Regulations (Preambles to Final Rules) - Table of Contents|
| Record Type:||Occupational Exposure to Cadmium|
| Title:||Section 7 - VII. Significance of Risk|
VII. Significance of Risk
In the 1980 benzene decision, the Supreme Court, in its discussion of the level of risk that Congress authorized OSHA to regulate, indicated when a reasonable person might consider a risk significant and take steps to decrease it. The court stated:
It is the Agency's responsibility to determine in the first instance what it considers to be a "significant" risk. Some risks are plainly acceptable and others are plainly unacceptable. If, for example, the odds are one in a billion that a person will die from cancer by taking a drink of chlorinated water, the risk clearly could not be considered significant. On the other hand, if the odds are one in a thousand that regular inhalation of gasoline vapors that are 2% benzene will be fatal a reasonable person might well consider the risk significant and take the appropriate steps to decrease or eliminate it. (I.U.D. v. A.P.I.., 448 U.S. et 655).
The Court further stated that "while the Agency must support its findings that a certain level of risk exists with substantial evidence, we recognize that its determination that a particular level of risk is significant will be based largely on policy considerations." The Court added that the significant risk determination required by the OSH Act is "not a mathematical straitjacket," and that "OSHA is not required to support its findings with anything approaching scientific certainty." The Court ruled that "a reviewing court [is] to give OSHA some leeway where its findings must be made on the frontiers of scientific knowledge [and that] ... the Agency is free to use conservative assumptions in interpreting the data with respect to carcinogens, risking error on the side of overprotection rather than underprotection" (448 U.S. at 655, 656).
As part of its overall significant risk determination, OSHA considers a number of factors. These include the type of risk presented, the quality of the underlying data, the reasonableness of the risk assessments, the statistical significance of the findings and the significance of risk (48 FR 1864; January 14, 1983).
Cadmium exposure causes a number of extremely serious adverse health effects. In 1971 OSHA adopted the ANSI standard with a TWA PEL of 100 ug/m(3) for cadmium fume and a TWA PEL of 200 ug/m(3) for cadmium dust to prevent the acute effects caused by exposure to cadmium at levels higher than the PELs. Since 1971, however, a body of evidence has developed which shows that exposure to cadmium, dust or fumes, at levels well below these PELs can also lead to very serious health effects such as kidney dysfunction and cancer. Because current occupational cadmium exposure levels generally are below 100 ug/m(3), the discussion of the significance of risk not does emphasize acute health effects, but rather focuses exclusively on two of the most common chronic adverse health effects related to cadmium exposure.
As indicated in the health effects section of this preamble, exposure to cadmium causes cancer, kidney dysfunction, reduced pulmonary function, and chronic lung disease indicative of emphysema. Other health effects, such as improper bone mineralization also have been reported. In addition to these major effects in humans, studies of experimental animals suggest that exposure to cadmium may also cause anemia, change in liver morphology, decrease in immunosuppression, and hypertension.
As discussed in the health effects section, there are numerous epidemiologic studies that show an elevated risk of lung cancer among cadmium exposed workers. Because lung cancer is almost certainly fatal, OSHA considers this disease to represent the greatest material impairment to health. A number of studies of workers also suggest an association between occupational cadmium exposures and increased deaths from other types of cancer, most notably prostate cancer. However, the relationship between cadmium exposure and prostate cancer is difficult to establish on the basis of more recent mortality studies. Most epidemiological cohort studies of workers use mortality rates to estimate risk of disease, but prostate cancer does not always lead to death. Consequently, the mortality studies probably underestimate the true incidence of the disease. In any event, although prostate cancer is not always fatal, OSHA nonetheless considers it to be a very serious material impairment to health.
Chronic exposure to cadmium is also known to cause renal dysfunction. This impairment of kidney function typically is manifested as proteinuria, a condition characterized by an excess of proteins in the urine. Early stage, cadmium induced proteinuria typically is tubular proteinuria, which is characterized by an excess of low molecular weight proteins in the urine. Chronic exposure to cadmium may also cause glomular proteinuria, a still more serious dysfunction, characterized by an excess of total proteins in the urine. The damage to the proximal tubules or glomerulus in the kidney indicated by proteinuria is likely to be irreversible in a substantial proportion of workers, except in its very earliest stages.
Because of the body's ability to accumulate and store cadmium over long periods of time, the loss of kidney function may develop even after a reduction or cessation of external cadmium exposure. Upon prolonged exposure, tubular proteinuria may progress to more severe forms of renal dysfunction such as glycosuria, aminoaciduria, phosphaturia and glomular proteinuria. Therefore, OSHA also considers tubular proteinuria to be a material impairment of health. As discussed in the health effects section of this preamble, this conclusion is consistent with OSHA's analyses and court decisions regarding the lead and air contaminants standards [FR 52952, 11/14/78, p. 52963; USWA v. Marshal 647 F. 2d 1189 (1980), p. 1251; 54 FR 2332-2983, 01/19/89; AFL-CIO v. OSHA, Nos. 89-7185 et al., 911th Cir. 7/7/92)].
