Human exposure to mercury and its hematological effects : a systematic review

Mercury is a metal found in the environment from natural and anthropogenic sources. It is highly toxic to ecosystems and living beings. Most human exposures come from ingestion of contaminated seafood, outgassing from dental amalgam or occupational exposure (e.g. gold mining), among other cases. Large populations are exposed to mercury, making it a very important issue from the public health perspective. Adverse health effects are commonly seen in the nervous system, but every organ is a potential target, such as the bone marrow. The main goal of this study was to assess the available evidence on human exposure to mercury and its hematological effects. A search strategy was constructed, including key terms (MeSH, text word and equivalents) for querying 2 repositories of master dissertation and PhD thesis (Fiocruz/ARCA and University of São Paulo) and 4 different electronic databases: BVS/ LILACS, MEDLINE/PubMed, Scopus and TOXLINE/NIH, for articles published from 1950 to February 2018. There was no language restriction and a tool (EPHPP) was used to assess the quality of included studies. According to pre-established criteria, 80 studies were retrieved, all of them observational (48 case reports, 24 cross-sectional, 6 case series and 2 cohorts), comprising 9,284 people. Despite the fact that most exposed ones (6,012) had normal blood cell count and mercury hematological effects did not seem very usual (1,914 cases: 14 severe and 29 deaths), three studies reported association (β) for anemia, lymphopenia, neutrophilia and basophilia. We concluded that the gathered information pointed to mercury hematotoxic effects, some of them may be serious and even fatal. Mercury Poisoning; Heavy Metal Poisoning; Mercury; Blood Cell Count This article is published in Open Access under the Creative Commons Attribution license, which allows use, distribution, and reproduction in any medium, without restrictions, as long as the original work is correctly cited. REVISÃO REVIEW


Introduction
Mercury is a heavy metal considered as the most toxic non-radioactive element in the world.It is ubiquitous, indestructible and exists in three forms in nature: inorganic, metallic and organic 1 .It is released to the atmosphere from four different sources: (i) primary natural (e.g.volcanic and geothermal activities), responsible for 10%; (ii) primary anthropogenic (e.g.mining and fossil fuel extraction, including oil, gas and coal); (iii) secondary anthropogenic (mercury-dependent artisanal and small scale gold mining sector [ASGM], several industrial processes including chlor-alkali industry), both anthropogenic responsible for 30%; and (iv) remobilization and re-emissions (wildfires, forest clearing, biomass burning), responsible for 60% 2 .Because it is a widespread environmental toxicant, humans are unable to avoid exposure to its forms 3 and the main sources are: fish and shellfish consumption, outgassing from dental amalgam, vaccines containing thiomersal and occupational exposure (agricultural products, industry and gold mining) 4 .Specifically in relation to the latter, ASGM is considered the number one anthropogenic mercury pollutant in the world, responsible for 37% (410 to 1,400 tones/year) of its emissions to air and water worldwide.It poses a risk not only to miners, estimated at 10 to 19 million workers, of which 5 million are women and children, in more than 70 countries, but also to the environment and general population by water and air 2,5 .Such large variation of human exposure to mercury makes it a very important issue from the public health perspective 6,7 .
All forms of mercury could poison cellular function by altering the tertiary and quaternary structure of proteins and membrane permeability due to its affinity for sulfhydryl and selenohydryl groups.As a consequence it can potentially impair function of any organ 8,9 .Its adverse effects on human health may induce over 250 symptoms.The main ones are from nervous, renal, cardiovascular, respiratory systems, and skin, but any organ may be a target, such as the bone marrow 9 .The hematological system, due to its intense cellular proliferation, is quite sensitive to the action of a variety of substances, such as benzene 10 .However, there is sparse information and research about mercury's hematotoxicity on humans, despite its wide exposure 11 .Most of them come from occupational settings 12,13 , in vitro 14,15 and animals studies 16,17 .
The aim of this systematic review was to assess the available evidence on human exposure to mercury and its hematological effects.

