Can parabens be added to cosmetics without posing a risk to human health? A systematic review of its toxic effects

Objective: To summarize evidence regarding the toxic potential of administering parabenscontaining cosmetics in humans. Methods: The systematic review followed the methodology proposed in Preferred Reporting Items for Systematic Reviews and Meta-Analyzes (PRISMA). Electronic searches of the PubMed, Virtual Health Library (BVS), and Science Direct databases were performed between October 1st and 31st, 2018. No language restriction was determined. Original articles reporting observational, in vitro and in silico studies of toxicity caused by parabens in human or human cells were considered for eligibility. Two independent reviewers performed data extraction and assessed the methodological quality and risk of bias of articles by using the Downs & Black Scale. Score levels greater than 70% were assumed to reflect good methodological quality. The Kappa coefficient was calculated. Results: A total of 254 studies were found. Following the eligibility evaluation, 22 studies were included for the qualitative synthesis. The concordance between the reviewers was substantial (Kappa coefficient = 0.650). The meaningful reported outcomes were: high concentrations of parabens in the body; apoptosis damage to sperm DNA; oxidative stress; DNA damage; irritative potential; interference in the control of adipogenesis; estrogenic activity; genotoxicity; necrosis; role in carcinogenesis of breast cancer; harmful effects on human skin when exposed to the sun; stimulation of oncogenes expression; and interference with DNA transcription. Despite most included articles presenting appreciable methodological quality, remarkable limitations were observed and the mechanisms by which parabens exert toxicity on humans remained unclear. Conclusions: The accumulation of parabens in the human organism following repeated cosmetics administration on the skin is noteworthy. However overall, the evidence so far does not make it possible to determine whether, and in what extent, the use of paraben-containing cosmetics can disturb human health. Further investigations are still required for clarifying these issues.


INTRODUCTION
Parabens are esters derived from parahydroxybenzoic acid used as preservatives in the manufacture of several cosmetic, pharmaceutical and food products 1 . The parabens class of preservatives was introduced in the 1920s due to their broad spectrum of action against fungi, yeasts, and bacteria; remarkable compatibility with several raw materials; a wide range of pH and temperature stability; affordability; regulatory acceptance; and the need of low concentrations to exert antimicrobial effects, i.e., from 0.01 up to 0.3% 2 .
Such class includes methyl, ethyl, propyl, butyl, pentyl, phenyl and benzylparabens. Methylparaben and propylparaben have been the first-choice preservative in cosmetic formulations, as they provide a synergistic and advantageous effect together with their attractive physicochemical properties (e.g. appreciable solubility in both aqueous and oily media). Due to these features, there are reports of their use in approximately 42% and 35% of products, respectively. The preservative properties of parabens are driven by a structureactivity relationship, and vary according to the extent of the carbon chain i.e., the greater the number of carbons and the length of the chain, the greater the antimicrobial activity, the greater lipophilicity, and the lower solubility in water 3,4 .
The Brazilian determinations on cosmetics allow a maximum concentration of 0.4% expressed as individual acid and 0.8% expressed as acid for mixtures of salts or esters 5 . The European Union considers it safe to use at a concentration of 0.4% for individual acid or 0.8% when used in combination, however, considering propylparaben and butylparaben, the sum of their individual concentrations should not exceed 0.14% 6 . Long-chain parabens (isopropylparaben, isobutylparaben, phenylparaben, benzylparaben and pentylparaben) are banned in Europe 7 . In Japan, up to 1% of total paraben is tolerated for any cosmetic product 8 . In contrast, the US Food and Drug Administration (FDA) does not have specific standards applied exclusively to preservatives in cosmetics. Hence, they do not undergo approval before entering into the market 9 .
From a safety point of view, typically the parabens incorporated in cosmetics do not lead to allergic reactions, and the majority of sensitizing reactions are due to the use of such products in previously injured skin. Even in the light of this low allergenic potential, there is growing concern regarding the estrogenic and endocrine disrupting activity caused by parabens. Despite the estrogenic activity in vivo and in vitro being considered weak, repeated exposures may lead to endocrine disruption and cancer development 10-12 . Recently, a study has shown that butylparaben is toxic to human trophoblastic cells, acting in the inhibition of cell proliferation, induction of apoptosis, and endoplasmic reticulum stress. Therefore, childbearing age women are easily exposed when using cosmetics; butylparaben may trigger problems related to early placental development 13 .
A recent study has shown that exposure to parabens is related to metabolic changes and an increased risk of metabolic disorders in pregnant women. The results indicated that exposure to parabens in early pregnancy was associated with purine metabolism, betaoxidation of fatty acids, metabolism of tryptophan and other pathways that were altered by parabens. Eighteen and three metabolites were correlated with exposure levels of methylparaben and propylparaben, respectively 14 .
To date there is no consensus regarding the toxicity of parabens following topical administration of cosmetics on humans, which reflects the broad differences in the acceptance range for these chemicals in the cosmetics marketed worldwide, and sheds light on a great public health concern. Motivated by the large cosmetics consumption in Brazil and abroad, we report in this paper a systematic literature review of the toxic potential of the use of parabens as preservatives in cosmetic products.

