Richard Wilson

Mallinckrodt Professor of Physics

Harvard University

Cambridge, MA

A.M. Langer, R.P. Nolan

Environmental Sciences Laboratory

Brooklyn College

Brooklyn, New York

May 7, 1996

EXECUTIVE SUMMARY

We have performed a risk assessment for installers of blown glass wool insulation. We make and justify the assumption that glass wool is less chemically active than chrysotile asbestos, or any other form of asbestos, and is less persistent in the lung. We also assume that the fiber sizes that lead to the risk are similar to the fiber sizes of the asbestos that are found to cause cancer, and that the glass wool fibers act synergistically with cigarette smoking in causing the lung cancer risk (if any).

The risk for a non-smoking installer of glass wool insulation who wears a respirator is 5.6 in a million for every year that he works. We compare this small risk with a variety of occupational risks ranging from being President of the United States (1.4 in a hundred) to police work (3 in ten thousand). The low risk we assign to installers of glass wool is consistent with the fact that no one has found such a risk in an animal study or an epidemiology study in spite of diligent searches.

The reason for interest in assessing the risk for workers installing blown glass wool insulation is two-fold. On the one hand, there is a general interest in assessing the risk associated with any industrial or work place activity, or exposure to some manipulated material in those environments. On the other, the material that blowing wool insulation largely replaces, asbestos, is well known for producing significant health hazards under certain exposure conditions. Obvious questions arise as to whether exposure to blown glass wool insulation is associated with some of the same problems as those caused by asbestos exposure; whether or not these problems are of the same nature and magnitude and whether or not risks associated with glass fiber inhalation, if they exist, can be quantified.

All scientists with whom we are acquainted start with the assumption that fibers are hazardous per se, and the differences observed among the fibers in terms of biological potential are in part related to the exposure characteristics associated with their use. There are three reasons for this default assumption:

If these assumptions are true then blown glass wool, which is fibrous, should also exhibit carcinogenic properties, under conditions of very high exposure, such as have not been seen in the industry.

There have been a number of review papers and monographs that address the above issue, i.e., the carcinogenicity of synthetic vitreous fibers (IARC, 1988; Brown et al 1991; Proceedings Symposium on Synthetic Vitreous Fibers: Scientific and Public Policy Issues, 1994). However none of these actually put into numerical terms the risk which the blown wool insulator incurs, and more importantly, no one has tried to put this risk in perspective with other risks encountered in other workplaces as well as in society in general. This is the topic of this study which was carried out on behalf of the Insulation Contractors Association of America (ICAA). For a brief elementary discussion of risk assessment see Wilson and Crouch (1987).

The most obvious, the most direct and therefore the most important way of assessing the risk to workers in any occupation is to perform an epidemiological study to see if their rate of morbidity and/or mortality from any disease is appreciably different from the rest of the unexposed population. Indeed this follows a much more general, and older, proposition that the proper study of (wo)man is (wo)man. If a statistically significant adverse health effect is evident, it is stated that an association between exposure and outcome has been found. If other information suggests that the association is causal (exposure to substance "x" is the etiological agent producing the outcome), then a numerical risk can be assigned to workers who might be similarly exposed elsewhere.

It is important to realize that while epidemiology is a direct approach to assigning a risk, it is often an insensitive one. If the ratio of the number of lung cancer cases observed to the number expected in the general population is not considerably greater than one, it becomes difficult to separate the observed increase from the confounding caused by other factors, e.g., unknown cigarette smoking habits. Without knowledge of the group's smoking habits the expected incidence of lung cancer cannot be calculated with a sufficient accuracy to be certain that any small increase in the SMR for lung cancer is real.

In many occupations where epidemiological studies have been carried out past exposures have been unnecessarily high, and for the future the exposure can be reduced. The risk at the lower exposure is presumed to be smaller, at least by the factor by which the exposure is reduced.

Neither the statistical accuracy nor the systematic errors due to confounders (alternative explanations) can be indefinitely reduced. It is always possible that a real effect may exist that is too small to show itself above these backgrounds or some critic may claim a small effect where none exists. Thus all one can logically say about a risk where there is no significant measured effect is that there is some upper limit value below which the risk must lie. The workers manufacturing synthetic vitreous fibers (SVFs) have rarely been exposed to the unnecessarily high, even carelessly high, exposures that were often common in other industries (such as those measured in the asbestos industry) in the early part of the twentieth century. Therefore, it is not easy to further reduce these exposures and thereby reduce the upper limit of risk. If the upper limit is not low enough to allow adequate reassurance of safety, it becomes useful to look at indirect estimates of the risk, which although intrinsically less reliable than the direct measures, can be considerably more sensitive.

There is no situation where an epidemiological study has found a SMR less than 200 (twice the expected number) and successfully used that to assign causality without strong supporting data. The epidemiological studies that we discuss in Section C have SMRs that are approximately 110 (an excess of 10%). Including uncertainties the SMRs might just as well be as high as 120, but might be 100 (the number of cases of disease found is exactly the number predicted). This makes it essential to examine the supporting information available from animal studies and the physico-chemical characteristics of the fibers. These provide a basis for biological plausibility and assists in the interpretation of data.

One of the most important considerations is the actual parameter defining or characterizing the exposure which is postulated to be a cause of the risk. If the wrong parameter is assumed, it might be reduced but have no effect on reducing the risk. In this report, therefore, we first discuss in Section B the fibers that are likely to be of concern, their properties, concentration, and the exposure of workers to them. It will be seen that they are not characterized as well as is desirable, and this leads to the principal recommendation at the end of this report.

We then proceed in Section C to discuss the epidemiological studies. We follow closely the recent review of Lee et al (1995). We conclude, in agreement with these authors, and with Lamm (1994), that there is little direct evidence to suggest, and none to prove, that SVFs (either fiber glass or rock wool) ever causes human cancer.

In Section D, we examine indirect evidence. First among the indirect evidence, we examine the results of experimental animal studies and secondly we compare these results with chrysotile asbestos, to show that the risk, if any, is likely to be considerably smaller than this upper limit determined directly. This indirect evidence is, in effect, discussing a possible mechanism by which fibers can cause cancer and demonstrating that glass fibers do not do so.

In Section E, we will perform a risk assessment from exposures in Section B and the mechanistic principles that are suggested in Section D. In this we will take the risk measured by epidemiology for those fibers where each of these determinants is unfavorable (chrysotile asbestos) and reduce by an appropriate factor. We expect to express the result in an upper limit of risk determined by this indirect mechanistic procedure, which will be much less than the upper limit from epidemiology. In this we take a pessimistic, scientific, viewpoint that everything poses a risk, however small. We will show that in this case the risk is small and those using other definitions might consider it to be zero.

In Section F, we will then comment upon the positions of scientists and others in regulatory agencies; the meaning of the IARC classification as a 2B carcinogen and the recent review by Infante et al (1994). In this section we go further than many reviewers in one respect; we discuss directly (and refute) the claims of some pessimists that these studies demonstrate a real effect and therefore glass fibers should be regulated as a (potent) carcinogen at high exposure. We will argue that such claims are unproductive and should be moot. If exposure continues to be controlled in the responsible manner that has characterized the industry so far, such as that during production of SVF, the risk will continue to be small and not measurable directly.