Methods
This systematic review followed the precepts established by the PRISMA model 18 and had a PROSPERO register: CRD42018086389.

Data sources, search strategy and study selection
The selection criteria were based on PICOS' acronymous 19 : "Does human exposure to mercury lead to hematological effects?" and included all studies (except textbook, author's opinion and review) regarding human exposure to mercury and hematological effects, published between 1950 and February 2018.Hematological effects were considered as any blood cell alteration concerning number 20 and the normal values of mercury on biological matrices were those presented by the authors.
We developed a search strategy including key terms (MeSH, text word and equivalents) for querying four different electronic databases (BVS/LILACS; MEDLINE/PubMed; Scopus; and TOXLINE/ NIH) and two Master dissertation/PhD thesis databases (Fiocruz/ARCA and University of São Paulo).There were four search strategies containing the descriptors according to database and repositories.For BVS/LILACS: "mercúrio" AND "anemia" OR "leucopenia" OR "basopenia" OR "eosinopenia" OR "neutropenia" OR "linfopenia" OR "monocitopenia" OR "trombocitopenia" OR "policitemia" OR "leucocitose" OR "basofilia" OR "eosinofilia" OR "neutrofilia" OR "linfocitose" OR "monocitose" OR "trombocitose" OR "hemograma completo".For MEDLINE/PubMed: "anemia" OR "leukopenia" OR "thrombocytopenia" OR "eosinopenia" OR "basopenia" OR "monocytopenia" OR "polycythemia" OR "leukocytosis" OR "thrombocytosis" OR "eosinophilia" OR "basophilia" OR "neutrophilia" OR Cad.Saúde Pública 2019; 35(2):e00091618 "monocytosis" OR "blood cell count" AND "mercury".For TOXLINE/NIH, the search was made in a binary way: anemia and mercury/leukopenia and mercury.For the two repositories: "mercúrio" and "efeitos hematológicos".Reference lists were also searched for relevant studies.No restrictions were applied concerning language and translation was done whenever necessary.Both authors (A.S.V. and E.P.M.) followed the same schedule independently: first they reviewed the title, then the available abstract, soon after the analysis of full text, and finally the search for reference.Any discrepancy in the search results not solved between A.S.V. and E.P.M. was planned to be discussed with a third author (C.I.R.F.A.).In order of priority, we excluded: non-human studies; without mercury exposure; lacking hematological effect; textbook, author's opinion and review papers (type of study); and those published before 1950.To determine agreement between the two raters, Cohen's kappa statistic was used for each step.

Data extraction
The extraction process was also done independently and included: author, year, place, journal, data base, type of study, substance, exposure (local and duration), population (number, exposed versus non exposed, age, sex), hematological outcome (primary or secondary), death, blood cell count, bone marrow biopsy, mercury (sample, level, method) and statistical analysis.Once more, Cohen's kappa statistic was used to evaluate an inter-rater agreement.

Study quality assessment
To assess the quality/bias risk of the selected studies, we chose a tool known as Effective Public Health Practice Project (EPHPP) 21 .This quality assessment tool for quantitative studies has eight components ratings: (a) selection bias; (b) study design; (c) confounders; (d) blinding; (e) data collection methods; (f) withdrawals and drop-outs; (g) intervention integrity; and (h) analyses.Items from "a" to "f" are rated as strong (1), moderate (2) or weak (3).There is a dictionary to correctly rate each section.These six items are included in global rating, ranked as follows: strong must not have no weak ratings; moderate may have one weak rating, and weak may have two or more weak ratings.At the end there is an item for discrepancy between both reviewers that indicate the reason for discrepancy: oversight, differences in interpretation of criteria and differences in interpretation of study.That will lead to a final decision of both reviewers.
The results were summarized in a descriptive manner for occupational and non-occupational exposure data, due to toxicological differences between them.