Long term exposure to cadmium appears to cause other adverse effects on the respiratory system in addition to lung cancer. Workers with prolonged exposure to cadmium dust or fumes have exhibited shortness of breath, impaired pulmonary function, and chronic lung disease indicative of emphysema. These diseases also constitute material impairment of health or functional capacity, but it has not been possible to determine a dose response relationship between them and occupational exposure to cadmium. Therefore, these diseases were not quantified in the quantitative risk assessment section of the standard.
Workers with progressive forms of proteinuria also have exhibited adverse bone effects associated with improper bone mineralization, such as osteoporosis and osteomalacia. These latter diseases are also serious, though not usually fatal. THese diseases also are not quantifiable with the data available, but they are likely to occur as a result of exposure levels above the old permissible limits. The discussion of significant risk concentrates on the quantifiable diseases, cancer and kidney dysfunction. OSHA concludes that the risk of contracting each of these diseases from occupational exposure to cadmium above the new PEL is significant. The other disease risks mentioned above, though not as readily quantifiable, add to the significance of the risk presented.
In the health effects section above, OSHA discusses at length its assessments of the various relevant animal and human studies, and in the quantitative risk assessment section above, OSHA discusses in great detail its own and other's risk assessments for cadmium, including the bases and criticisms of those assessments.
The underlying epidemiologic and experimental animal studies that provide the basis for this quantification of risk are of reasonable quality and demonstrate a relationship between cadmium exposure, on the one hand, and cancer and kidney dysfunction, on the other. There is a reasonable basis for determining the exposed population, estimating dose, and excluding other potentially causal agents of the observed diseases. The Environmental Protection Agency (EPA) has concluded that the available data are adequate to quantify the risk of cancer due to cadmium exposure. This is OSHA's conclusion as well.
OSHA used two types of data for its quantitative assessment of the risk of death from cancer. One is animal data, the rat bioassay by Takenaka and others (Ex. 4-67) and the long term bioassay by Oldiges and others (Exs. 12-10-i; 12-10-h; 12-35; 8-694). The other is human data, the human mortality study by Thun and others (Ex. 4-68). In its proposal for its preliminary quantitative risk assessment, the Agency relied on the rat data for its best estimate of total risk of cancer death, because OSHA believed that the measures of exposure were more accurate in the rat study and because the rat study can be used to predict all cancer deaths attributable to occupational exposure to cadmium. By contrast, the Thun data can be used to predict only lung cancer deaths attributable to occupational exposure to cadmium. This use of animal data to predict total cancer deaths is consistent with risk assessments conducted for other standards and upheld in the Courts (e.g. ethylene oxide).
Nevertheless, in this final standard OSHA relies on both the animal data, especially the Takenaka study, and the Thun human mortality study for its estimate of excess cancer risk. Each of these studies could, in the Agency's expert opinion, be used independently to establish the significant cancer risk associated with excess exposure to cadmium. However, OSHA recognizes that at the current state of the art reliance on either animal or epidemiolgical studies to determine human risk entails its own set of associated problems/limitations, and it is therefore prudent, where possible, to evaluate risk using both animal and human data.
For example, to determine human risk from animal studies it is necessary to extrapolate across species, and humans may be more or less susceptible than the animals studied. Moreover, animal experiments are typically carried out using relatively few animals (typically about 50 for each dose and sex group), which are often exposed for short periods of time and sacrificed often before cancer may manifest itself. With only 50 animals, it may be difficult to identify a carcinogen so potent as one that would double the background mortality rate in humans for lung cancer (overall age adjusted rate of 64 per 100,000 in white males in the USA). With a doubling of this rate, the testing of 50 animals would result in less than one additional animal per dose group developing lung cancer. So, even in this extreme situation, it is unlikely that an increased risk would be observed. To be assured of observing such an increase in cancer risk, the number of test animals would need to be tremendously increased. However, managing large numbers of animals is impractical. The more managable alternative is to increase the dose of the material being tested, often to the maximum dose that can be tolerated without causing early mortality in the animals from diseases other than cancer. However, in order to use these data to estimate human risk requires another extrapolation, from high dose in animals to low dose in humans, which in turn creates additional uncertainty.
On the other hand, epidemiological studies cannot be controlled nearly as carefully as animal studies, and information about factors potentially relevant to these studies is typically less than complete. Thus, in epidemiological studies there are inevitably confounding factors, like cigarette smoking or exposures to toxic substances other than the test substance, that may raise some doubt about findings of association between the substance under study and disease. Furthermore, detailed, complete, and accurate exposure data needed to precisely determine dose often are unavailable in such studies and must be reconstructed by applying reasonable assumptions to imperfect available data.