Study design
A systematic review of the scientific literature was conducted according to the methodology proposed in Preferred Reporting Items for Systematic reviews and Meta-Analyses -PRISMA 15 . Guiding question and definition of "PICOS" The following guiding question was defined: "What are the toxic effects of parabens in cosmetics?" For the construction of the systematic review and elaboration of the guiding question the strategy of PICOS was used, being: "P" (population), humans or human cells; "I" (intervention), parabens-containing cosmetics; C (comparator), without comparison; "O" (outcome), toxic effects, side-effects, adverse reactions; "S" (study type), observational and experimental studies (in vitro and in silico).

Database and search strategy
The PubMed, Virtual Health Library (BVS), and Science Direct databases were accessed. The searches were performed by two authors (DFR/ GCSA) between October 1 st and 31 st , 2018. For all databases, the searches were conducted in the "advanced search" interface. The search terms were combined using the boolean logic terms 'OR' and 'AND'. The search strategy in Pubmed was: (((("Parabens"[MeSH Terms]) OR ((4-Hydroxybenzoic Acids) OR (4 Hydroxybenzoic Acids) OR (para-Hydroxybenzoic Acids) OR (para Hydroxybenzoic Acids)))) AND (("toxicity [Subheading]") OR ((toxic potential) OR (margin of safety)))) AND (("Cosmetics"[MeSH Terms]) OR ((Care Product*, Personal) OR (Personal Care Product*) OR (Product*, Personal Care))). For both BVS and Science Direct, the search strategy was "Parabens" AND "toxicity" AND "cosmetics".

Additional Analyzes
A manual search was also performed in the list of references of the included articles throughout the search for eligible publications, since they could not have been identified in the selection of the studies (gray literature). The authors of the unavailable articles were contacted twice by e-mail, through which access to these articles was requested.

Eligibility criteria
The inclusion criteria for the selection of articles were observational studies (performed in humans), experimental studies (in vitro), computational simulation studies (in silico), published up to September 30 th , 2018. No language restriction was imposed. The starting date of the collection was not restricted, since the aim was to recover the maximum number of articles dealing with the toxicity of parabens. Reviews, comments, expert opinions, letters to the editor, supplements, conference abstracts, dissertations, editorials, theses, and studies on animals or on animal cells were excluded. The exclusion reasons were allocated into the following categories: wrong drug, wrong population, wrong study design, wrong publication type, and wrong outcomes.

Studies selection
The search and selection of articles were performed independently by two authors (DFR/ GCSA). Initially, duplicate articles were deleted. Subsequently, the titles related to the topic were evaluated and those that were not related to the subject were excluded. Afterwards, a detailed reading of the abstracts of the remaining articles was conducted, in order to select those that approached the proposed subject. Abstracts that did not address the topic in question were excluded. Thenceforward, articles were read in full, and those that were within the inclusion criteria were inserted as search results. If there were any disagreements between the two authors, a third author was involved when necessary, to make the final decision. The Kappa coefficient 16 was determined to assess the degree of agreement between the two evaluators (DFR/ GCSA). In this pursuit, we considered a 95% confidence interval and used the Stata 11.0 software package (StataCorp LLC, Texas, USA).

