You are here

  • Home
  • Archive
  • Volume 69, Issue 8
  • Increased lung cancer mortality among chrysotile asbestos textile workers is more strongly associated with exposure to long thin fibres

Article Text

Article menu
Workplace
Original article
Increased lung cancer mortality among chrysotile asbestos textile workers is more strongly associated with exposure to long thin fibres
  1. Dana Loomis1,
  2. John M Dement2,
  3. Leslie Elliott1,
  4. David Richardson3,
  5. Eileen D Kuempel4,
  6. Leslie Stayner5
  1. 1Department of Epidemiology, College of Public Health, University of Nebraska Medical Center, Omaha, Nebraska, USA
  2. 2Division of Occupational & Environmental Medicine, Department of Community & Family Medicine, Duke University Medical Center, Durham, North Carolina, USA
  3. 3Department of Epidemiology, School of Public Health, University of North Carolina, Chapel Hill, North Carolina, USA
  4. 4National Institute for Occupational Safety and Health, MS C-15, Cincinnati, Ohio, USA
  5. 5Division of Epidemiology and Biostatistics, School of Public Health, University of Illinois, Chicago, Illinois, USA
  1. Correspondence to Professor Dr Dana Loomis, Department of Epidemiology, College of Public Health, 984395 University of Nebraska Medical Center, Omaha, NE 68198, USA; dana.loomis{at}unmc.edu

Abstract

Background Animal data and physical models suggest that the carcinogenicity of asbestos fibres is related to their size and shape.

Objectives To investigate the influence of fibre length and diameter on lung cancer risk in workers at asbestos textile mills in North Carolina and South Carolina, USA.

Methods Men and women (n=6136) who worked ≥30 days in production and were employed between 1940 and 1973 were enumerated and followed for vital status through 2003. A matrix of fibre size-specific exposure estimates was constructed using determinations of fibre numbers and dimensions through analysis of 160 historical dust samples by transmission electron microscopy. Associations of lung cancer mortality with metrics of fibre exposure were estimated by Poisson regression with adjustment for age, sex, race and calendar year.

Results Exposure to fibres throughout the range of length and diameter was significantly associated with increased risk of lung cancer. Models for fibres >5 μm long and <0.25 μm in diameter provided the best fit to the data, while fibres 5–10 μm long and <0.25 μm in diameter were associated most strongly with lung cancer mortality (log rate about 4% per IQR, p<0.001). When indicators of mean fibre length and diameter were modelled simultaneously, lung cancer risk increased as fibre length increased and diameter decreased.

Conclusions The findings support the hypothesis that the occurrence of lung cancer is associated most strongly with exposure to long thin asbestos fibres. The relationship of cancer risk and fibre dimensions should be examined in cohorts exposed to other types of fibres.

  • Asbestos
  • chrysotile
  • fibres
  • lung cancer
  • epidemiology
  • cancer
  • hygiene/occupational hygiene
  • epidemiology
  • public health
  • asthma
  • respiratory

https://doi.org/10.1136/oemed-2012-100676

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    What this paper adds

    • Exposure to asbestos fibres of all types increases the risk of cancer in humans.

    • Animal data and physical models suggest that the carcinogenicity of asbestos fibres may be related to their size and shape.

    • Exposure to chrysotile fibres of all sizes is associated with excess lung cancer in asbestos textile workers.

    • The association is strongest and most consistent for long thin fibres.

    • The relationship of cancer to fibre size and shape should be studied in cohorts exposed to other types of fibres.

    Introduction

    Observations from experiments with animals beginning in the 1970s have led to a hypothesis that long thin asbestos fibres are more carcinogenic than shorter thicker fibres1–6 but that hypothesis had not been tested in studies of humans until recently due to a lack of data on the distributions of asbestos fibre sizes in exposed populations.

    Two recent epidemiologic studies of workers at asbestos textile mills in South Carolina and North Carolina, USA, used analysis of historical dust samples by transmission electron microscopy (TEM) to estimate individual exposures to fibres defined by length and diameter.7 ,8 Both studies found lung cancer risk to be associated with exposure to fibres of all sizes but most strongly with exposure to long thin fibres, particularly those >10 μm long and <1 μm in diameter.9 ,10 While the principal findings of those studies are consistent with the hypothesis that long thin fibres have greater carcinogenic potency, uncertainties remain because of the use of different models and analytical approaches in the two studies, the small numbers of TEM samples analysed for each cohort and the strong correlations among indicators of exposure to fibres of different dimensions. In this study, we analyse pooled data from the North Carolina and South Carolina cohorts in an attempt to resolve those uncertainties.