Notwithstanding the respective sets of problems/limitations that seem to be endemic to animal bioassays and epidemiological studies, in particular cases animal and/or human studies will provide the best available evidence of the toxicity of a substance and either or both may prove to be quite reliable. In relying in this final cadmium standard on both animal and human studies to determine risk, OSHA seeks to answer the criticism that stereotypically arises from reliance on either one alone. Thus, when one relies on the Takenaka animal study, the need to extrapolate across species may be a problem, but there is no such need and therefore no such problem when one relies upon the Thun study of cadmium exposed workers. On the other hand, when one relies upon the Thun study, co-exposure to arsenic or cigarette smoking may be potentially confounding factors, but there are no such co-exposures and therefore no such confounding when one relies on the Takenaka study.
As a result, to the extent that the two studies are in basic agreement about the nature and extent of the risk, the concerns generated by exclusive reliance upon one type of study should be substantially alleviated by reliance upon the other as well. Such agreement, OSHA believes, would also strongly suggest that the Agency's risk determinations for cancer are realistic. It is highly unlikely that such agreement could have been produced by mere coincidence.
With regard to its quantitative risk assessment for cancer based upon the animal data, OSHA has relied upon the Takenaka (Ex. 4-67) and the Glaser/Oldiges studies (Exs. 8-694-B; 8-694-D). The Takenaka study involved exposure of male rats to cadmium chloride, while the Oldiges study involved exposure of male and female rats to cadmium chloride, cadmium oxide, cadmium sulfate, and cadmium sulfide. OSHA applied the multistage model and two variants of it to the ten data sets from these two studies to estimate excess lung cancer risk from an occupational lifetime (45 years) exposure to each of the various exposure levels for each of the cadmium compounds.
OSHA has relied upon the Takenaka rat study to derive its best estimate of risk based upon data from experimental animals. This study is particularly suitable for quantitative risk assessment, because exposure levels were well documented, the study was run with concurrent controls, there was no opportunity for confounding exposures, and the route of exposure, inhalation, is the same as the primary route of exposure in occupational settings. Two possible drawbacks to this study raised in the proposal are that the animals were exposed continuously and to cadmium chloride. By contrast, workers are exposed mostly to camium compounds other than cadmium chloride and generally for eight hours a day.
With regard to the carcinogenicity of particular cadmium compounds, analyses of dose response data for several cadmium compounds show a similar carcinogenic potency. With regard to length of exposure, although rats in the Takenaka study were dosed continuously, and workers are not, cancer risk assessments show a similar dose response in relation to total cadmium dose, whether the animals were exposed continuously or in an exposure pattern simulating the workplace mode of exposure (Ex. 31; See also Table VI-6 of the Quantitative Risk Assessment section). Thus, there appears to be no dose rate effect.
To quantify risk from cadmium exposure using the Takenaka rat data, OSHA in its proposal examined five low-dose extrapolation models. The choice of model involves scientific judgment. There is no certain way to determine which model is correct. The statistics that allow us to measure goodness of fit cannot provide help in judging "best" fit among the models. Consequently, the best (correct) model must be chosen on the basis of some other criterion.
OSHA prefers the multistage model as its best model because the Agency believes the multistage model has the best empirical and theoretical justification of all the models for estimating carcinogenic dose-response. The multistage model is a nonthreshold model that is linear at low doses. The Agency believes that this model conforms most closely to what we know of the etiology of cancer. OSHA's preference is consistent with the position of the Office of Science and Technology Policy, which recommends that "when data and information are limited, and when much uncertainty exists regarding the mechanisms of carcinogenic action, models or procedures which incorporate low-dose linearity are preferred when compatible with limited information. In addition, there was good general support during the rulemaking for using the multistage model to estimate cancer risk with animal data.
OSHA applied three multistage models to the data. The results are shown in Table 6 of OSHA's risk assessment. With the exception of cadmium oxide (CdO) fume (the results which might be explained on the basis of lower lung deposition) at an occupational lifetime (45 years) exposure to cadmium of 100 ug/m(3), the current PEL for cadmium fume, most of the 28 maximum likelihood estimates (MLEs) for all of the cadmium compounds and for both male and female rats project excess cancer deaths well above 100 per 1000. Even for CdO fume, the excess risk of death from cancer associated with 100 ug/m(3) ranges from 6.8 to 37 per 1000.
By contrast, with an occupational lifetime exposure to the new PEL of 5 ug/m(3), the models project dramatically reduced risks in all categories. Thus, the new PEL significantly reduces the risk of cancer among cadmium exposed workers. Nevertheless, of the 28 MLEs associated with 5 ug/m(3) for the various cadmium compounds and models using data for both male and female rats, all but three project risks of greater than 5 excess cancer deaths per thousand workers, and nearly two-thirds project 15 or more excess cancer deaths per thousand.
