The Québec miners and millers cohort study (1–4) is one of the most important sets of epidemiological evidence for the study of asbestos health effects, including potential differences between chrysotile and the asbestiform amphiboles’ carcinogenic potential and mesothelioma potency.

At the same time, methodology and quality of exposure estimations for this cohort has been questioned (5). Recently, the U. S. EPA (6) excluded Québec data from its potency estimation for chrysotile, potentially causing significant bias in the meta-analysis because the largest data source was ignored. Darnton (7) demonstrated that exclusion of Québec information from mesothelioma meta-analysis can artificially elevate the potency factor for chrysotile. Burdett (8) argued that exclusion of the Québec cohort does not seem reasonable given the plausible extent to which this data point could be errant, and doing so would likely bias the chrysotile risk estimation upwards considerably”.

In fact, the Québec cohort has consistently been that with the largest absolute numbers of mesothelioma deaths among chrysotile-exposed cohorts. Given the historic nature of the cohort, “comprising all those born between 1891 and 1920 who had ever been employed in any part of the industry for more than 1 month” (1), it is not surprising that the confidence with which deaths were ascribed to mesothelioma varied, as did possibilities for alternative asbestos exposures. As of 1972, seven cases had been identified, but one was rejected by the Canadian Mesothelioma Panel of Pathologists, a second probably had domestic exposure exceeding his cohort exposure, and a third had an uncertain history of having ever worked in the industry at all. All were nonetheless included in cohort study and carried forward as mesothelioma deaths in the cohort. By the end of 1975, 10 male and one female death due to pleural mesothelioma had been identified, nine from death certificates and one at necropsy (2), of a total number of 4,463 male deaths in a cohort of 10,939 and 84 deaths among 440 eligible women. By the end of 1988, there were 33 deaths from mesothelioma in the cohort, 28 in miners and millers and five employees of a small asbestos products factory where commercial amphibole asbestos had also been used (3). As of final publication, there were overall 8,009 deaths in the cohort to end 1992, and five additional mesothelioma deaths had been added (4).

Like the 1973 assessment, the 1997 cohort analysis was inclusive for all possible cases of mesothelioma. The total of 38 cases (including five from the Johns-Manville factory where crocidolite was used) included all deaths in which the term mesothelioma was mentioned on the death certificate, and all deaths coded from 1978 to ICD 163 (cancer of pleura); by ongoing ascertainment of malignant mesothelioma through all Canadian pathologists, 1966–1984; and by including data from more recent studies by other Québec investigators with special interest in mesothelioma. Diagnosis was “confirmed by full autopsy in 27 cases, but in eight of these the pathologist expressed an element of doubt.” Overall, the “admittedly subjective assessment of the diagnosis being correct was high in 19, moderate in 14, and low (though more likely than not) in five”.

As noted by Goodman et al. (9), diagnostic error for mesothelioma can contribute to misclassification of risk in either direction. Although, as noted, pathology reports—most often autopsy-based—were available for the Québec cases, they relied histologically on Hematoxylin and Eosin (H & E) staining, since no cases were diagnosed after 1992. More recent studies have been improved by the advent, in the late 1990s, of immunohistochemical (IHC) markers of mesothelial cell origin, although these vary in sensitivity and specificity. The gradual evolution and improvement in certainty of pathology diagnosis attributable to IHC advances is described elsewhere (10).

In addition to diagnostic doubt, a significant impediment to data quality assessment for the Québec cohort lay in the fact that the majority of the exposure assessment for chrysotile miners and millers had necessarily been performed based on the original midget impinger dust measurements available at the time, and not on actual fiber counts. Berman and Crump (11, 12) assumed that there is a constant ratio that could be applied to the Québec impinger concentrations to derive the fiber concentration in f/cc. This coefficient was assumed to be equal to 3. However, as Burdett (8) noted, the utilization of a ratio value for midget impinger and Phase Contrast Microscopy (PCM) fibers would require additional substantiation. Also, Robock, (13), expressed doubts about the methodological feasibility of recalculating midget impinger concentration to PCM (f/ml, or f/cc) estimations. Robock correctly asserted that the range of the ratio between PCM and midget impinger concentrations varied significantly by industry, or even at the same location. Certainly, a relationship between fiber concentration by PCM and midget impinger measurements exists objectively: one should increase with another. However, there are numerous confounding factors for this relationship. It is expected that additional studies are beneficial to see if the objective relationship between the two methods can be approximated using available data, at least for a certain location (as in the Québec mining and milling cohort study).