Quality assessment
The Downs and Black Scale 17 was used in the analysis of quality and risk of bias. This is a scale which includes items that assess internal and external validity, reporting standards, and power of the studies. Such tool is considered prompt, useful, and can be used either to assess the quality of original or primary source research articles, as well as to synthesize evidence from quantitative studies. The "Checklist for Measuring Quality" contains 27 'yes' or 'no' questions distributed in five sections concerning: i) reporting (n = 10; questions 1 -10) -the overall quality of the study; ii) external validity (n = 3; questions 11 -13) -the ability to generalize findings of the study; iii) internal validity / bias (n = 7; questions 14 -20) -to assess bias in the intervention and outcome measure(s); iv) internal validity / confounding and selection bias (n = 6; questions 21 -26) -to determine bias from sampling or group assignment; and v) power of the study (n = 1; question 27) -to determine if findings are due to chance.
Herein, for observational studies in humans, five items from clinical trials were excluded, therefore, 22 items were evaluated (i.e., questions 1 -13, 16 -18, 20 -22, 26 and 27). For in vitro and in silico studies, 15 specific items for studies performed in human populations were excluded, thus 12 items were considered for scoring (i.e., questions 1, 2, 5 -7, 10, 16 -18, 20, 25 and 27). In all cases, for a study to be individually considered as having good methodological quality, the sum of scores should be ≥ 70% of the total scores of evaluated items. The quality assessment was performed independently by two researchers (DFR / GCSA), and divergences between assessments were resolved by consensus between a third researcher.

Data extraction
After applying search criteria, the selected studies were arranged in electronic spreadsheets, and the data extracted from each study were: author and year, country, objectives, design, substance and / or cosmetic, and toxic effect caused by paraben. Moreover, the authors critically analyzed all the included papers and where described, the limitations were collected and summarized in the review. Data were divided into (i.e., Tables I and II), corresponding to human studies (observational) and in vitro / in silico studies.

RESULTS
A total of 254 studies were found. The total of articles resulting from the research in each database can be seen in Figure 1, which also summarizes the numbers and motivations for exclusion of the papers in each step of this review. Throughout the eligibility evaluation, the degree of agreement between the two researchers was considered substantial (good agreement) since the Kappa coefficient was 0.650 16 . Full analysis of these articles revealed that 22 studies fulfilled the inclusion criteria. From the selected articles, twelve were available in the Pubmed database, eight in BVS, one in Science Direct, and one was selected through search references of included studies. 5/18  Table I and II. From the analyzed studies, 31.8% (n = 7) were performed in humans, 63.6% (n = 14) were performed in vitro and 4.6% (n = 1) was performed in silico.
The publication time span of studies was between 2004 and 2018. The majority of the studies, 54.6% (n = 12), were published in the last seven years before the completion of the search. Regarding the country, the United States had the greatest number of published studies (n = 6), followed by Sweden, Poland, United Kingdom (UK), China (n = 2); and Denmark, Norway, India, Japan, France, Brazil, Romania and Germany (n = 1). From the included papers, 95.5% (n = 21) were published in the English language and 4.5% (n = 1) in Polish.
Concerning the methodological quality of the articles, of the seven studies in humans included (Table I), 85.7% (n = 6) presented scores greater than 16 points, with a response rate above 70%. They were therefore considered of great methodological quality according to the Downs and Black 17 score.
Considering the included studies performed in vitro (n = 14), 71.4% (n = 10) had a response rate greater than 70% (Table II). The other studies were considered well elaborated, however, as they were in vitro and in silico studies they had lower scores in some items evaluated, namely: description of the main confounding factors and possible adverse effects; characteristics of included and excluded patients; representative sample of the majority of the population; description of unplanned analyzes at baseline; adequate primary statistical analysis; and tests over the same period of time. No study was excluded due to methodological quality.

