Introduction

Bisphenols comprise a family of chemicals with a common chemical structure formed by two phenol rings bonded with a variety of bridges, including a linear or branched alkyl hydrocarbon chain that can also contain other heteroatoms, such as sulphur or oxygen (Fig. 1). These chemicals are widely used to manufacture polycarbonate plastics, epoxy resins and thermal papers. The most representative compound of this family is bisphenol A (BPA). The European Food Safety Agency (EFSA) determined that exposure to BPA by handling thermal paper is the second largest source of exposure to this substance (EFSA 2015).

Fig. 1
figure 1

Chemical structures of bisphenol A (left) and bisphenol S (right)

Full size image

BPA is a chemical that has caused considerable social and scientific concern about its capability to act as an endocrine disruptor. However, the capability of BPA to alter the oestrogen cycle is very limited because its potency as such is usually several orders of magnitude lower than the positive controls used in studies (Rochester and Bolden 2015).

According to European CLP Regulation No. 1272/2008, BPA is classified as a known or presumed human reproductive toxicant (reproductive toxicant category 1B) on the basis of its capability to cause fertility and sexual function impairments (Table 1) and is also classified as a skin sensitizer and as agent that cause serious eye damage. This classification is probably the reason why one of the main concerns of BPA effects lies in its effects on reproduction. However, alterations to reproduction are not the most sensitive effect reported for BPA. Indeed BPA is not considered a selective reproductive or developmental toxicant in mice (Tyl et al. 2008) or rats (Tyl et al. 2002).

Table 1 Classification of BPA and BPS according to European CLP Regulation No. 1272/2008
Full size table

BPA also seems able to induce some kind of disruption in organs such as pancreas, adipose tissue and muscle (Rahmani et al. 2018). In its assessment of the risks to public health related to the presence of BPA in foodstuffs, EFSA took the proliferative and morphological changes potentially related to mammary gland carcinogenicity, along with liver and kidney impairments, to be the most likely hazards (EFSA 2015). EFSA was unable to derive a Benchmark Dose Level (BMDL) for the first effect but determined a BMDL10 (BMDL lower confidence limit of 10%) of 8960 µg BPA/kg bw/day. This BMDL10 was further transformed into a human equivalent dose (HED) of 609 µg/kg bw/day and allowed the estimation of a temporary tolerable daily intake of 4 µg/kg bw/day. Finally, EFSA concluded no health concerns due to BPA dietary exposure (EFSA 2015).

BPA is also used in thermal paper receipts at concentrations between 2 and 4% (w/w). For example, Thayer et al. (2016) reported concentrations of 19.6 ± 4.7 mg BPA/g paper [19.3 mg BPA/g paper (median), 7.0–36.0 mg BPA/g paper (range)] in 33 samples collected in USA. Rocha et al. (2015) reported either BPA or BPS in 98% of 190 Brazilian thermal paper receipt samples with up to 4.3% (w/w) (geometric mean of 1.6%). BPA was also found in 44 of 50 thermal paper receipts from the Italian market with concentrations up to 1533 µg BPA/100 mg paper (mean concentration of 107 µg/100 mg paper) (Russo et al. 2017).

Despite the EFSA’s assessment for dietary exposure of BPA contained in foodstuff, upon the proposal of France the Risk Assessment Committee of the European Chemical Agency (RAC) determined that the Risk Characterisation Ratio (RCR) [the ratio between exposure and the Derived No Effect Level (DNEL)] in cashiers occupationally exposed to BPA through thermal paper was above one in several reasonable worst cases. RAC proposed to the European Commission the restriction of the use of BPA, which shall not be placed on the market in thermal paper at concentrations equal or exceeding 0.02% per weight (RAC 2015). This restriction proposal was adopted by the European Commission and will be fully operative as from 2020.

One potential alternative to BPA in thermal paper is bisphenol S (BPS), which has no use restrictions to date and is being widely used in thermal paper receipts (Table 2). Liao et al. (2012a) reported that 100% of the 111 thermal receipt paper samples collected randomly in four different countries (USA, Japan, Korea, and Vietnam) contained BPS at concentrations ranging from 0.0000138 to 22.0 mg/g (geometric mean 0.181 mg/g) (Table 2). Japan and the USA were the countries with the highest BPS concentrations in thermal paper receipts, whereas the concentration in Korea and Vietnam was several orders of magnitude lower (Table 2). In another study, 31 of 50 samples of thermal paper receipts collected in Italy contained BPS at a mean concentration of 41.97 µg/100 mg of paper (Russo et al. 2017). The median BPS concentration obtained from 32 thermal paper receipts collected in the USA was 14.6 mg/g paper (mean 15.0 ± 2.6, range 11.9–26.2) (Thayer et al. 2016).

