A FORMULA FOR PREDICTING THE TUMOR INCIDENCE IN INTRAPERITONEAL EXPERIMENTS WITH MINERAL FIBERS

Abstract

We present the results of a statistical analysis based on 175 intraperitoneal fiber experiments coming from numerous reports, published and internal as well. The reference, geometry and chemical composition of each fiber, number of exposed animals, injected quantities, and resulting observed tumors have been recorded in a single database. Following a logistic regression model, we show that the predicted number of tumors matches in most cases the observed number at an acceptable precision level. The formula includes fiber biosolubility, length, diameter, and injected mass in place of number of fibers, which would achieve less reliable results with this database. To represent the chemical composition, the dissolution coefficient at pH 7.4 is used; a chemical index, such as KI, developed in Germany (TRGS, 1994), could have been used but it does not apply to natural fibers and so would have limited the size of the database. Moreover, the KI index and the Kdis dissolution coefficient are known to be closely related within synthetic fibers. A slight adaptation of the model allows one to classify any fiber experiment in at least two distinct categories reflecting some tumor incidence level (e.g. higher or lower than 10%), given the experiment parameters. We notice that a 10% tumor incidence cutpoint as proposed earlier (Pott et al., 1990) allows a reliable classification of fiber experiments, and that a 25% cutpoint is not as reliable.