2011 Volume 14 Issue 1 Pages 93-107
In cancer therapy, achievement of cure and delaying of death make enormous differences in patient survival benefit and quality of life. Unfortunately, conventional survival analyses such as the log-rank test and Cox regression have failed to distinguish between these two outcomes, occasionally misguiding clinicians in the evaluation of cancer therapies and prognostic factors. Cox himself has recently acknowledged the limitations of his model. These problems have spurred the present body of work which has included the development of computer programs for cancer survival analysis based on the Boag model and its extensions, and the distribution of CDROMs of the programs among clinical oncologists so that they can share relevant survival information with their patients. The present paper explains the Boag model and its extensions, and instructs how to run the programs. Boag assumed that, in a group of cancer patients, a fraction c are cured of the disease (cancer under study) while the remaining uncured patients (1-c) die from the disease at times whose logarithms follow a normal distribution with mean m and variance s 2. The first task is to calculate the Boag three parameters for a given group of patients. The second task is to predict the mean survival time (life expectancy) for the whole group using the competing risk model, taking into consideration the risk of deaths from all causes. The third task is to evaluate the effects of prognostic factors (e.g., treatments, laboratory data, clinicopathological variables) on the Boag parameters using the Gamel three regressions, in which prognostic factors are predictor variables and c, m and s are dependent variables. The computer programs were first written in HTBasic for Windows (by S.M.) and translated into Visual Basic for Application (by H.B). It is hoped that this survival analysis will provide more meaningful survival information to clinical oncologists and patients.
