Knowledge acquisition from historical data: application to breast cancer patients.

The thesis of this paper is that knowledge acquired from historical data collected in a clinical research program may be used for planning future clinical trials and evaluating the effect of treatment. In a breast cancer study comparing adjuvant, treatment with FAC-BCG to control treatment Cox's regression model including number of involved nodes, stage of disease, and menopausal status was used to demonstrate that FAC-BCG patients have superior disease-free survival. A logistic regression model for acute leukemia involving age, cytogenetics, prior hematologic abnormality, temperature, serum creatinine, and hemoglobin was obtained by fitting to 325 previously untreated adult patients who received treatment between March, 1973 and April, 1977. The model was validated by prospective application to 107 patients treated between May, 1977 and December, 1978. Results in this trial will be compared with those in similar patients for the historical control series.