COM_HD_DNEL_Working Document_vs2
Comments
by
Peer Review Panel
on
proposed Guidance for DNEL & DMEL derivation based on Human Data
(Working Document version 2, July 8th 2008)
Name / Affiliation / e-mail / commentsMihaiela / Abulescu / MA / expert NADSP / / Not received
Ulrike / Bernauer / UB / DE BfR / / By August29th; see below
Celine / Boudet / CB / FR INERIS / / Not received
Susy / Brescia / SB / UK HSE / / By August 22nd; see below
Michel / Falcy / MF / FR INRS / / Not received
Sten / Flodström / SF / SE KemI / / By September 5th; see below
Peter / Kujath / PK / DE BAuA / / Not received
Erich / Pospischil / EP / AU AMZ / / Not received
Emanuela / Testai / ET / IT ISS Italy / / By August 27th; see below
Section /
PRP member / No / Comment
General comments
SB / UK DC / I see no reason to use different numerical values for the default intra species assessment factor when using human data, whether they come from a small volunteer study or from a large cohort study.
When using epidemiological studies as a starting point for risk assessment, I think it is not simply a question of identifying the exposure range in which the first effects are observed. Account needs to be taken of statistical uncertainty and the overall pattern of risk estimates across different exposure levels.
Good-quality extensive negative human data should certainly be capable of overriding positive animal data for the same end point.
ET / IT / I am fully aware that in the REACH process is necessary to find some ‘standardised procedure’ in order to evaluate a high number of chemicals, and that the use of ‘default’ procedures, such as the application of more or less fixed AFs, goes exactly in this direction. The problem is to give to people approaching the evaluation process the right message from the very beginning. It is my opinion that the concept that the evaluation of the risk associated to the exposure to a chemical is not simply the combination of numbers, but a process during which the expert judgement is frequently necessary should be stressed. Providing that appropriate justification, supported by scientific data, are given in a transparent way, deviations from the default AFs based a case-by-case approach are fully acceptable. Actually, a very long discussion on this issue (use of default values and application of standard AFs) went on within the SCHER during the preparation on the opinion on the revised version of the TGD. I remember that Sharon (Munn) was not very happy, and starting from my initial consideration I can understand it. But, from a scientific point of view, we all know that to be ‘too’ standardised is not fully correct.
SF / SE / We appreciate the work done on this 2nd draft guidance and that some of our comments have been taken onboard. However, we also note that a large number of our comments made on the previous version still apply, as we are not convinced by the responses. We will not repeat our previous comments in detail, but just reiterate the main points. We also note that similar comments have been raised by the UK and DE.
Thanks for mentioning case reports in the guidance. However, more guidance is needed with regard to the use of other human data than epidemiological data, i.e., published case studies/case reports and unpublished company reports with data on short-term effects in humans. Case reports/observations may be small studies, showing e.g. skin irritation in a worker(s) among those (few) involved in the production or use of the substance. For low production volume chemicals, with little data of other types, this may be very important data that should be used despite possible shortcomings with regard to methodology or reporting. Considering that this may apply to perhaps 10.000-15.000 LPV substances under REACH, we think the guidance should focus more on this type of human data.
SF / SE / Quality is important, but also context-dependent, i.e., observations from occupational settings may be very important even if not very scientifically robust. We would advocate a weight-of-evidence approach. Obviously, a high-quality study should carry more weight that a low-quality one. However, findings in studies of lower quality cannot be dismissed, but must be weighed-in in the overall assessment (using a weight-of-evidence approach), especially if the findings are alarming (e.g. serious health effect). Similarly, an effect found in an animal study cannot be dismissed only because it was not found in a human study of similar or higher quality. A plausible explanation to any alleged species difference must be given, in order to dismiss such an effect.
UB /DE1 / Concerning the different possible assessment factors, it is suggested to check thoroughly, whether their values are adequate, consistent with the relevant literature and also with other parts of chapter 1. In case that different opinions exist, they should be discussed (e.g. in a meeting or telephone conference). Furthermore, subdivision of the intra- and interspecies factor into a toxicokinetic and also a toxicodynamic part is strongly recommended, not only for for the part “DNEL/DMEL-derivation based on human data, but also in the whole chapter 8.
Concerning the adequacy of human data: more emphasis should be put on the question, whether and to which extent human health-related parameters are adequate as starting point for DNEL/DMEL derivation. The document focuses too much on epidemiological studies, however, as our own experience has demonstrated, more and more human data will be obtained from observational studies in the general population. This aspect is underrepresented in the document.
Further, more emphasis should also be put on the discussion of probable differences of modes of actions of effects in animals and humans.
R.8.1.2.8 / Human data as source for derivation of DNEL and/or DMEL
SB / UK HD / 1st para – Dose descriptors for non-threshold chemicals are the exposure levels associated with specific RR, SMR or SIR and not the other way around. This should be corrected throughout the text. For example this applies to the introduction of R.8.5.1.
UB / DE1 / 1st para, last line: “criterion” might be used instead of “criterium”
SB / UK DC / 2nd para - BMD 10 or its lower 95% confidence limit can also be used in risk assessment for threshold chemicals.