Finally, we will give some recommendations on further studies and experiments that might be undertaken to further reassure the workers in the industry, their employers and the public.

The fibers of interest are variously called man-made mineral fiber (MMMF), man-made vitreous fibers (MMVF), and synthetic vitreous fibers (SVFs). We will use SVFs in the pages that follow. The properties of these materials, especially those with control persistence in biological hosts are discussed in Appendix I.

Our task would be made simpler if there were excellent measurements of exposure. We need to know, and characterize, the exposure in the past over the time period covered by the epidemiological studies and take into account, present and anticipated future exposures. For comparison take asbestos fibers as indicators of potential hazard, therefore information concerning chrysotile asbestos exposures is relevant.

Likewise in the literature often the distinction between airborne concentrations, exposures, and doses has not been clearly stated. One can measure airborne concentrations at specific locations. Until someone goes into that location there is no exposure. The exposure is the concentration multiplied by some occupancy factor. What actually goes inside the body (into the lung in this case) and reaches target tissues is the dose. It is considered to be obvious that it is the dose that is important for determination of potential human risk.

Unfortunately the historical measurements available to describe the early exposures or even concentrations in the SVF industry are limited in that few measurements are available and the methodologies used did not characterize the respirable fibers -- the potential dose -- present in the air. Prior to 1970, the concentration measurements (most often described as, "exposures") were made gravimetrically (by determining the weight of the dust in air) or by light microscopy (the total number of particles present in a given volume of air). The results were expressed as milligrams of dust per cubic meter of air (mg/m³) or millions of particles per cubic foot of air (mppcf). These early indices of concentration were designed for determining the effectiveness of dust control technologies.

As further research identified the health hazards associated with the inhalation of inorganic fibers, these indices were replaced by methodologies which characterized respirable fibers, and non-fibrous particles, present in the air of workplaces (Johnson et al, 1969). Concentration measurements to describe the number of microscopic and sub-microscopic fibers per cubic centimeter of air (fibers/cm³) and a description of the lengths and diameters of individual fibers were not determined for SVFs until a decade later (Esman et al, 1978). In the early 1980s, a reference method using phase-contrast optical microscopy (PCOM) and membrane filters to determine the airborne concentrations of SVF became available (WHO, 1985).

Although the historical exposure measurements are limited, sufficient information exists to describe the aerosols created during the production and use of SVF and to compare these to airborne chrysotile asbestos fibers.

The asbestos minerals are naturally occurring fibrous crystalline minerals while the synthetic vitreous fibers are synthesized. The precursor starting materials are melted to form a viscous liquid at high temperatures, as described previously (IARC, 1988 and Appendix I). The molten liquid can then be made into fibers using one of several industrial processes.

The SVFs used as blown thermal insulation are either rock wool, or more commonly fiber glass, both of which are generically referred to as mineral wool. The elemental composition and the morphological characteristics of the SVFs have varied as the industry developed improved technologies for producing vitreous fibers for specific applications. In addition to controlling the morphology (length and diameter) to enhance commercial characteristics of the fiber, the surfaces have been coated with a range of organic compounds, referred to as binders. Generally, fibers prepared with binders have a greater diameter and are markedly less dusty then fibers without binders. Although the insulation wools blown into place are generally uncoated.

The SVFs are commonly thousands of times greater in length than in diameter (Christensen et al, 1993). The characterization of the airborne fibers generated by manipulating SVFs are quantitatively different than those of the bulk material (Jacob et al, 1992). The mean diameter and lengths of the fibers in a SVF aerosol are substantially smaller than in the bulk materials. For example, the median airborne fiber diameters were found to be 4Fm, 1.8Fm and 0.2Fm for aerosols generated from bulk SVFs with nominal fiber diameter of 14Fm, 6Fm and 1Fm respectively (Esman et al, 1979). These are important considerations for blown glass wool where the fibrous particles in the aerosol represent a poorly-defined sub-population of the bulk fiber.

The airborne fiber concentrations measured today during the manufacturing of SVFs are substantially lower than the high concentrations of asbestos measured in asbestos factories in the past. These latter situations are associated with diseases in workers. In contrast, the mean concentrations measured by phase contrast optical microscopy (PCOM) during the manufacture of the SVFs used for blown insulation are generally between 0.01 and 0.05 fibers/cm³ (Esmen et al, 1982). The mass of the total dust exposures are typically of the order of 1mg/m³ although the mass of total dust is a poor index of the number of fibers in the air (due to the non-fibrous fragments which exist in the aerosol as well, and the fiber population consists of respirable and non-respirable particulates).

An important fact to keep in mind is that the exposures which occur during blowing of loose thermal insulation wool are consistently higher than those which occur during the manufacture of SVFs and installation of most other vitreous fiber products such as batt insulation (Esmen et al, 1982; Lees et al, 1993). Therefore the exposures are higher than those associated with the morbidity and mortality reported in all of the epidemiological studies (Bunn et al, 1993; Lee et al, 1995). The exception to this is the fine diameter fiber glass used in aircraft insulation (Head and Wagg, 1980; Esman et al, 1982).

Loose insulating wools, which may be either fiber glass or rock wool, are used to insulate attic spaces. Frequently, these products are called blowing wool because they are installed using a pneumatic system. Historically, loose fiber glass or rock wool used for blowing was made of chopped up batts or trimmings from the batts produced in the manufacturing process. During the 1980s, a loose fiber glass specifically designed for blowing was introduced and it currently has a considerable share of the blown glass wool insulation market. In 1995, to the total loose-fill insulation market in the United States was 700,000,000 lbs. Industrial hygiene data indicates the newer blowing wools have a considerably smaller diameter than the older blown insulation wools (Lees et al, 1993). Particularly the loose fiber glass blowing wool insulation without binder where virtually all of the fiber in the aerosol are respirable.

Typically the loose insulation wools are installed into attics by a two person crew performing three operational tasks. The feeder cuts open the bags and dumps the contents into the hopper of the blowing machine. The pneumatic blower device is distant from the site where the insulation is going to be installed, usually in a truck. The loose wool is fluffed in the blower and then passed through a hose to the attic, where a worker prepares the attic. The blower operator directs the location where the insulation is to be applied. The worker who prepares the attic generally performs the tasks of the blower; the SVFs are blown from the pneumatic hose to cover the attic floor to a specified depth.

Exposure during installation of SVF insulation products other than blown wool insulation is generally comparable to the exposures during the manufacture of SVFs (Esman et al, 1982; Lees et al, 1993). The results of two studies designed specifically to characterize fibrous glass and mineral wool exposures of insulation workers during the application of blown wool insulation products in both well-ventilated and poorly-ventilated locations will be reviewed in detail. The first U.S. study reported that during blowing of loose wool insulation in attics, exposures to all fibers ranged from 0.09-5.3 fibers/cm³ (Table 1a,b Esman et al, 1982). The airborne concentrations for individual workers ranged from 0- to 20.0 fibers/cm³ by phase-contrast optical microscopy (PCOM) and from 0.0019 to 23 fibers/cm³ using scanning electron microscopy (SEM). The additional resolution available from SEM allowed the thinner submicroscopic fibers to be visualized and thereby counted, increasing the exposures reported for all tasks using both types of blowing wool. While the exposures decreased counting respirable fibers only using a combination of light and electron microscopy (Table 1b).

