Background briefing on process of EFSA draft health assessment of BPA
EFSA conducted a comprehensive and systematic weight-of-evidence approach
EFSA looked at over 450 studies in a fully transparent process: the large guideline studies that already
formed the basis for their 2006 opinion, those that were specially pointed out to them by critical groups
such as the French NGO Réseau Environnement Santé (RES), as well as studies that became
available more recently. A table lists each individual study with their strengths and weaknesses, any
additional points to be mentioned, and the influence on likelihood of evidence as judged by the EFSA
experts (page 425ff). Based on the evaluation of these studies EFSA defines the high quality guideline
studies and a recent comprehensive NCTR/FDA study as the most reliable studies for their risk
assessment, and uses these studies as a point of departure for their assessment. In the subsequent
evaluation, EFSA chose the most conservative approach, and applied a sophisticated, and highly
conservative, assessment method on the data.
EFSA assessed toxikokinetic study data on BPA
Toxikokinetic studies investigate the „behaviour“ of a substance in the body, for example, what parts of
the body or organs the substance may enter (e.g. fat, liver etc.), whether or not it is expected to be
metabolized or excreted, and at what rate (absorption, distribution, metabolism and excretion).
Toxicokinetic studies are also used for establishing relationships between experimental exposure in
animals and the corresponding exposures in humans. Such studies are complex and costly, and not
normally available for chemical substances. However, for BPA, a number of such toxikokinetik studies
have been done in the past years. EFSA referred specially to three studies done by scientists of the
US FDA NCTR in mice, rats, and monkeys, confirming efficient metabolism of BPA in all species
investigated (*Doerge, 2010a, 2010b, 2011).
EFSA ranked all studies based on potential association of BPA to a particular health effect
EFSA looked at all endpoints investigated in the studies, and grouped them into areas. Then EFSA
looked into each of these clusters in more detail, to assess, based on all available studies, whether a
potential association between exposure to BPA and particular effects could be identified. The
likelihood of such association of effects was classified in a six-staged ranking system, reaching from
„very likely“ to „very unlikely“. This system is quite comparable to the one used by the US Center for
Risks of Human Reproduction (CERHR), or to weight of evidence assessments in the medical arena.
In context of the overall weight of scientific evidence EFSA judged association with reproduction and
fertility, brain and behavioural development, immune, cardiovascular, metabolic, genotoxic and
carcinogenic effects to be “less than likely”. These areas were therefore not specifically considered in
the following human health assessment, but are explicitly mentioned to be covered by the
Three areas were classified to show „likely“ associations: liver, kidney and mammary gland (the latter
with a larger amount of uncertainty on the likelihood) – although not at dose levels relevant to human
exposure. It is important to note that with this assessment EFSA does not indicate a “likely” risk for
humans but only defines the starting point for their risk assessment. The EFSA experts indicated that
the “likely” associations on liver, kidney and mammary gland should be further risk assessed. This is
because they considered that it is “likely” that an effect on the above mentioned organs might occur if
animals are exposed to very high doses, i.e. doses above 3.6 mg/kg/day, 1000-fold above the t-TDI.
EFSA deliberately selected the most sensitive data for its assessment
EFSA identified general toxicity as the most sensitive endpoint (like in their previous assessments).
For the three „likely associations“ EFSA identified the lowest effect levels seen in the mouse (kidney
3.6 mg/kg bw/d,) and defined this dose as a starting point for their risk assessment. Effects on liver
and mammary gland are only observed at substantially higher doses. This starting point is therefore
considered to be conservative and covers all other endpoints. In effect, this low baseline ensures that
any remaining uncertainty is covered in the evaluation.
It is interesting to note that the level defined by EFSA was the lowest level measured in a
comprehensive two-generation study in one generation of adult male mice only, and only in the parent
generation, not in their offspring. The values of the offspring or of the pregnant mother are
substantially higher in this study, indicating that EFSA took a very low value as a starting point for their
In addition, if EFSA would have taken toxikokinetic data from neonatal mice, or other species such as
rat or monkey – in each case, the resulting TDI would have been even higher than the current one of
50 microgram/kg bw/d. However, again, EFSA deliberately selected the very most sensitive datapoints
available to base their assessment on.