UB / DE3 / Concerning Comment No. 9 (previous comment)
We are of the opinion, that the comment has not been addressed adequately.
Reason for the comment was a first practical experience: industry has prepared a chemical safety report (CSR) of a substance, where a DNEL was derived based on human data. The DNEL was derived based on clinical-chemical blood parameters analysed in humans. However, industry did not state transparently how and to which extent these parameters were linked to adverse effects of the substance.
Based on these experiences, it is important, that the questions raised in comment No. 9 should be addressed in the guidance document on DNEL/DMEL derivation based on human data. As the example has demonstrated, it is not true, that “Large panels of lab parameters can normally not be collected from study subjects”. This is definitively not true. There are activities/organizations, where a large panel of different parameters are collected from workers or the general population (e.g. the center for disease control in the US). The main question is: whether and to which extent may parameters, that can be analyzed and investigated in humans (and which may differ considerably from parameters that can be investigated in laboratory animals) be linked to the adverse effect of the substance of interest and whether and to which extent it is justified to use the respective parameter(s) as a surrogate for the adverse effect.
ET / IT / A note related to the above comment 9 from UBDE: I agree with your response that is not possible to describe a set of parameters (even simple haematological/clinical chemistry parameters) to be considered/measured in human studies. However, I partly agree with the comments in the sense that the assessor should given his/her opinion on the relevance of the parameters/results described in the available studies, and state if they are suitable or not to define an appropriate NOAEL. Again, the expert judgement on the stage. I think that is important to stress the importance of the relevance of the studied parameters with respect to the end-point.
R.8.4.1 / Derive DNEL(s) for threshold endpoints, with adequate human data
b) Modification of the starting point to account for differences in exposure conditions
SB / UK HD / 1st para – Editorial: delete “exposure” (repeated twice).
Also replace “(e.g. corrected dose associated with a specific RR)”
SB / UK DC / Last para on p.5 - Where human exposure is evaluated on the basis of biological monitoring data, it may be necessary to take account of the time interval between exposure and measurement of the biomarker.
UB / DE1 / Ad 1. 2nd paragraph, sentence beginning with “In general, it is difficult to quantify differences in metabolism…” Disagree with this sentence, because in the past, a lot of efforts have been made in order to quantify interindividual human variability of metabolism/toxicokinetics, for example on the basis of interindividual variability of xenobiotic metabolizing enzymes (see publications of Renwick et al.). Therefore, it is not true that “in practice only differences between the different routes as determined by the percentages of absorption into the systemic circulation can be accounted for”.
Ad 1. Sentence starting with “Default absorption values have been proposed…”: this statement is not correct. Default absorption values have only been proposed for the dermal uptake rate (De Heer, C., Wilschut, A., Stevenson, H. & B.C. Hackkert: Guidance document on the estimation of dermal absorption according to a tiered approach: an update. TNO Report V98 1237, 27p, January 1999, Zeist, The Netherlands). In chapter R.7.12, however, no default absorption values are given for the oral, dermal or inhalation uptake pathway.
Ad 1. - 4th para: 2nd sentence: “Such information…”: not only chemical structure and physico-chemical properties of a substance but also available toxicity data can give information about absorption (see section R.7.12).
SB / UK HD / Ad.2 – last sentence – Which corrections are needed for continuous exposure and how to perform them? HSE recommends a factor of 5in EH75/2 to convert an 8-hr TWA OEL to a value that can be applied in continuous exposure situations such as divers living in a diving bell for several weeks at a time. This is based on the adjustment approach used by the Navy for submariners.
SB / UK DC / Ad.2 - para 1 - Even for toxic effects that are driven mainly by exposure concentration, the duration of exposure above a threshold concentration may determine the severity of the effect.
c) Assessment factors
SF / SE / The starting point for the AFs should be the ones agreed for animal data, and any deviations need to be discussed and explained in the guidance. It should be borne in mind that the default AFs were not eligible issues of discussion in the DNEL-group but were simply taken over from earlier agreed work in the revision of the old TGD. We can not agree with the present proposals, e.g., with respect to stating that kinetic differences do not need to be accounted for when starting from HD (p 8). All our previous comments on AFs still apply and we support those from the UK and DE.
SB / UK DC / I would question the features listed under the heading in the last bullet point (quality of available human database). The major factors to be considered when interpreting the available human database are bias (of which healthy worker effect is just one example), confounding, statistical power (which matters more than the size of the study as such), exposure-response relation, and consistency with evidence from other sources such as animal toxicology.
Intraspecies differences
SB / UK HD / 1st para – It is stated that provided the sample size is sufficiently large, human variability is already taken into account and no AF should be applied. How large is sufficient? This is too vague and rather unhelpful. We do not believe there are such studies where all the potential human variability of a population is accounted for by the study sample.