The insulation of attics with blown mineral wool produces both the highest fiber concentrations and the highest mass of total particulate in the air. These concentrations occur over 20-60% of the work day, and the work is generally divided between the crew members. Therefore, the time weighted average exposure (TWA) is considerably lower, about 1 fiber/cm³. On some occasions, TWA exposures as high as 14.33 fibers/cm³ were reported, while exposures using all other insulation products were less than 1 fiber/cm³. The fraction of fibers less than 3Fm in diameters (generally considered to be respirable), and lengths $9Fm, are slightly greater for loose blowing insulation as compared to other SVF insulation products (44-92% compared to 55-87%). While the fraction <1Fm in diameter and >5Fm in length is 20.5-54.8% for loose blowing insulation and 5.7-38.5% for other SVF insulation products (Table 2).

Lees et al (1993), reported an additional 1200 concentration measurements, using gravimetric methods, PCOM, and electron microscopy, for workers installing residential blown glass insulation products. The workers were divided into eight homogenous exposure groups depending on the type of SVF product used, and the task involved. The results showed that the mean task length concentrations, determined by PCOM for all the homogeneous exposure groups, except the installer (blower) and feeder, are less than 1 fiber/cm³.

The concentrations to which the installer (blower) and feeder of loose fill fiber glass with and without binder are exposed to are -- 0.55 and 0.18 fibers/cm³; 7.67 and 1.74 fibers/cm³ -- respectively, while for installer (blower) and feeder of loose fill mineral wool (without binder) concentrations are 1.94 fibers/cm³ and 0.31 fibers/cm³ respectively (Table 3a). This group of workers spent a similar amount of time installing glass insulation, 12.5-50%, to those workers studied by Esmen et al (1982) but the exposures levels are significantly different.

The TWA exposures for the eight homogeneous exposure groups, except for installers of loose blowing insulation, was <0.3 fibers/cm³. While the 8-hr TWA exposures to which the installer (blower) and feeder of loose fill fiber glass (without binder) were exposed were 1.96 fibers/cm³ and 0.85 fibers/cm³ respectively, while the concentration to installer (blower) of loose fill mineral wool was only 0.97 fibers/cm³. Unlike the exposures reported by Esman et al (1982), the blown wool fiber applications are lower than those of blown glass. Analysis of the filters by scanning electron microscopy found installer (blower) of fiber glass insulation to be 12.7 fiber/cm³ (Table 3b).

Due to the presence of other fibers, the concentration of SVFs resulting from SEM analysis of the air samples were slightly lower than the value determined by PCOM for every homogeneous exposure group (except for blowers and feeders of loose fill fiber glass, without binders). The increase in the concentration, and hence exposure, was due to the presence of fibers with diameters too fine to be resolved by PCOM. The geometric mean diameter (GMD) for the homogeneous exposure groups ranged from 1-2Fm, except blowers and installers of loose fill where the GMD was found to be 0.6Fm (see Table 4 for size distribution). Further analysis by transmission electron microscopy found 50% of the blown fiber glass to have diameters <0.1Fm (of these fibers ~18% were greater than 5Fm). The results indicate that the exposures to the workers installing loose fill fiber glass increased significantly during the 1980s. The aerosol also contained a significant proportion of fibers less than 0.1Fm in diameter and greater than 5Fm in length. The total dust (mg/m³) for Lees et al (1993) and the earlier Esman et al (1982) study are given in Table 5.

More recent sampling of airborne respirable fibers are shown in Tables 6 - 10. The exposures were 1.45 " 2.27, 0.35 " 0.38 and 0.14 " 0.21 fibers/cm³ for the installer (blower), feeder and travel respectively (Table 6). The total dust and respirable dust in the air are given in Table 7a, b (personal and area sampling) and 8. Additional air samples of Product C are given in Table 9. The ranges of exposure are similar to those reported in earlier studies (Table 1 and 5) although the presence of non-respirable fibers and particles limited the determination to minimum number of fibers present. The available industrial hygiene data indicate that the type of product used can make a significant difference in the fiber levels (Table 10).

Evidence of the possible adverse effects produced by exposure to a substance often starts with a case report. The publication of an observation often encourages further reports. Because only positive reports are considered and negative ones ignored, case reports have an inherent publication bias and are very far from proof of a problem in themselves, but serve as an impetus to further study.

Murphy (1961) published just such a case report in which an individual with occupational exposure to fibrous glass clinically presented with a pneumoconiosis. The accompanying descriptive pathology reported the presence of focal abscesses involving both the terminal bronchiolus and peribronchiolar parenchyma. Glass particulate matter was identified within the tissues. Some slight fibrosis was noted as well. The author concluded that the clinical syndrome and pathology resulted from fiber glass exposure and suggested that fiber glass pneumoconiosis was a clinical entity. The physico-chemical properties of the glass as well as dust measurements were not included in the report. This individual was thought to have been exposed to high levels of dust.

Lee et al (1995) reviewed five case-control studies and eleven cohort studies described in many more papers. Only three of the studies were among workers solely associated with glass fibers, or had subgroups identified with exposure solely to glass fibers. In the other studies, exposures to rockwool or slag wool were included.

In the case-control studies, a specific medical end-point was targeted. Since an increase in lung cancer is the most difficult to quantify in the mortality study particularly with limited smoking histories, this was the only end point in three of the studies. Also, since fatalities represent a more definitive end point than disease, mortality rather than disease incidence was the end point studied in four of the five case-control studies.

In case-control studies, a group of individuals with the selected lung cancer fatality (usually called cases) is compared with a sample of the population from which the cases are drawn but who do not have the cancer (usually called controls). Attempts are made to match the medical history of the controls to that of the cases in all respects other than the cancer itself. In particular, since it is well known that the predominant worldwide cause of lung cancer is tobacco smoking, the smoking history of all participants must be ascertained. This was done in all but one study (Bayliss et al, 1976a, b). Even so, variation and uncertainties in the assessment of cigarette smoking inherently limit the ability of these studies to find small effects.

It is important in any epidemiological study that the participation rate of people in the selected group (or cohort) be high. There is a potential for bias when only some of the cohort (perhaps only those with cancer) participate. It is well known that cases may recall exposure better than controls (recall bias). The participation rate was high in all of the studies, so that this does not apply.

None of these case control studies found a statistically significant increase in cancers caused by specific exposure to glass fibers. However, Enterline et al (1987) found a small, but statistically significant increase in cancers associated with exposure of production workers to rock and slag wool fibers but causality remained unproven.

In a cohort study, a group of individuals who are exposed to the fibers of concern are studied without any prior regard to their medical history. Then the number of cancers seen among the exposed groups is compared to the number in the unexposed group. Ideally, such a study would be prospective in nature (and therefore in a statistical sense a blind study); that is, the cohort would be selected well in advance of the actual outcome, and then selection biases would be minimized. That is rarely possible and due to choice.