Transfer of data from mouse to human: EFSA used a more conservative baseline compared to
previous assessment to calculate human equivalent dose
The toxikokinetic data provide a deeper understanding of the behaviour of BPA in the body, and also
enable a more concrete calculation of the human equivalent level. Based on the most protectively
defined low benchmark dose level (3.6 mg/kg bw/d), and the most conservative toxikokinetic factor
reported in animal experiments (of 32) to convert the animal data into human relevance, EFSA
calculated the so-called „human equivalent dose“ to be 113 microgram/kg bw/day.
EFSA applied a high safety factor to manage uncertainties
In every scientific process, uncertainties or open questions will – by nature of science – always remain.
Given the large amount of available scientific information on BPA, such remaining uncertainties
however do not give any rise to concern for human health – otherwise EFSA would clearly have
reacted much differently. The standard safety factor to account for uncertainties is 100 (10 for inter-,
and another 10 for intra-species-variabilities). This safety factor is also included in the current TDI of
50 microgram/kg bw/d.
In its new assessment, EFSA applied a more conservative overall assessment factor of 800; after final
rounding to derive a one-digit value for the t-TDI the overall resulting safety factor is 740:
EFSA uses a factor of 32 for toxikokinetic differences between mice and humans to define the human
equivalent dose, a remaining inter-species factor of 2.5 and an unchanged factor 10 for intra-species
variability, resulting in a remaining uncertainty factor of 25. Both values multiplied lead to an overall
assessment factor of 800 and a very protective TDI of 4.52 microgram/kg bw/d, which is rounded up
to 5 microgram/kg bw/d, thus including a final safety factor of ~740.
EFSA derived a highly conservative and overprotective t-TDI to cover all uncertainties, even
the „less than likely“ ones
In 2006, EFSA defined 5 mg/kg bw/d as lowest level to work with (the „starting point“, also called
NOAEL – no adverse effect level). Applying a safety factor of 100 resulted in the TDI of 50
The new temporary-TDI was derived by applying the remaining uncertainty factor of 25 on the „human
equivalent dose“ (HED), which was derived using a benchmark model in animal studies.
HED 113 microgram/kg bw/day divided by 25 = 4.52, i.e. ≈ t-TDI 5 microgram/kg bw/d.
After having derived the t-TDI, EFSA cross-checked with the values of each of the other endpoints to
make sure no single endpoint was overlooked, and to confirm that the human equivalent dose used as
starting point for all discussions would cover all other areas of potential toxicity. Given the extremely
conservative approach in deriving such a highly protective TDI, EFSA concludes that this t-TDI covers
also remaining uncertainty related to the endpoints liver and mammary gland as well as those
endpoints which were considered „less than likely“.
Therefore, the substantially reduced t-TDI is neither a result of increased uncertainty, nor of new
scientific data that would have risen the concern. On the contrary: EFSA derived the t-TDI using worst
case data (the most sensitive endpoint: kidney toxicity, in the most sensitive species: mice, at the
lowest reported benchmark level derived in the first generation in a comprehensive study investigating
three generations), and applying a modern, sophisticated mathematical model to define a benchmark
dose to calculate the human equivalent dose (very conservative conversion factor from animal to
humans). Still, the exposure of all groups of the population, including the most sensitive and the most
highly exposed ones, is well below even this substantially reduced t-TDI. We therefore understand the
fact that EFSA decided to amend the established safe intake levels for BPA as a sign of an
abundance of caution at EFSA.
If there is reduced uncertainty based on the toxikokinetic data, why is the t-TDI so substantially
Uncertainties in the previous data could be reduced by new toxikokinetic data – therefore the
previously used uncertainty factor of 100 could be reduced to 25. The new t-TDI is therefore more
solid and provides high security for all age groups and exposure scenarios.
