Peer Review Comments on the IRIS Assessment of Benzene
Response to the Peer Review, January 19, 2000
Reference: Extrapolation of the Benzene Inhalation Unit: Risk Estimate to the Oral Route of Exposure (EPA/NCEA-W-0517, July 1999)
Four external expert reviewers were requested to review the document referred to above dealing with the extrapolation of the inhalation unit risk to that of the oral unit risk of cancer from exposure to benzene. The external peer reviewers who provided comments on this document are as follows:
- Patrick W. Beatty, Chevron Research and Technology Company, Richland, CA
- David Eastmond, University of California - Riverside, CA
- David Ross, University of Colorado Health Sciences Center, Denver, CO
- Robert Snyder, Rutgers University, Piscataway, NJ
On May 12, 1999, the U.S. Environmental Protection Agency's Office of Research and Development announced in the Federal Register (Fed. Reg. 64(91):25502) the availability of a draft document entitled Extrapolation of the Benzene Inhalation Unit Risk Estimate to the Oral Route of Exposure (EPA/NCEA-W-1999, April 1999) for a 30-day public review and comment period. Only two comments were received. Public comments were received from:
- Consultants in Toxicology, Risk Assessment and Product Safety, Washington, DC
- American Petroleum Institute, Washington, DC
Overall the comments received from the peer reviewers and the public were supportive of the Agency's position, while public comments out about issues addressed in the document were well thought out. Most commenters complimented the Agency for doing a good job in covering the issues that mattered. The following text consists of a compendium of the main comments from all sources followed by appropriate responses discussing their disposition by the Agency.
Comment: The document says that the oral unit risk is a risk factor for ingestion and not a unit risk number for drinking water. This seems inconsistent. It was pointed out in the background document, page 11, line 17, that "For development of a drinking water safe concentration, the risk due to inhalation of volatilized drinking water and risk due to dermal uptake must be added to the ingestion risk."
Response: The document refers to the oral unit risk as a risk factor for ingested benzene. Benzene ingestion risk is assumed to be the same as the drinking water risk. As pointed out in the document in Section 5.1, page 11, EPA agrees with the comment, but the development of a corrected intake factor to account for total exposure to drinking water is beyond the scope of this report. While the Agency agrees with the comment on dermal uptake, a review of literature on dermal absorption of benzene (Section 3) indicates that dermal absorption is minimal (less than 1%), and is, therefore, not likely to have contributed significantly to the overall oral risk estimation.
Comment: A key reference was left out: Matthews et al. (1998) Dose-dependent metabolism of benzene in hamsters, rats and mice. Toxicol Sci 44:14-21.
Response: This reference has been added to the bibliography and has been addressed in Section 2, line 36.
Comment: The extrapolation from inhalation to oral risk is based upon physiological values that have been historically used by the agency such as 20 m3/day, 70 kg person and 2 L/day. It might be more accurate to use updated values from the U.S. EPA's Exposure Factors Handbook for drinking water consumption, inhalation rates, etc.
Response: The Agency chose to use the values recommended by the Office of Drinking Water, EPA. However, as pointed out by the commenter, the Exposure Factors Handbook provides alternative physiological values that can be substituted for the ones referred to above. However, calculations using variations of the different values suggested results that are little changed from the estimates used in this document.
Comment: This document refers to its content as being the "current state of the art" in dose extrapolation.
Response: The phraseology "current state of the art" has been deleted.
Comment: The discussion does not take into consideration the differences that would occur due to first-pass effects that occur after oral intake but not after inhalation.
Response: The revised draft considered the differences due to first-pass effects; however, their influence on the subsequent 50% absorption factor and the development of the unit risk are considered to be minimal.
Comment: Several comments were made that relate to metabolism and pharmacokinetic information and suggest that appropriate metabolism data and pharmacokinetic modeling can provide a basis to extrapolate between routes of exposure.
Response: While the Agency agrees with the commenters, the current knowledge of the metabolism of benzene and the physiologically based pharmacokinetic (PBPK) models proposed may provide insights (as discussed in the revised draft) as to the putative toxic metabolites and potential biochemical mechanisms, but they are not sufficiently developed and validated to predict the acute myeloid leukemia (AML) risk to humans from exposure to benzene.
