CORESTA Congress, Paris, 2006, SS 14

A mixture experimental design with 6 variables and the effects on smoke components:2 - Hoffmann analytes behaviour through a mixture experimental design

FIGUERES G.; VIDAL B.; BIESSE J.P.; BREGEON B.; LOUVET F.; MUZELLEC L.
Altadis Research Centre, Fleury-les-Aubrais, France

Today in the tobacco industry, one of the main scientific challenges is to understand the formation of some smoke compounds ("Hoffmann analytes"). For several years, work has been conducted in order to know the influence of cigarette design parameters or the effect of tobacco types onto the generation of smoke constituents such as Hoffmann analytes and also to search for precursors from pyrolysis studies. Here we propose an experimental approach based on a complex mixture experimental design (MED) under constraints already described in the previous paper. This MED took into account six tobacco components classically incorporated in an industrial US blend (without casing): Burley, Oriental, Virginia, reconstituted tobaccos, stems and expanded tobaccos. In order to carry out this study, experimental cigarettes were manufactured using a constant design and the weight of tobacco blends was chosen to get the same pressure drop (around 110 mm). Single tobacco and cigarette blends were characterized chemically and physically (burning rate, hardness, filling power). The contribution of each blend tobacco component to the generation of smoke constituents was evaluated by a statistical approach based upon a linear model consisting of first-order items for each of the mixture components. The fitted model is shown with a P-value that tests to determine if it is statistically significant when compared with the mean square for the error term and is illustrated by the polynomial prediction equation. Furthermore we have explored the relationships between Hoffmann analytes yields and the tobacco chemical compounds that could be considered as precursors or would enable predicting smoke constituents from a multi-linear regression such as the Partial Least Squares considering also physical variables of cigarette blends in the modelling conception. In the last investigation we considered the possibility to estimate analytes in the smoke of cigarette blends from those measured in the mainstream smoke of cigarettes made up of single tobacco component.