Table 2 BPS content (mg/g) in thermal paper receipts.
Full size table

The chemical structure of BPS is similar to that of BPA (Fig. 1) and it still has no harmonised classification set out in European CLP Regulation No. 1272/2008 but has so far been considered a suspected human reproductive toxicant (reproductive toxicant category 2) (Table 1). Therefore, the concern about reproductive impairments seems lower for BPS than for BPA. However, there is also a body of scientific literature that presents BPS as a potentially similar endocrine disruptor to BPA (see for example Rochester and Bolden 2015).

It seems plausible that, although cashiers’ occupational exposure to BPA is considered relevant as RAC (2015) highlighted, the exposure of these workers to BPS by a similar route might also be relevant. Therefore, an assessment of the risk associated with such exposure might be urgently needed.

Methods

In this paper, we developed DNELs for BPS on the basis of reliable publically available toxicological information following similar methodologies to those used by RAC in its opinion of year 2015. We also compared several occupational exposures to BPS reported in the scientific literature with the derived DNELs to make a preliminary assessment of the risk associated with the exposure to BPS of the general population and cashiers through skin contact with thermal paper receipts. We also compared the results with the situation indicated for BPA.

Results

Background information: the BPA case

In 2015, EFSA made a very detailed assessment of the risk associated with exposure to the BPA contained in foodstuffs. This assessment also served RAC (2015) as the main basis to support its view about the need to restrict BPA contents in thermal paper. In the next paragraphs, we present the main conclusions drawn from both scientific opinions as regards the risk associated with occupational exposure to the BPA contained in thermal paper. See both opinions for specific details.

Dermal exposure of cashiers and the general population to BPA via thermal paper

RAC (2015) used several procedures to model the exposure to BPA of the general population and workers via dermal exposure to thermal paper. These models were the “percutaneous absorption flow model” and the “absorption rate model”. The first was a probabilistic model, while the second was a deterministic model.

The probabilistic model (percutaneous absorption flow model) was subcategorised by RAC into three different reasonable sub-scenarios according to distinct input parameters. We present here for simplicity reasons the results of the worst scenario, which considers the following parameters (for both workers and the general population): (1) an absorption flow of 0.022 µg/cm2/h (the maximum value obtained in an in vitro determination with seven human skin explants from two different donors); (2) an area of skin contact sized 6 cm2 (based on the USEPA default surface area of 2 cm2 for the thumb and of 1 cm2 for all the other fingers); (3) a discrete distribution of probabilities to illustrate body weight. The differences in inputs parameters lay in the duration of exposure; uniform distribution was considered up to 2 h/day as the maximum for the general population and a triangular distribution with minimum, mean (mode) and maximum values of 3, 5.5 and 8 h/day, respectively, for workers. The results of the exposure assessment modelled according to this percutaneous absorption flow probabilistic model are summarised in Table 3.

Table 3 Exposure assessment to BPA via dermal exposure in thermal paper according to the percutaneous absorption flow model.
Full size table

The deterministic model (absorption rate model) was subcategorised by RAC into four different reasonable sub-scenarios according to several input parameters for the workers and to two sub-scenarios for the general population. Differences were addressed with distinct variations in absorption flow, exposure duration and surface area term. For simplicity reasons, the results of the worst reasonable scenario are presented, which considers (for both workers and the general population): (1) an absorption flow of 0.258 µg/cm2/h (the 95th value obtained in an in vitro determination with 15 human skin explants from six different donors); (2) an area of skin contact of 12 cm2 (cumulated surface area of the pads of the ten fingertips); (3) a body weight of 70 kg. The differences in the input parameters lay in the exposure duration, which was considered 2 h/day for the general population and 10 h/day for workers. The results of the exposure assessment modelled according to this percutaneous absorption flow probabilistic model are summarised in Table 4.

Table 4 Exposure assessment to BPA via dermal exposure in thermal paper according to the absorption rate model.
Full size table

RAC also reviewed the data obtained from biomonitoring BPA in urine. Using the correlation between oral daily intake and urinary excretion given by Krishnan et al. (2010), a representative worst case BPA daily intake was estimated in occupationally exposed people to be 400 ng/kg bw/day, which is similar to the values estimated in Tables 3 and 4.

