May 15, 2000
Volume 52, No. 33
Software detects cancer recurrence
By Lillian Kim
A new piece of software developed by researchers from several institutions can help identify prostate cancer patients who are likely to develop recurrence of cancer after surgical treatment.
Emory is among four institutions that participated in a large, multicenter trial aimed at developing a "genetic adaptive neural network model," essentially a program that uses data from the medical histories of 840 previous patients to predict outcomes in new patients. The new prognostic method was presented May 1 at the American Urological Associ-ation's 95th annual meeting in Atlanta.
The model is designed to help determine a patient's chances of recurrence and whether surgery is the best treatment option. It predicts the likelihood that a patient will develop elevated prostate-specific antigen (PSA) levels within five years of radical prostatectomy, or removal of the prostate. Elevated PSA level is a hallmark of cancer recurrence after treatment.
According to the study, the genetic adaptive model--so called because it continually uses new patient information to modify and improve its predictive abilities--has an accuracy rate of 76 percent.
"It helps patients to know more about their prognosis, and it helps doctors make management decisions appropriately," said Emory urologist Muta Issa, the study's lead author. "If the chance of cure is low in a patient considered at high surgical risk, an alternative treatment that is less invasive may need to be considered. But if the chance of cure is high in an otherwise healthy patient, surgical therapy would be the best treatment option.
"Such information is vital for both the patient and the urologist in making decisions regarding choice of treatment," Issa said.
The study was supported by a departmental grant from the Josephine Ford
Cancer Center in Detroit.