If OSHA were to choose a best estimate of risk based upon the animal data, for the reasons presented in the preamble to its proposed cadmium standard and in the health effects section of this preamble, it would continue to choose the risks generated by applying the multistage model to the data for CdCl2 from the Takenaka study. Because the strengths and weaknesses of that study have been vetted in this rulemaking, OSHA feels assured of its reliability and appropriateness. OSHA now calculates the best estimate of risk associated with a PEL of 5 ug/m(3) from this study to be 15 excess cancer deaths per thousand. (See Table 6 in the quantitative risk assessment section.) This risk is slghtly higher than the 10.6 per 1,000 risk that OSHA projected as the best estimate in its preliminary risk assessment from the Takenaka data. The reason for the difference is that OSHA calculated the dose slightly differently for its final risk assessment. In any event, both the estimates of 10.6 and 15 excess deaths per thousand are an order of magnitude above risks that previously have been considered at least minimally significant by OSHA.
With regard to its quantitative risk assessment for cancer based upon the human data, OSHA continues to rely upon the Thun study presented in the proposed cadmium rule. The cohort has been updated to include six additional years of follow-up. As discussed at length in the health effects section of this preamble, OSHA believes that the Thun study with updated information is an excellent epidemiological study. While subject to the limitations inherent in such studies, it provides a reasonably reliable basis for quantitative risk analysis. Furthermore, it is the only epidemiologic study available that has reliable data on dose and it has undergone extensive peer review. In addition, as discussed in OSHA's quantitative risk assessment in this preamble, extensive additional analyses of its data have been performed and several additional models have been used to model the data. OSHA further believes that the various challenges to the Thun study (Exs. 19-43; 12-41) have forced its authors and others, like NIOSH and OSHA, who choose to rely upon it, to thoroughly consider and respond to the questions raised. These responses, in OSHA's judgment, have been more than adequate.
OSHA requested public comment on the uncertainties involved in using the Thun et al., epidemiological data to perform its quantitative assessment of the cancer risk associated with occupational exposure to cadmium. OSHA further requested public comment on how the Agency might resolve the issue of basing its final quantitative risk assessment on either the Thun study or the Takenaka study. Based upon additional follow-up and new analyses of the Thun cohort, OSHA has concluded that confounding from cigarette smoking and arsenic exposure played little role in the excess of lung cancer observed among the cohort members. With such an extended and comprehensive assessment of the strengths and weaknesses of the Thun study as part of this rulemaking, OSHA feels comfortable with its reliance on that study.
The Thun study is an historical prospective study of 602 white men employed in production areas of a smelter for at least six months between 1940 and 1969 and followed through 1984. It provides the strongest evidence of the carcinogencity of cadmium in humans. For workers with two or more years of employment at the smelter, mortality from lung cancer was statistically significantly elevated (SMR=229). Dividing the cohort of workers into those with low, middle and high cumulative exposures to cadmium, a significant dose-response relationship between cadmium exposure and lung cancer was observed.
The methods used to quantify risk from the Thun data closely follow those used by EPA (Ex. 4-04). In its final risk assessment, OSHA applied a relative risk model, adjusted for Hispanic ethnicity, to the updated Thun data. Because the new estimates are based upon more complete data and more reliable quantitative methods, OSHA prefers them over those in the proposal.
As shown in Table 12 of OSHA's final risk assessment, with an occupational lifetime exposure to cadmium at the new PEL of 5 ug/m(3), OSHA projects from the Thun data a risk of three excess deaths from lung cancer per 1000 workers based on the MLE. This estimate of risk at the new PEL is based exclusively on the reduction in exposure to cadmium achieved by the new PEL and does not take into account the additional risk reduction arising from the ancillary provisions of this standard. Nonetheless, this estimate constitutes a 95% reduction from the comparable estimated risk of 58.3 deaths per thousand at an occupational lifetime exposure to cadmium of 100 ug/m(3). It also represents greater than an 85% reduction of risk from the risk at an occupational lifetime exposure to 40 ug/m(3) and a 75% reduction of risk from the risk at an occupational lifetime exposure to 20 ug/m(3), both of which levels are closer than 100 ug/m(3) to typical existing occupational exposure levels in many industries with current exposures above 5 ug/m(3).
Since OSHA published its proposed rule, Dr. Leslie Stayner and others of the National Institute for Occupational Safety and Health (NIOSH) have also developed an independent quantitative risk assessment based on the updated Thun cohort. That risk assessment differs in ways discussed in the risk assessment section of this preamble from OSHA's own preliminary and final risk assessments. NIOSH, in response to criticisms of OSHA's risk assessment in the cadmium proposal and to recommendations made by various scientists, made methodological adjustments and applied three separate models to the Thun data. The results of NIOSH's risk assessment are shown in Table 9 of OSHA's final risk assessment.
