A reanalysis of the Stanton et al. pleural sarcoma data

Our plain-English summary of A reanalysis of the Stanton et al. pleural sarcoma data, by G.W. Oehlert (Environmental Research, 1991).

The study

• Author: G.W. Oehlert
• Journal: Environmental Research
• Published: 1991;54(2):194-205
• Cited by: 13 articles, as counted by NIH iCite
• Read it: PubMed · Publisher

Abstract

An analysis of the Stanton et al. (1981, J. Natl. Cancer Inst. 67, 965-975) data reconfirms number of index particles as the primary dimensional predictor of tumor incidence. Fitting separate intercepts and/or slopes to each mineral type results in substantial significant improvement in fit, indicating the importance of mineral type. This contrasts with the "Stanton hypothesis," which states that dimensional properties alone determine carcinogenesis. Log mean aspect ratio is not as good a predictor of tumor incidence as the number of index particles; among those samples that did not have index particles, log mean aspect ratio is not a significant predictor of tumor incidence.

In plain English

This paper was a statistical reanalysis, using logistic regression, of data published by Stanton et al. in 1981. In those 72 experiments, fibrous minerals and other materials, including crocidolite, amosite, tremolite asbestos, glass and talc, were surgically placed in the pleura of rats. Each sample went into 30 to 50 rats, followed for 2 years for pleural sarcoma. The number of 'index particles' (very long, very thin particles) was a highly statistically significant predictor of tumors. No alternative size measure alone was nearly as good. Log mean aspect ratio fit much worse (deviance 538.7 versus 315.8) and was not a significant predictor among materials with no index particles. Allowing a separate intercept for each material significantly improved the fit (reduction in deviance 122.7 with 11 degrees of freedom). Adding the fraction of index particles improved it further. Crocidolite was poorly fit by all models. The author concluded that dimensional properties are not the sole determinants of carcinogenicity. The author named the effective sample size calculation as the weakest link and cautioned against extrapolating to other samples.

We wrote this plain-English summary from the full article with the help of AI, and a second AI model checked each sentence against the article. It is not the authors' wording; the abstract and the article are their own account.

What the study did and found

• Design: Statistical reanalysis, using logistic regression, of previously published rat pleural implantation data (Stanton et al., 1981)
• Who was studied: Rats from the 72 experiments reported by Stanton et al. (1981), with 30 to 50 rats per mineral sample followed for 2 years. Where and when the original experiments were done is not stated.
• Exposure: Surgical implantation into the pleura of 12 material types: crocidolite, glass, dawsonite, aluminum oxide, silicon carbide, attapulgite, potassium octatitanate, talc, tremolite (tremolite asbestos), wollastonite, halloysite and amosite. Particles were characterized by size, including 'index particles', which the original study defined as longer than 8 micrometers and narrower than 0.25 micrometers.
• Main results:
  • None of the alternative size measures alone was nearly as good a predictor of tumor as the log number of index particles (lnip).
  • In a single model for all materials, lnip was highly statistically significant (coefficient 0.8287, SE 0.0395; deviance 315.8 with 70 degrees of freedom), but considerable lack of fit remained.
  • Allowing a separate intercept for each material significantly improved the fit.
  • Allowing a separate lnip slope for each material further improved the fit (reduction in deviance 26.8 with 6 degrees of freedom, P-value 0.00016).
  • Comparing glass and talc samples with the same lnip, the article states the odds ratio of tumor for glass was about six times that for talc.
  • Crocidolite and dawsonite fit the separate-lines models badly. All other mineral types fit tolerably well.
  • Adding the fraction of index particles to the separate intercepts model improved the fit (new deviance 163.1).
  • Log mean aspect ratio was not significant once both lnip and the fraction of index particles were in the model (P-value 0.62).
  • Log mean aspect ratio alone fit much worse than lnip (deviance 538.7 versus 315.8).
  • Among materials with no index particles, log mean aspect ratio was not a significant predictor of tumor probability.
  • The author wrote that the models reinforce the idea that tumor probability is related to the count (or fraction) of very long, very thin particles.
  • Particles outside the index class did not appear to have significant predictive power.
  • The negative lnip coefficient for talc came from a single talc sample with index particles but no tumors. The author stated it is not evidence that talc reduces tumor incidence.
• The authors' conclusion: The author concluded that particle dimensions are not the sole determinants of carcinogenicity in these data and that mineral type is also important.
• Limitations the authors state:
  • The computation of effective sample size was described as the weakest link of the analysis.
  • Crocidolite was poorly fit by all models, and fits for several other materials were marginal.
  • The models may not apply to other samples of the same materials, possibly processed in other ways.
  • Fitting additional covariates separately by mineral type risked overfitting.
  • The data were not precise enough to establish whether non-index particles contribute to carcinogenicity.

Topics

The National Library of Medicine indexes this study under subjects our site covers: Actinolite, Anthophyllite, Tremolite.

Medical terms in this study, explained in our Lexicon: Sarcoma, Stanton Hypothesis, Carcinogenesis, Aspect Ratio (Length-to-diameter ratio).

Related studies

• Lung asbestos content in chrysotile workers with mesothelioma — A. Churg, B. Wiggs, et al. (1984)
• Nuclear Brachyury Expression Is Consistent in Chordoma, Common in Germ Cell Tumors and Small Cell Carcinomas, and Rare in Other Carcinomas and Sarcomas: An Immunohistochemical Study of 5229 Cases — M. Miettinen, Z. Wang, et al. (2015)
• Ordinal regression models for epidemiologic data — B.G. Armstrong, M. Sloan (1989)
• The morbidity and mortality of vermiculite miners and millers exposed to tremolite-actinolite: Part I. Exposure estimates — H.E. Amandus, R. Wheeler, et al. (1987)
• Pulmonary fibrosis in workers exposed to non-asbestiform tremolite asbestos minerals — H.Y. Yang, R.H. Shie, et al. (2013)
• Sarcomas in organ allograft recipients — I. Penn (1995)

Browse more research in Environmental Research, or return to medical research.