Results
The search yielded 1,297 citations as of February 14th 2018, 323 from BVS/LILACS, 142 from MEDLINE/PubMed, 525 from Scopus, 110 from TOXLINE/NIH, 197 from Fiocruz/ARCA and none from the University of São Paulo.After 78 duplicates were removed, 1,219 records were screened based on review of titles and abstracts.Thereafter 63 full texts of articles were assessed for eligibility.The search to identify any missed report or citations resulted in selection of 80 articles: 61 from electronic databases search and 19 from reference lists.The reasons for 1,158 articles exclusion were: no mercury (457), without hematological effect (381), non-human (215), study type (104) and year of study (1) (Figure 1).The Cohen's kappa statistic was considered as almost perfect (0.98, 95%CI: 0.91; 1.0, p < 0.001) during screening title and abstract and substantial (0.76, 95%CI: 0.54; 0.98, p < 0.001) during extraction process.There was no more disagreement at other steps.

Figure 1
Flowchart showing the selection of studies for the systematic review.
the main language was English (63).However, hematological effect was the primary outcome only in 14 studies (17.5%).
A total of 9,284 people were evaluated: 6,601 from non-occupational (60 studies) and 2,605 from occupational (23 studies) exposure.There was no report on 78 times.Three articles had both types, so there was a split between persons, according to it.There were differences of age and sex distribution between exposure: at non-occupational, children and teenagers were the majority (4,982/75.47%)while at occupational, adults were (1,752/67.26%).According to sex: women were predominant (3,640/55.14%)at non-occupational and men were at occupational settings (1,402/53.84%).
Three distinctive groups that are more susceptible to chemical substances, due to physiological characteristics and proportional high exposure levels, received attention on 33 studies: 23 for exposed occupational populations (13 cross-sectional of which seven with exposure and control groups), eight for children and teenagers (six cross-sectional and two cohorts), one for pregnant, neonates and children (cross-sectional) and one for pregnant (cross-sectional).
Chronic exposure was the main type for both, comprising 33 non-occupational (6,127 persons) and 16 occupational (2,041 persons) studies.
Hematological effects were described 2,376 times, in 69 studies comprising 1,914 cases (20.62%): 479 children and teenagers, 476 adults, 13 elderly and 946 not classified.Non-occupational exposure was the most frequent (1,111).Blood cell count was done in all cases and bone marrow biopsy in 13 times.Anemia (875), lymphocytosis (361) and lymphopenia (306) were the top three, although only anemia was the most common in both type of exposures.In fact, there was alteration of all bone marrow cellular series in mercury's presence (Table 1).In 1,567 of all cases (81.87%), mercury's exposure biomarker was above the recommended threshold.The blood cell count was normal in 7,250 times, where 6,012 individuals were exposed to mercury (75.85%).
The characteristics of these studies are summarized in Tables 2 and 3.
Cad. Saúde Pública 2019; 35(2):e00091618 The quality assessment for these studies was considered weak (3) according to EPHPP for 75 of them.Almost all component ratings were considered as weak (selection bias, study design, confounders and blinding) or not applicable (withdrawals and dropouts).There was no discrepancy between the two reviewers concerning component ratings.