    Methods

    Study population

    The study population included 3717 men and 2419 women who were employed at any of the four asbestos textile mills in North Carolina and South Carolina, USA, between 1 January 1950 and 31 December 1973 in North Carolina or 1 January 1940 and 31 December 1965 in South Carolina and who worked at least 30 days in textile production departments. The plants and workers included in this study have been the subject of previous publications11–14 and the population we studied is identical to the combined cohort of North and South Carolina asbestos textile workers described by Elliott,15 who studied lung cancer mortality in relation to fibre exposures determined by the conventional phase contrast miscroscopy (PCM) method. All four mills are engaged in the full process of converting raw fibres into yarn and woven goods. Operations at these plants used only chrysotile asbestos with the exception of a small insulation operation in one North Carolina plant, which processed limited amounts of amosite between 1963 and 1976 and use of limited amounts of crocidolite yarn at the South Carolina plant from the 1950s until 1975. Processing of amphiboles was confined to specific well-defined operations and work areas, minimising the number of workers exposed and the potential impact on the study results. North Carolina workers potentially exposed to amphibole fibres were identified, and none had died of lung cancer.11

    The vital status of the South Carolina workers was determined as of 31 December 2001 and that of North Carolina workers through 31 December 2003. Causes of death, including underlying cause, immediate causes and other significant conditions, were coded to the International Classification of Diseases in effect at the time of the death. Research procedures involving human subjects were approved by the Institutional Review Board of the University of North Carolina, Chapel Hill.

    Exposure assessment

    The assessment of exposure for this study has been described in detail elsewhere.16 Briefly, TEM was used to estimate the distribution of fibres for combinations of plant and department in categories defined by diameter (four categories) and length (six categories). This assessment was performed on a stratified random sample of 160 historical dust samples captured on membrane filters collected in surveys of the study plants in 1964–1971, as described previously.7 ,8 The TEM fibre-counting protocol was based on the ISO direct-transfer method17 and procedures for combining the data by plant and operation and deriving size-specific exposure estimates followed the procedure described by Dement et al.7 ,8 ,16 Data from all the study plants were modelled simultaneously by Poisson regression to estimate adjustment factors as proposed by Quinn et al18 for each length–diameter category as described by Dement.7 ,8 ,16 These were applied to a matrix of plant-, department- and time-specific fibre concentrations estimated by the standard PCM method to generate fibre size-specific estimates of exposure.16 Estimated exposures to fibres of different dimensions were linked to workers' occupational histories for assignment of individual cumulative exposure measured in fibre-years per millilitre.

    To integrate information across categories of fibre size and reduce collinearity, we used indicators of the mean length and diameter of the fibres to which workers were exposed over their careers.10 Cumulative mean fibre length-years were estimated by ∑L–icidi/∑ci, where L–i is the mean length of fibres in length–diameter category i and ci and di are the concentration of fibres and the duration of exposure to fibres in category i, respectively. Cumulative mean fibre diameter-years were estimated similarly by ∑D–icidi/∑ci, where D–i is the category-mean diameter. Both indicators have units of micrometre-years (μm-y). The concentration-weighted means of fibre length and diameter over each worker's career (with units of μm) were also estimated by omitting duration, di, from the preceding equations.

    Data analysis

    Exposure-response analyses were based on deaths with any mention of lung cancer on the death certificate. Lung cancer mortality rates were modelled by Poisson regression as in the previous internal analyses of the North Carolina cohort.11 The ungrouped form of Poisson regression, equivalent to the discrete-time proportional hazards model, was used to allow predictors to be entered in continuous or categorical form in the same model.19 ,20

    Potential confounders and alternative variable specifications were evaluated by assessing model fit and the effect on the coefficient for exposure. The final model was of the form ln(rate)=α+βX+γ1age+γ2sex+γ3race+∑γidecade, where X is a quantitative indicator of fibre exposure, age is a continuous variable for attained age in years, sex and race are binary indicators and decade is a categorical variable with categories covering the decades 1940–2000. The parameter of interest, β, describes the change in the log lung cancer mortality rate for an increment in exposure. To facilitate comparisons among fibre size indices with markedly different distributions, we scaled β for these indices by the indicators' intraquartile ranges (IQRs). Exposure indicators were not lagged, as previous analyses showed lags of up to 30 years had negligible effects on the results.9 ,11 ,15 Regression analyses were carried out using R V. 2.12.1 for Mac OS.21

    The fit of the models was assessed using the Akaike Information Criterion (AIC),22 and dispersion was measured by the ratio of model deviance to degrees of freedom. The contribution of exposure terms was evaluated by likelihood ratio χ2 test. Additional models in which exposure was entered as a smoothed function using penalised splines were evaluated to assess the shape of exposure response. Comparisons of alternative models were based on model goodness of fit and the likelihood ratio for the exposure term, as well as on the magnitude and precision of the regression coefficients.