With an occupational lifetime exposure at the new PEL of 5 ug/m(3), NIOSH estimates an excess risk of death from cancer ranging from 3.9 to 5.5 to 9.0 per thousand workers for the Cox Regression Analysis, the multistage model, and Poisson Regression model, respectively. These estimates of risk are statistically all very close to one another and strikingly similar to OSHA's own, independently derived estimate of 3 excess lung cancer deaths per 1000 workers.
The NIOSH risk estimates for occupational lifetime exposure at the new PEL, like the OSHA risk estimates for that level, represent a very substantial reduction of risk from risks estimated for comparable exposures at the higher, currently allowable levels and at existing levels. For example, the estimated excess risk at an occupational lifetime exposure of 100 ug/m(3) ranges from 73 to 102.2 to 157 per thousand workers according to the Cox Regression, multistage, and Poisson Regression models, respectively. Thus, under all three models the risk at 5 ug/m(3) represents a nearly 95% reduction of risk from the risk at 100 ug/m(3), a nearly 87% reduction of risk from the risk at 40 ug/m(3), and a nearly 75% reduction of risk from the risk at 20 ug/m(3). Again, the NIOSH estimates of risk at the new PEL do not take into account the additional reductions in risk arising from the ancillary provisions of this standard. OSHA expects these additional reductions to eliminate significant risk of cadmium associated cancer at the PEL of 5 ug/m(3).
Summing up the results of these various risk assessments based on animal and human data, all indicate a very high excess risk of death from cancer arising from an occupational lifetime exposure to cadmium at the current PEL (for fume) of 100 ug/m(3). All also show very high excess risks at levels much lower than 100 ug/m(3). Further, the results based upon all of the models show a very high reduction of risk associated with the new PEL of 5 ug/m(3). And all show that the risk that OSHA is seeking to regulate, without regard to the ancillary provisions, remains significant at least down to the new PEL.
Indeed, at the PEL of 5 ug/m(3) the best estimate of excess risk from the animal data, 15 deaths per thousand, and all the estimates from the human data, 3 per thousand under the OSHA model, and 3.9 to 9 per thousand under the NIOSH models also all reflect continuing significant risk without regard to the ancillary provisions. If OSHA were relying exclusively upon the PEL to reduce risk and there were no ancillary requirements that effectively eliminated remaining significant risk at the new PEL, and if there were no other circumstances that further mitigated the risk, OSHA might well have to set the PEL still lower than 5 ug/m(3) if that were feasible.
These estimates of remaining risk at the new PEL are all very similar statistically. They are all within one order of magnitude. This similarity is even more striking when one realizes that estimates based on the animal data are for total cancers, whereas, the estimates derived from the human data are based on lung cancer only. (The possibility exists that lifetime studies of the occupational cohorts might identify additional cancer sites in humans related to cadmium exposure.) Thus, OSHA feels assured by these mutually confirming results that its risk estimates for cancer are realistic and reasonably accurate. As stated above, by implementing the new PEL along with the ancillary provisions of the standard, OSHA expects the significance of the risk to be eliminated.
For its final quantitative assessment of the excess risk of kidney dysfunction associated with occupational exposure to cadmium, OSHA applied a logistic regression model to the five independent studies that have relevant quantifiable data. These studies, discussed at length in the health effects section above and analyzed for quantitative estimates of kidney dysfunction in relation to dose in the quantitative risk assessment section, were conducted by by Falck and others, Ellis and others, Elinder and others, Jarup and others, and Mason and others. In these studies, the authors investigated the association between levels of low molecular weight proteins in the urine of workers and cumulative occupational exposure to cadmium. The low moleculator weight proteins being measured are Beta 2 microglobulin (B(2)-M) or retinol binding protein (RPB), excessive levels of either of which are taken as indicative of kidney dysfunction, The logistic model that OSHA applied to the data from each study was modified from the model presented in the proposal to take account of background levels of kidney dysfunction unassociated with occupational cadmium exposure. This explains why the results projected from the model in the final risk assessment for the Falck and Ellis data sets are somewhat lower than those projected in the preliminary risk assessment, which relied exclusively upon those two data sets.
