ABSTRACT
Rodenttumorigenicityexperimentsplayanimportantroleinassessing thecarcinogenicriskofvariouscompoundstowhichhumansmaybe exposed.Theanalysisofdatafromsuchstudiesiscomplicatedbyseveral features,includingthefactthattumorsvaryintheirlethality.Although commonlyusedinpractice,methodsthatmakeextremelethalityassumptionscanbebiasedandinefficient.Theonlyalternativescurrentlyavailablerequireeithercause-of-deathinformationorsacrificedatatoestimate theunknownlethality.Bothoftheseapproacheshaveshortcomingsin appliedsettings.Inthischapter,weproposeamethodofanalysisthat drawsontheknowledgeofpathologistswhotypicallyhaveextensiveexperienceinthisareaandcanreliablypredictafeasiblerangefortumor lethalities.Weusethisinformationtoconstructasubjectivepriordistributionforthelethalityparameterinamultiplicativethree-statemodel.The modeliscomparedtothestandardanalysesthatmakeextremeassumptionsaboutlethalityandonethatestimateslethalityonlyfromthecurrent data.