In principle, cohort studies can be superior to case control studies especially if they are large enough. The number of people involved in the major studies of SVFs is large enough that the statistical accuracy is high and can, in principle, detect effects as small as a 105, i.e., a five percent increase in the rate of respiratory cancer. However, there are other sources of uncertainty in addition to the statistical ones. The cohort studies for fiber exposure were all retrospective in nature. That is, the investigators initiated the study after both the exposure and the outcome had occurred. The investigators had to use existing historical data to perform their study which did not include a history of exposure to other carcinogens of importance. Most of them did not, for example, include cigarette smoking, which is a very important confounder. If, for example, a higher proportion of glass fiber workers smoked than the comparison group, artifactually elevated rates would been seen of all cancers that are normally attributed to cigarette smoking.

The results of a cohort study are presented as a standard mortality ratio (SMR) this number being the ratio of cancers observed in the subject cohort to those in a population standardized to have the same demographic characteristics of age and gender in that particular period of time. The choice of comparison groups is of vital importance. We here discuss two studies only and refer to the reviews of Lee et al (1995) for more detailed information. In the European collaborative study (Simonato et al, 1987) the SMR was higher if a national population group was used for comparison than a local population group. The local population group is normally considered to be superior for comparison, but when there is a large difference between the two there is reason to be concerned about the reliability of either.

When the cases in the European group are limited to those showing up 20 or more years after the onset of exposure ($20 years latency), the SMR was 139 (significantly different from 100) based on the national mortality group, and 111 (not insignificantly different from 100) if the local mortality group was used for comparative purposes.

In the Marsh et al (1990) study, the SMR was 120 for glass wool production workers compared to national mortality rates (see their Table 3) but dropped to 108 when compared to local rates (see their Table 4) In both cases there is an inverse relationship of mortality with duration of employment (see their Table 6), a relationship counter to the principle of dose response. And, according to the text, no indication of an increased mortality with an increase in intensity of exposure. An increase in the later follow up periods suggests an increase in mortality with age. Since only 497 cancers were observed in the population, any SMR produced under 100 is insignificantly different from 100. The comparable SMR of 112 for glass filament production workers was also not insignificantly different from 100.

However, in one study (Shannon et al, 1984), an SMR of 199 was found, but was based upon observing 19 lung cancers where only 9.53 were expected, a very few number of cases. The effect is drowned by the higher statistical accuracy of the other studies based on more deaths. There seems no reason to believe that the differences observed could be accounted for by higher exposure to glass fibers in the Shannon et al (1994). We conclude that the finding resulted from chance fluctuation.

The difference between the results using national controls and local controls deserves further mention. It is well known that there are differences in death rates in general, and death rates in diseases associated with pulmonary function in particular, among cities and geographic regions (e.g., in the Gulf coast and eastern seaboard areas of the U.S., Blot and Fraumeni, 1979). In a recent study a difference of 47% was found in pulmonary diseases between the most polluted city and the least polluted city (Dockery et al 1993). While we have not examined this issue in great detail, we note the location of the plants which are listed in Table 1 of Enterline et al (1987). All are in the industrial heartland east of the Mississippi River (including Kansas City on the Mississippi River) and north of the Richmond latitude. The largest plant is in Newark located in the Ohio valley. Steubenville, Ohio, was the most polluted city in the six cities cohort study of Dockery et al (1993) with a 50% increase in lung cancer, attributed to fine particles (PM10, probably acid particulates). A difference of 12% between national and local controls is therefore easily accounted for by the differences in air quality. The increase with the period of follow-up is consistent with the increase observed in the 5 cities study.

These facts are important for this study as to whether or not one believes that the increase in mortality in Steubenville, reported by Dockery et al (1993) is due to air pollution (the fine particulate fraction is usually implicated) or the workplace exposures, or both. This crucial conclusion can be expressed in (at least) two ways. Firstly, as noted by Enterline et al (1987), the increased SMR can be accounted for if one uses national rates for comparison which makes the results from local controls even more uncertain than the statistical error suggests, and secondly the effect of working in a factory manufacturing SVFs, if any, is smaller than the effect of merely living in the community.

Mesothelioma is a much rarer cancer and any sign of that would be a real indication that fibers are causing problems rather than one of the other factors in particulate air pollution. But no mesotheliomas have been observed in the epidemiological studies except in groups with strongly suspected (concomitant) asbestos exposure.

From these studies we conclude, in agreement with Lee et al (1995) and Lamm (1994), that no association between glass fiber exposure and respiratory cancers has been demonstrated. However, it is inherently impossible to prove a negative and the precise statement of the risk is that any effect of glass fiber exposure (at the levels of the studies) does not increase the normal lung cancer risk by more than a factor 1.2.

D. INDIRECT EVIDENCE: EXPERIMENTAL STUDIES

A number of in vivo and in vitro experimental studies have been carried out to determine the biological activity of SVFs (see Ellouk and Jaurand, 1994 for a review). The majority of these studies have concentrated on glass fibers (comprising glass wool or glass filament), rock wool and slag wool. Blown insulation can be described as mineral wool which can be either rock wool or glass wool. Experimental animal studies have also been carried out on refractory ceramic fibers, fibers with increasing, but specialized, applications.

A major research program to study the pathological potential of SVFs by inhalation has been sponsored by the manufacturers of these products. The results of two of these studies, Hesterberg et al (1993) and McConnell et al (1994) were focused on for this risk estimate. Inhalation studies in animals for any agent are not often done because they are technically far more difficult and expensive to do than comparable studies utilizing other routes of administration.

For comparison, the earlier inhalation studies in rats of chrysotile asbestos are shown in Table 11. The total lifetime exposure is determined by multiplying the concentration of fibers (fiber/cm³) by the number of hours the rats were at this level. The total lifetime exposures ranged from 250,000 to 318,200,000 fibers/cm³ x hrs. No mesotheliomas were reported to occur among the control rats while lung cancers rarely occur in the controls. However, in the studies where the McConnell and Hesterberg are senior authors, a slight increase in lung cancers has consistently occurred. Mesothelioma ranged from 1.4 to 7.5% among the 4 of 12 experiments in which the disease occurred. While lung cancer occurred in 11 of the 12 experiments and when present ranged from 10.0 to 64.7%. Among the rats with these lifetime exposures, lung cancer occurred more frequently and to a higher percentage than mesothelioma.

The results of the inhalation of synthetic vitreous fibers by rats are shown in Table 12. The total lifetime exposures ranged from 90,000 to 25,000,000 fibers/cm³ x hrs. The experiments with the type of SVFs used for blowing were consistently less than 1,000,000 fibers/cm³ x hrs. These exposures produced no mesothelioma except for refractory ceramic fiber (RCF). The lung cancer incidence was also elevated among the RCF exposed rats although the increase in lung cancer in the other groups of rats exposed to the other types of SVFs did not clearly indicate an increase in lung cancer (Table 12). With the exception of refractory ceramic fibers (RCF) no mesotheliomas occurred.

In order to obtain sufficient sensitivity to observe effects in the small number of animals used it is conventional to expose the animals at the Maximum Tolerated Dose (MTD). This is the dose that is as high as can be achieved without causing toxic effects which influence the induction of cancer. There is considerable debate over whether these two studies made the case that their animals were exposed at or near the MTD and if exposure to higher levels would induce cancer in the experiment animals (Hesterberg et al, 1996). If they did not reach the MTD, the results are not invalidated; but are merely less sensitive than is possible. Of course all fibrogenic agents may cause cancer, through an indirect effect (Merewether, 1938; Hughes and Weill, 1991; Weill, 1994, see discussion in Section E).