Hazard identification

EFSA (2015) assessed the available evidence to determine the likelihood of a specific hazard occurring in humans (Table 5). EFSA determined as “likely” the general toxicity (mainly nephrotoxicity and hepatotoxicity) reported in two and three generation reproductive toxicity studies performed in mice and rats. EFSA derived a BMDL10 of 8960 µg/kg bw/day for these effects. This hazard was considered by EFSA and RAC to be the risk characterisation end-point.

Table 5 Hazard identification associated to exposure to BPA. EFSA (2015) categorised the likelihood of occurrence of each hazard in humans on the basis of weight of evidence and considered 7 different categories
Full size table

In addition, the general toxicity proliferative and morphological changes potentially related to carcinogenesis in mammary gland were also considered “likely” by EFSA. However, no dose-response could be derived and this effect was accounted for in the risk assessment using additional assessment factors.

Five hazards were labelled by EFSA with the likelihood of “as likely as not likely”, which were: immune effects, neurological, neurodevelopmental and neuroendocrine effects, reproductive and developmental effects, metabolic effects and cardiovascular effects (Table 5). Once again, no dose-response could be derived for these hazards and the corresponding risk was addressed using additional assessment factors.

Finally, the likelihood of carcinogenicity and genotoxicity occurring was considered “unlikely to as likely as not” and “unlikely”; respectively (Table 5). These two hazards, together with cardiovascular effects, were not taken into account by RAC (2015) in its assessment of the risk associated with BPA exposure via thermal paper.

Overall, the critical point considered for the risk assessment of the effects to human health was nephrotoxicity, an effect considered not associated with endocrine disruption, fertility or developmental impairments (effects that lead to more social concern). However, with less likelihood, a variety of additional effects was also be accounted for and was considered in the risk assessment using additional assessment factors to cover any uncertainties in the database.

Oral DNEL derivation

The process for DNEL derivation is overall in Table 6. EFSA (2015) considered the HED approach to transform animal doses into human doses, which was also accepted and adopted by RAC (2015) in its assessment. The HED represents dose (D) in an animal species that a human would require to obtain an equivalent area under the curve of toxicokinetic experiments. Therefore;

$${\text{HED}} = D \times {\text{HEDF;}}$$
(1)

where the HEDF (human equivalent dose adjustment factor) is defined by a common relationship between the external dose given to an animal and the resultant area under the curve, and the external dose given to a human and its area under the curve.

Table 6 Overall of oral DNEL derivation performed by RAC for assessment of the risk associated to dermal BPA exposure via thermal paper.
Full size table

The area under the curve for oral BPA administration to adult mice was 0.244 nmol/h/l, while the area under the curve for humans was estimated by physiologically-based pharmacokinetic modelling in monkey, and was estimated as 3.6 nmol/h/l (RAC 2015). Thus the HEDF for estimating the HED according to Eq. 1 starting with a mice dose was a factor of 0.068 (0.244/3.6).

EFSA considered the HED approach to replace the default uncertainty factor for the interspecies extrapolation of toxicokinetics, although the 2.5 assessment factor (out of 10) for toxicodynamics is still needed. RAC (2015) considered an extra assessment factor to cover the uncertainties associated with accounting for effects on mammary gland, reproductive, neurobehavioral, immune and metabolic systems. EFSA (2015) indicated that these effects could occur starting with doses from 100 µg/kg bw/day, which corresponds to an HED of 6.8 µg/kg bw/day according to Eq. 1 (100 µg/kg bw/day × 0.068). RAC finally set this extra assessment factor at 6.

Dermal DNEL derivation

The DNEL derivation process is overall in Table 7. RAC (2015) derived a DNEL for dermally absorbed BPA using toxicokinetic information about the fraction of an external dermal dose reaching systemic circulation. RAC used to convert the starting point into the dermal HED areas under the curve of 0.244 nmol/h/l for an oral dose of 100 µg BPA/kg bw in mice, and of 329.5 nmol/h/l for a dermal dose of 100 µg BPA/kg bw in humans. Therefore, the HEDF for estimating the dermal HED according to Eq. 1 starting from a mice dose was a factor of 0.00074 (0.244/329.5).

