Table 1.

Primary components for a causality determination based on the epidemiologic database for TCE.

Consideration	Summary of weight of evidence
Consistency of observed association	Strong evidence of consistency for kidney cancer (consistently elevated RRs). Meta‑analysis yielded robust, statistically significant summary RR, with no evidence of heterogeneity or potential publication bias.
	Moderate evidence of consistency for NHL (consistently elevated RRs); RR estimates more variable compared with kidney cancer. Meta-analysis yielded robust, statistically significant summary RR, with some heterogeneity (not statistically significant) and some evidence for potential publication bias.
	Limited evidence of consistency for liver cancer (fewer studies overall, more variable results). Meta-analysis showed no evidence of heterogeneity or potential publication bias, but the statistical significance of the summary estimate depends on the large study by Raaschou-Nielsen et al. (2003).
Strength of observed association	Strength of association is modest. Other known or suspected risk factors (smoking, body mass index, hypertension, or coexposure to other occupational agents such as cutting or petroleum oils) cannot fully explain the observed elevations in kidney cancer RRs. The alternative explanation of smoking was ruled out by the finding of no increased risk of lung cancer. Indirect examination of some specific risk factors for liver cancer or NHL did not suggest confounding as an alternative explanation.
Specificity	Limited evidence suggesting that particular von Hippel-Lindau mutations in kidney tumors may be caused by TCE (Brauch et al. 1999, 2004; Brüning et al. 1997; Nickerson et al. 2008; Schraml et al. 1999); additional research addressing this issue is warranted.
Biological gradient (exposure–response relationship)	Only a few epidemiologic studies examined exposure–response relationships. Studies with well-designed exposure assessments reported a statistically significant trend of increasing risk of kidney cancer (Charbotel et al. 2006; Moore et al. 2010; Zhao et al. 2005) or NHL (Purdue et al. 2011) with increasing TCE exposure. Further support was provided by the meta-analyses; higher summary RR estimates for kidney cancer and NHL were observed for the highest exposure groups than for overall TCE exposure, taking possible reporting bias into account. Liver cancer studies generally had few cases, limiting the ability to assess exposure–response relationships. The meta-analysis for liver cancer did not provide support for a biological gradient (lower summary RR estimate for highest exposure groups than for overall TCE exposure, taking possible reporting bias into account).
Biological plausibility and coherence	TCE metabolism results in reactive, genotoxic, and/or toxicologically active metabolites at target sites in humans and in rodent test species.
	The active GSTT1 enzyme in humans was associated with increased kidney cancer risk, whereas the lack of active enzyme was associated with no increased risk (Moore et al. 2010).
	TCE is carcinogenic in rodents; cancer types with increased incidences include kidney, liver, and lymphohematopoietic cancers.
	A mutagenic mode of action is considered operative for TCE-induced kidney tumors, based on mutagenicity of GSH-conjugation metabolites and the toxicokinetic availability of these metabolites to the target tissue.
	Modes of action are not established for other rodent cancer findings; human relevance is not precluded by any hypothesized modes of action due to inadequate support.
NHL, non-Hodgkin lymphoma. Data from U.S. EPA (2011d).