    Results

    During follow-up of the cohort, 218 631 person-years at risk, 3356 deaths from all causes and 361 deaths with a mention of lung cancer were observed.

    Cumulative exposures to total fibres counted by TEM and to fibres in every length and diameter category were significantly associated with lung cancer mortality (table 1). Model fit was similar for total TEM fibres and for cumulative exposure estimated by the standard PCM method (AIC 4677 and 4680, respectively). The log rate of lung cancer increased by 0.039 per IQR (95% CI 0.020 to 0.057, p=0.001) for total TEM fibres and 0.032 per IQR (95% CI 0.014 to 0.049, p=0.002) for PCM fibres.

    View this table:
    Table 1

    Model goodness of fit and association of lung cancer mortality rate with cumulative exposure to asbestos fibres by size category, estimated by Poisson regression with adjustment for age, sex, race and calendar time and standardised to the IQR; North Carolina and South Carolina asbestos textile workers, 1940–2003

    When cumulative exposures to fibres in the 21 length-by-diameter categories were included one at a time in the model, the best model fits were obtained for several lengths of very thin fibres <0.25 μm in diameter, notably for fibres 5–10 μm long and 20–40 μm long (both AIC 4675). The variation in fit among fibre size categories was relatively modest, however, and approximately similar fits were obtained for several other size classes (table 1). The associations with lung cancer mortality with the greatest magnitude were observed for fibres <0.25 μm in diameter, with a regression coefficient of 0.041 per IQR (p<0.001) for fibres 5–10 μm long and coefficients between 0.03 and 0.04 per IQR for other lengths in that diameter range, including very small fibres ≤1.5 μm long (table 1 and figure 1). Coefficients for thicker fibres were smaller with the exception of fibres 0.25–1.0 μm in diameter and 5–10 μm long (0.032 per IQR) and fibres >3.0 μm in diameter and 5–10 μm long (0.031 per IQR). Among the 10 marginal categories of fibre length and diameter (table 1), the associations of greatest magnitude were observed for fibres <0.25 μm in diameter (0.039 per IQR, p<0.001) and fibres 5–10 μm long (0.039 per IQR, p=0.001).

    Figure 1
    Figure 1

    Coefficients for the association of lung cancer mortality rate with a 1 IQR increment in cumulative exposure to asbestos fibres by size category. Estimated by Poisson regression with adjustment for age, sex, race and calendar time and standardised to the IQR; North Carolina and South Carolina asbestos textile workers, 1940–2003.

    Cumulative exposures in all categories of fibre length and diameter were mutually intercorrelated (Pearson correlation coefficients 0.75–0.98). A model that included all 21 length-by-diameter categories simultaneously fit better (AIC 4659) than the models for single size categories or for total fibres (table 1), but 19 of the 21 estimated coefficients had SEs larger than the coefficients (supplementary table 1), suggesting that multicollinearity induced instability in the estimates of individual parameters.

    Models for the association of lung cancer mortality with indicators of cumulative mean fibre length-years and diameter-years fit better than those for associations with individual fibre size categories (table 2). Both measures were positively and significantly associated with lung cancer when modelled singly, but when both dimensions were entered simultaneously, the coefficient for diameter-years became negative and its SE increased (table 2). The pattern of associations was similar for mean fibre length and diameter without the duration component, although model fit was reduced (table 2). Terms for the interaction of cumulative fibre exposure with fibre length and separately with fibre diameter were not statistically significant.

    View this table:
    Table 2

    Model goodness of fit and association of lung cancer risk with alternative indicators of exposure to asbestos fibres, estimated by Poisson regression with adjustment for age, sex, race, and calendar time in North Carolina and South Carolina asbestos textile workers, 1940–2003

    Substituting smoothed functions of fibre length and diameter indicators that allowed exposure response to take on a non-linear form (supplementary figure 1) did not significantly improve the fit relative to models using linear exposure terms (data not shown).