Table 7 Overall of dermal DNEL derivation performed by RAC for assessment of the risk associated to dermal BPA exposure via thermal paper.
Full size table

RAC used the same assessment factors to derive the dermal DNEL but noted how published papers described for low dermal exposures the dermal metabolism can reduce up to 40% of the original dose. Thus, a compromise approach was used and a biotransformation rate of 50% was considered (i.e., only 50% of the dose reached the potential target organs in an unconjugated form) to correct the dermal DNEL.

Risk characterisation

The RCRs (the ratios between exposure and DNELs) for workers and consumers based on both probabilistic exposure modelling and deterministic modelling are summarised in Tables 8 and 9, respectively. Both approaches yielded similar results, with an RCR for workers slightly above 2 in a reasonable worst case, and with RCRs of 0.1 and 0.88 for the general population.

Table 8 Risk characterisation for workers and consumers exposed to BPA through dermal contact with thermal paper
Full size table
Table 9 Risk characterisation for workers and consumers exposed to BPA through dermal contact with thermal paper
Full size table

The BPS case

We reviewed the scientific literature to assess the risk associated with dermal BPS exposure through contact with thermal paper in cashiers and the general population. One of the main inconveniences was that available information is considerably reduced compared with BPA, especially in terms of exposure assessment, toxicokinetics and risk characterisation after skin exposure and, consequently, the risk assessment will necessarily have greater uncertainties.

Dermal exposure of cashiers and the general population to BPS via thermal paper

Two well-defined strategies were found in the literature to estimate BPS exposure. One was based on biomonitoring BPS excretion in urine and the second on estimations of transfer rates of BPS from the thermal paper matrix to skin.

Biomonitoring

The literature has clearly determined that BPS can be biomonitored in the urine taken from both the general and worker populations. Ndaw et al. (2018) found significant differences (3.7-fold) between BPS contents in the urine of the general population and cashiers, and was consistently detected in the general population (Table 10).

Table 10 Urinary concentrations of BPS (µg/l) in general and worker population in France.
Full size table

Liao et al. (2012b) found BPS in 81% of 315 urine samples collected in the USA, China, India, Japan, Korea, Kuwait, Malaysia and Vietnam. The BPS concentrations ranged from below the limit of quantitation (0.02 ng/ml) to 21 ng/ml (geometric mean 0.168 ng/ml) (Table 11). A clear correlation between high BPS content in thermal paper in some countries (Table 1) and high urinary BPS in the citizens of these countries (Table 11) was observed.

Table 11 BPS content in urine (µg/l) of general population of 8 different countries.
Full size table
Estimation of daily BPS exposure by biomonitoring urinary excretion

This methodology estimates the total daily exposure as:

$${\text{Exposure}} = {\text{urinary BPS concentration }} (\upmu {\text{g}}/{\text{l}}) \times {\text{urine excretion rate }}({\text{l}}/{\text{day}}).$$
(2)

Using this methodology, Liao et al. (2012a) concluded that the mean and median exposures of BPS for the general populations in eight different countries were collectively 0.930 and 0.248 µg/day, respectively, with Japan, US, China and Kuwait having the highest exposures (Table 12).

Table 12 Estimated exposure (µg BPS/l urine) in general population of eight different countries.
Full size table

The two highest exposures estimated in Table 12 for the general population were similar to the exposure estimated for two different cashiers in France after collecting urine at 24 h, with 3.42 and 1.5 µg BPS/day (Ndaw et al. 2018).

Estimation of daily BPS exposure starting with BPS contents in thermal paper

Several papers have estimated exposure to BPS as:

$${\text{Exposure (ng}}/{\text{day)}} = k \times C \times {\text{HF}} \times {\text{HT}} \times {\text{AF}}/10^{6}$$
(3)

where k is the paper-to-skin transfer coefficient for BPS (21,522.4 ng/s) (the same value reported for BPA because this parameter is unknown for BPS); C is the BPS concentration in paper samples (µg/g); HF is handling frequency (2–3 times/day for the general population and 150 times/day for the workers population); HT is the handling time of paper (5 s/handling); AF is the absorption fraction of BPS (set at 27% in all the reviewed studies). The estimations of the exposures found in the recent literature are summarised in Table 13.

