Re-evaluation of the risks to public health related to the presence of bisphenol A (BPA) in foodstuffs

Abstract

In 2015, EFSA established a temporary tolerable daily intake (t-TDI) for BPA of 4 μg/kg body weight (bw) per day. In 2016, the European Commission mandated EFSA to re-evaluate the risks to public health from the presence of BPA in foodstuffs and to establish a tolerable daily intake (TDI). For this re-evaluation, a pre-established protocol was used that had undergone public consultation. The CEP Panel concluded that it is Unlikely to Very Unlikely that BPA presents a genotoxic hazard through a direct mechanism. Taking into consideration the evidence from animal data and support from human observational studies, the immune system was identified as most sensitive to BPA exposure. An effect on Th17 cells in mice was identified as the critical effect; these cells are pivotal in cellular immune mechanisms and involved in the development of inflammatory conditions, including autoimmunity and lung inflammation. A reference point (RP) of 8.2 ng/kg bw per day, expressed as human equivalent dose, was identified for the critical effect. Uncertainty analysis assessed a probability of 57-73% that the lowest estimated Benchmark Dose (BMD) for other health effects was below the RP based on Th17 cells. In view of this, the CEP Panel judged that an additional uncertainty factor (UF) of 2 was needed for establishing the TDI. Applying an overall UF of 50 to the RP, a TDI of 0.2 ng BPA/kg bw per day was established. Comparison of this TDI with the dietary exposure estimates from the 2015 EFSA opinion showed that both the mean and the 95th percentile dietary exposures in all age groups exceeded the TDI by two to three orders of magnitude. Even considering the uncertainty in the exposure assessment, the exceedance being so large, the CEP Panel concluded that there is a health concern from dietary BPA exposure.

Keywords: BPA; Bisphenol A; TDI; food contact materials; hazard; health risks; toxicity.

Figure 1

Diagram of the conceptual model for the uncertainty analysis of the hazard assessment for HOCs other than genotoxicity, the uncertainty analysis for which was conducted separately and is described in Section 2.3.4.2. See text for further explanation

Figure 2

In vitro chromosomal aberration and micronucleus studies: evaluation and summary of test results from 15 studies

Figure 3

In vivo chromosomal aberration and micronucleus studies: evaluation and summary of test results from 11 studies

Figure 4

In vitro comet assay: evaluation and summary of test results from 22 studies

Figure 5

In vivo comet assay: evaluation and summary of test results from 21 studies

Figure 6

Effects in germ cells: summary of evaluation and results from five studies

Figure 7

Cumulative probability functions (cpfs) quantifying uncertainty about the estimated lowest BMD across the 21 primary clusters considered in the uncertainty analysis

1. The dashed black curves show the lower and upper envelope for the cpf resulting from the range of judgements between WG experts (lower and upper refer to relative position on the vertical axis). The red solid curves show the lower and upper bounds of the averaged cpf, where judgements of different experts for each cluster were aggregated by averaging. Blue dotted lines show probabilities for the estimated lowest BMD across all clusters being below the RP of 8.2 ng BPA/kg bw per day (HED). Green solid lines show probabilities for the estimated lowest BMD across all clusters being below the RP when an additional UF of 2 is applied, as discussed in Section 3.2.4. Subtracting these probabilities from 100% gives the corresponding probabilities for the estimated lowest BMD across all clusters being above the RP.

Figure B.1

Criteria for evaluating if there is a significant dose–response

Figure B.2

Results from fitting a restricted cubic spline using 3 and 4 knots to the data from figure 8A (Montévil et al.,  [RefID 13788])

Figure C.1

Application of PBPK model of Karrer et al. (2018) [RefID 12289] to check linearity

Figure D.1

Experts’ approximate probabilities, for each cluster, that there is at least one endpoint in the WoE table for the cluster that occurs in animals tested with BPA and is relevant and adverse in humans

1. Each bar represents the range of % probability for the specified cluster and expert (expert A, B etc.).

Figure D.2

Two examples, for different clusters and experts, of elicited probability distributions for the estimated lowest BMD of those endpoints in the cluster that occur in animals tested with BPA and is both relevant and adverse for humans

1. Each graph shows the histogram provided by a single expert, and the parametric distribution that was subsequently fitted to their judgements.

Figure D.3

Key endpoints and effect sizes considered by the experts when assessing Question 2 for the cluster cellular immunity

1. Labels on the horizontal axis comprise study reference identification (RefID) number, endpoint, tier of study (from WoE assessment), sex and species, and developmental stage or chosen BMR value. Percentages shown by BMDU and BMDL symbols refer to the BMR on which they are based; percentages shown by NOAEL and LOAEL symbols refer to the effect size at that dose as % change from the control group, estimated by EFSA from the original study.

Figure D.4

Key endpoints and effect sizes considered by the experts when assessing Question 2 for the cluster allergic lung inflammation

1. Labels on the horizontal axis comprise study reference identification number (RefID), endpoint, tier of study (from WoE assessment), sex and species. Percentages shown by BMDU and BMDL symbols refer to the BMR on which they are based; percentages shown by NOAEL and LOAEL symbols refer to the effect size at that dose as % change from the control group, estimated by EFSA from the original study (na = not available).