    Discussion

    In pooled analyses of observations on workers at four US asbestos textile plants, we found evidence that lung cancer mortality is associated with exposure to fibres of all sizes but associated most strongly with exposure to thin long fibres. This was evident from the generally larger effect sizes and better model fits observed for structures <0.25 μm in diameter and structures >5 μm long. Moreover, when indicators for both dimensions were modelled simultaneously, the rate of mortality from lung cancer rose as the mean length of fibres to which workers were exposed increased and the mean diameter decreased.

    While the effects were strongest and most consistent for long thin fibres, exposures to fibres in all categories of length and diameter were significantly associated with lung cancer. The specific categories of fibre size that best predict risk are difficult to identify because every worker was exposed to fibres throughout the range of length and diameter leading to strong correlations among exposure indicators defined by those characteristics. Cumulative exposure to fibres <0.25 μm in diameter, and particularly those 5–10 and >20 μm long, had the most consistent pattern of stronger association with lung cancer mortality and better model fit compared with other size classes. The parameter estimates for very short thin fibres <0.25 μm in diameter and ≤1.5 μm long were similar to those for longer thin fibres. This strong association between the thinnest fibres and lung cancer may have implications for exposures to nanofibres or nanotubes because it suggests that carcinogenic potential of fibres increases as the size decreases. Current sampling and analytical methods based on PCM are insensitive to detecting the thinnest individual structures, however, so TEM-based methods would be needed to assess these exposures.16 ,17

    This pooled analysis found some evidence suggesting that fibres >3 μm in diameter and >5 μm long are associated more strongly with lung cancer risk than fibres between 0.25 μm and 3 μm in diameter. Similar associations were observed previously in the North Carolina cohort but not in the South Carolina cohort.9 ,10 Fibres >3 μm in diameter were a small proportion of all fibres (<1.5%), resulting in small numbers when stratified by plant and department, and the relative concentration of these fibres correlated with the proportion of all long fibres regardless of diameter. For this reason, it was not possible to determine if the observed relationship between thick fibres and lung cancer is merely an artifact of their general correlation with other long fibres in the aerosols. Given the low respirability of these thick fibres,23 a correlation with the presence of other long but thin, and more respirable, fibres is the more plausible explanation.

    Variation in the strength of association by fibre size is unlikely to be a result of confounding by job categories, exposure patterns or co-exposures. While the relative proportions of fibres in different size categories varied by plant and process, all processes were found to produce fibres in every size category.16 The only other notable exposure potentially experienced by workers in these mills would have been to mineral oils used for dust control, but prior case–control analyses have shown that mineral oil exposures did not confound the association between cumulative fibre exposures and lung cancer.24 Contamination with amphibole fibres is also unlikely to confound the results. Fibres were identified by morphology using TEM, and those lacking distinctive chrysotile morphology were further reviewed using selected area electron diffraction. Of 38 940 fibres counted, only 16 (0.04%) were identified as amphiboles by selected area electron diffraction and 14 of those were identified as tremolite–actinolite by electron beam microchemical analysis. These were largely short fibres, while virtually all thick structures >3 μm in diameter were identified as chrysotile bundles.

    Associations of disease with fibre dimensions have not been studied in other cohorts of workers exposed to chrysotile asbestos or other types of fibres. However, our findings here are largely consistent with those of separate analyses of associations of lung cancer with fibre dimensions in the North Carolina and South Carolina cohorts.9 ,10 The previous studies of these cohorts used similar methods of exposure assessment,7 ,8 but we revised and refined the estimates for this analysis, taking advantage of the data from both cohorts to inform the estimates.16 The earlier study of the relationship of lung cancer and fibre dimensions in the South Carolina plant focused on associations with specific fibre size categories and found the strongest associations with cumulative exposures to very thin fibres (diameter <0.25 μm) and to long fibres, particularly those in the 20–40 μm range.9 The effects of fibre length and diameter were not analysed simultaneously because of collinearity. A separate analysis of data from the North Carolina mills10 found associations with lung cancer that were strongest with cumulative exposure to long fibres but inconsistent for diameter when fibre size categories were modelled one at a time. Collinearity again precluded simultaneous modelling of multiple fibre size categories. However, as in the present study, lung cancer risk increased with greater mean fibre length and smaller mean fibre diameter when both parameters were modelled simultaneously. The findings of the current pooled analysis suggest that this relationship holds in the South Carolina cohort, as well.