Table 13 BPS exposure (ng/day) estimated assessing transfer rate from thermal paper to skin according to Eq. 3
Full size table

By the same procedure, Rocha et al. (2015) estimated the 95th percentile of bisphenol exposure to be 2000 and 101,000 ng/day for the general population and the worker population, respectively, while the median exposure for these populations was, respectively, 1420 ng/day and 71,000 ng/day. These values were considerably higher than those reported in Table 13. However, it was remarkable that these authors determined total bisphenols, and that it was not possible to discriminate the specific exposure to both BPA and BPS separately. Therefore, these figures were considered unsuitable for our purposes.

Discussion about BPS exposure

Exposure estimated by BPS excreted in urine might be more representative of systemic (internal) exposure, but the respective figures can be underestimated if the substance bioaccumulates. This is not expected if we consider: (1) the water solubility of BPS (1.1 g/l) and a log Pow of 1.2 (BPS dossier for registration according to the REACH Regulation); (2) the high excretion in humans after a single oral administration (92% in males; 72% in females) (Oh et al. 2018); (3) the slight retention in tissues following a single gavage dose in mice and rats, which anticipates a lower accumulation potential with repeated dosing (Waidyanatha et al. 2018). Therefore, we used the exposures estimated by this procedure to make a comparison with the DNEL derived from the oral dose to be assimilated to the systemic DNEL by assuming 100% oral absorption, which is typically considered for the pharmacokinetic modelling of bisphenols (Karrer et al. 2018).

For our assessment, we took the reasonable worst cases to be the highest exposure that derived from one cashier after the 24-h urine collection employed in the study of Ndaw et al. (2018) (3.4 µg BPS/day) (Table 14) and, for the general population, the mean exposure derived from monitoring urine in eight different countries (0.93 µg/day) (Tables 12, 14).

Table 14 Exposure figures taken for risk assessment in this paper. It was considered a corporal default weight of 60 kg for estimating µg/kg bw/day
Full size table

The exposures derived from assessing BPS paper-skin transference (Table 13) were clearly higher than those obtained from BPS urinary excretion (Table 12). As we see it, this estimation procedure might overestimate real exposure basically for two reasons: (1) the paper-to-skin transfer coefficient for BPS is unknown and the coefficient for BPA was taken by default; (2) the absorption fraction of BPS through skin was considered by several authors to be 27%, while the potentially absorbed dose determined in one ex vivo study using human skin stratum corneum samples was considerably lower (8.79 ± 3.24%) (BPS dossier for registration according to the REACH Regulation). For our assessment, we considered the 95th percentile for the general population and the workers in the study of Liao et al. (2012a) to be reasonable worst cases (Table 14).

The general population exposures shown in Table 14 were lower (for both scenarios), but of the same order of magnitude as the BPA exposures estimated by RAC (2015) (Tables 3 and 4). However, the exposure estimated by assessing the amount of BPS that came into contact with the workers’ skin approximately doubled the estimated BPA exposure. The workers’ exposures shown in Table 14, estimated by urine excretion, are around eightfold lower than the BPA exposures estimated by RAC (2015) (Tables 3 and 4).

Hazard identification

A 90-day repeated dose toxicity study

The results of this repeated dose toxicity study are summarised in Table 15. The highest dose had to be reduced by day 70 and onwards due to excessive body weight reduction in animals. At this high dose, in addition to this body weight and body weight reduction, other effects were reported: (1) slight haematological impairments (red blood cell, haemoglobin, haematocrit, mean corpuscular volume and relative reticulocyte counts); (2) changes in clinical chemistry (increases in creatinine and alkaline phosphatase and reductions in bilirubin; (3) increases in organ weights (adrenals, brain, epididymis, heart, kidneys, liver, spleen, thyroid, thymus, testes and ovaries) in males and/or females; (4) histopathological alterations in caecum, spleen, mammary gland, liver and uterus of females.

Table 15 Summary of the 90-days repeated dose toxicity study with BPS.
Full size table

Increases in organ weights were found for kidney in males, and for adrenals, kidney, liver and spleen in females. The only relevant effect noted at the lowest dose was a focal squamous cell metaplasia of glandular epithelium in the uteri of two females, which was also found in two females at the mid dose and in five females at the highest dose, but not in the control females (Table 15).

According to the registrant, the No Observed Adverse Effect Level (NOAEL) of this study should be 100 mg/kg bw/day, based on the alterations of organ and body weight and body weight gain reported at 300 mg/kg bw/day [dose considered to be the Lowest Observed Adverse Effect Level (LOAEL)]. We noted that the focal squamous cell metaplasia of glandular epithelium in the uterus reported at 100 mg/kg bw/day was not contemplated because a similar incidence was reported for 300 mg/kg bw/day. However, we also noted that this effect was not reported in the control animals and, based on a more conservative approach; we considered that 100 mg/kg bw/day should be taken as the LOAEL of this 90-day repeated toxicity study, while no NOAEL could be set.