Figure D.5

Parametric distributions fitted to the judgements of 16 experts for Question 2 for the cluster cellular immunity

Figure D.6

Parametric distributions fitted to the judgements of 16 experts for Question 2 for the cluster allergic lung inflammation

1. Note: The odd‐looking distribution with a sharp peak near the centre of the graph is mixture of two normal distributions which provided the best fit to the judgements of this expert (expert C), shown in Annex K. Sensitivity analysis showed that using these fitted distributions or the histograms provided by the experts made no material difference to the results (see later).

Figure D.7

Summary of distributions fitted to each experts’ judgements, for each cluster, for the estimated lowest BMD in the cluster for endpoints that occur in animals tested with BPA and is relevant and adverse in humans

1. Each bar represents the 90% probability interval of the distribution for the specified cluster and expert (expert A, B etc.).

Figure D.8

Example of the results of combining one expert's judgements on Questions 1 and 2 for five immunotoxicity clusters; see text for explanation

Figure D.9

Example illustrating how different experts’ judgements for the same cluster were combined by two alternative methods: enveloping (black curves) and unweighted averaging (blue curves). See text for explanation

Figure D.10

Combination of the enveloped cpfs for the 21 individual clusters to obtain lower and upper enveloped cpfs for the estimated lowest BMD across all clusters

1. The black curves in the graphs show the lower and upper bounds of the enveloped cpf for the estimated lowest BMD across all clusters for endpoints that occur in animals and are relevant and adverse for humans, assuming that the experts’ judgements for different clusters are independent. The coloured curves show the lower and upper bounds of the enveloped cpfs for the individual clusters, from which the combined cpfs are calculated. The curves for each cluster are identified by the combination of line type (solid, dashed, etc.) and colour, as shown by the legend on the right.

Figure D.11

Lower and upper bounds for the overall enveloped and averaged cpfs quantifying uncertainty about the estimated lowest BMD across the 21 primary clusters considered in the uncertainty analysis

1. The solid black curves show the lower and upper bounds for the overall enveloped cpf resulting from the range of judgements between WG experts (lower and upper refer to relative position on the vertical axis). The red dashed curves show the lower and upper bounds of the overall averaged cpf, where judgements of different experts for each cluster were aggregated by averaging.

Figure D.12

Effect of including (solid curves) or excluding (dashed curves) allergic lung inflammation on the enveloped cpfs (left‐hand graph) and averaged cpfs (right‐hand graph) for the estimated lowest BMD across all the assessed clusters. Lower and upper bounds are shown for each cpf. See text for further explanation

Figure D13A

Impact of excluding different clusters on the 5th percentile of the overall enveloped cpf (upper panel) and overall averaged cpf (lower panel) for the estimated lowest BMD across clusters

1. Each bar shows the lower and upper bound for the fifth percentile when the indicated cluster was excluded. The bottom bar (‘all’) shows the result when no clusters were omitted.

Figure D13B

Impact of excluding different clusters on the 50th percentile of the overall enveloped cpf (upper panel) and overall averaged cpf (lower panel) for the estimated lowest BMD across clusters

1. Each bar shows the lower and upper bound for the fifth percentile when the indicated cluster was excluded. The bottom bar (‘all’) shows the result when no clusters were omitted.

Figure D.14

Results of excluding allergic lung inflammation plus one additional cluster (indicated on the vertical axis) on the fifth percentile of the overall enveloped cpf (upper panel) and overall averaged cpf (lower panel) for the estimated lowest BMD across clusters Each bar shows the lower and upper bound for the fifth percentile when allergic lung inflammation plus the indicated cluster were excluded. The bottom bar (‘all’) shows the result when only allergic lung inflammation was omitted.

Figure D.15

Sensitivity analysis assessing the impact on the overall enveloped cpf (left hand graph) and overall averaged cpf (right hand graph) of using fitted parametric distributions (red curves) or the histograms provided by the experts (black curves) for Question 2

Figure D.16

Sensitivity analysis assessing the impact of hypothetical dependencies between allergic lung inflammation and cellular immunity on the overall enveloped cpf (left hand graph) and overall averaged cpf (right hand graph). See text for explanation

Figure D.17

Derivation from the overall enveloped and averaged cpfs of probabilities for the estimated lowest BMD across the 21 primary clusters being below the RP (see text for explanation)

1. The percentages shown in blue are probabilities for the estimated lowest BMD across the 21 primary clusters being below the RP derived from the lower and upper bounds of the overall enveloped cpf (probabilities 5% and 100%) and the lower and upper bounds of the overall averaged cpf (probabilities 56% and 72%).

Figure D.18

Judgements of the experts for Question 3: their probability that the lowest BMD of all those endpoints that occur in animals tested with BPA and are both relevant and adverse for humans is below the RP of 8.2 ng BPA/kg bw per day (HED), taking account of the 21 clusters assessed for Questions 1 and 2 and the additional uncertainties listed in Table D.5

Figure D.19

Comparison of probability ranges provided by the experts for Question 3 with the probability ranges calculated from their assessments for Questions 1 and 2, showing the adjustments made by the experts to account for the additional uncertainties listed in Table D.5