SB / UK DC / Although intra species differences may influence the overall findings of an epidemiological study, it is generally only the overall average risk that is assessed, and not the risk in an unusually susceptible subset of the population. Unless the susceptible subset can be identified (e.g. because they have a different genetic constitution or exposure to an environmental effect modifier), there is no way that the exposure-response relation for susceptible people can be characterised. And even where a susceptible sub-group can be identified, if they only constitute a small proportion of the total population, studies will normally lack the statistical power to characterise their risks. For these reasons, allowance for intra species differences in susceptibility should normally be made when conducting risk assessments based on human data.
UB / DE1 / 1st para. Further factors affecting e.g. toxicokinetics and metabolism such as medication and life-style factors (smoking, alcohol consumption) should be added.
1st para. Suggest adding (“gene-environmental interactions”) at the and second sentence (“These differences can be the result of genetic and/or environmental influences).
1stpara - Third sentence: what is a sufficiently large sample size ? More information should be given.
SB / UK HD / 2nd para – It is stated that kinetics can be considered comparable across populations. This may be true. However, the problem is that in any epidemiology study it is not an entire population that is investigated but small samples (and not always representative) of the population of interest. Therefore it cannot be assumed that an AF for kinetic intraspecies differences is always not needed. Also, we find no specific evidence in Barton et al (1998) to indicate that a kinetic intraspecies AF is not required when human data is the starting point for derivation of exposure standards.
SB / UK DC / 2ndpara - I do not understand what is meant by the second sentence. How can “data” be a “population”? Also, at the end of the paragraph, I would again contest the assertion that no additional assessment factor is required when a study is based on the general population.
SB / UK HD / 2ndpara - Later on, it is stated that the dynamic component should be taken into account provided differences in dynamics exist. This should be rephrased to indicate that the default assumption/position is that differences in dynamics exist and not that no differences in dynamics exist.
It is stated that in instances where the human data derive from populations without sensitive individuals, a large default AF is required. However, no value has been proposed. Furthermore, the authors suggest that the study size is an additional uncertainty for which a separate AF has been proposed in the guidance. This is therefore confusing. Should a default AF of 10 for a general population DNEL and of 5 for a worker DNEL (the defaults used for intraspecies differences when starting from animal data) be applied when the starting point has been identified from a volunteer study (tens of healthy subjects of similar age)? Or should it be 5 or 3 (intraspecies – see below for our proposals) x 3 (study size/power)?
Overall, the text concludes that where representative worker data are available a default AF of 3 should be applied for a general population DNEL and a default of 1 for a worker DNEL. First of all, we are of the opinion that a factor for dynamics only is not justified (see above). We propose that where representative worker data are available a reduced default intraspecies AF should be applied because some human variability in kinetics and dynamics is already accounted for. If the standard default is 10 (and 5 for workers), a pragmatic value when deriving a DNEL for the general population would be 5 and a pragmatic value for deriving a DNEL for workers would be 3. We disagree with the proposal that no AF is needed from a worker study to a worker DNEL and from a general population study to a general population DNEL, as the variability of the target population is much larger than that of the study sample. In both instances an AF of 3 should be applied. Then, for a small study with low power, on top of these factors, the additional default of 3 should be considered.
Also, some indication should be given in the text of what a sufficiently large study is. Are we talking about tens, hundreds or thousands of individuals? Or what?
UB / DE1 / 2ndpara (p.8): disagree with the statement “If human data are used as both the source and target population, then the pharmacokinetics (metabolism) can be considered comparable across populations”. Reason: There can be differences in pharmacokinetics (metabolism) in case that the source population comprises (healthy) workers and the target population comprises humans including probably sensitive target populations such as very young children, elderly people and sick persons (diabetics, people with kidney diseases)., The ontogenicity of drug metabolizing enzymes as well as the disease state may have substantial influence on metabolism and toxicokinetics (e.g. Hines, R.N. (2008) The ontogenicity of drug metabolizing enzymes and implications for adverse drug events. Pharmacology and Therapeutics 118, 250-267).
Second sentence: Account must therefore be taken of the pharmacokinetic component; a default of 3.16 is usually taken for interindividual variability of pharmaco- (or better: toxico-) dynamics in case that no data-derived (i.e. substance-specific) values have been obtained.
The sentences mentioning sensitive individuals are in contradiction to the second sentence of this paragraph. As suggested above, sensitive target populations should be mentioned in the second sentence.
SB / UK HD / 3rd para – In the first sentence please add that these are effects which depend on direct reactivity at the initial site of contact.
Regarding acute CNS effects, we are not aware of conclusive evidence showing that kinetic considerations do not apply.
We disagree that a factor for dynamics is not needed here. We would suggest that the defaults proposed above for systemic effects of 5 and 3 are reduced to 2.5 and 1.5 for local effects.
General - The text as it stands does notaddress the issuethatcertain genetic polymorphismshave been foundtobe associated with certain ethnic groups conferring particular sensitivity to particular chemicals on that group.We think the default assumption should be that different ethnicgroups will react the same unless there aregood data to demonstrate otherwise but we thinkthe issue should be raised here (we note it is raised on p19).
SB / UK DC / Paragraph 3 - Again, I disagree with the last sentence. Limitations of statistical power mean that one can rarely be confident that an apparent NOAEL in an epidemiological study would apply to the most sensitive subset of the population.