|
Prognostic tools that are commonly used to predict postsurgical outcomes in patients with prostate cancer dramatically underestimate the efficacy of surgery, according to a study reported by Dr. Brant Inman, of the Mayo Clinic, Rochester, Minn.
"The prognostic models we commonly use to predict outcomes after radical prostatectomy may not be as accurate as previously reported," he said in an interview about the review of several leading models.
To determine how well these rapidly accumulating tools actually perform, the researchers used the Mayo Clinic Radical Prostatectomy Registry, a prospective database of radical prostatectomy outcomes, to evaluate the following commonly used models:
• 1999 and 2005 Kattan nomograms.
• 1998 and 2001 Center for Prostate Disease Research (CPDR) scores.
• 2001 GPSM score, which encompasses Gleason score, prostate specific antigen level, seminal vesicle status, and margin status.
They analyzed data on 13,313 patients who were treated between 1990 and 2005. During that time, there were 3,256 PSA failures, 566 metastases, and 1,599 deaths, 301 of which were due to prostate cancer.
The five tools were assessed for discrimination (the ability to tell the difference between individual patients) and for calibration (the ability to accurately predict outcome). In general, the tools were better at discrimination than at calibration, Dr. Inman said.
"Most tools, with the exception of the GPSM, were very poorly calibrated, and their predictions were off. They could discriminate very well, so they could be useful for stratifying patients in a clinical trial, but for the doctor in his or her office who is trying to tell the patient what the chances are that the treatment is going to work, some of these nomograms were very poor performers," he said.
This was particularly true at the extremes of risk, where patients who had more advanced cancer actually did four times better than predicted (10% predicted good outcome vs. 40% actual good outcome).
"This is important, because if your doctor tells you that you only have a 10% chance of having a successful operation and being cured of your cancer, you might choose not to have the operation. But if your doctor gives you an almost 50-50 chance, most patients are going to take those odds of being cured," Dr. Inman commented.
"The Kattan is best for use in a clinical trial, for stratifying patients, and the GPSM is best to use for patients in your office, which is what the majority of physicians are using these tools for," he said.
Inman B.A. et al. A validation/evaluation of five postoperative prognostic tools for prostate cancer. Abstract 5117.
Commentary |
"What are my chances, Doc?" This is a question clinicians face daily. New research from the Mayo Clinic suggests many new prognostic tools—including some of the most commonly used ones, fail to perform as expected. Of greatest concern, the models appear to work most poorly in high-risk men, pointing to an especially ominous prognosis that may not be warranted. On the positive side, the models were able to separate higher-risk men from lower-risk men reliably. PSA, stage, and grade can provide only so much information. Clearly, the future of prognostication will require standard clinical data as well as molecular markers. Until the future arrives, however, we must live with the tools we have—just with a grain of salt. — Stephen J. Freedland, M.D., and Judd W. Moul, M.D.
|