Comment: The same pattern of metabolites should occur via the oral route as via the inhalation route. The liver is the dominant site of metabolism either by inhalation or by ingestion. Therefore, benzene should have the same qualitative toxicity in the same species by either route. Consequently, EPA cannot attribute a risk of acute myeloid leukemia (AML) to oral human benzene exposure, because of a lack of direct data about orally exposed humans.
Response: EPA agrees with the comment that the same pattern of metabolic activity should occur by either route of exposure. Although the liver is the primary site of first-order metabolism, there is evidence that second-order metabolism may take place in bone marrow. But the difference between the oral versus the inhalation route in terms of metabolite formation involves the quantity of benzene actually absorbed and how much actually reaches the liver versus how much is expelled unchanged. Since metabolic profile and potential mode of action at the cellular level are the same by either route, there would be no reason to assume that the risk of AML differs by anything other than the dose that reaches the target organs.
Comment: The EPA also strangely states the "...the kinetics of metabolite formation and clearance after inhalation and ingestion of benzene are not known for humans." However, the human inhalation data cited in Section 3.2 are sufficient for EPA to calculate the maximum velocity, the dose of half-maximum velocity, and the uncertainties in both estimates for human benzene metabolism.
Response: The Agency does not find this comment "strange." The information regarding the kinetics of metabolite formation and clearance remains theoretical. Despite this, it is still possible for the Agency to make predictions about what the unit risk is likely to be, given the knowledge that is available and making minimal assumptions. If the words "maximum velocity" and "half-maximum velocity" mean the 100% absorption factor and 50% absorption factor, respectively, then the answer is yes. It is sufficient to calculate unit risks from such data.
Comment: EPA should relate the risk of AML to the total dose of metabolites after different benzene exposures. By the oral route, most absorbed benzene will undergo first-pass metabolism in the liver, whereas by the inhalation route only approximately 20% of absorbed benzene will pass through the liver each time the blood recirculates. The circulation of blood is independent of benzene until exposures reach very high (anesthetic) levels. Thus, appropriate pharmacokinetic modeling can provide a basis to extrapolate between routes and a better basis for the Agency's route-to-route extrapolation.
Response: The Agency does not disagree with the idea that pharmacokinetic modeling can provide a more accurate basis upon which to decide route-to-route extrapolation. In the absence of an appropriate and validated model, EPA assumed that the amount of inhaled benzene absorbed is 50% on the basis of a number of studies described in the background document where the levels of exposure are similar to or not much higher than those found in the ambient environment. The Agency believes that this is a reasonable assumption to use in the absence of an appropriate and validated PBPK model.
Comment: EPA's conclusions in the draft document support absorption as the rate-limiting factor in potency between the oral and inhalation routes. This conclusion only appears to contradict EPA's previous finding that the metabolites of benzene (not benzene itself) are responsible for human AML. In estimating the difference in absorption between routes, EPA relies on data from Sabourin and co-workers (1987). Yet the same data show that production of benzene metabolites is a nonlinear function of the dose absorbed. A more persuasive approach would relate AML (or animal risks) to total metabolites and show how absorption is the rate-limiting factor in the production of metabolites, a point that the draft document does not address directly.
Response: The Agency believes that absorption is possibly just one of several rate-limiting factors in calculating potency from the inhalation to the oral route, not necessarily the only one. This does not contradict a purported position that the commenter ascribes to the EPA that the metabolites minus benzene are responsible for human AML. The Agency concluded that it is possible that any one, or all of, or any combination of benzene and its metabolites may be responsible for AML and other blood disorders. The Agency did not depend upon only the Sabourin and co-workers study (1987) to support its choice of 50% as the absorption factor in the extrapolation. In Table 2 of the document will be found several human epidemiological studies that suggest that 50% is a reasonable value around which the other calculated absorption rates seem to hover. These same studies do not offer evidence that the absorption rate increases or decreases substantially with increasing or decreasing exposure concentrations.