Discussion
We identified 80 out of 1,219 studies of mercury exposure and hematological outcomes, including environmental studies of children, teenagers, adults and elderly, as well as occupational ones.However, there were only 14 studies that aimed at hematological effect as the primary outcome.All were observational comprising a total of 9,284 studied people, although 42 were case reports of just one person.* From normal to high levels of mercury (range).
It is important to emphasize the growing articles publication involving human exposure to chemical substances over the past decades.This is a consequence of the efforts made by many countries, through their agencies and institutions, in order to improve health by reducing environmental exposure to toxic substances 55,56 .Mercury is no exception, as in this review, we reported 26 studies published between the 1950s and the 1980s and 54 studies in the last three decades 57 .
Distribution by age and sex presented the results expected in the literature, where children/ teenagers and women were more commonly exposed non-occupationally, while adults and men were occupationally exposed.According to the report on human exposure to environmental chemicals (National Health and Nutrition Examination Survey IV -NHANES), women are the most exposed on non-occupational setting 8 .The main exposure pathway for children/teenagers was food consumption, mainly fish and shellfish, although rice may be another methylmercury source for Asians 58 .Another concern regarding this group is the fact that this silver liquid metal -found at home, at school and at others sites where it is not adequately stored -is seen by them as an amusing substance to play with, which may cause health problems.Lee et al. 59 addressed this subject by reviewing the sources of mercury exposures in children, the location and proportion of children affected and also making recommendations to prevent them.
There was a wide range of exposure pathway, from food and medicine intake, suicide attempt to industrial process and gold mining, among others.Many of them are supposed to be prohibited by 2020, according to Minamata Convention, an international treaty signed in 2013 by more than 140 countries, including Brazil, which have committed to eliminate the use of mercury in different products, such as batteries, light bulbs and health equipment 2 .Two of them deserve a special attention: gold mining/ASGM and fish/shellfish consumption, since they play a role in both types of exposure, occupational and non-occupational.
The first exposure pathway, gold mining/ASGM, is the main anthropogenic mercury pollutant in the world, affecting not only the miners but also the neighboring population, mainly in Southeast/East Asia, Sub-Saharan and South America 2 .It is impressive that only seven research articles (four occupational and three non-occupational) have addressed hematological effects among people working or living nearby the gold mining sector, as more than 10 million ASGM miners, most of them informally or even illegally 5 , are exposed to mercury through both direct inhalation of mercury vapor and consumption of material taken from contaminated areas (e.g.fish).One example addressing this topic was the research, a purposive field sampling, conducted in Indonesia by Ekawanti & Krisnayanti 60 among non-miners (29) and miners (71), who showed lower levels of hemoglobin and hematocrit.In non-occupational situations, houses near gold mining put the surrounding population at risk due       to contamination of soil (children playing outdoor), water (fish consumption) and air (amalgamation process or re-burning it at gold shops) 5 .An example of the latter was a study carried out in Poconé, a town in Mato Grosso State/Brazil 61 .They evaluated the levels of exposure to metallic mercury emissions by gold dealers and its health effects.It was reported higher mercury levels and referred morbidity among downtown residents.
The second route of exposure, consumption of fish and shellfish, is a major concern for regulatory agencies around the world, because although it is an important part of a healthy diet (presence of omega-3 fatty acids and low in saturated fat), it is also cited as the most significant source of methylmercury.One of the agencies is the U.S. Environmental Protective Agency (EPA), who sets a recommendation to limit or avoid certain species of fish and shellfish for general public and for specific groups of people at risk, such as: high consumers of fish (e.g.coastal dwellers, riverside communities), women of childbearing age, pregnant and breastfeeding women and young children.For example, the threshold for tuna consumption, a carnivorous fish, is one can (226-340g) per week for groups of people at risk 62 .The risk of contamination of this kind of food is usually high (especially for the species at the top of food chain), because of the bioavailability of this metal in the aquatic environment from different sources such as geothermal activities, fossil fuel burning, hazardous waste incineration, industrial processes, gold mining and so forth.The ASGM, despite its decline in Amazon Basin, continues to contribute