The best way one can really get quantitative information from an animal study is to calibrate it with a positive control - that is to demonstrate that a study with the same protocol can find a significant effect when there is exposure to an agent known to produce tumors in humans.

Putting aside the question of whether the Hesterberg et al (1993) experiment was performed at the MTD, there is disagreement about the analysis and interpretation of the data. We believe that the reanalysis by Infante et al (1994) is numerically correct, but we argue that their far reaching conclusions are not justified.

We use a "maximum likelihood method" to fit the data to either a linear (proportional) plot or to a quadratic (fourth degree) equation. This was done using the MSTAGE program of Crouch (1983). The data are collected in Table 12 and 13, both for exposure to MMVF10 and MMVF11 insulation. There was a control population of rats in which there was zero exposure, but the rats were otherwise similar. Infante et al (1994) point out that one could use a pooled control or historical control in which unexposed rats from previous experiments were included.

We first analyzed the data similarly to the way Hesterberg et al (1993) do. We take the experiment control and use data for only one fiber type, MMVF11 (the exposures are 0,41,153 and 246 fiber/cm³ shown in Table 13). The data are consistent (chi² = 4 with 2 degrees of freedom and P = 0.13) with a constant effect independent of dose (no significant linear term). However, using all the points together (assuming that both types of fiber are equivalent), but still using only the experiment control, we find a not very consistent fit with a constant plus a linear term (chi² =13 with 5 degrees of freedom P = 0.025) and with a linear term not significantly different from 0 (at a 95% one sided value = 90% confidence limit)

If we go further and still using all the data points, we take the pooled control instead of the experiment control, we all find a slightly more consistent set (Chi² = 12 with 5 degrees of freedom, and P= 0.03).

Then the best fit to the data, which can be considered to be an equation giving the risk (R) as a function of the exposure (e), is given by:

R = 0.00168 + 0.0000134 e

with an upper 95th percentile confidence limit for the linear term 0.0000238 e. For chrysotile we have a positive control of R = 0.18 at e = 30 or approximately :

R = 0.0017 + 0.0055 e

It is interesting but the upper limit on the linear term is about the same whether one considers each type of insulation separately, or takes them together, or whether one takes pooled (historical) controls or experiment controls.

We believe that the argument about the analysis of these data is misplaced, and both sides are largely wrong and partially right. On the one hand it is conventional to take experiment controls as being more reliable, and to consider each agent separately. Therefore the argument of Infante et al (1994) is an ex post facto one, and they are urging a decision on how to analyze the data with knowledge of the result and the statistical analysis. Making such a decision after seeing the data renders the statistical analysis invalid and adds an uncertainty which has not been quantified (although one could, with some work, quantify it by a Monte Carlo simulation). Therefore the claim by Infante et al (1994) of statistical significance has not been, and probably cannot be, justified.

On the other hand, it is logically impossible to prove that SVFs are not carcinogenic: only that the carcinogenic potency is less than the limit of sensitivity of the experiment. This is important, since the implantation and injection experiments have made a case to be addressed (Ellouk and Jaurand, 1994). We emphasize that all that can be concluded from these experiments of inhalation by rats is that the risk (conservatively calculated by using a linear term in a dose-response relationship) is less than an amount given by these data. In addition, the tumors which occur in the rodents are often histopathologically different from those which occur in man (Kuschner, 1995).

This then leads to the all important question: "How does one compare the inhalation dose of these rats to the inhalation dose of glass insulation workers?" In all uses of animal toxicology to calculate human risk and protect public heath this is a crucial question which has never been fully resolved. Here we are in some disagreement with Hesterberg and Hart (1994) who argue by comparing lung burdens that the exposure is equivalent to many hundreds of times the exposure dose of workers. While that seems superficially a reasonable argument, the issue is important enough that the argument must be tested against a situation where we know the answer. That is the purpose of a positive control. Here we must distinguish between the weight of the inhaled material and the number of fibers.

Since we have chrysotile as a positive control we can make a direct comparison. Chrysotile did not produce as many tumors as the simple argument of Hesterberg and Hart (1994) would suggest. The chrysotile exposure of approximately 100,000 fibers/cm³ produced 18.9% lung tumors (see Table 14 for a summary of fiber characteristics of MMVF10, MMVF11 and chrysotile). What would an exposure similar to the highest blowing insulation wool (~275 fibers/cm³ counting total fibers) produce? One should keep in mind that the current occupational exposure standard for asbestos is 0.1 fibers/cm³. The lifetime exposures to chrysotile asbestos by inhalation and the percentage of lung cancer for all the animal studies are presented in Table 11 and the number of tumors are plotted against this in Figure 1. We note that the highest cumulative exposure from Hesterberg et al (1993) produced fewer tumors in his experiment than in previous ones. If the high-exposure number in Hesterberg et al (1993) is correct, the percentage of lung cancer is not comparable to the exposure and the positive control is of less value.

All of the experimental animal exposures to the types of SVF used in blowing insulation have lifetime exposures on the left hand side of Figure 1 below 1 million fibers/cm³ x hrs. Our reluctant conclusion, therefore, is all that can be said from these animal experiments is that SVFs are not appreciably worse, fiber for fiber, than chrysotile. To make a more definite statement about the safety of SVFs we must make mechanistic arguments.

E. MECHANISTIC RISK ANALYSIS

The mechanistic risk analysis starts with the assumption that we know what is happening when a lung cancer is caused by a large dose of fibers (always asbestos fibers because these are the only fibers for which high doses have existed in humans). The distillation of many years of work by hundreds of investigators has led us to the following tentative conclusions.

There is a fourth item which many scientists believe, but we suggest is uncertain and we decide not to take into account. Merewether (1938) first asked the question (here slightly paraphrased): "Is it asbestos or the asbestosis that it induces, that is the cause of the lung cancer?" Hughes and Weill (1991) report evidence for the latter, in which case the absence of fibrosis among SVF installers is an indication of zero risk. We conservatively assume the former and also assume a linear dose response relationship. It is important to realize that by so doing we calculate a risk where there may in fact be none.

There is yet another factor to consider in this calculation. The fiber counting techniques in the past employed optical microscopy (usually phase contrast). Determination of the chrysotile content of workplace air was only a rough index of its actual presence in that most of the fiber in the air could not be seen. The fibers were beyond the resolution of the technique (fiber diameter # 0.2Fm). The actual chrysotile concentration could have been 5-100 times greater than the value determined microscopically, depending on how the asbestos fiber was being used. On the other hand, blown wool insulation fiber have a larger diameter and a higher proportion of the entire population may be visualized using optical microscopy. A direct comparison of older data obtained by optical microscopy, of chrysotile levels with glass levels, is invalid.

We tentatively assume that these factors will give a value for the potency of SVFs expressed in fibers/cm³ (where surface chemistry, durability, and retention in the lung matter) a factor of 5 reduction below the potency for chrysotile asbestos. We emphasize that this factor is not provable or disprovable either by epidemiology or animal tests. For the carcinogenic potency of chrysotile, the risk calculation is confined to lung cancer.

We start with the assumption that the excess Relative Risk of lung cancer from exposure to chrysotile asbestos can be related to the accumulated exposure by the formula:

Excess Relative Risk = (K_L) x (cumulative exposure)

where the cumulative exposure is measured in fibers/cm³ years.