Comment: The commenter disagrees with EPA's statement in Section 5.1 that absorption of benzene might remain constant as a function of exposure. The human data in Section 3.1 of the draft document rule out an independence of absorption on exposure. However, EPA's choice of a constant ratio of oral to inhalation absorption, even if not precisely correct, could prove a practical approximation for benzene exposures that elicit observable toxicity.
Response: The statement above attributed to the Agency in Section 5.1 that "absorption of benzene might remain constant as a function of exposure" cannot be found in Section 5.1. The assertion that the "human Section 3.1 rules out independence of absorption on exposure" is a misinterpretation of the content of Section 3.1, which deals with rats and mice. However, the commenter then chooses to agree with the choice of a constant absorption ratio because it is a "practical approximation for benzene exposures that elicit observable toxicity." This is a contradiction from the premise stated at the beginning of this comment.
Comment: As the draft document notes, no currently acceptable models of benzene metabolism integrate the two different routes to estimate a dose of metabolites to the bone marrow. However, several scientists have modeled dose to the bone marrow, and the final document should describe them. The lack of an accepted model also supports the use of a practical approximation for route-to-route extrapolation. However, EPA should characterize the approximation as "interim" or "provisional" pending the development and validation of an accepted pharmacokinetic model, with the implication that the Agency will change its practice when such a model becomes available.
Response: See earlier response on the use of pharmacokinetic modeling.
Comment: The available epidemiological data show that the human risk of AML is not a linear function of benzene doses. Mode-of-action data support these observational human data. The existing animal data are not appropriate to estimate the shape of the dose-response curve because animals do not exhibit AML after benzene exposure and because the animal bioassays were designed to elicit qualitative indications of toxicity, not quantitative dose-response relationships. Thus, the fatal error in the draft document is the conversion of an incorrect inhalation potency to an incorrect oral potency.
Response: The Agency disagrees with the commenter concerning the shape of the dose-response curve. There is insufficient evidence at this time to determine the shape of the dose-response curve. Even the mode-of-action information that is available now cannot guide scientists in distinguishing how the metabolites of benzene and benzene itself interact to cause AML and other forms of leukemia and myelodysplastic disease. The shape might be supralinear, sublinear, or just linear. Estimates of exposure to humans that are available depend to a great extent on assumptions that can't be validated and that vary considerably. Animal data were not used to develop any exposure scenarios, but yes, contrary to the commenter's statement, animal studies by Cronkite and Snyder have produced cancers of the hematopoietic system in animals as referred in the U.S. EPA (1998).
Comment: EPA should withdraw the current draft document and revise it (1) to relate toxicity via different routes to the total dose of metabolites, (2) to account for nonlinear potency as a function of benzene doses by both routes, and (3) to make the revised draft document available for additional comments.
Response: EPA disagrees with the first two comments for the reasons stated above. With respect to the third comment, an ample period of time has been provided for potential commenters to send their comments to the Agency. Agency policy is to seek additional comments if substantial changes are made in the revised final document. The revised final document addressed the final comments of external reviewers and the public that were scientifically relevant. These did not change the overall discussion or conclusions; therefore, the document would not be available for additional comments.
Response to the Peer Review, October 14, 1998
Reference: Carcinogenic Effects of Benzene: An Update (EPA/600/P-97/001A, June 1997)
On June 30, 1997, the National Center of Environmental Assessment (NCEA), within the U.S. Environmental Protection Agency's Office of Research and Development, announced in the Federal Register (Fed. Reg. 62 (125): 35172-35173, 1997) the availability of a draft document entitled Carcinogenic Effects of Benzene: An Update (EPA/600/P-97/001A, June 1997) for a 60-day public review and comment period. The comment period closed on August 29, 1997.