In its final risk assessment OSHA performed analyses on seven data sets from five studies using a modified logistic model. The five studies themselves were quite different from each other in many material ways. For example, authors chose different levels of B(2)-M or RPB as indicative of kidney dysfunction; some used spot urine samples, others used 24-hour samples; and the number of subjects in each ranged from 33 to 440. In part because of the small size of some of the studies and in part because of the uncertainty in extrapolating results to low TWA exposures, some of the confidence intervals in Table 19 of the risk assessment section are fairly wide. Considering this and the differences among the underlying studies, which doubtless affect the results, the analyses produced reasonably consistent results. Thus, for example, as shown in Table 19, all of the seven analyses project high rates of proteinuria at an occupational lifetime exposure of 100 ug/m(3) (24-99.8%); all continue to project relatively high rates of proteinuria down to exposures as low as 20 ug/m(3) (2.1-47.2%); and all project a risk greater than one per thousand at an exposure of 10 ug/m(3) (2.7-234 per thousand). Moreover, at the new PEL of 5 ug/m(3), all but two of the analyses show an excess risk of proteinuria greater than 1 per thousand (1.9-95). With regard to those two, both show a risk greater than two per thousand (2.7-2.8) at an exposure of 10 ug/m(3). Thus, even with regard to the two lowest results, the analyses indicate a risk greater than one per thousand at exposure levels somewhere between 5-10 ug/m(3). Furthermore, for reasons described in OSHA's risk assessment, the Agency no longer considers one of the two studies that provide the source for these low estimates, the Falck study, as reliable a basis for OSHA's quantitative risk assessment as the other studies.
In the other six risk estimates, the results range at the extremes from .37 to 95 estimated excess cases of proteinuria per 1000 workers exposed to cadmium at 5 ug/m(3) for an occupational lifetime. The four results between the extremes range between 1.9-27 cases per thousand, a range only slightly greater than one order of magnitude.
From all the analysis reflected in Table 19, OSHA's best estimate of risk at 5 ug/m(3) is 14-23 excess cases of proteinuria per thousand. OSHA arrived at this best estimate by determining the upper and lower 95% confidence intervals for each of the risk estimates at 5 ug/m(3) reflected in Table 19. With the exception of the Mason 1 analysis, the estimate of 14-23 excess cases falls within the 90% confidence intervals (95% upper bound and 95% lower bound) of the six data sets analyzed. To put this in other words, 14 represents the highest of the 95% lower bounds and 23 represents the lowest of the 95% upper bounds. So 14-23 is within the 90% confidence intervals for each of the six analyses. For example, the 90% confidence interval for Jarup 1 is 8.3-23; 14-23 falls within that interval. Similarly, the range for Elinder is 0-99; 14-23 falls within that interval. This is true for Ellis, as well: the 14-23 range falls within the 90% confidence intervals of 14-288. OSHA therefore believes that its best estimate of 14-23 excess cases reflects the central tendency of the relevant data. This risk estimate, like comparable estimates for cancer, does not take into account the additional reduction in risk arising from the ancillary provisions of the standard. OSHA expects these additional reductions to eliminate significant risk of cadmium associated kidney dysfunction at the PEL of 5 ug/m(3).
In response to a comment in the rulemaking (Ex. 17-D), OSHA also applied several types of models to the continuous data from the Mason study. The results of that analysis are shown in Table 21. At an exposure of 5 ug/m(3), with the exception of the results for the threshold model (Model IV), the results are all greater than OSHA's best estimates of 14-23 cases per thousand workers. By contrast, the threshold model predicts 0 risk at 5 ug/m(3), and, using the matched analysis, 0 risk at 10 ug/m(3) and even at 20 ug/m(3). These analyses provide estimates that are both higher and lower than the results produced by the other models and from OSHA's best estimates. For reasons discussed in OSHA's risk assessment, OSHA gives greater weight to results shown in Table 19 of the quantitative risk assessment section that were derived from the logistic models modified to incorporate background response.
With regard to reduction in risk of kidney dysfunction, the best estimate of risk, 14-23 per 1000, for occupational lifetime exposure to cadmium at 5 ug/m(3) represents a 90-94% reduction in risk from the lowest estimated risk (242 per thousand) associated with 100 ug/m(3) based on the modified logistic model. For five of the seven data analyses for exposure at 100 ug/m(3), 14-23 represents greater than a 96-98% reduction in risk. For occupational lifetime exposures to 20 ug/m(3), with the exception of the estimates derived from the Elinder data, the 14-23 best estimate of occupational liftime risk at the new PEL represents a 62-76% reduction from the lowest estimate of risk (60 per thousand). For the next lowest estimate of risk (80 per thousand) at 20 ug/m(3), 14-23 represents greater than a 70-82% reduction in risk, and for the other four risk estimates (91, 186, 236, and 472 per thousand) at 20 ug/m(3), the reduction in risk is still greater. Even at an exposure of 10 ug/m(3), with the exception of the analyses based upon the Elinder and Falck data sets, the best estimate of 14-23 represents a 4-40% reduction in risk from the lowest estimate (24 per thousand), a 22-55% reduction in risk from the next lowest estimate (31 per thousand), and greater than a 40-60% reduction from the other risk estimates.
