How good are the numbers for cigarette tar at predicting deliveries of carbon monoxide, hydrogen cyanide, and acrolein?

Abstract

Simple and multiple linear regression analyses were performed on the tar, nicotine, CO, HCN, and acrolein deliveries of 98 brands of commercially available Canadian cigarettes. Simple linear regressions of CO, HCN, and acrolein on tar and of acrolein on CO were not significant at the 5% level for plain cigarettes; all other regressions were very strongly significant (p less than 0.01). In addition there was strong statistical evidence that filter, vented filter, and plain cigarettes have different regression relations and different levels of precision associated with the regression parameters. With respect to CO, tar level could predict the corresponding CO value only within 3.9 mg (filter), 1.9 mg (vented filter), and 3.1 mg (plain). These results show that tar delivery alone is an inadequate predictor of gas-phase constituents such as CO. Information concerning these constituents is best conveyed by an analytically determined value for each component.