Data from three academic institutions were used to develop a model to predict pathologic stage in a group of men with clinically localized prostate cancer. The model combined serum prostate-specific antigen (PSA) level, clinical stage, and Gleason score. The data were used to generate nomograms that present the probability of a patient having organ-confined cancer, isolated capsular penetration, seminal vesicle involvement, or pelvic lymph node involvement. [Oncol News Int 6(Suppl 3):14-15, 1997]
ABSTRACT: Data from three academic institutions were used to develop a model to predict pathologic stage in a group of men with clinically localized prostate cancer. The model combined serum prostate-specific antigen (PSA) level, clinical stage, and Gleason score. The data were used to generate nomograms that present the probability of a patient having organ-confined cancer, isolated capsular penetration, seminal vesicle involvement, or pelvic lymph node involvement. [Oncol News Int 6(Suppl 3):14-15, 1997]
The ability to better predict pathologic stage of localized prostate cancer and so make more informed treatment decisions may be improved by a model combining prostate-specific antigen (PSA), clinical stage, and Gleason score. How and why the model was developed and its results among more than 4,000 patients were described at the First Sonoma Conference on Prostate Cancer by Alan Partin, MD, of Johns Hopkins Hospital.
The model was used to generate nomograms. The numbers within each cell of the nomogram represent the percent probability of a final pathologic stage based on the three variables (PSA, Gleason score, and clinical stage). Validation analyses showed that 72.4% of the time, the nomograms predicted the pathologic stage correctly to within 10%: for organ-confined disease, they were correct 67.3% of the time; for capsular penetration, 59.6%; for seminal vesicle involvement, 79.6%; and for lymph node involvement, 82.9% of the time they were correct. Clinicians can use these nomograms in counseling patients about the probable stage of their tumor and treatment options.
“Prediction of pathologic stage is very important for our patients,” said Dr. Partin, in explaining the rationale leading to development of the model. “It’s what they want to know. I think that if we’re trying to predict that a patient has regional disease versus clearly defined, local disease, that’s where the decision point is going to be made in the form of external beam radiation versus brachytherapy or surgery.”
The data presented by Dr. Partin are an update of a previous study conducted among 800 patients at Johns Hopkins Hospital. That study was expanded in response to criticisms that the number of patients was too small and that the study was limited to a single surgeon and a single institution.
The new study was broadened to involve 4,133 patients at three academic institutions that serve as centers of excellence for the surgical treatment of clinically localized prostate cancer-3,116 at Johns Hopkins Hospital, 782 at Baylor College of Medicine, and 235 at the University of Michigan School of Medicine. None of the patients had received preoperative hormonal or radiation therapy.
Presurgical clinical variables were measured as follows:
“The reason we put this together was that the patients all walk in with these three tests. A man sits down across from me, he hands me his biopsy score, he tells me what his PSA was, and I examine him. And he says, well, what’s it look like? How am I going to do?”
“And we felt the best thing to tell them was, rather than ‘this looks pretty good, I think you’ll do okay,’ was to let them know how 1,000 other men who walked in with the same tests that they have, fared after surgery, with respect to the surrogate end point of pathologic stage, and then try to show them some data.”
Pelvic nodes removed at the time of surgery were examined for micrometastatic disease and the surgical specimen (prostate and seminal vesicles) was prepared for histologic examination. The pathologic stages of the patients reported by Dr. Partin were: 48% organ-confined; 40% capsular penetration; 7% positive seminal vesicle involvement; and 5% pelvic lymph node involvement.
Statistical analysis was performed simultaneously for each of the four outcomes. Nomograms were validated by bootstrap samples and have 95% confidence intervals.
Dr. Partin discussed the weaknesses of using a single test.
“Our old friend, the digital rectal exam, although it’s the oldest method for predicting prognosis, was excellent in predicting volume, back when patients had palpable cancer. But sensitivity and specificity are very poor. However, it continues to be recommended in initial screening programs.” Today the TNM classification system is the most commonly used method for predicting prognosis.
He said that there was “an incredible overlap” in PSA levels in the 4 to 10 range. “It’s very hard to use PSA alone, just to predict what you’re going to find on pathological stage.”
“And then, lastly, the third variable we put into this patient aid algorithm was the Gleason grading system,” Dr. Partin added. If you go back and look at the original work that Gleason was looking at, it was a different prostate cancer, back in early 1970. Back then, not 80% of the patients were falling in 5, 6, 7, like we’re seeing now.”
As in the earlier study, the combination of the three clinical variables was a better predictor than any single variable.