TLV Recommendation
Occupational benzene exposure is an established human leukemogen;(137,thus, assignment of the A1 (confirmed human carcinogen) category is appropriate. For most benzene cohorts (i.e., refinery employees, gasoline truck drivers, retail motor vehicle filling station attendants, leather and other workers), concomitant exposures to complex mixtures or other volatile solvents and other chemicals confound the determination of the carcinogenic potency of benzene. It is the judgement of the TLV Committee that benzene exposure estimates from the Pliofilm cohort were not confounded by concomitant exposures to other chemicals. However, as with all retrospective epidemiologic studies where personal sampling was not carried out, dosimetry is "admittedly incomplete."(137) No exposure reconstructions and calculation of dose from these exposures that occurred 50 years ago for an operation that does not and would not be permitted today can be considered precise; thus, criticism of methods, assumptions, and conclusions are inevitable in any reconstruction of past exposures.
The exposure estimates for the Pliofilm employees are a matter of intense controversy, but Goodyear corporate log books and industrial hygiene records mention two issues: 1) "when exposure is above 35 ppm, even momentarily, employees are required to wear a chemical cartridge respirator" and 2) "the following levels of benzene were determined: 4 ppm, on elevated platform; 7 ppm, over the filter; 8 ppm, general area, nearfilters.,,(136) These log books are the only empiric data on workplace air benzene concentrations and represent area sampling; as such, it is not possible to link the presence or absence of any individual worker's disease directly to a particular dose. In addition, respirators, if used, would lower the dose received. Therefore, the estimates of leukemogenic risk posed by occupational exposure to benzene for the Pliofilm cohort carry a large uncertainty.
The recommended TLV-TWA for benzene of 0.5 ppm with a 2.5 ppm STEL is based on the TLV Committee's interpretation of three papers authored by Paxton et al.,(194) Crump,(213) and Schnatter et al.(208) All three of these studies are analyses of the NIOSH Pliofilm Cohort. These analyses have considered three exposure matrices (Rinsky et al.,(137) Crump and Allen,<188) and Pausten-bach et al.(195)) and three statistical models (Lifetable Analysis, Proportional Hazards Regression, and Maximum Likelihood Method). In addition to deaths from all leukemias, two of these analyses (Crump(213) and Schnatter etal.(208)) considered the subset of acute myelogenous leukemia (AMML).
The analyses that have used cumulative exposure (ppm-years) as the dose metric (Paxton et al.(193,194) and the Linear Model in Crump(213)) produce results which are well understood and readily translated to a TLV-TWA. A further strength of these models is that they produce results which are consistent regardless of choice of exposure matrix. However, the cumulative exposure approach may not accurately reflect the myelotoxic process by which benzene affects bone marrow cell proliferation (see the Pharmacokinetic/Metabolism Section above).
Alternatively, two of the statistical models (Schnatter*208* and the Intensity-dependent model of Crump(213,214)) are based on the concept that dose rate (measured as ppm) is more important than cumulative dose, and results from those studies are interpreted by the TLV Committee as consistent with a TLV-STEL. This latter approach is consistent with the evolving metabolic and molecular mechanistic theory of benzene leukemo-genesis proposed by Irons and Stillman<185) and Ross.(84) However, these statistical models are sensitive to choice of exposure matrix, possibly because the industrial hygiene monitoring results used to create the exposure matrices may not document peak exposures accurately.
Table 1 (reproduced from Paxton et al.<193)) provides the estimates of SMR by categories of cumulative exposure based on three sets of exposure estimates (Rinsky et al.,(137) Crump and Allen,(188) and Paustenbach et al.,(195) respectively). The results exhibit a dose-response relationship between the cumulative benzene exposure (ppm-years) and the corresponding SMRs which is reasonably consistent for all three exposure estimates, implying that the statistical method is robust. The interpretation of these results by the TLV Committee suggests that exposure for a working lifetime at the current OSHA PEL of 1 ppm, equivalent by this metric to 45 ppm years, results in an unacceptable risk of leukemia. However, the categories of cumulative exposure given in Table 1(193) do not allow for a more refined estimate of the dose-response relationships at exposures less than 45 ppm-years.