to an increase in the mercury load, becoming a major risk for indigenous groups and riverside communities, who have fish as their main source of protein 63 .On other hand, the general urban population has a low fish ingestion as a result of its cultural and social characteristics, in such a way that they do not face significant health effects from this pathway exposure 64,65 .However, the fish resources for urban centers may come from a contaminated water body, as reported by Hacon et al. 64 in a study carried out in Alta Floresta, a town in Mato Grosso State/ Brazil.The assessment of the impact of fish and shellfish contamination on the exposure of human beings and on their health through food deserves special attention, specifically, but not only, for those who are large consumers, such as indigenous groups, riverside communities (e.g.Amazon Basin), coastal (e.g.Florida/Puerto Rico) and island dwellers (e.g.Faroe Islands/Denmark).In this review, 3 articles have targeted this population and hematological effects: 1 sectional study with an indigenous group from the Peruvian Amazon near ASGM, that reported anemia among children under 12 years (83 persons) 66 ; 1 cohort with children from Faroe Islands (56 persons), that reported leukopenia and lymphopenia 52 , and other sectional with children from Jeju Island/South Korea (311 persons), that reported lymphocytosis 50 .
Anemia and less commonly leukopenia, eosinophilia, thrombocytopenia and pancytopenia have been reported due to mercury toxicity 67 .In this review, most exposed people (75.85%) had a normal Cad.Saúde Pública 2019; 35(2):e00091618 blood cell count, however, hematological effects were reported 2,376 times, mainly at non-occupational settings, comprising 1,914 cases.All bone marrow cellular series were affected and the most common, for both exposures, was the erythroid series with anemia (875).Out of five studies that addressed this subject using statistical analysis with significant p-value, there was mean difference in two 43,47 , none in one study 42 ; negative correlation 42 was reported in one and inverse association 53 in another.On the other hand, polycythemia was also reported in the mercury exposure group and a mean difference was found 45 .Other two hematological effects were also reported: lymphocytosis and lymphopenia.For the first outcome, there were three studies that reported a weak to moderate positive correlation 22,24,50 between lymphocytes and mercury.For the latter, one reported mean difference 12 and other two described inverse association between lymphocytes and mercury exposure 51,52 .One of these studies also described an association between mercury exposure and an increased neutrophils and basophils percentage 51 .These results confirm there are relation or association between mercury exposure and hematological effects, especially for anemia, lymphopenia, lymphocytosis, neutrophilia and basophilia.However, none of these studies could determine a causal relationship, as they were not designed for this purpose.
Recently, researches shed some light on the role of heavy metal exposure at anemia, which is estimated by the World Health Organization (WHO) in 1.62 billion cases (95%CI: 1.0; 1.74 billion) 47,48,49 .More than half of the cases are caused by iron deficit (51%) and current data point to relation between heavy metals, such as lead and mercury, and iron metabolism (positive correlation for mercury and inverse for lead) 49,68 .
There is some scientific debate about mercury effect on lymphocytes.It is suggested that the difference observed (lymphopenia x lymphocytosis) could be explained by mercury's level and form, in a way those exposed to methylmercury would be prone to lymphopenia and those to metallic or inorganic mercury to lymphocytosis 50,52,69,70 .The latter effect on lymphocytes was observed in six out of seven studies, which might corroborate this theory.Some hematological effects are considered quite severe according to preestablished criteria and can lead to a number of critical clinical conditions.They were seen at this review as a consequence of mercury's direct effect on blood cell and their corresponding clinical pictures, mainly as severe bleeding, but also as renal insufficiency, multiple organ dysfunction syndrome and sepsis.Despite the reports of other potential severe hematological effects, such as polycythemia, thrombocytosis and lymphocytosis, there were no cases of thrombosis or hematological cancer.
Mercury is a toxic substance that can lead to death.Its lethal dose is defined at 150 to 300mg/ 70kg 6 .There were 29 reports of it mainly on non-occupational exposure, especially due to use of medicine in a chronic way.In the past it was prescribed as laxative, diuretic and antiseptic.Nowadays, mercury is still present in some traditional therapies and religious practices (e.g.Santería, Espiritismo or Ayurvedic medicine) 71,72 as well as in vaccine preservative.All the five occupational deaths were related to higher level of mercury exposure at acute setting.