    Relative to previous separate analyses of the two cohorts, this pooled analysis benefited from increased power and removed uncertainties due to differences in methods. However, we were not able to fully resolve several uncertainties arising from the exposure data. Although we had access to over 9000 PCM measurements of fibre concentrations in air from the 1930s to the 1980s, the number of samples that could be analysed by TEM was much smaller due to the labour-intensive protocol and the limited amount of material available for some job categories. We attempted to minimise the effect of between-sample variability by analysing as many dust samples as possible and pooling the data by plant and department.7 ,8 The expectation of variability in fibre size proportions from sample to sample motivated our prior decision to allocate resources to analysing as many samples as possible rather than studying a few in detail. For this pooled analysis, we also combined TEM fibre data for the North Carolina and South Carolina plants and estimated size proportions and adjustment factors using Poisson regression models; in addition, we estimated 95% CIs to describe the precision of these estimates.16 We expect this approach captured TEM size variability by operation and plant reasonably well but acknowledge that there will be statistical noise in our results and expect this would dampen rather than enhance associations by fibre size.

    Another limitation of the TEM analysis is that the samples were taken only between 1964 and 1971 and consequently covered a small proportion of the exposure period. While the lack of samples from the remainder of the period is a source of uncertainty, it is unlikely to have biased the estimated distributions of fibre sizes to a significant degree because methods of operation did not change appreciably during the study period. Finally, workers were exposed to fibres of a wide range of sizes throughout their careers, resulting in multiple correlations among length–diameter categories. These correlations hampered efforts to identify size ranges independently associated with lung cancer risk despite the large amount of data available with the combined cohort. Our analyses using indicators of average fibre length and diameter bypass this limitation by integrating all categories of length or diameter into a single parameter, but the effects of fibres of specific sizes are still of interest.

    Conclusions

    Our findings suggest two major conclusions. First, the tendency for lung cancer risk to increase as fibres become thinner and longer may extend to other types of fibres, including amosite and crocidolite asbestos and engineered nanofibres and nanotubes. The evidence for such a relationship should be examined in cohorts exposed to other types of fibres in various environments. Second, the association of lung cancer with fibres <0.25 μm in diameter in the combined cohort supports the use of TEM and other improved methods to assess asbestos fibre exposure for research and regulatory purposes; these fibres are too small to be counted with the PCM method that has been the standard since the 1960s.