Paxton et al.(194) attempted to refine their risk estimate by the use of proportional hazards modeling, a statistical method which can take into account time-dependent covariates, but the authors failed to correct for the back¬ground levels of leukemia in the United States (7.1 deaths per thousand). This omission was detected by Crump(214) in an unpublished presentation to the TLV Committee; the nature of this error was to underestimate the risk of leukemia throughout the observed range of benzene exposure. The corrected proportional hazards model pro¬vided by Crump (utilizing exposure estimates only from Paustenbach(195)) predicted a relative risk of approxi¬mately 1.5 (Linear Quadratic Model) for exposure at 45 ppm-years, which is more consistent with the SMR re¬sults (Table 1) presented by Paxton et al.(193) The proportional hazards model also allows for more refined estimates of the exposure-response relationship.
Crump published a more sophisticated analyses of the same data set,(213> involving a maximum likelihood approach. The most important results predicting excess deaths are given in Table 2 (reproduced from Crump<213)). Crump(213) presents statistical arguments favoring intensity-dependent models over cumulative exposure models. The results of statistical analyses presented by Crump(213) (Table 2) were generally comparable across exposure matrices and were comparable to the life table results, but interpretation of the Crump results are dependent upon the incomplete exposure records for the cohort.
There are consistent epidemiologic results from three separate types of statistical analysis of the 1987 update of the data from the Pliofilm workers cohort: life table analyses (Paxton et al.(193) [Table 1] and Crump(213) [Table 2]), proportional hazards modeling (Crump(214) [Ta¬ble 2]), and the linear model results from the maximum likelihood procedure developed by Crump{213) (Table 2). These results imply that exposure to benzene at a level of 1 ppm or greater over a working lifetime results in an excess leukemia mortality.
These same epidemiologic data were analyzed by Schnatter et al.<208) using exposure data from Rinsky et al.,(136) Crump and Alien,(188) Paustenbach et al.(195) and dose-rate metric rather than cumulative dose. Schnatter and coworkers(208) found that the risk of benzene-induced leukemia was no greater than 1.0 at peak concentrations always equal to or less than 1, 5,10, 15, or 20 ppm, but leukemia risk increased progressively at higher peak concentrations.
The results from studies of the Pliofilm cohort which use a ppm-years dose metric(208) suggest that, at a TWA of 0.5 ppm, the odds of death from leukemia due to occupational benzene exposure would be indistinguishable from the odds of death from leukemia for a worker who is not exposed to benzene. It is these analyses and interpretation of the Pliofilm cohort of benzene exposure and deaths from leukemia that provide the basis for a TLV-TWA of 0.5 ppm benzene. The Pliofilm cohort has the advantage that benzene exposure occurred in isolation, thus yielding a risk estimate for benzene exposure alone. This fact has an advantage for unequivocal des¬ignation of occupational benzene exposure as a known human carcinogen since these exposures were not confounded by concomitant exposures to other chemicals in the workplace.
A STEL of 2.5 ppm is recommended to protect against excess risk of leukemia due to the dose-rate dependent hematopoietic toxicity of benzene.(208) The hypothesis supporting an approach to control peak expo¬sures suggests that bone marrow toxicity occurs only after the critical delivered (threshold) dose to target hematopoietic progenitor cells is exceeded. Table 3 pre¬sents the results of the Pliofilm dose-rate-based anaylses<208) for all leukemias using the ppm-dose metric [to compare to the results of Paxton et al.(193) and Cmmp,{213,214) who used the ppm-year dose metric]. Results for both total leukemias and for AMML imply that measurable leukemogenic risks exist for the Pliofilm co¬hort at peak benzene concentrations of 20 to 25 ppm or higher. While percutaneous absorption of liquid benzene through intact human skin can be limited (e.g., 0.05% of the applied dose),(103) the absorbed dose via direct dermal contact combined with that received from body surface
exposure to benzene in workplace air is such that a substantial fraction (20%-40%)(105) of the total exposure is due to skin absorption; thus, the skin designation is assigned to the benzene TLV.