The factor K_L varies from industry to industry as shown, for example, in Table 2 of Stayner et al. (1996). In the textile industry it is 0.031, but in the friction product manufacturing industry it is 0.00058 or less. The number chosen here is that used by the US EPA and is called by them the unit carcinogenicity factor (K_L = 0.01).

This is defined as a Relative Risk, relative to the risk in a normal, unexposed population. We derive this from the numbers in Table 126 of the Statistical Abstract of the United States (1993). Fifty-nine respiratory malignancies (mostly lung cancer) develop per 100,000 persons in the USA or 7% of the 855 deaths in the same population.

It is generally believed, and we here assume, that smoking and exposure to fibers are synergistic in causing lung cancer, and it is the risk relative to the lung cancer incidence in the appropriate population (smokers or non-smokers) that is the same. We must therefore calculate this risk separately for these two groups. We do this by assuming that 90% of these lung cancers occur among the half of the population that smokes cigarettes and only 10% among the half that are non-smokers. Taking this into account, 53 of the respiratory malignancies are among the 50,000 smokers, with 427 deaths or 12% of the deaths among smokers. Only 6 of the 50,000 non-smokers developed respiratory malignancies and therefore only 1.4% of the deaths among non-smokers were from respiratory malignancies.

Then the absolute excess risk of lung cancer for someone exposed for a year to 1 fiber/cm³ of chrysotile becomes:

R = 0.12 X 0.01 = 0.0012 = 1.2 X 10^-3 (among smokers)

R = 0.014 X 0.01 = 0.00014 = 1.4 X 10^-4 (non-smokers)

It is this risk that must be compared with other typical occupational risks (expressed as a risk per year of the occupation) in Table 15.

For installers of blown wool without the use of respirators we assume that a worker is exposed to a concentration of 1 fiber/cm³. This is derived from the suggestions in Section B of this report that glass insulation blowers are exposed to an average of 1.7 fibers/cm³ but for crew members who alternate tasks the average exposure is less. Reducing the carcinogenic potency of blown glass wool fibers by a factor of five compared to chrysotile, the risk becomes:

R = 0.00024 = 2.4 X 10^-4 for smokers and

R = 0.000028 = 2.8 X 10^-5 for nonsmokers.

These are entered into Table 15.

We further estimate that a 8710-NIOSH type respirator can further reduce this by a factor of 5 to 10. We therefore divide this risk by a further factor of 5 to find the appropriate risk for a user of respirators as also shown in Table 15. (See Boehlecke et al. 1996 for a detailed discussion of respirators).

The occupational risks listed in Table 15 are calculated per year of engaging in the occupation. For a typical 25 year working period, the cumulative risk becomes 25 times larger. These risks are compiled from a variety of sources. Some are from Crouch and Wilson (1993) but most are updated from historical records in the Statistical Abstract of the United States (1993). For example, there has been a President of the United States for 207 years. During that time, three Presidents have been assassinated. The risk is then very simply 3/207 = 0.014 or 1.4 X 10^-2.

This list can guide us, but should not force us, to take action about specific risks. The highest occupational risk in Table 15, that of being President of the United States is well calculated from objective historical data and has only the uncertainty of extrapolation from the past to the future. Yet the national pastime in which we are engaged every four years show that this risk does not discourage people from actively seeking the job, and even in some cases, paying millions of dollars of their own money to get it.

We note that the risk to installers of blown wool insulation using these pessimistic (or conservative) assumptions is smaller than many occupational risks. We do not include the risk to the public, but concentration measurements in houses are very low and even though the occupancy is high, the risk should be negligible. Even the more pessimistic authors of the paper of Infante et al (1994) agree with this last point.

Further, what are the risks associated with use of asbestos insulation as compared to glass? Using data from Selikoff and Seidman (1991), almost 16% of the observed asbestos insulator mortality was due to excess lung cancers and about 10% of their mortality was due to mesothelioma. For ten thousand such workers, 25 years of occupational pursuit produced an unanticipated excess of 1,600 lung cancers and about 1,000 mesotheliomas. Asbestos insulators possessed vastly greater risk of death due to cancer than risks found among glass insulators. When simply comparing trades (not the exposure, dose, or working conditions), glass trades carry significantly less risk.

F. REGULATORY AND OTHER POSITIONS

There are a number of claims that glass fibers may be carcinogenic and should be treated, from a regulatory standpoint, as carcinogens (see Appendix II for current labeling).

For example, the International Agency for Research in Cancer (IARC), a division of the World Health Organization (WHO), classified glass wool, rock wool, slag wool and ceramic fibers possibly carcinogenic to humans (Group 2B) (IARC 1988). This was based upon the studies where the SVFs were injected into experimental animals (see Table 16-18 for summary of the results of the animal studies for chrysotile asbestos and SVFs) and the limited epidemiological data most of which were discussed earlier. The Environmental Protection Agency has a slightly different classification procedure and labels man-made mineral fibers "Class C" materials.

It is important to understand the criteria for classification. "Possibly carcinogenic" means that there exist data from animals, exposure circumstances, and limited epidemiological data, that suggest that at some dose (perhaps very high), the materials might cause cancer in humans. Critics of the classification process argue that IARC has no classification for those materials where no data have yet been produced, although Group 3 (not classifiable) and Group 4 (probably not carcinogenic) also exist. Some critics argue that it is prudent to assume that any substance can cause cancer if the exposure, and hence the dose, is large enough. In Section D and E a discussion of the mechanisms by which it is believed that fibers operate and sometimes produce cancer. But there are great differences in fiber potency which may not show up in animal studies, and therefore the analogy with asbestos, if not properly done, can be highly misleading. It can call attention to issues that are almost irrelevant to the health risk of those occupationally exposed and contribute to regulatory and legal impasses.

Whatever the scientific merit, or demerit, of the IARC classification, there is no doubt that it has a strong influence on regulation. It is not, therefore surprising that there is a lot of attention to the detail of a classification especially the substance is a borderline case. While it might well be argued from the information earlier in this report that SVFs should not have been classified as 2B, the classification has been made. Such a classification has never been reversed. For the reasons noted in this report, we doubt whether that will be done at least in the foreseeable future. However, it should be possible, with effective work practices, to control this low risk. It is important to note that, while there is slight direct evidence for carcinogenicity of slag wool fibers, which tend to support the IARC classification, there are important differences of glass wool fibers from the slag wool that make the likelihood that glass wool poses a carcinogenic risk at similar exposures much smaller. If effective dose is taken into consideration, as we believe it should be, the data suggest that glass fibers will not produce cancer at the doses of interest.

Infante et al (1994) purport to demonstrate unequivocally that glass fibers cause cancer. However, this seems to be based on a reanalysis of the animal data which, while numerically correct, is a post hoc selection of criteria and therefore statistically invalid.

The question still arises, "What is the appropriate safety standard for exposure of workers to glass wool fibers?". The comparative table of risks suggests that the present procedures have kept the risk (at least) a factor of ten below typical, occupational risks and could therefore be deemed adequate.

But unfortunately common sense is often replaced by rigorous regulation. In this connection we note that Infante et al (1994) do not recommend any drastic action: merely that workers should be properly informed. They argue that the classification is the crucial part of such information. We prefer our table of comparative risks. For the sophisticated both are useful.