    References

    1. ↵
      1. Pott F,
      2. Huth F,
      3. Friedrichs KH
      . Tumorigenic effect of fibrous dusts in experimental animals. Environ Health Perspect 1974;9:313–15.
      OpenUrlPubMed
    2. ↵
      1. Stanton MF,
      2. Layard M,
      3. Tegeris A,
      4. et al
      . Relation of particle dimension to carcinogenicity in amphibole asbestoses and other fibrous minerals. J Natl Cancer Inst 1981;67:965–75.
      OpenUrlAbstract/FREE Full Text
    3. ↵
      1. Lippmann M
      . Effects of fiber characteristics on lung deposition, retention, and disease. Environ Health Perspect 1990;88:311–17.
      OpenUrlCrossRefPubMedWeb of Science
    4. ↵
      1. Berman DW,
      2. Crump KS,
      3. Chatfield EJ,
      4. et al
      . The sizes, shapes, and mineralogy of asbestos structures that induce lung tumors or mesothelioma in AF/HAN rats following inhalation. Risk Anal 1995;15:181–95.
      OpenUrlCrossRefPubMedWeb of Science
    5. ↵
      1. Davis JM,
      2. Beckett ST,
      3. Bolton RE,
      4. et al
      . Mass and number of fibres in the pathogenesis of asbestos related lung disease in rats. Br J Cancer 1978;37:673–88.
      OpenUrlCrossRefPubMedWeb of Science
    6. ↵
      1. Davis JM,
      2. Jones AD
      . Comparisons of the pathogenicity of long and short fibres of chrysotile asbestos in rats. Br J Exp Pathol 1988;69:717–37.
      OpenUrlPubMedWeb of Science
    7. ↵
      1. Dement JM,
      2. Kuempel ED,
      3. Zumwalde RD,
      4. et al
      . Development of a fibre size-specific job-exposure matrix for airborne asbestos fibres. Occup Environ Med 2008;65:605–12.
      OpenUrlAbstract/FREE Full Text
    8. ↵
      1. Dement J,
      2. Meyers D,
      3. Loomis D,
      4. et al
      . Estimates of historical exposures by phase contrast and transmission electron microscopy in North Carolina, USA asbestos textile plants. Occup Environ Med 2009;66:574–83.
      OpenUrlAbstract/FREE Full Text
    9. ↵
      1. Stayner L,
      2. Kuempel E,
      3. Gilbert S,
      4. et al
      . An epidemiological study of the role of chrysotile asbetos fibre dimensions in determining respiratory disease risk in exposed workers. Occup Environ Med 2008;65:613–19.
      OpenUrlAbstract/FREE Full Text
    10. ↵
      1. Loomis D,
      2. Dement J,
      3. Richardson D,
      4. et al
      . Asbestos fibre dimensions and lung cancer mortality among workers exposed to chrysotile. Occup Environ Med 2010;67:580–4.
      OpenUrlAbstract/FREE Full Text
    11. ↵
      1. Loomis D,
      2. Dement JM,
      3. Wolf SH,
      4. et al
      . Lung Cancer Mortality and Fiber Exposures among North Carolina Asbestos Textile Workers. Occup Environ Med 2009;66:535–42.
      OpenUrlAbstract/FREE Full Text
    12. ↵
      1. Hein MJ,
      2. Stayner LT,
      3. Lehman E,
      4. et al
      . Follow-up study of chrysotile textile workers: cohort mortality and exposure-response. Occup Environ Med 2007;64:616–25.
      OpenUrlAbstract/FREE Full Text
    13. ↵
      1. Dement JM,
      2. Harris RL,
      3. Symons MJ,
      4. et al
      . Exposures and mortality among chrysotile asbestos workers. Part II: mortality. Am J Ind Med 1983;4:421–33.
      OpenUrlCrossRefPubMedWeb of Science
    14. ↵
      1. Stayner L,
      2. Smith R,
      3. Bailer J,
      4. et al
      . Exposure-response analysis of risk of respiratory disease associated with occupational exposure to chrysotile asbestos. Occup Environ Med 1997;54:646–52.
      OpenUrlAbstract/FREE Full Text
    15. ↵
      1. Elliott L,
      2. Loomis D,
      3. Dement J,
      4. et al
      . Lung cancer mortality in North Carolina and South Carolina asbestos textile workers. Occup Environ Med 2012;69:385–90.
      OpenUrlAbstract/FREE Full Text
    16. ↵
      1. Dement JM,
      2. Loomis D,
      3. Richardson D,
      4. et al
      . Estimates of historical exposures by phase contrast and transmission electron microscopy for pooled exposure-response analyses of North Carolina and South Carolina, USA asbestos textile cohorts. Occup Environ Med 2011;68:593–8.
      OpenUrlAbstract/FREE Full Text
    17. ↵
      International Standards Organisation. Ambient Air—Determination of Asbestos Fibres—Direct Transfer Electron Microscopy Method. Geneva: International Standards Organisation, 1995.
    18. ↵
      1. Quinn MM,
      2. Smith TJ,
      3. Eisen EA,
      4. et al
      . Implications of different fiber measures for epidemiologic studies of man-made vitreous fibers. Am J Ind Med 2000;38:132–9.
      OpenUrlCrossRefPubMedWeb of Science
    19. ↵
      1. Richardson DB
      . Discrete time hazards models for occupational and environmental cohort analyses. Occup Environ Med 2010;67:71.
      OpenUrl
    20. ↵
      1. Loomis D,
      2. Richardson D,
      3. Elliott L
      . Poisson regression analysis of ungrouped data. Occup Environ Med 2005;62:325–9.
      OpenUrlAbstract/FREE Full Text
    21. ↵
      R Development Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing, 2008.
    22. ↵
      1. McCullagh P,
      2. Nelder JA
      . Generalized Linear Models. 2nd edn. London: Chapman & Hall, 1989.
    23. ↵
      1. Timbrell V
      . Deposition and retention of fibres in the human lung. Ann Occup Hyg 1982;26:347–69.
      OpenUrlAbstract/FREE Full Text
    24. ↵
      1. Dement JM,
      2. Brown DP,
      3. Okun A
      . Follow up study of chrysotile asbestos textile workers: cohort mortality and case-control analyses. Am J Ind Med 1994;26:431–47.
      OpenUrlPubMedWeb of Science

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

      Files in this Data Supplement:

    Footnotes

    • Funding Support for this research was provided by grant R01-OH007803 from the National Institute for Occupational Safety and Health (USA).

    • Competing interests None.

    • Ethics approval Ethics approval was provided by University of North Carolina Public Health Institutional Review Board.

    • Provenance and peer review Not commissioned; externally peer reviewed.