Beryllium and lung cancer: A weight of evidence evaluation of the toxicological and epidemiological literature

Abstract

The potential carcinogenicity of beryllium has been a topic of study since the mid-1940s. Since then, numerous scientific and regulatory bodies have assigned beryllium to various categories with respect to its carcinogenicity. Past epidemiologic and animal studies, however, have been marked with notable methodological shortcomings. Because it has been about 16 yr since IARC evaluated beryllium and approximately 50 relevant papers on the topic have been published since that time, we conducted a weight-of-evidence analysis of the historical as well as recent animal and human literature. We also assessed whether recently published studies improved upon methodological shortcomings or shed light upon uncertainties in prior studies. Thirty-three animal studies, principally designed to evaluate the cancer hazard or related mechanisms, and seventeen epidemiologic studies were considered in this assessment. Based on this analysis, the evidence for carcinogenicity of beryllium is not as clear as suggested by previous evaluations, because of the inadequacy of the available smoking history information, the lack of well-characterized historical occupational exposures and shortcomings in the animal studies. We concluded that the increase in potential risk of lung cancer was observed among those exposed to very high levels of beryllium and that beryllium’s carcinogenic potential in humans at exposure levels that exist in modern industrial settings should be considered either inadequate or marginally suggestive.

Keywords::

• Beryllium
• epidemiology
• IARC
• lung cancer
• toxicology

Acknowledgements

Funding for the preparation of this article and the underlying research was provided by Brush Wellman Inc.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the article.

Notes

¹ The U.S. EPA announced the release of the updated cancer assessment in the Toxicological Review of Beryllium and notice of the External Peer Review Panel Meeting in the May 12, 2008, Federal Register notice. The draft document that was available for external peer review and the Peer Review Summary Report were reviewed in this analysis (U.S. EPA 2008; Peer Review Summary Report 2008).

² The 2008 ACGIH draft documentation for beryllium was reviewed for this publication. The ACGIH has adopted the draft documentation and it will be available March 13, 2009. (http://www.acgih.org/resources/press/TLV2009list.htm).

³ Although not published as a peer-reviewed paper, Morgareidge (1976) was reviewed by the U.S. EPA for the 2001 Toxicology Review of Beryllium in support of summary information on the Integrated Risk Information System (IRIS) (U.S. EPA 1998).

⁴ The U.S. EPA noted that this methodology likely resulted in an underestimation of expected lung cancer deaths, and a subsequent overestimation of the SMR (U.S. EPA 1998). A correction factor of 10% (referred to as the Saracci adjustment by IARC), has been utilized to correct for differences in the mortality experience of the U.S. population during these two time frames; however, it has also been reported that the expected lung cancer deaths were underestimated by as much as 11% (Saracci 1985; IARC 1993; MacMahon 1994).

⁵ The effect of smoking was analyzed using a procedure described by Axelson and Steenland (1988) based on information from a 1968 U.S. Public Health Survey of the beryllium workers. A smoking correction factor of 1.13 was calculated to account for differences in smoking rates between the cohort and the U.S. population. The SMR is divided by the smoking correction factor to take into account the smoking habits of the cohort.

⁶ Ward noted that smoking data were collected during a 1968 Public Health Service medical survey of four beryllium facilities (Reading, Hazelton, Elmore, and St. Clair) and that 94% of the lung cancer cases occurred among workers hired in the 1940s and 1950s. Thus, the validity of the adjustment for smoking depends on the assumption that the difference in smoking habits between the cohort and U.S. population was the same in the 1940s and 1950s as in the late 1960s, and that the distribution of smoking at the four facilities adequately represented the distribution of smoking at all plants. At the four facilities, 1466 subjects contributed data on smoking; this comprised 16% of the total cohort (9225). Levy et al. (2002) noted that these data did not adequately represent smoking habits of the entire cohort and most importantly did not characterize smoking habits at the Lorain plant, the only plant that showed significantly elevated lung cancer rates in subsequent analyses after adjustment for this confounder.

⁷ Statistically significant increased risk was reported for latency of 15–30 yr, but when recalculating the risk estimate to obtain confidence intervals, we found that the confidence interval for the reported risk included 1 (reported SMR 1.20, calculated 95% CI 0.99–1.44, p = .057).

⁸ Based on personal communication with Dr. Neil Roth.