The purpose of this paper is to explore whether use of an expanded set of now available original data for parallel measurements of asbestos concentrations in the Québec mines and mills in MPCF and in fibers/cc supports average estimation of cumulative exposure as used in the meta-analysis recently updated by Darnton (7, 14). While there are difficulties in reconstructing cumulative exposure in PCM units for every person in the cohort (because of limited numbers of parallel measurements and the presence of confounders), we can demonstrate that estimations of average exposure for all cohort members can be made with good predictive quality.

We demonstrate that application of strict quality-based criteria to mesothelioma cases in the study also helps to confirm estimations of chrysotile potency as published by Darnton. We use the original set of chrysotile datapoints from Darnton (7, 14), supplementing it with recently published updated exposure values for Balangero miners and millers (15), as well as from the new IARC publication on mesothelioma in chrysotile asbestos miners and millers in Russia (16). We test the changes observed in the meta-analysis performed by Darnton if newly-derived estimations of average cumulative exposure in Québec miners and millers are used.

We also explore estimations of minimal cumulative exposure in confirmed mesothelioma cases and show that there is no data on any individual mesothelioma in the Québec cohort with chrysotile exposure lower than a certain value. Fully proving the existence of a threshold for mesothelioma is an impossibility so long as a linear no-threshold model is adhered to; in the words of Hodgson and Darnton (17), this is a “logical nonsense.” However, a lowest observed adverse effect level for diagnostically confirmed mesothelioma cases. Without known alternative asbestos exposures and with the enhanced exposure assessments now available among the 8,009 deaths of chrysotile workers in this cohort may serve as a “practical” guide, informing toxicological studies and regulatory efforts.

We also attempt the use of several different models that also support the concept of the existence of a chrysotile exposure range not associated with excess mesothelioma risk.

It should be emphasized that this paper is focused on a very important topic of historical exposure data reliability. Human epidemiology provides the most critical part of evidence for carcinogenicity of various agents. Burdett (8) reviewed the reliability of risk estimations based on historical data for workers’ exposure to various mineral types of asbestos. Burdett demonstrated that consistency of the results of meta-analysis for crocidolite, amosite, and chrysotile, as demonstrated in major studies during the recent decades, can be considered the evidence of reliability of exposure information. In this paper, we will explore the role of method-to-method conversion (in this case, midget impinger to optical microscopy) for deriving mesothelioma slope factors. We will use several new approaches (including statistical simulation of possible uncertainty of average exposure estimates vs. individual datapoints) to substantiate the reliability of historical data assessment. We will also demonstrate how minimal reported exposure associated with the cases of disease can be estimated and how it can be used for derivation of “practical threshold” for chrysotile asbestos.

The potency of chrysotile for the production of mesothelioma is a long-contested issue. Meta-analyses performed over the last few decades have suggested that chrysotile mesothelioma potency is significantly lower than that for commercial crocidolite and amosite, as well as for Libby asbestiform amphiboles (LAA) (7, 12, 14). Berman and Crump (28) demonstrated that the mesothelioma potency of chrysotile might not be statistically different from zero. The “tremolite asbestos hypothesis” that suggested that much of the mesothelioma potency of chrysotile can be attributed to amphibole asbestos contamination was never disproved (29). Korchevskiy et al. (30), nevertheless, demonstrated that non-zero, though very low, chrysotile potency can be derived from chemical composition and dimensions of fibers.

At the same time, some authors have questioned the quality of available exposure data for asbestos-related disease mortality in chrysotile-exposed cohorts (5). PCM-estimated exposure levels in the Québec chrysotile cohort were questioned as to possible bias caused by conversions from the available midget impinger “dust” data expressed in MPCF to membrane-filter fiber/cc counts.

