneity rather than chance [20]. An I² value of 25% is considered low heterogeneity, a value between 50 and 75% is moderate, and a value above 75% is considered high heterogeneity [20].

The random effect meta-analysis model with restricted maximum likelihood (REML) method [21] was used to calculate the overall SMD and its 95% confidence interval (CI). The REML method performs well with a small number of studies and produces an unbiased estimate of the between-study variability owing to differences in study designs and interlaboratory reproducibility. The REML assumes a normal distribution of the random study effect sizes [22]. Forest plots were used to present the results of the meta-analysis.

To further explore heterogeneity between studies, subgroup analyses were performed based on study setting, i.e., whether studies were conducted among e-waste workers vs. non-e-waste workers (occupational) or studies were conducted among residents of an e-waste exposed town/village vs. residents in a neighbouring town/village without e-waste exposure (ecological), and quality, (high quality vs. low quality) as determined by the NOS for cross-sectional studies. All statistical analyses were conducted using Stata version 16.1 (StataCorp LLC, College Station, TX, USA).

Study characteristics

Narrative synthesis of study outcomes

Of the 20 studies included, six (6) biomarkers of DNA damage (micronuclei, comets assay, 8-OHdG, telomere length, apoptosis rate and chromosomal aberrations) were measured in seven (7) different biological matrices (buccal cells, blood, umbilical cord blood (UCB), placenta, urine and semen) (Table 3). Telomere length (n=1) and apoptosis rate (n=1) were the least biomarkers measured, while the micronuclei frequency rate (n=9) was the most commonly measured biomarker. Finally, whereas six studies measured more than one biomarker, the remaining 14 studies measured one biomarker each. Most studies used blood (n=9) and urine (n=7) as biological matrices in which DNA damage was measured.

Risk of bias (quality) assessment

All studies included in this review were cross-sectional. The risk assessment scores ranged between 3 and 7 (maximum of 10). Of the 20 studies, 12 studies [2, 14, 15, 26], [27], [28], [29], [30], [31, 33, 36, 37] scored above the median score of five and were considered high quality. The appraisal details are summarized in Table 4.

Estimate of variance across studies and pooled outcomes

Nine out of the 20 included studies used Micronuclei (MN) frequency assay to measure the risk of DNA damage associated with e-waste exposure. Seven of these studies were included in the meta-analysis. One study [14], which measured MN in exfoliated buccal cells, was excluded from the meta-analysis because it was considered an outlying study that significantly affected the meta-analysis results. The other excluded study was conducted in the Philippines by Berame et al. [24], which did not report the mean and standard deviation (SD) values. Studies that reported medians and interquartile range (IQR) or range [29, 33, 34] were converted to means and SD using the formulas provided for different sample sizes by Wan et al. [43].

Despite high heterogeneity observed between studies (I²=95.81%), the meta-analysis showed a significantly higher MN frequency among e-waste exposed population than the control population with an overall SMD of 2.30 (95% CI: 1.36, 3.24, p-value<0.001) (Figure 2a). Potential publication bias was not explored due to the limited number of studies (<10 studies) included in the meta-analysis.

Figure 2a:

Standardized mean difference of micronuclei frequency for e-waste exposed population vs. control.

In a subgroup analysis considering only ecological studies, four studies were included in the analysis. The pooled estimate of the SMD was 1.68 (95% CI: 0.85, 2.51, p<0.001), with high variability between studies (I²=90.94%) (Figure 2b). Considering only studies conducted among e-waste workers (occupational), only three studies were included in the analysis. The combined SMD showed that e-waste recyclers had higher MN than controls (SMD: 3.09, 95% CI: 1.53, 4.66, p<0.001) (Figure 2b). In a further sensitivity analysis stratified by study quality, five studies adjudged “high quality” by the NOS confirmed that MN frequency was higher among the e-waste exposed group compared to the controls (MSD: 2.80, 95% CI: 1.79, 3.81, p<0.001) (Figure 2c). Only two studies were considered “low quality” and did not show a significant difference in MN frequency between the two groups (p=0.35).

Figure 2b:

Sub-group analysis by study setting.

Figure 2c:

Sub-group analysis by study quality.

Discussion

To the best of our knowledge, this systematic review with meta-analysis is the first to specifically assess the risk of DNA damage associated with e-waste processing/disposal in human populations. We identified and evaluated 20 studies that investigated associations between e-waste exposure and various biomarkers of DNA damage. Nine of these studies measured DNA damage by MN assay of which seven were deemed combinable for meta-analysis. The review provides ample evidence of the deleterious effects of e-waste processing on DNA integrity, as evaluated through well-known DNA damage biomarker assays.

Despite high heterogeneity between studies, the overall SMD estimate showed higher MN frequency among the e-waste exposed group compared with the control group with an effect size (SMD) of 2.30 (95% CI=1.36, 3.24). Subgroup analysis by study setting revealed that workers who directly recycle e-waste (occupational) had higher MN frequency (SMD: 3.09, 95% CI: 1.53, 4.66, p<0.001) compared to occupationally unexposed individuals (SMD: 1.68, 95% CI: 0.85, 2.51, p<0.001). A possible explanation for these results might be that e-waste workers in the informal sector are continuously involved in multiple tasks and work in the open using rudimentary tools with little or no use of personal protective equipment [44, 45]. This practice exposes the recyclers to higher levels of toxic chemicals compared to the general population [46]. The frequency of MN is generally used as a biomarker of effect associated with exposure to genotoxic chemicals [47, 48]. Therefore, chronic exposure to these toxic chemicals may result in some degree of DNA damage, as explained by the increased levels of MN in e-waste recyclers compared to the non-occupational exposed group.