The FR notice also announced that Eastern Research Group, Inc. (ERG), an EPA contractor, would conduct a peer-review workshop on July 16, 1997, to review the draft document. An expert panel, consisting of six reviewers, was charged to review pertinent literature on the mode of action and carcinogenic effects of benzene. The panel was asked to determine whether new information that has become available since the 1985 interim quantitative cancer risk estimates due to inhalation of benzene would substantially modify the estimates of cancer risk contained in the earlier document. The panel members and their affiliations were as follows:
- David Eastmond - University of California, Riverside
- David Hoel - Medical University of South Carolina (chair)
- George Kalf - Jefferson Medical College, Thomas Jefferson University
- Michele Medinsky - Chemical Industry Institute of Toxicology
- Robert Schnatter - Exxon Biomedical Sciences, Inc.
- Lauren Zeise - California Environmental Protection Agency
Attendees at the workshop included representatives from industry, trade organizations, and EPA and other interested Federal agencies.
The summary report of the workshop entitled Report on the Expert Panel Peer Review of Benzene Risk Assessment Update (September 23, 1997) is available. The report included questions posed by EPA about the draft document and answers provided by the Peer Review Panel. The view of the panel members with respect to these questions is the basis of the summary report.
The following summarizes the comments received from the expert panel, and provides EPA's responses as reflected in the final document, Carcinogenic Effects of Benzene: An Update (EPA/600/P-97/001A, June 1997).
The peer-review panel complimented the Agency for the review draft document (EPA/600/P-97/001A dated June, 1997) and felt that the Agency did a very good job in covering the large body of scientific literature on the carcinogenic effects of benzene. The panel generally agreed that the abbreviated format and the focused discussion of the report provided a good overview of the current understanding of benzene's mode of action as well as the human risks associated with exposures to benzene. The panel further noted that the Agency focused on critical issues involved in developing a new potency estimate and has made an attempt to incorporate methodology consistent with the newly proposed cancer risk assessment guidelines.
The panel noted that the primary focus of the document should be on an assessment of the risk of benzene at environmental exposure levels and that the intent of the report should be more clearly stated. The panel questioned why the EPA's ongoing efforts to develop a non-cancer risk assessment of benzene are not being integrated into the report. The panel noted that a number of pertinent recent references had been omitted and should be added. The panel raised concern about the citation of unpublished information. The recommendation was to delete the citations that are not peer-reviewed. The panel was in agreement that many published epidemiologic studies indicate that benzene does not pose an increased risk of multiple myeloma and therefore the discussion should reflect this fact. The panel questioned the scientific evidence for benzene as a risk factor for leukemia cell types other than acute myeloid leukemia (AML) and recommended that the Agency should reflect this uncertainty. There was also substantial disagreement among the panelists regarding the use of non-linear versus linear extrapolation models to estimate cancer risk to humans at environmental levels. The panel agreed with EPA that the Pliofilm workers cohort (Rinsky et al., (1981, 1987), is the best available cohort for dose-response analysis and derivation of unit risk estimates for exposure to benzene. The panel also noted a recent publication containing dose-response information from the Chinese cohort (Hayes et al., 1997) and recommended discussing its use prior to releasing the final risk analysis. In addition, the panel raised the issue of the use exposure estimates of Rinsky et al (1981) and recommended that consideration be given to the use of the more recent exposure estimates of Paustenbach et al. (1992, 1993) rather than the Crump and Allen (1984) exposure estimates when calculating the unit risk. However, some panel members subsequently determined that these estimates were biased on the high side, while others disagreed.
The panel noted that ANLL provides the strongest basis for dose-response analysis and to characterize uncertainty, the potency estimated should be supplemented by those for all leukemia. In general, the discussion of the various modes of action of benzene was supported by the panel, but recommended that the Agency identify areas of significant controversies. There was agreement that the use of biologically-based modeling to estimate leukemia risk was premature. The panel members agreed that there were still considerable uncertainties and complexities regarding the mode of action of benzene. In general, the panel members felt that the document fairly presented the two contrasting perspectives that there is evidence that the leukemogenic effects of benzene exhibit a linear (non-threshold) as well as non-linear (threshold) response, and the Agency's decision to present both approaches and selection of linear model is consistent with the spirit of the proposed cancer risk assessment guidelines.
There was agreement among panel members that more scientific research is necessary to establish the shape of the dose-response curve, particularly, in the moderate to low-dose range.