The conclusions of OSHA's quantitative analysis of risk for lung cancer and kidney dysfunction are further supported by a very recent review of the scientific basis for regulating cadmium in the workplace. The article (Ex. L-140-50), which was written by three of OSHA's expert witnesses in the cadmium rulemaking, applies a very different approach to the analysis. The article's analysis compares OSHA's preliminary quantitative risk estimates derived from mathematical modelling of data from several studies with the results derived from other models and with published, empirical data on kidney dysfunction and lung cancer. Although the authors find that "modelling generally implies greater certainty than exists at low doses....," they also find that OSHA's risk estimates generally are in line with the empirical data and the results of other modelling.
With regard to kidney dysfunction, for example, the authors find that "the empirical data and models...all show a similar pattern. The prevalence of kidney dysfunction increases sharply at cumulative exposures above 500 ug/m(3)-year....[However,they point out,] the studies are too small to estimate prevalence at low[er cumulative] exposures...." It is therefore "impossible," the authors conclude, "to identify a no-effect level with certainty."
For cancer, the authors find that the epidemiological data provide more plausible estimates of risk than the animal data. The rat data, they find, overpredict risk.
Rather than relying upon mathematical modelling, the authors suggest using a safety margin. Based upon the analysis of both kidney dysfunction and lung cancer, the authors conclude that "occupational exposure to cadmium should be controlled as stringently as is technically feasible, with the PEL not to exceed 5 ug/m(3)." OSHA finds these conclusions broadly confirmatory of the results of its own analysis.
Consequently, based upon the best estimates of excess risk associated with each of various occupational lifetime exposures to cadmium, whether OSHA relies upon the cancer or the kidney data, and in connection with cancer whether OSHA relies upon the animal data or the human data, it consistently appears necessary for OSHA to set the PEL at least as low as 5 ug/m(3).
Even at the new PEL of 5 ug/m(3), most of the analyses and all of the best estimates of risk indicate a continuing risk of death from cancer and cases of kidney dysfunction somewhat greater than one per 1000 workers. Thus, the real problem for OSHA when it sets the PEL at 5 ug/m(3) lies not in establishing that the Agency is regulating a significant risk. Rather, the problem lies in establishing that, to the extent feasible, the PEL should not be set still lower in order to eliminate what appears, without regard to the reductions in risk arising from the ancillary provisions of this standard and other factors, to be a continuing significant risk.
OSHA thinks the decision to set the PEL no lower than 5 ug/m(3) involves complex policy determinations that draw upon OSHA's experience and expertise and also reflect a delicate balancing of counterveiling factors. The reasons for the decision are wide ranging.
First, OSHA fully expects the medical surveillance and other requirements in the standard ancillary to the PEL (e.g., MRP, action level, regulated areas, training, etc.) to substantially lower the risk of kidney dysfunction and the risk of cancer from the estimates in the risk assessment. Although OSHA cannot quantify the reductions in risk that may be expected from these and other similar provisions of the standard, OSHA believes that the effect of including the ancillary provisions in the final standard will eliminate the remaining significant risk estimated at the new PEL.
Second, industry has stated that the best way to assure that the new PEL will be met consistently is for industry to implement work practice and engineering controls to achieve a mean exposure considerably below 5 ug/m(3) (Ex. 144-6). Some industries maintain that it would be best to establish a mean 40% below the PEL. When these levels are achieved, much, if not most of the time exposure levels will be controlled to below 5 ug/u(3). As a consequence, the risk estimated at an occupational lifetime exposure level of 5 ug/m(3) will overstate the actual risk, which would decline linearly.
Third, well over half the exposed workforce already is exposed below the new PEL, so that the actual risk to these employees already is below the risk estimated for an occupational lifetime exposure at that PEL.
Fourth, the vast majority of exposed employees work in industries/occupations where cadmium and its compounds are not the primary product produced or processed. Of these employees, most are not exposed above 5 ug/m(3). Of the remainder who are currently exposed above 5 ug/m(3), most are exposed only intermittently and will continue to be exposed only intermittently. Their cumulative exposure will be lower than the cumulative exposure used to derive the estimated risks, which are based upon an assumption that the employee is exposed to a TWA exposure at 5 ug/m(3) for eight hours every day. With a lower cumulative exposure, the actual risk for employees intermittently exposed at 5 ug/m(3) will be lower than the estimated risk.