There are some explanations for some mercury hematological effects, such as: pancytopenia due to direct toxic effect on bone marrow 11,67 ; anemia due to apoptosis 14,15 , loss of blood from direct effect on gastrointestinal mucosa 73 and hemolysis 14,15,36 ; polycythemia from increased level of erythropoietin 45,74 ; leukopenia, neutropenia, lymphopenia and basopenia due to passed inflammatory reaction 46 and apoptosis 69,70 ; leukocytosis and neutrophilia due to lung inflammatory reaction (pneumonitis) 75,76 ; eosinophilia related to hypersensitivity 77,78 and idiosyncrasy 79 ; lymphocytosis due to increased calcium content in cytoplasm 23 , and; thrombocytopenia immunologically mediated 39,80 .
The actual dimension of mercury's hematologic effects is unknown for many reasons that come from the lack of studies that could evaluate this topic as a primary goal, which was discussed by two studies 39,81 to the lack of knowledge of mercury role on this subject.In the latter, two situations were observed: the physician did not request mercury biomarker when evaluating an hematological effect or he did not request blood cell count when evaluating a case of mercury intoxication, merely because of the lack of knowledge.In this review, only 14 studies had hematological effect as main outcome of mercury exposure and 381 out of 1,158 studies were excluded due to the fact of not requesting blood cell count.
A meta-analysis was not pursued because the only hematological effect that had a sufficient number of comparable groups and a statistical measure (correlation coefficient r) was lymphocytes Cad.Saúde Pública 2019; 35(2):e00091618 alteration (lymphocytosis or lymphopenia).There were differences in reporting this measure: one study did not report the r value and four, its statistical significance.
We have tried to mitigate the publication bias by a comprehensive, sensitive, unrestricted search for language, with a long period of time (more than 70 years) and search in the gray literature (e.g.congress, master and thesis).We were able to retrieve a significant amount of normal blood cell count results between exposed people as a consequence of a more sensitive search that included blood cell count as a key term.As a meta-analysis was not done, both the visual evaluation of the funnel plot and the statistical tests of hypothesis were not performed.
The quality assessment of these studies was considered weak according to EPHPP, a quality assessment tool (global ratings: 3) for 75 studies out of 80.This poor quality of most studies, mainly due to the study design, absence of possible confounders' evaluation and presence of bias risk, limited the power of the epidemiological studies included.However, these data were able to identify, in absolute terms, 20% of hematological effects on the presence of mercury exposure, in particular for anemia, lymphopenia, neutrophilia and basophilia, as statistical tools with significant p-value were used.Without any doubt, this should stimulate further researches with special attention to studies of methodological elaboration.All steps of this process must be thoroughly thought, including random selection, comparison between exposure and non-exposure groups, control for confounders according to bone marrow cell affected (e.g.micronutrients, enteroparasitosis, malaria, others infections, glutathione S transferase deletion polymorphisms) and data collection methods that should be reliable and valid.For obvious ethical reasons, no clinical trials will be conducted to study this potential association.However, there are some others observational studies that can be done aiming, for example the frequency of this outcome (cross-sectional with comparing groups; case-control; multicentric cohort studies), the risk assessment of this exposure, as well to ascertain the clinical significance of this relationship.

Conclusion
This review was able to retrieve a significant number of studies for an issue with sparse information, although only few of them have evaluated hematological effect as the primary outcome.Despite the fact that the majority of exposed individuals had normal blood cell count and mercury hematological effects do not seem very usual, few studies reported association from refined observational study designs including robust statistical analysis, especially for anemia, lymphopenia, neutrophilia and basophilia.In this way, the effects of mercury on health should receive worldwide attention because of its toxicity and wide source of human exposure.Researchers, as well as health practitioners, should be aware of the potential hematological effect as sometimes it can be severe and even lethal.
confidence interval; EPHPP: Effective Public Health Practice Project; NA: not applicable; ND: not done; NI: not informed.
Cad. Saúde Pública 2019; 35(2):e00091618 Contributors A. S. Vianna, E. P. Matos, and C. I. R. F. Asmus contributed in the elaboration of study and search terms, assessment of the articles reviewed, and article writing.I. M. Jesus contributed in the elaboration of study and article writing.V. M. Câmara contributed in the elaboration of study and search terms, discussion of results, and article writing.

Table 1
Hematological effects found in the studies according to exposure.

Table 3
Mercury's occupational exposure studies.