DISCUSSION
One of the great challenges of cosmetic conservation is the choice of safe and effective preservatives. This systematic review reports that there are few conclusive studies on the toxicity of parabens, explained by some factors such as a) the majority were conducted in vitro or in silico, making difficult a conclusive relation to humans, despite their quality; b) a wide diversity of substances evaluated; and c) diversity of toxic effects described.
The main route of exposure to parabens was dermal and only 4% of the aggregate exposure was from parabens in foodstuff. Estimates of aggregate exposure to parabens were less than 10 mg/kg/day. Methyl, propyl, and ethylparaben are present in 0.04% to 0.35% of the cosmetics and represent 74%, 75% and 70% of the aggregate exposure to parabens, respectively. Butylparaben was detected at very low concentrations after dermal use of cosmetics 4 .
Topical allergic reactions to parabens are uncommon, ranging from 0.2 to 1.2% 22,39,40 . Fransway et al. 41 advocate that preservatives can be used safely in cosmetic formulations for application on the whole skin, since the possibility of allergy of contact is rare. Furthermore, significant plasma concentrations of parabens have been described after topical administration of hand cream, body lotion, and face cream with the repetitive use of these products increasing exposure 42,43 .
For undergoing transdermal absorption, chemicals and drugs are supposed to equilibrate several physicochemical properties, i.e. (i) modest molecular weight (MW up to 500 Da); (ii) balanced lipophilicity (log octanol-water partition coefficient}, log P, ideally around 2 to 3); (iii) significant solubility in both oil and water; and (iv) be mostly in non-ionized form 44 . Most parabens satisfy those requirements such as methyl, ethyl, propyl and butylparaben 45,46 .
Hence it is reasonable to consider that once administered through cosmetics, such parabens have prospects for breaching the lipophilic stratum corneum and resorption into the aqueous central compartment of systemic circulation. These features, together with that fact that vehicles (or co-vehicles) and botanical oils often used in the design of cosmetic formulations are assumed to enhance the permeability coefficients of methyl, ethyl, propyl and butylparaben 45 , may support the trends observed for the plasmatic levels of these preservatives and cumulative effect in the body. Besides the effects of intrinsic physicochemical properties in the permeation of parabens, it would be of great relevance for future work devoted to predict and determine the role of other formulation aids with recognized permeation enhancement properties (e.g., moisturizers, surfactants, flavoring, etc.) in the bioavailability of these chemicals.
Moreover, skin integrity is a crucial issue concerning transdermal drug absorption 44 , so that the injuries that skin could undergo day by day such as sun exposure, chemical peeling, depilation, friction, stetic treatments (e.g., infrared, sonophoresis and iontophoresis, etc.) may also contribute to favor the permeation of parabens through the skin following the exposure to cosmetics over a long time spam.
Concerning the hypothesis of sexual hormones or sperm alterations, Scinicariello and Buser 20 observed no significant relationship with serum levels of total testosterone by children and adolescents using total parabens (methylparaben and propylparaben). In animal models, exposed orally to parabens, no effects caused by methyl and ethylparaben [47][48][49] were observed, whereas propyl and butylparaben were associated with somewhat toxic effects, such as variations in spermatogenesis and dose-dependent reduction of serum levels of testosterone and activity-related estrogenic effects 48,49 . Conversely, in a similar investigation 47 , the results did not express modifications for butylparaben. In an animal model, when exposed to butylparaben, sperm DNA methylation was observed in mitosis and post meiosis 50 . According to Buttke et al. 24 , parabens were not significantly associated with the age of menarche in adolescents. However, James-Todd et al. 51 demonstrated the association between early menarche and use of hair products before 13 years of age, but did not specify the ingredients involved. Therefore, sexual hormones or sperm alterations documentation seems to be poor related to parabens, and a slight evidence of butylparaben alteration should be investigated.
In the evaluation of the risks to the fetus of women exposed to parabens, despite the presence of parabens in plasma and urine of women or in the newborn, none of the parabens were consistently associated with maternal diseases or risk for fetal development 18,19,52,53 . The analyzes were not conclusive and more studies will be needed to characterize a more comprehensive scenario in relation to exposure while in the uterus.
Some studies have related the use of parabens in cosmetics with the presence of these compounds in breast tumor, which may not determine the cause of the disease but may be pre-conditions in its advance 25,28,52,54,55 . Darbre et al. 27 , quantified methylparaben in 62% of tumors, at the mean concentration of 100 ng/g tissue. In vitro tests found that parabens activated estrogen receptors and that the activity increased with the size of the alkyl chain as described in trials with breast cancer cells 30 . Although parabens are considered estrogens in the environment, they are converted to parahydroxybenzoic acid, which is not an estrogenic substance 56 .
In agreement with Pop et al. 33 , by means of an androgen receptor mediated transcriptional activity assay, Chen et al. 57 found that at concentrations between 10 −3 and 10 μM (1.0 nM to 10 μM), p-hydroxybenzoic acid and its derivatives revealed no androgenicity; and no statistically significant inhibition of the transcriptional activity of testosterone was detected for p-hydroxybenzoic acid, the major paraben metabolite. In contrast, at the highest concentrations tested (10 μM), methyl-, butyl-and propyl-4-hydroxybenzoate significantly inhibited the transcriptional activity of testosterone by 40%, 33%, and 19%, respectively (P<0.05).