Lenters et al. (5), in a study limited to lung cancer risk, proposed five “quality criteria” for chrysotile asbestos cohorts, suggesting that analysis restricted to studies with “few quality limitations of the exposure assessment component, the epidemiological evidence base is too sparse to draw conclusions about potency differences per fiber type.” Berman and Case (31), however, argued that there is no evidence that applying the Lenters et al. “quality criteria” to exposure values in asbestos epidemiology would change the estimated risk of lung cancer. Hodgson (32) also noted that limiting meta-analysis to the five “quality criteria” chosen by Lenters et al. (5) potentially erase important epidemiological evidence.

Of greatest importance for lung cancer, meta-analyses of chrysotile asbestos potency are most affected by two studies, one of which is the Québec cohort, the other, a study of South Carolina asbestos textile workers. The South Carolina cohort was assumed in some publications not to have had other amphibole asbestos exposures, despite evidence to the contrary from multiple lung-retained fiber studies. Nevertheless, omission of either of the two cohorts leads to “overreliance on a single study” (31, 33).

For mesothelioma, our study helps to ascertain both disease and exposure quality measurement factors and significance in the Québec study, which remains the chrysotile cohort with the largest number of mesothelioma cases recorded. Following the same logic as for lung cancer, we believe that exclusion of this cohort from meta-analysis (as was recently attempted by the U. S. EPA in TSCA risk evaluation for chrysotile) (6), would lead to a loss of significant epidemiological information and bias the outcome of risk analysis beyond any practical applicability.

The Québec cohort is also of interest when the question of threshold for exposure is considered. The existence of a threshold for asbestos exposure in mesothelioma is supported by various studies and approaches (as reviewed by Goodman et al., 2025) (9). The level of threshold exposure in the Québec study can be assessed from various positions, as well. In particular, the minimum level of exposure in mesothelioma cases (or lowest observed adverse effect level) can be useful for the overall argument, though not being a “proof” for threshold, but providing scientists and regulators with reasonable practical benchmarks of mesothelioma risk.

Our paper provides additional arguments for inclusion of data on mesotheliomas in the Québec miners and millers cohort in meta-analysis for chrysotile potency. We utilized the originally published datapoint pairs (22) for parallel analysis of midget impinger and PCM measured concentration for the Québec asbestos industry that were not previously available for this type of analysis. Regression equations were determined between fiber concentrations as dependent variable and midget impinger concentration as predictor. If all original datapoints from Dagbert were utilized as published, the regression is statistically significant, but with comparatively low correlation coefficient (R = 0.46, R² = 0.21, p < 0.00001). However, if concentrations by midget impinger and PCM values are averaged by groups of 30 datapoints, the correlation improves (R = 0.91, R² = 0.82, p < 0.000001). We assumed that both regressions in our study can be used in comparisons, providing estimations for fiber concentration based on MPCF values.

We suggested that the application of log–log regression equation based on the original data provides a cumulative exposure estimation for the cohort that is not substantially different from what was used previously, and likely even higher than previous estimations. The average cumulative exposure of 528.5 f/cc-years (95% CI 510.3, 546.7) based on Equation 4 in our study and 979 f/cc-years (95% CI 945, 1,013) based on Equation 5 are of the same range as previously published estimates of 600 f/cc-years.

We performed a statistical simulation and demonstrated that both Equations 4, 5 can serve for estimation of average potency for selection of datapoints, similar to the dataset we used on our study. In particular, all average values of chrysotile estimated by Equation 4 are expected to be lower than observed by a factor of 2 and greater. The corresponding ratio for Equation 5 will be closer to 1, with 0.8 being the lower bound. This means that risk estimations based on Equation 4 will be always conservative, with at least a safety factor of 2; however, Equation 5 provides a better estimation of mesothelioma risk factors.