Toxicity to reproduction: screening for reproductive/developmental toxicity

An experimental study performed, which observed OECD Guideline 421 and GLP compliance, was found in the BPS dossier for registration according to the REACH Regulation. This study was conducted with Crj: CD(SD) rats and presented a deviation from the standard protocol because the high dose group only included seven pregnant females, which diminished the study’s reliability. The total exposure in this study was 45 days in males (including a 14-day pre-mating period) and 40–46 days in females (from mating to the gestational period and parturition to day 3 of lactation). Animals without delivery were administered until day 25 after confirming copulation. The study results are summarised in Table 16.

Table 16 Summary of the study for screening for reproductive/developmental toxicity with BPS.
Full size table

Severe body weight gain reduction, together with alterations in the relative weight of liver, pituitary gland, and the absolute weight of seminal vesicles and histopathological alterations in caecum and liver were reported at the highest dose. At this same dose, reproductive performance, the implantation index and the total number of offspring at birth were also statistically reduced, as observed in the controls. The same gross pathology and histopathology were described for the mid dose, but with lower incidences than for the top dose.

The NOAEL for parental toxicity was set at 10 mg/kg bw/day based on the gross pathology and histopathology reported at 60 mg/kg bw/day (the dose considered to be the LOAEL). The NOAEL for reproductive toxicity was 60 mg/kg bw/day based on alterations in oestrous cycles, the fertility index, the implantation index and the number of live offspring reported for 300 mg/kg bw/day (the dose considered to be the LOAEL).

Developmental toxicity

One prenatal developmental toxicity study, performed by observing OECD Guideline 414 and GLP compliance, was found in the BPS dossier for registration according to the REACH Regulation. Twenty-five pregnant Wistar rats per dose were treated by gavage with 30, 100 or 300 mg BPS/kg bw/day from gestation day (GD) 6 through to GD 19.

Reversible salivation was found in 7/25 females at the highest dose. This top dose caused reductions in body weight gain from gestation day 8 to 10 (− 29%) and from gestation day 6 to 19 (− 8%). The corrected body weight of the animals treated with 300 mg BPS/kg bw/day was 10% lower than the corrected body weight of the control animals. This difference was not statistically significant. No signs of maternal toxicity were reported for the animals dosed with 30 and 100 mg BPS/kg bw/day.

Neither significant differences in the mean placental weight among the various groups, nor incidences of external variations and soft tissue malformations, were observed. The incidences in soft tissue variations, external malformations, skeletal malformations, skeletal variations and foetuses with skeletal malformation/litter did not statistically differ from the concurrent control or fell within the historical control data or no dose response was observed. Therefore, these effects were not considered dose-related.

The maternal NOAEL was set at 100 mg/kg bw/day based on the severe alterations in body weight gain reported at 300 mg/kg bw/day (dose considered to be the LOAEL for maternal toxicity). No dose-related teratogenicity was detected in this study. Therefore, the NOAEL for developmental toxicity was the highest tested dose (300 mg BPS/kg bw/day) and no LOAEL could be derived.

Setting critical end-points for assessment

The repeated dose toxicity studies caused severe reductions in body weight and bodyweight gain, alterations in the weight of several organs, and histopathological alterations in caecum, spleen, mammary gland, liver, and uterus and fertility impairments. No teratogenicity was reported for BPS. Table 17 summarises the risk assessment end-points.

Table 17 End-points for risk assessment of BPS according to data provided in Tables 15 and 16
Full size table

Given the different duration of the studies and the various types of end-points (one NOAEL and one LOAEL), we derived DNELs for both to later consider the lowest one for the risk characterisation.

DNEL derivation

RAC used a procedure for transforming animal doses into the HED. This approach is more questionable for BPS because there was much less toxicokinetic available information than for BPA. Indeed the basic toxicokinetics in vivo study is not presented in the BPS dossier for registration according to the REACH Regulation because it is “ongoing” (checked in December 2018).

Nevertheless, we decided to derive two DNELs (with and without HED transformation) to be as conservative as possible. For such transformations, we used the same factors employed for BPA (0.068 for the oral DNEL and 0.00074 for the dermal DNEL) by EFSA (2015) and RAC (2015). Other assessment factors were considered according to the ECHA procedures for deriving DNELs for threshold end-points (ECHA 2012).