Even though all studies included in the meta-analysis were conducted in China and had similar exposure profiles, substantial heterogeneity still existed. The high variability between studies could be attributed to participant’s characteristics such as age, gender, diet, and lifestyle factors (e.g., smoking, alcohol intake, and recreational drugs) which may influence MN frequency in peripheral blood leukocytes (PBL) [49]. Except for the study done by Li et al. [35], which did not control for any confounders, the remaining seven studies controlled for confounding by smoking and other important risk factors through data collection and analysis. However, four studies [28, 29, 33, 37] that controlled for confounding by smoking did not provide quantitative data about smoking, which could result in misleading conclusions when comparing smokers and non-smokers [49]. Given the causal role of MN in cancer development [50] and the evidence of their increased levels shown in this review, MN could be a primary biomarker for assessing DNA damage in e-waste-exposed populations.

The review also found that other DNA damage biomarkers, including chromosomal aberrations, tail length (TL), percent tail DNA (TDNA%), tail moment (TM) and 8-OHdG showed a consistent higher frequency among the e-waste exposed group than the controls. For instance, three studies in this review assessed chromosomal aberrations (CAs) associated with e-waste exposure. Wang et al. [28] demonstrated increased CA levels in the PBL of e-waste workers compared to controls in China. Similarly, Liu et al. [37] showed that individuals living in e-waste polluted villages in northern China had higher CA levels than those living in a neighbouring village with no e-waste exposure. A similar study conducted in Palestine by Khlaif et al. [23] reported increased CA levels in an e-waste-exposed population compared with controls. Induced chromosomal aberrations are useful biomarkers of occupational and environmental exposures to genotoxic agents and are predictive of future cancer risk [51, 52]. This review’s results are consistent with a recent meta-analysis that concluded that occupational exposure to genotoxic agents such as benzene was associated with CA and MN frequencies [53]. Other studies have found increased frequencies of CA in foundry workers [54] and farmers [55].

Six studies adopted the comet assay (single-cell gel electrophoresis assay) to assess DNA damage associated with e-waste disposal. Primary comet assay measurements, including tail length (TL) and a fraction of DNA in the tail (% tail DNA), and derived indices, including tail moment (TM) (tail length × % tail DNA) and olive tail moment (OTM) (the distance between the centres of gravity of the head and the tail along the x-axis of the comet × TDNA%) [56], were used to determine DNA damage levels associated with e-waste exposure. The comet assay is a versatile, economical, and fast technique that is widely used in biomonitoring human exposure to mutagenic agents and is considered one of the most reliable biomarkers of early biological effects [57]. However, the available literature does not consider the comet assay to be predictive of cancer risk [58]. Overall, all the studies reviewed demonstrated strong and consistent relationships between e-waste exposure and DNA damage, represented by TL, TDNA%, TM and OTM, as determined by the comet assay. These results may be explained by the fact that the majority of the studies were conducted in and around Guiyu, China. Guiyu is noted for informal e-waste recycling and other industrial activities contributing significantly to environmental pollution [46, 59]. E-waste recyclers and residents may be exposed to high levels of potential clastogens and aneugens, which may damage the DNA, as observed in this systematic review’s findings. Genotoxic chemicals released during e-waste recycling such as metals and other persistent organic pollutants (POPs) are suggested to induce DNA damage through direct interaction with the DNA [60]. The results of this review are similar to those reported by Villarini et al. [61] and Cayir et al. [62], where DNA damage measured by the comet assay was higher among welders exposed to magnetics and farmers exposed to pesticides, respectively, than the general population.

Three of the seven studies [2, 27, 29] that examined DNA damage using 8-OHdG did not find any significant differences between e-waste-exposed populations and the reference populations. Several factors could account for the lack of differences observed in these studies. First, Yuan et al. [29] recruited only 49 (23 exposed and 26 controls) participants in their study. This small sample size may lack the power to detect any differences between the groups. In addition, the biological matrix used to measure 8-OHdG concentration could affect the results of these studies. For instance, urine was widely used to measure 8-OHdG in e-waste workers and controls. Although useful, urinary 8-OHdG is considered a less sensitive and accurate biomarker compared to peripheral blood leukocytes (PBL) 8-OHdG levels [63]. None of the studies in this review measured 8-OHdG in PBL, representing a long-term response to oxidative stress and a more accurate measure of the body burden of DNA damage lesions [63]. However, direct involvement in e-waste recycling was consistently associated with 8-OHdG levels. For example, neonates of mothers who were directly involved in the processing of e-waste had a significantly higher umbilical cord blood (UCB) plasma 8-OHdG than neonates whose mothers were non-occupationally exposed to e-waste [27]. This could be attributed to the higher concentrations of metals detected in mothers who recycle e-waste, as evidenced by 8-OHdG been positively associated with Cd, Cr, and Ni concentrations (p_all<0.05) [27]. In addition, Neitzel et al. [16] found a significant association between increased work duration and urinary 8-OHdG concentration, whiles Sheng et al. [32] observed a significant difference between preworkshift and postworkshift urinary levels of 8-OHdG in e-waste workers.