Our review of the various studies bring us to the conclusion that it is vital to keep concentrations of, and hence exposure low to, fibers of the relevant fiber sizes, surface chemical activity and lung clearance time. However we are disappointed in the amount and quality of the data on the relevant exposures. If further reassurance of workers, public or regulators is needed, we recommend that considerably more attention be given to a proper characterization of exposure.

While it should be clear from Table 15 that even installers of loose wool have an annual risk that is not larger than risks of other occupations, common sense suggests that care be taken. While we think that regulation is probably inappropriate but good work practices, e.g., wearing an approved respirator, should be a normal precaution. We suggest any approved EPA respirator for asbestos work will suffice (see Boehlecke, 1996 for a discussion of respirators). For installers of glass batts, a lower efficiency rating is adequate, but for installers of blown loose insulation a better design is required. This should reduce the exposure of glass insulators to that of the manufacturers without respirators.

There is some indication that fiber manufacturers over the last twenty years had made fluffier products. For a given amount of material, this leads to a higher fiber count and therefore increased exposure. Although the total mass of suspended particulate remains similar (Table 5 and 7). We urge that the industry monitor this trend which can increase concentrations and exposures, and make efforts to reverse the trend.

At the meeting at Sydney, Australia in October 1995, Eastes, Hadley and Bender (1995) discussed noted efforts to decrease the chemical sensitivity and the durability of the product. We recommend that such efforts continue and be intensified.

REFERENCES

Bayliss DL, Dement JM, Wagoner JK, Blejer HP (1976a). Mortality patterns among fibrous glass production workers. Ann NY Acad Sci 271: 324-335.

Bayliss DL, Dement JM, Wagoner JK (1976b). Mortality patterns among fibrous glass production workers - provisional report. In: Occupational Exposure to Fibrous Glass. Proceedings of a Symposium, Maryland, 1974, 349-363. HEW Publication No. (NIOSH) 76-151, U.S.D.H.E.W., Washington, D.C.

Blot WJ, Fraumeni, JF Jr. (1979). Lung cancer mortality in the United States: Shipyard correlations. Ann NY Acad Scis 330, 313-315.

Boehlecke BA (1996). Respirators. Chapter 59. In: Occupational and Environmental Lung Disease. Harber P, Schenker MB, Balmes JR (Editors). Mosby, St. Louis, MO, p. 963.

Brown RC, Davis JMG, Douglas D, Gruber UF, Hoskins JA, Ilgren EB, Johnson NF, Rossiter CE, Wagner JC (1991). Carcinogenicity of the insulation wools: reassessment of the IARC evaluation. Reg Tox Pharmarcol 14: 12-23.

Bunn WB, Bender JR, Hesterberg TW, Chase GR, Konzen JL (1993). Recent studies of man-made vitreous fibers. JOM 35: 101-113.

Christensen VR, Eastes W, Hamilton RD, Strauss AW (1993). Am Ind Hyg Assoc J 54: 232-238.

Crouch EAC (1983). A program for analysis of carcinogenic potency: MSTAGE. Privately circulated.

Crouch EAC, WIlson R (1983). Risk Benefit Analysis. Ballinger Press, Cambridge, MA.

Davis JMG, Beckett ST, Bolton RG, Collings P, Middleton AP (1978). Mass and number of fibers in the pathogenesis of asbestos-related lung disease in rats. Br J Cancer 37: 673-688.

Davis JMG, Addison J, Bolton RE, Donaldson K, Jones AD and Smith T (1986). The pathogencity of long versus short fiber samples of amosite asbestos administered to rats by inhalation and intraperitoneal injection. Br J Exp Pathol 67: 415-430.

Davis JMG, Jones AD (1988). Comparisons of the pathogenicity of long and short fibers of chrysotile asbestos in rats. Br J Exp Pathol 69: 717-737.

Davis JMG, Bolton RE, Miller BG, Niven K (1991). Mesothelioma dose response following intraperitoneal injection of mineral fibers. Int J Exp Pathol 72: 263-274.

Dockey DW, Pope CA, Xu X, Spengler JD, Ware JH, Fay ME, Ferris BG, and Speizer FE (1993). An association between air pollution and mortality in six U.S. cities. N Engl J Med 329: 1753-1759.

Eastes W, Hadley, JG, and Bender, JR (1995). Biological approaches to the design of safe synthetic vitreous fibers. Symposium on the Health Effects of Fibrous Materials (excluding asbestos) used in Industry Sydney, Australia.

Ellouk SA, Jaurand MC (1994). Review of animal and in vitro data on biological effects of man-made fibers. Environ Health Perspect 102 (Suppl 2): 47-63.

Enterline PE, Marsh GM, Henderson V, Callahan C (1987). Mortality update of a cohort of U.S. man-made mineral fiber workers. Ann Occup Hyg 31: 625-656.

Esmen NA, Corn M, Hammad YY, Whittier D, Kotsko N (1979). Summary of measurements of employee exposures to airborne dust and fiber in sixteen facilities producing man-made mineral fibers. Am Ind Hyg Assoc J 40: 108-117.

Esmen NA, Hammad Y, Corn M, Whittier D, Kotsko N, Haller M, Kahn RA (1978). Exposure of employees to man-made mineral fibers: Mineral wool production. Environ Res 15: 262-277.

Esman NA, Sheenan MJ, Corn M, Engel M and Kotsko N (1982). Exposure of employees to man-made vitreous fibers: Installation of insulation materials. Environ Res 28: 386-398.

Head IWH, Wagg RM (1980). A survey of occupational exposure to man-made mineral fiber dust. Ann Occup Hyg 23: 235-258.

Hesterberg TW, Miiller WC, McConnell EE, Chevalier J, Hadley JG, Bernstein DM, Thevenaz P, and Anderson R (1993). Chronic inhalation toxicity of size separated glass fibers in Fisher 344 rats. Fundam Appl Toxicol 20:464-476.

Hesterberg TW and Hart GA (1994). A comparison of human exposures to fiberglass with those used in a recent rat chronic inhalation study. Reg Tox and Pharmacol 20: S35-S46.

Hesterberg TW, McConnell EE, Miiller WC, Chevalier J, Everitt J, Thevenaz P, Flessner H, Oberdorster G (1996). Society of Toxicology, Abstract.

Hill RJ, Edwards RE, Carthew P (1990). Early changes in the pleural mesothelioma following intrapleural inoculation of the mineral erionite and the subsequent development of mesothelioma. J Exp Pathol 71: 105-118.

Hughes JM, Weill H (1991). Asbestosis as a precursor of asbestos-related lung cancer: Results of a prospective mortality study. Br J Ind Med 48: 229-233.

IARC (1988). Man-made mineral fibers and radon. Monographs on the Evaluation of Carcinogenic Risks to Humans International Agency for Research on Cancer (IARC), WHO, Lyon, France, pp. 39-171.

Infante P, Schuman LD, Dement JM, and Huff J (1994). Fibrous glass and cancer. Amer. Journ. Indust. Med. 26:559-584.

Jacob TR, Hadley JG, Bender JR, Eastes W (1992). Airborne glass fiber concentrations during installation of residential insulation. Am Ind Hyg Assoc J 53: 519-523.