The panel recommended that the Agency identify sensitive groups in the general population who may have differential exposure and differential risk. The panel complimented the Agency on the work it has done in bringing scientific considerations to the process of quantitative risk estimation.
With regard to the panel members' recommendations for integrating the ongoing non-cancer risk assessment on benzene, the Agency intends in the near future to develop and integrate cancer with noncancer risk assessment and submit it to the Agency's Integrated Risk Information System (IRIS) process (Fed. Reg. 63 (1): 75-77, January 2, 1998).
The Agency agrees with the peer reviewers that all citations of unpublished literature should be eliminated and has done so. The peer reviewers have also recommended that several additional citations should be added. The Agency has included and discussed several additional studies.
The peer reviewers also noted that more recent epidemiological data indicates that multiple myeloma is no longer associated with exposure to benzene. Although multiple myeloma in earlier studies (Decoufle et al., 1983; Rinsky et al., 1981, 1987) was suggested as a consequence of exposure to benzene, this finding has not been substantiated in more recent epidemiological studies. The Agency agreed with the commenters and has changed the text to reflect this fact.
Some members of the panel questioned the scientific evidence for benzene as a risk factor for leukemia cell types other than AML. The evidence for assuming that other leukemia cell types are not associated with exposure to benzene is insufficient. The exact underlying mechanism(s) of leukemia induction from exposure to benzene is (are) not fully known (Smith and Fanning, 1997). There appears to be enough evidence in the literature to suggest that there may be more than one mechanism involved. Furthermore, the epidemiological evidence to date does not entirely support the concept that only the risk of AML is increased. Savitz and Andrews (1996) after reanalyzing the Rinsky et al (1981,1987) cohort by excluding AML found that the recalculated relative risk for the remaining aggregated three leukemic cell types was still elevated. On the other hand, recent studies (Hayes et al., 1997, Yin et al., 1996) have implicated exposure to benzene as responsible for an increase in the risk of non-Hodgkin's lymphoma as well as that of myelodysplastic syndrome and hematologic neoplasms. The Agency also does not believe that these data should be factored in at this time for derivation of the risk of exposure to benzene. Potential weaknesses in the Chinese cohort, as have been pointed out in the document, preclude its use in risk estimation.
There was substantial disagreement among the panelists regarding the use by the Agency of non-linear extrapolation models versus linear extrapolation models to estimate cancer risk to humans at environmental levels. The Agency has found that a current scientific consensus does not exist to support a decision to exclude either linearity or nonlinearity. The Agency has evaluated the evidence and concluded from published studies (Smith, 1996: Smith and Fanning, 1997) that the underlying mechanism(s) of leukemogenesis are too controversial to permit predicting the shape of the dose-response curve. Therefore, the Agency chose to use a linear response model as a default approach because no new data is available that will allow the Agency to proceed with a risk characterization using available exposure information.
The panel members agreed with the Agency that the best available cohort is still the Pliofilm workers (Rinsky et al. 1981, 1987) and it should be used to develop a dose-response analysis and derivation of unit risk estimates for exposure to benzene.. The panel members did feel that the Agency needed to better justify its use. The Agency agreed with the commenters and provided a better justification.
There was some disagreement amongst the panel members concerning which exposure estimates were the best to use for unit risk estimation. Some panel members preferred the exposure estimates of Rinsky et al., (1981), while others wanted the Agency to use Paustenbach's much higher exposure estimates (1992,1993. The Agency continued to rely on the exposure estimates of Crump and Allen, 1984 because they appear to be the least biased.
With respect to the recently published Chinese cohort, the Agency Reviewed the recent publications and has discussed these new findings in detail in the final document. Among cancer or related conditions that were found by the authors to be associated with exposure to benzene are non-Hodgkin's lymphoma, hematologic neoplasms and myelodysplastic syndromes as well as acute nonlymphocytic leukemia. Several of these conditions were found to be significantly elevated at 10 ppm-years. Several potential methodological problems exist that need to be resolved prior to its use in derivation of cancer unit risk estimates as discussed in the document. The Agency has provided in the text a detailed discussion of this study including its strengths and weaknesses.