Oral systemic DNEL

The DNEL derivation for the critical end-point of the 90-day repeated dose toxicity study is summarised in Table 18. We considered an allometric factor of 4, an interspecies remaining differences factor of 2.5, an intra-species difference of 5 (for the workers) or 10 (for the general population), adjustment sub-chronic to chronic of 2, a LOAEL to NOAEL extrapolation of 3 and an additional factor due to poor quality (only two studies by a single route) of the database of 10 (RAC used a factor of 6 to cover BPA assessment uncertainties but we considered that the uncertainties for BPS were larger than for BPA). We considered 100% oral absorption for the transformation of oral DNEL into systemic DNEL, as did Karrer et al. (2018).

Table 18 Derivation of systemic oral DNEL for BPS from the reproductive/developmental toxicity in rats. The starting end-point was taken from Table 17 and deduced from data presented in Table 15
Full size table

The DNEL derivation for the critical end-point of the reproductive/developmental toxicity study is summarised in Table 19. We considered similar assessment factors to those discussed above but took into account that the critical end-point was a NOAEL. Therefore, no additional assessment factor was needed, exposure was 45 days and we assigned this duration to a subacute study to consider an assessment factor of 6 to extrapolate from subacute exposure to chronic exposure.

Table 19 Derivation of systemic oral DNEL for BPS from the reproductive/developmental toxicity study in rats. The starting end-point was taken from Table 17 and deduced from data presented in Table 16
Full size table

In both cases, the transformation of animal doses into the HED yielded lower DNELs than when using an allometric factor for derivation. The DNEL derived from the reproductive/developmental toxicity study in rats was one order of magnitude lower than that derived from the 90-day oral repeated dose toxicity study in rats. Therefore, to characterise the worst possible scenario, we made our risk characterisation of systemic DNELs and obtained 0.91 and 0.45 µg/kg bw/day for the workers and the general population, respectively (Table 19). These DNELs were around one order of magnitude lower than the oral DNELs derived by RAC (2015) for BPA (Table 6), which suggests that our assessment might be conservative.

Dermal DNEL

The dermal derivation was performed using the same factor as for BPA, while the derivation without transformation into the HED yielded the same DNELs outlined in Tables 18 and 19 (not shown here for simplicity reasons). The dermal DNEL estimation is summarised in Tables 20 and 21. As we did not consider biotransformation factor as RAC (2015) did for BPA, our work is as conservative as can be reasonably assumed.

Table 20 Derivation of systemic oral DNEL for BPS from the reproductive/developmental toxicity in rats. The starting end-point was taken from Table 17 and deduced from data presented in Table 15
Full size table
Table 21 Derivation of systemic oral DNEL for BPS from the reproductive/developmental toxicity study in rats. The starting end-point was taken from Table 17 and deduced from data presented in Table 16
Full size table

Following the same conservative approach used for the systemic DNELs, we considered the data derived from the reproductive/developmental toxicity study in rats (respectively 0.0098 and 0.0049 mg/kg bw/day for workers population and general population). These DNELs were around 20-fold lower than the dermal DNELs derived by RAC (2015) for BPA, which once again suggests that our assessment is conservative.

Risk characterisation

The RCR was estimated, as in the case of BPA, to be the ratio between exposure and the DNEL. We estimated RCR for both considered cases: the exposures estimated from urinary excretion versus the derived systemic DNEL, and the exposure estimated from dermal contact versus the derived dermal DNEL (Table 22).

Table 22 Risk characterisation for dermal exposure to BPS in thermal papers. RCR was estimated as the ratio between exposure and DNEL. Both exposures and DNEL were expressed in µg BPS/kg bw/day. Exposures were taken from Table 14. DNELs were taken from Tables 19 and 21
Full size table

For dermal exposure, the RCR was higher than for both workers and the general population, which suggests that the risk, especially for workers, might not be adequately controlled.

Discussion

RAC (2015) released a scientific opinion which demonstrated that the risk associated with the dermal exposure of cashiers to BPA via thermal paper receipts might not be adequately controlled. Consequently, a restriction in BPA contents in thermal paper will come into force in Europe as of 2020. Here we made a preliminary assessment of the risk associated with the exposure of the general population and workers to BPS via dermal contact with thermal paper receipts. We concluded that, for one of the scenarios that we considered, and as in the case of BPA, once again the situation might not be adequately controlled because exposure was around threefold the DNEL for the general population and around 100-fold for workers.