New and better comprehensive toxikokinetic animal data enabled EFSA to define a human equivalent
level. The availability of new toxikokinetik data (which were not available in the previous EFSA
assessment), provided a better understanding of the behaviour of BPA in the metabolism. EFSA
evaluated these new toxikokinetic data and lowered the remaining overall uncertainty in their
assessment by applying a worst-case animal-human conversion factor based on animal experiments.
Toxikokinetik studies on BPA are available in three animal species: in mouse, rat and monkey. The
monkey test-system is the one most near to the human system, while the mouse system is the most
far away. Authorities have a scope of discretion as to which system/data to select. EFSA choose the
mouse, thus being consistent with previous European risk assessments of BPA which also used the
mouse test-system as reference.
The specific data starting point used by EFSA (adult mouse) lead to an adjustment of the previously
used animal-human conversion factor. (4->32).
By using the most sensitive data point from the mouse test system as starting point and calculating a
benchmark level using the increased animal-human conversion factor, the new t-TDI was derived.
EFSA used in each assessment step the most protective scenario, resulting in a „very conservatively
The new t-TDI is scientifically very conservative and provides solid protection, as it is further
substantiated by new scientific data. Earlier uncertainties due to lacking scientific knowledge have
been substantially reduced. During the evaluation of the available studies an extremely conservative
approach was taken by EFSA and therefore, the new t-TDI should be regarded as highly protective.
Still, all age groups and exposure scenarios are well below even the t-TDI.
The new TDI for BPA is based on a large safety factor and is temporary
A „temporary TDI“ is established for a specified, limited period of time to enable additional
biochemical, toxicological or other data to be obtained as may be required for estimating a tolerable
daily intake. A temporary TDI involves the application of a safety factor larger than that used in
establishing a permanent TDI. http://www.codexalimentarius.net/pestres/data/reference/glossary.html
EFSA defined the new safety value as temporary TDI pending the outcome of the long-term study in
rats involving prenatal as well as postnatal exposure to BPA, which is currently being undertaken by
the US National Toxicology Program (NTP). This study will also clarify open questions re mammary
gland effects. We would expect that EFSA would re-consider and decide on a permanent TDI once
these data have become available for regulatory assessment.
EFSA default (2006) New EFSA draft assessment (Jan 2014)
Interspecies variability: 10
x 2.5 (toxicodynamic, also called „additional
uncertainty“) = 10
Interspecies variability: 80
32 (toxicokinetic data adult mouse)
x 2.5 (toxicodynamic, also called „additional uncertainty“) = 80
Intraspecies variability: 10 Intraspecies variability: 10
=> 10 x 10 = 100
overall assessment factor
=> 80 x 10 = 800
EFSA default (2006) New EFSA draft assessment (Jan 2014)
starting point: 5000 microgram (NOAEL)
(NOAEL = no adverse effect level)
starting point: 3600 microgram (BMDL10 adult mouse)
(BMDL = benchmark dose low)
Human equivalent dose (HED): 113 microgram
(3600 ./. 32 = 112,5 -> ≈ 113),
already includes the toxikokinetic data
Uncertainty factor: 100 Remaining uncertainty factor 25
(800 ./. 32 = 25; due to availability of toxicokinetic data)
5000 ./. 100 = TDI 50 microgram/kg bw/d 113 ./. 25 = 4.52 i.e. ≈ t-TDI 5 microgram/kg bw/d.
Safety factor: 100 Safety factor: 740
- Doerge DR, Twaddle NC, Vanlandingham M and Fisher JW, 2010a. Pharmacokinetics of bisphenol A in
neonatal and adult Sprague-Dawley rats. Toxicology and Applied Pharmacology, 247, 158-165.
Doerge DR, Twaddle NC, Vanlandingham M and Fisher JW, 2011. Pharmacokinetics of bisphenol A in neonatal
and adult CD-1 mice: inter-species comparisons with Sprague-Dawley rats and rhesus monkeys. Toxicology
Letters, 207, 298-305.
Doerge DR, Twaddle NC, Woodling KA and Fisher JW, 2010b. Pharmacokinetics of bisphenol A in neonatal and
adult rhesus monkeys. Toxicology and Applied Pharmacology, 248, 1-11.