lung cancer, these variables accounted for 20 of the 25

variables used in the equation. In the present study, two

treatment variables were conceptualized. These were initial

treatment and subsequent treatment. The combination and

sequence of treatment modalities was recorded for both of

these treatment variables through coding of the various

categories in ones and zeros. This method not only reduces

the number of variables for evaluation but results in an

increase in the amount of important information included in

the treatment variables. Two kinds of information are gained

by this method. The first is related to the difference

between initial and subsequent treatment. Because the

initial treatment is the "definitive" attempt to cure the

cancer, this is an important characteristic of the patient's

treatment plan. The second kind of information gained by

this method is related to the sequence in which various kinds

of treatment are given.

 The identification of variables in the present study

also involved exploratory regression analyses for all three

cancer sites. These analyses revealed differences between

the cancer sites in terms of the variables that had

predictive power and the degree to which certain variables

contributed to the explanation of variability in survival.

The stage of disease was the best predictor of survival for

all three sites. These findings are in keeping with the