Our work allows us to update the estimation of chrysotile potency that Darnton published in 2023–2024 (7, 14). In addition, we included a cohort from Russia published by IARC to meta-analysis (this cohort was not considered in Darnton’s work). Utilization of Equation 5 (that was demonstrated to better approximate exposure values) put the estimated chrysotile potency at the level of 0.0010–0.0011%, vs. 0.0014% in Darnton (7, 14). If Equation 4 were used, the overall chrysotile potency would be about 0.0017%, but this would include the potency of Québec chrysotile over-estimated by a factor of 2 and higher.

We also explored implications of our approach for determination of a “practical threshold” related to chrysotile exposure and mesothelioma. The LOAEL estimation of minimal cumulative exposure in mesothelioma cases demonstrated values that ranged from 4.8–6 MPCF-years, or 18–50 f/cc-years.

We also performed a statistical simulation to generate larger intervals of cumulative exposure values in correspondence with mesothelioma rate. The linear-log model was utilized. Based on the developed models we described, the average threshold value is 6.38 MPCF-years (standard deviation 2.6), or in the range from 35.9 f/cc-years (standard deviation of 15.4) to 65.9 f/cc-years (standard deviation of 29.6).

A simulation with random combination of datapoints would suggest even higher threshold levels for average cumulative exposure. In particular, no mesothelioma cases in the Québec cohort were determined for average cumulative exposure levels of 113 MPCF-years by impinger, 309 f/cc-years by PCM Equation 4, or 577 f/cc-years by Equation 5.

Analysis of lung fiber burden, with values derived from McDonald et al. (4) and using the methodology therein, also confirmed that concentrations of fibers in lung tissue can be related to cumulative exposure, estimated based on the proposed log–log regression equation. Our study has demonstrated that the central tendency estimation when chrysotile lung burden concentration is used in this way as exposure proxy is not expected to be elevated for cumulative exposure below about 178–300 f/cc-years, nor tremolite asbestos below about 5 f/cc-years.

It should be noted that chrysotile does accumulate in the lungs of miners and millers, as does tremolite. Unlike the situation in experimental animals, where it has been shown that rapid clearance of chrysotile takes place after exposure, these workers were largely exposed to very high concentrations of chrysotile for decades, with resulting lung accumulation of chrysotile fibers (4).

Table 6 contains data on clearance rates for chrysotile asbestos in human lungs in comparison to other types of minerals, based on the summary from various sources as demonstrated in Korchevskiy and Wylie (34).

As we can see, chrysotile is expected to be quickly cleared from human lungs, and tremolite to stay there much longer. It is noteworthy that in the lungs of Québec workers, chrysotile was observed at substantial levels, that can be the indicator of slowed clearance rate, or continuous exposure up to the moment that the biopsy was performed. It is possible, however, that even if chrysotile fibers remained in the lungs of workers, the level of lung burden might be below a threshold level that would be sufficient to increase the risk of disease.

Interestingly, meta-analysis by Goodman et al. (9) suggested that, if threshold for chrysotile exists, it can be found at the level of 90 f/cc-years, and tremolite asbestos “threshold” in the range from 4.3 to 10.8 f/cc-years.

Additional studies would be needed to perform a full sensitivity analysis for the relationship between threshold level and various assumptions. However, consistency of threshold estimations between various recent publications is encouraging that at least “practical thresholds” for various mineral types of asbestos can be derived from available data.

As we can see, it was demonstrated in our study that more robust estimations of cumulative exposure for the Québec cohort are not expected to significantly change estimations of meta-analytical potency factor as calculated by Darnton, even if new IARC estimation of mesothelioma mortality in Russian chrysotile cohort would be included to the analysis. We demonstrated that mesothelioma potency for all types of chrysotile can be estimated as 0.0014% (95% CI 0.0010, 0.0.0018), and for non-textile chrysotile 0.0011% (95% CI 0.0008, 0.0015), fully confirming Darnton’s analysis.