We based our assessment on publically available information in the scientific literature as we recognised many uncertainties, which we attempted to overcome using an assessment factor. This means that our assessment might be extremely conservative and the situation is probably not as bad as suggested by the data displayed in Table 22.

Uncertainties associated with exposure assessment

We contemplated two well-defined exposure scenarios. One was based on an exposure assessment based on urine-excreted BPS, while the other one was based on an exposure assessment by estimating the BPS transfer rate between thermal paper and skin. We considered that the first scenario was much more reliable as it suggested that the risk could be adequately controlled. We based our reliability in this scenario on the fact that urine-excreted BPS was necessarily absorbed and could theoretically reach target organs. This assessment could underestimate exposure by not considering the bioaccumulation of this substance. However, such bioaccumulation was not expected given the basis of both the substance’s physical properties and the toxicokinetic results obtained with humans (Oh et al. 2018). Moreover, the potential bioaccumulation, if any, would probably not be able to increase exposure by a factor of 17, which is what will be needed to reach a systemic DNEL for the workers.

The exposure scenario characterised based on the transfer of BPS from the matrix to skin has, to our understanding, several great uncertainties as: (1) the percentage of dermal absorption used in the literature (27%) is considerably higher than the dermal absorption rate found in the BPS dossier for registration according to the REACH Regulation (9%); (2) dermal absorption rates are not the best way to estimate dermal absorption because it assumes that transfer is constant with time and independent of the amount of substance that comes into contact with skin, which is not completely true; (3) the paper-skin transfer rate for BPS was unknown and the figure for BPA was used as a surrogate, which could also introduce severe uncertainty because BPS is presumably much more polar than BPA and, therefore, BPS would cross membrane barriers by diffusion at much lower rates than BPA.

The uncertainties-associated database

BPA is the most widely studied bisphenol and BPS has attracted interest only in recent years because several restrictions on the use of BPA have been imposed in regulations in developed countries. However, many reports warn that BPA and BPS might exhibit a comparable endocrine disruption potential. For example, it has been demonstrated that BPS has similar anti-androgenic effects to those of BPA (Eladak et al. 2015), and that BPS has similar toxic and estrogenic effects to BPA insofar as inducing developmental deformities (cardiac oedema, spinal malformation and craniofacial deformities) in fish larvae (Moreman et al. 2017). A review has also pointed out that BPS is as hormonally active as BPA (Rochester and Bolden 2015). This suggested that we should include an additional larger assessment factor than the assessment factor used by RAC for BPA assessments, which could lead to a lower DNEL than that needed.

Moreover, the use of the HED still cannot be as appropriate as BPA because lack of toxicokinetic information for BPS brings about additional uncertainties in transformation as we assumed that BPS would display similar toxicokinetic behaviour to BPA in both humans and rats, which has not yet been demonstrated.

The inappropriateness of the transformation of animal dose into HEDs can be particularly notable in the dermal DNEL because it is well-known that dermal absorption is markedly influenced by the polarity of the substance, and that the solubility of BPS in water is 3.7-fold higher than the solubility of BPA, while the log Kow for BPA is 2.8-fold higher than it is for BPS (BPA and BPS dossiers for registration according to the REACH Regulation). These data suggest a very different polarity. In our assessment, we considered that both behave similarly way in dermal terms, which might be the largest uncertainty introduced into calculations.

Conclusions

We created two different BPS dermal exposure scenarios: the risk of that considered more realistic seemed to be adequately controlled, while the worse scenario implied greater uncertainties, which suggests a very worrying situation. Moreover, it should be assumed that the situation might presumably get worse with the European Union restriction fully operative if the reduction in BPA contents causes an increment in the BPS in thermal paper receipts.

This work highlights that more studies on BPS are urgently needed, specially related to toxicokinetics but also related to test the substance in a second species and in a 2-generation reproductive toxicity study, to properly assess the risk associated to dermal exposure via thermal paper.

This work also warns that it might be dangerous to ban the use of certain substances when alternatives might not be safer than banned substances. However, the RAC scientific opinion (2015) did not propose banning BPA in thermal paper but suggested the need to reduce it. This might be a good option as the risk associated with BPS is still not well assessed and controlled, as we reflect in this paper.