Regulatory Toxicology and Pharmacology 35, 80–94 (2002) doi:10.1006/rtph.2001.1508, available online at http://www.idealibrary.com on The Use of Mechanistic Data and the Handling of Scientific Uncertainty in Carcinogen Risk Assessments The Trichloroethylene Example Christina Rud´ en Philosophy Unit, Royal Institute of Technology, Fiskartorpsv 15A, S-100 44 Stockholm, Sweden; and Institute of Environmental Medicine, Karolinska Institutet, Stockholm, Sweden Received June 1, 2001 The purpose of this paper is to explore how risk assessors actually use mechanistic data in carcinogen risk assessment and to discuss how the handling of scientific uncertainty may affect the outcome of the risk assessment. The analysis is performed by compar- ing 29 trichloroethylene risk assessment documents in general and 2 of these, namely the ECETOC (1994, Trichloroethylene: Assessment of Human Carcinogenic Hazard, Technical Report No. 60) and the OECD/EU (1996, Initial Assessment Report for the 4th SIAM (Screening Information Data Set Initial Assessment Meeting), May 1996: Trichloroethylene, sponsor coun- try, United Kingdom [Draft]), in more detail. It is concluded that in this example the ECETOC required less evidence for considering a carcinogenic mech- anism irrelevant to humans than did the OECD/EU risk assessors. There are examples of when two risk assessors have selected different primary data for their argumentation and also examples of how one and the same primary publication was interpreted differently. Biased data selection and evaluation of primary data that correlate to the risk assessor’s overall conclusions have also been identified. The general comparison of all 29 TCE risk assessment documents indicates that the assessment of scientific uncertainty in the mechanistic data affects the overall conclusions. C  2002 Elsevier Science (USA) 1. INTRODUCTION AND BACKGROUND This is the third report from a case study in which the chlorinated solvent trichloroethylene (TCE) is taken as a model substance for a detailed study of how risk as- sessments of chemicals are performed by different risk assessors. TCE has been chosen as a model substance for this study since unusually many risk assessments have been made of this substance and since the sci- entific database is relatively rich in data and contains observations that are complex and prone to divergent interpretations. The study focuses on the assessment of the potential carcinogenic properties of TCE. In particu- lar, the part of the risk assessment usually referred to as “hazard identification,” i.e., an assessment of whether TCE has an inherent potential to cause cancer. In the first of these papers, 29 TCE risk assessors’ conclusions are described and compared within the framework of a proposed cancer risk assessment index (CRAI), and the influence that data availability, data selection, and data evaluation have had on the overall conclusions is discussed. In this paper it is concluded that the data sets utilized by the trichloroethylene risk assessors are surprisingly incomplete and that biased data selection may have influenced some of the risk as- sessors’ conclusions. Furthermore, risk assessors were shown to often interpret and evaluate one and the same study in different ways and there were also indications of both interpretation bias and evaluation bias for some of the risk assessors (Rud´ en, 2001a). In the second paper, the differences between risk as- sessors in their interpretations of primary carcinogenic- ity data results (bioassays and epidemiology) are inves- tigated. It is shown that about one-fourth of the primary carcinogenicity data referred to in the TCE database have been interpreted differently by different risk as- sessors. The main reasons for differences in the inter- pretations of bioassay results are different assessments of statistics and different assessments of the toxicolog- ical relevance of the results obtained (Rud´ en, 2001b). In the present paper the role of mechanistic data in the TCE cancer risk assessments is described and dis- cussed. (The TCE risk assessment documents are listed in the References.) A default assumption used in regulatory toxicology is that positive effects in animal carcinogenicity stud- ies indicate that the agent under study can have car- cinogenic potential in humans. This is a scientifically based, but policy influenced “rule of thumb” on how to overcome lack of data or scientific uncertainty when extrapolating from animal toxicity to human hazard 80 0273-2300/02 $35.00 C  2002 Elsevier Science (USA) All rights reserved.