Several approaches allowed to suggest different estimations of possible threshold on mesothelioma risk depending on cumulative exposure. Our analysis cannot demonstrate existence of a threshold but can suggest what the estimation should be if existence of threshold is proven on a biological basis. The range of estimation is quite wide. An especially important observation is that the detected threshold on average exposure in a cohort can be significantly higher than for individuals. The reason for this difference is related to the fact of uncertainties of possible alternative exposures (as well as variability in spontaneous mesothelioma rates) in the cohort members that is smoothed out by average cumulative exposure estimated for a group of people. Analyzing Table 1, we can see that among non-textile chrysotile cohorts, the Russian study shows the lowest average exposure producing elevated mesothelioma level in a cohort (33.6 f/cc-years). However, even much higher average exposure (120 f/cc-years) apparently has not provided elevated risk of mesothelioma (zero excess mesothelioma cases in Qinghai). In this context, we can see that true threshold for chrysotile cumulative exposure in mesothelioma can be quite substantial (if its existence would be fully demonstrated, as it is possible to hypothesize based on biological mechanisms of carcinogenicity).

Our study has uncertainties and limitations. We have mentioned “tremolite,” which is an amphibole that occurs in varying concentration in the mines in Québec in both asbestiform and nonasbestiform habits. There has been much written as to whether the tremolite—rather than the chrysotile—is responsible to a greater or lesser extent for mesothelioma and other asbestos-related diseases in the cohort, and indeed McDonald et al. (4) have shown this to be true for cohort members with more than 20 years’ service in those areas with higher tremolite content. However, as recently pointed out by Chatfield (27), while there is considerable variation between different mines, “It appears that “tremolite-free Canadian chrysotile” does not exist,” and that all Canadian sources analyzed contain tremolite/actinolite, at least “a proportion of which is asbestiform in morphology.” Therefore, it is not possible to assign all the risk for mesothelioma only to tremolite; the possibility of chrysotile itself providing risk—at very high dose—remains.

Our study demonstrated that a relationship between cumulative exposure to chrysotile and lung fiber burden can be established, and that the corresponding regression equation may have a threshold, with a certain accumulated number of fibers needed to increase the probability of disease. Relationship between cumulative exposure and lung fiber burden has been well established in scientific literature (34–37). However, for chrysotile workers in Québec, only limited data exists to evaluate this type of relationship. It seems that a lung burden study for the Québec population could be beneficial to confirm the conclusion of our modeling attempt.

A major question addressed in this study was our effort to improve the degree to which historic dust measurements taken by midget impingers, in MPCF, can be reliably used as proxy for fibers within the dust. Overall, it remains the case that the proportion of fibres among total dust “particles” is unknown. However, the new analyses and methods presented here using Dagbert’s paired sample data (23) from a variety of mines, mills, and jobs within them have improved our understanding.

Mesothelioma diagnosis has advanced considerably since the days during which every one of our cases occurred. Immunohistochemical tests specifically identifying mesothelial cell origin only became available in the late 1990s (10). We have therefore taken a conservative approach in our analysis and excluded three cases (of 30 overall) with low diagnostic probability. The possibility of inclusion of cases which were not mesothelioma is also balanced by the fact that mesothelioma has historically been under-diagnosed due to lack of awareness by pathologists and clinicians as well as anomalies in death certification.

Additional efforts are needed to explore the correspondence between Québec cohort study datapoints and the newly acquired specific measurements published in a governmental report by Dagbert in 1976 (23) and used here. However, our simulation was intended to cover various random combinations of exposure locations that can be attributed to the parallel measurements performed by Gibbs (38). The empirical relationship between mesothelioma mortality and cumulative exposure appears to depend on the selection of exposure categories (ranges). Further work is needed to test this relationship for different ranges of exposure. It may be beneficial if the Québec cohort can be followed for an additional time period, since the youngest surviving members were aged 73 at last follow-up. Nevertheless, our study addressed various aspects of the reliability of historical exposure data for chrysotile miners and millers and confirmed that possible statistical variability of exposure estimates should not significantly affect slope factors used for scientifically based risk assessment. Asbestos science continues to bring important methodological outcomes that would be valuable for epidemiology, toxicology, and other areas of research.