Fifth, OSHA has already made an important policy decision to sever the PEL from, and set it lower than the separate engineering control air limit(s) (SECAL(s)) for six of the cadmium producing/processing industries, which is set at 15 ug/m(3) and/or 50 ug/m(3) because of feasibility constraints. In setting the PEL that low, OSHA has inevitably required a substantial number of employees to wear respirators full time. OSHA did this with serious reservations about the advisability of requiring full time respirator use and in the face of a NIOSH recommendation against requiring such use (Ex. 57). OSHA understands that full time respirator usage poses certain safety and health risks but, on balance, has decided that the risks of not requiring some protection for employees from airborne cadmium levels above 5 ug/m(3) are more serious than those attaching to full time respirator use.
However, the vast majority of employees exposed to cadmium do not work in industries to which a SECAL applies. For them, the PEL is set at or very close to the limits of feasibility. If OSHA were to set the PEL still lower than 5 ug/m(3), large numbers of additional employees would have to wear respirators full time. OSHA is loathe to go further in this direction, especially since the actual risk to employees under the new PEL in practice is likely to be considerably less than the estimated risk. For all these reasons, OSHA has exercised its professional judgment and discretion in determining that the PEL for cadmium shall be established at 5 ug/m(3). As a result, OSHA concludes that its cadmium standard will protect employees and that employers who comply with the provisions of the standard will be taking reasonable steps to protect their employees from the hazards of cadmium.
OSHA's conclusion that the risk of death from cancer and the risk of kidney dysfunction resulting from exposure to cadmium at 100 ug/m(3) over a working lifetime are both significant is consistent with OSHA's determination of significance of risk at the previously existing TWA PELs for two carcinogens. The two carcinogens are inorganic arsenic (Jan 14, 1983; 48 FR 1864, 1986) and ethylene oxide (Apr. 21, 1983; 48 FR 17284). The risk estimates per 1000 employees for a working lifetime exposure at the prior PEL to these carcinogens ranged from 148 to 425 lung cancer deaths from inorganic arsenic and from 63 to 109 cancer deaths from ethlyene oxide.
In addition, for both carcinogens, OSHA concluded that, if it were feasible, OSHA would seek to further reduce the predicted remaining risk at the new PELs. That remaining excess risk of death for a working lifetime exposure per 1,000 workers was 8 for inorganic arsenic and 1 to 2 for ethylene oxide.
Further guidance for the Agency in evaluating significant risk is provided by an examination of occupational risk rates, legislative intent, and language of the Supreme Court of the United States. For example, in the high risk occupations of mining and quarrying (Division B), the average risk of death from an occupational injury or an acute occupationally-related illness from a lifetime of employment (45 years) is 15.1 per 1,000 workers. Typical occuptional risks of deaths for all manufacturing (Division D) are 1.98 per 1,000. Typical lifetime occupational risk of death in an occupation of relatively low risk, like retail trade, is 0.82 per 1,000 (Division G). (These rates are averages derived from 1989-1990 Bureau of Labor Statistics data for employers with 11 or more employees, adjusted to 45 years of employment, for 50 weeks per year.) There are relatively few data on risk rates for occupational cancer, as distinguished from occupational injury and acute illness. The estimated cancer fatality rate from the maximum permissible occupational exposure to ionizing radiation is 17 to 29 per 1,000 (47 years at 5 rems; Committee on Biological Effects of Ionizing Radiation (BEIR) III predictions). However, most radiation standards require that exposure limits be reduced to the lowest level reasonably achievable below the exposure limit (the ALARA principle). Consequently, approximately 95% of radiation workers have exposures less than one-tenth the maximum permitted level. The risk at one-tenth the permitted level is 1.7 to 2.9 per 1,000 exposed employees.
Congress passed the Occupational Safety and Health Act of 1970 because of a determination that occupational safety and health risks were too high. Congress therefore gave OSHA authority to reduce above-average or average risks when feasible. In discussing the level of risk that Congress authorized OSHA to reduce, the Supreme Court stated that "if the odds are one in a thousand that regular inhalation of gasoline vapors that are 2% benzene will be fatal a reasonable person might well consider the risk significant and take the appropriate steps to decrease or eliminate it." (I.U.D. v. A.P.I., 448 U.S. et 655).
Within this context, OSHA's best estimates of risk from occupational exposure to cadmium at the current TWA PELs are substantially higher than other risks that OSHA has concluded are significant, are substantially higher than the risk of fatality in high-risk occupations, and are substantially higher than the example presented by the Supreme Court. Consequently, OSHA concludes that its best estimates of risk associated with the current TWA PEL of 100 ug/m(3) are significant. Based on this reasoning, OSHA's best estimates of risk remain significant down to levels as low as the new PEL of 5 ug/m(3). As previously stated, these estimates do not take into account the additional reductions in risk that are attributable to the ancillary provisions of the cadmium standard, which OSHA fully expects will eliminate any remaining significant risk at the new PEL.
- [57 FR 42102, Sept. 14, 1992; 58 FR 21778, April 23, 1993]
|Regulations (Preambles to Final Rules) - Table of Contents|