It is noteworthy that when paraben-related activities were assessed via competitive antagonist binding on the human estrogen receptor, the results showed that they behaved as an inverse antagonist on the activities of the estrogen receptor, which may be an inducer of the development of breast cancer and a pharmacologic antagonist to tamoxifen 32 . It was found that human epidermal growth factor receptor (HER) increased the estrogenic capacity of butylparaben by increasing the expression of oncogenes and the proliferation of MCF-7 cells by means of estrogen receptor 36 . Although it was considered estrogenic, butylparaben was considered 10,000-fold less potent than estrogen 58 . Therefore there is a lot of evidence showing the estrogenic properties of parabens, but claims of carcinogenic activities are less argued in the literature and no human studies have certified these biological effects as relevant 59 .
By using human dermal fibroblasts, it has been possible to analyze some cytotoxic mechanisms triggered by parabens and other preservatives. The parabens tested induced necrosis in 65% of the cells and apoptosis in 95%, when at the concentration of 1%, and in smaller concentrations, necrosis and apoptosis decreased. Methylparaben caused genotoxicity in 0.8% of cells at all concentrations and propylparaben 1.5% in the presence of parabens (1%) 31 . According to Martín et al. 60 , propylparaben caused changes in cellular proliferation values, but not in cell viability, and led to DNA breaks. These effects confirm that oxidative damage is implied by the cytostatic effect on the cultured cells.
An in silico approach showed that butylparaben bound the terminal part of the DNA with 20% probability, which resulted in a stable binding energy with DNA, estimated to be 80 to 300 times weaker than the positive control, benzopyrene, considered to be a carcinogen 38 . However, there is evidence pointed out by other authors that parabens interact with the Golgi complex and potentially with DNA 61 .
Regarding the effects on adipocyte differentiation, parabens have been shown to modulate and activate glucocorticoid receptors, which are involved in the mechanism of adipogenesis 28 . The adipogenic potential increases as the length of the linear chain increases and the presence of an aromatic ring further increases adipogenic capacity. Butyl and benzylparaben, besides having the most notable adipogenic effects, caused toxicity when used in the concentration of 100 μM in the cells tested. Hu et al. 34 have presented results in agreement with these previous findings.
Handa et al. 35 investigated the effects of methylparaben at concentrations ranging from 0.003% to 0.3%. It was observed that at concentrations of 0.003% there was no effect on the cell viability of keratinocytes exposed to sunlight, but at higher concentrations the cell viability significantly decreased in 6 h, demonstrating that the harmful effect is dose and time dependent. Similarly, Dubey et al. 25 demonstrated that methylparaben presented timedependent photodegradation and reduced its antimicrobial activity by 40%.
Despite the great relevance of the findings reported by the included articles, there were several remarkable limitations in their methodological designs, which definitely impair the robustness of most of their conclusions. Assessing the studies for the quality of their findings is within the main strengths of a systematic review over traditional narrative reviews. This may provide a more precise estimate of a treatment effect and explain heterogeneity between the results of individual studies 62 . Accordingly, it may guide defining limits of what is known and unknown and helps to formulate hypotheses for further investigation. For example, even with these findings it was not possible to describe a mechanism of toxicity to the parabens, and further studies would be required.
Overall, the conflicting results obtained, and the limitations of the studies performed so far, do not make it possible to draw an accurate answer to the guiding question of this review. There remain many inquiries to be clarified and research to be completed, since there are insufficient supporting studies on the toxicity of parabens to patient health. In turn, this issue exposes a great gap to be solved by upcoming generations of pharmaceutical scientists.
Consumers' concerns regarding the safety and toxicity of cosmetics are genuine and understandable. From a market point of view, as long researchers do not move forward to obtain high-level scientific evidence as regards the safety of parabens, this may increasingly shift prescriber and consumer preferences for paraben-free products. In practice, by reading this report patients may increase their empowerment concerning the issues involved in the self-care process. Also, prescribers, health surveillance agents, and cosmetics manufacturers are invited to be aware and much more conscious of this uncertain and challenging scenario. Finally, regardless of the composition, the safety and quality of cosmetics must be utmost sought and achieved, rather than only market share and financial gain.

CONCLUSION
This work gathers, summarizes, and critically analyses scientific evidence as regards the risks for human health following the exposure to parabens used in cosmetics. There is evidence that after being repeatedly applied on the skin, parabens can permeate, reach the systemic circulation and accumulate in the human organism. However, it is still premature to determine whether the use of paraben-containing cosmetics must be avoided or contraindicated for humans. Considering the worldwide market growth trends for cosmetics and the prevalence of using parabens in their composition, addressing responses to this issue remains crucial to assure rational use of these products and to promote human health and wellbeing. Therefore, the immediate, conscious, and systematic toxicological evaluation of these products is fully justified.

ACKNOWLEDGMENTS
We thank the Federal University of São João del-Rei (UFSJ), Dona Lindu Center -West Campus (CCO) for infrastructure and institutional support. The present work was carried out with the support of the Coordination of Improvement of Higher Education Personnel -Brazil (CAPES) -Financing Code 001.