Johnson DL, Healey JJ, Ayer HE, Lynch JR (1969). Exposure to fibers in the manufacture of fibrous glass. Am Ind Hyg Assn J. 30: 545-550.

Kuschner M (1995). The relevance of rodent tumors in assessing carcinogencity in human beings. Reg Tox and Phar 21: 250-251.

Lamm, SH (1994). Is fiberglass exposure a carcinogenic risk to humans? Report to National Association of Home Builders.

LeBouffant L, Daniel H, Henin JP, Martin JC, Normand C, Tichoux C, Troulard F (1987). Experimental study on long-term effects of inhaled MMMF on the lungs of rats. Ann Occup Hyg 31: 765-790.

Lee I-Min, Hennekens CH, Trichopoulos D, Buring JE (1995). Man-made vitreous fibers and risk of respiratory system cancer: A review of the epidemiological evidence. JOEM 37:725-738.

Lees PSJ, Breysse PN, McArthur BR, Miller ME, Rooney BC, Robbins CA, Corn M (1993). End-user exposures to man-made vitreous fibers: I. Installation of residential insulation products. Appl Occup Environ Hyg 8: 1022-1030.

Marsh GM, Enterline PE, Stone RA, Henderson VL (1990). Mortality among a cohort of U.S. man-made mineral fiber workers: 1985 follow-up. J Occup Med 32: 594-604.

McConnell EE, Wagner JC, Skidmore JW and Moore JA (1984). A comparative study of the fibrogenic and carcinogenic effects of UICC chrysotile B asbestos and glass microfiber (JM100). In: Biological Effects of Man-Made Mineral Fibers Vol 2, pp. 234-250. WHO Regional Office, Copenhagen.

McConnell EE, Kamstrup O, Musselman R, Hesterberg TW, Chevalier J, Miiller WC, Thevanaz P (1994). Chronic inhalation study of size-separated rock and slag wool insulation fibers in Fischer 344/N rats. Inhalation Toxicology 6: 571-614.

Merewether (1938). Report of the Chief Inspector of factories, UK.

Muhle H, Pott F, Bellmann B, Takenaka S, Ziem U (1987). Inhalation and injection experiements in rats to test the carcinogenicity of MMMF. Ann Occup Hyg 31: 755-764.

Murphy GB (1961). Fiber glass pneumoconiosis. Arch Environ Health 3: 102-108.

Musselman R, Miiller WC, Eastes W, Hadley JG, Kamstrup O, Thevenaz P, Hesterberg TN (1994a). Biopersistence of crocidolite versus man-made vitreous fibers in rat lungs after brief exposures. In: Toxic and carcinogenic effects of solid particles in the respiratory tract. Mohr U, Dungworth DL, Mauderly JL and Oberdoster G, International Life Sciences Institute/ILSI Press, pp. 451-454.

Musselman R, Miiller WC, Eastes W, Hadley JG, Kamstrup O, Thevenaz P, Hesterberg TN (1994b). Biopersistence of man-made vitreous fibers and crocidolite fibers in rat lungs following short-term exposures. Environ Hlth Perspet 102 (Suppl 5): 139-143.

Nolan RP and Langer AM (1993). Limitations of the Stanton hypothesis. In: Health effects of mineral dusts. Guthrie GD, Mossman BT (eds). Reviews in Minerology 28: 309-326.

Pigott GH, Gaskell BA and Ishmall J (1981). Effects of long-term inhalation of alumina fibers in rats. Br J Exp Pathol 62: 323-331.

Pott F, Zeim U, Reiffer FJ, Huth F, Ernst H, Mohr U (1987). Carcinogencity studies on fibres, metal compounds, and some other dusts in rats. Exp Pathol (Jena) 32: 129-152.

Proceedings Symposium on Synthetic Vitreous Fibers: Scientific and Public Policy Issue (1994). Reg Tox Phar 20: 51-5224.

Selikoff IJ, Seidman H (1991). Asbestos-associated deaths among insulation workers in the United States and Canada, 1967-1987. Ann NY Acad Scis, 643, 1-14.

Shannon HS, Hayes M, Julian JA and Muir DCF (1984). Mortality experience of glass fibre workers. Br J Ind Med 41: 35-38.

Simonato L, Fletcher AC, Cherry JW, Andersen A, Bertazzi N, Charnay N, Claude J, Dodgson J, Estwe J, Frentzel-Beyme R, Gardner MJ, Jensen OM, Olsen JH, Saracci R, Teppo L, Winkelmann R, Westerholm P, Winter PD, Zocchetti C (1987). The International Agency of Research on Cancer historical cohort study of MMMF production workers in seven European countries: Extension of the follow-up. Ann Occup Hyg 31: 603-623.

Smith DM, Ortiz LW, Archuleta RF, Johnson NF (1987). Long-term health effects in hamsters and rats exposed chronically to man-made vitreous fibers. Ann Occup Hyg 31: 731-754.

Statistical Abstract of the United States (1993). U.S. Bureau of Census (113th Edition), Washington, D.C.

Stayner LT, Dankovic DA, Lemen RA (1996). Occupational exposure to chrysotile asbestos and cancer risk: A review of the amphibole hypothesis. Am J of Public Hlth 86: 179-186.

Wagner JC, Berry G, Timbrell V (1973). Mesotheliomas in rats following inoculation with asbestos. Br J Cancer 28: 173-179.

Wagner JC, Berry G, Skidmore JW, and Timbrell V (1974). The effects of inhalation of asbestos in rats. Br J Cancer 29: 252-269.

Wagner JC, Berry G, Skidmore JW (1976). Studies of the carcinogenic effects of fiber glass of different diameters following intrapleural inoculation in experimental animals. In: Occupational Exposure to Fibrous Glass. Proceedings of a Symposium, Maryland, 1974, 193-7. HEW Publication No. (NIOSH) 76-151, U.S.D.H.E.W., Washington, D.C.

Wagner JC, Berry G, Hill RJ, Munday DE and Skidmore JW (1984). Animal experiments with MMM(V) -- Effects of inhalation and intrapleural inoculation in rats. In: "Biological effects of man-made mineral fibers" Vol 2, pp. 209-233. WHO Regional Office, Copenhagen.

Wagner JC, Skidmore JW, Hill RJ and Griffith DM (1985). Erionite exposure and mesothelioma in rats. Br J Cancer 51: 727-730.

Wagner JC, Griffith DM and Munday DE (1987). Experimental studies with palygorskite dust. Br J Ind Med 44: 749-763.

Wagner MMF, Wagner JC, Davies R, Griffiths DM (1980). Silica-induced malignant histiocytic lymphoma: Incidence linked with strain of rat and type of silica. Br J Cancer 41: 908-917.

Weill H (1994). Biological effects: Asbestos-cement manufacturing. Ann Occup Hyg 38 (No.4): 533-538.

Wilson R and Crouch EAC (1987). Risk assessment and comparisons: An introduction. Science 236: 267-270.

Wong O, and Musselman RP (1994). An epidemiological and toxicological evaluation of the carcinogenicity of man-made vitreous fiber with a consideration of co-exposure. Journ. Env Path. Tox. and Onc. 13: 169-180.

World Health Organization (1985). Reference methods for measuring airborne man-made mineral fibers (MMMF). Environmental Health Series 4, Copenhagen.