[CANCER RESEARCH 45, 5173-5179, October 1985] PrognosticFactorsin Patientswith AdvancedStage ProstateCancer1 Lawrence J. Emrich, Roger L. Priore, Gerald P. Murphy, Mark F. Brady,2 and the Investigators of the National Prostatic Cancer Project3 Department of Biomathematics [L J. E, R L P.], the National Prostatic Cancer Project Headquarters Office [G. P. M.¡,and the National Prostatic Cancer Project Statistical Office [M. F. B.], Roswell Park Memorial Institute, Buffalo, New York 14263 ABSTRACT protocols since 1973 (Table 1). Details concerning the treatment regi mens used in these protocols, together with comparisons of the various The relationships of 13 potential prognostic factors to objective regimens, have been discussed previously (2-11). response to treatment and survival time were investigated, using Information gathered on the patients at the time of entry into group data gathered on 1,020 patients with advanced stage prostate protocols included the factors listed in Table 2. These 13 variables were cancer who have participated in the clinical trials of the National thought to be among the most important prognostic factors for advanced Prostatic Cancer Project. Multivariate statistical analyses re stage disease. One thousand, six hundred one (83%) of the patients vealed that previous hormone response status, analgesics, pain, were évaluablefor objective response to treatment, and 1020 (64%) of these 1601 patients had information recorded for all 13 of the factors elevated acid phosphatase, and anemia were the important, (Table 1). This report is restricted to these 1020 patients. independent prognostic factors for objective response to treat Definition of Prognostic Variables and End Points. Each of the ment. For survival time, the significant prognostic factors were potential prognostic variables listed in Table 2 was treated as a categor previous hormone response status, anorexia, elevated acid ical variable in all analyses, with categories defined as follows. Previous phosphatase, pain, elevated alkaline phosphatase, obstructive hormone response was classified as a failure if the disease was in symptoms, tumor grade, performance status, anemia, and age progression after hormonal or ablative therapy (3-5, 7-11), as stable if at diagnosis. It is recommended that future treatment protocols the disease had been stabilized by orchiectomy or hormone therapy for for advanced stage prostate cancer take into account hetero at least 3 mo (2), or as previously untreated. Performance status was geneity of the treatment groups with respect to these factors, classified as normal activity, symptomatic but ambulatory, in bed less than 50% of the time, in bed more than 50% of the time, or 100% either through the design of the protocol, or at the time of bedridden. Pain was classified as no pain, mild (occasional use of analysis. analgesics), moderate (controlled use of analgesics), or severe (uncon trollable pain). Tumor grade was classified as I, II, III, or IV, according to INTRODUCTION published criteria (12,13). Age of the patient was classified as less than or equal to 60, 61-65, 66-70, or greater than 70. The remaining factors Prostatic carcinoma is one of the most common forms of listed in Table 2 were classified as either yes or no according to whether cancer and a leading cause of cancer deaths of males in the the condition was present or absent, respectively. United States (1). Prognosis, as measured by objective response NPCP response criteria (2) were used to evaluate objective response to treatment and survival time, is known to be particularly poor to treatment as being either objective progression (treatment failure), for patients with advanced (Stage Oil) disease, as compared objectively stable, partial response, or complete response (treatment with patients with early stage disease. However, little is known success). about which factors within the former group of patients are most Survival time was measured from the time of randomization to death from any cause. important with regard to prognosis. Such knowledge is invaluable Statistical Methods. The prognostic importance of the 13 variables for planning and analyzing the results of future clinical trials with respect to objective response to treatment was assessed using a involving advanced stage patients and for assessing the prog linear logistic regression model (14). First, each variable was assessed nosis of individual patients. separately for its prognostic importance (univariate analyses). Then, to This paper presents the results of a detailed, mult ivariate study the relative prognostic importance of the variables, a forward statistical study of prognostic factors in patients with advanced selection procedure was used to develop a multivariate model which stage prostate cancer. included all prognostic factors which, after adjusting for the effects of all factors entered into the model previously, were found to be significantly (P value less than 0.05) related to objective response. MATERIALS AND METHODS The prognostic importance of the 13 factors with respect to survival time was assessed using Cox's proportional hazards regression model Patient Population. One thousand, nine hundred forty patients with advanced stage prostatic carcinoma have been entered into NPCP4 (15). Again, each factor was assessed separately for its prognostic importance, and then a forward selection procedure was used to develop 1This work was supported in part by Research Grants CA16056, CA37010, a multivariate model. and CA34903 awarded by the National Cancer Institute. Estimates of survival time distribution curves were obtained by the 2 To whom requests for reprints should be addressed. 3 Investigators of the National Prostatic Cancer Project include Edson Pontes, method of Kaplan and Meier (16). Roswell Park Memorial Institute; Raju Thomas, Tulane Medical Center; Robert P. All P values correspond to likelihood ratio tests of hypotheses for Gibbons, The Mason Clinic; Stefan A. Loening, University of Iowa Hospitals and regression coefficients of zero. A P value of 0.05 or less was considered Clinics; David G. McLeod. Walter Reed Army Medical Center; Jean B. deKemion, to be statistically significant. University of California at Los Angeles; James M. Pierce, Jr., Wayne State Univer sity; George R. Prout, Jr., Massachusetts General Hospital; Peter T. Scardino, Baylor University; Joseph D. Schmidt, University of California at San Diego; William W. Scott, Johns Hopkins Hospital; Mark S. Sotoway, University of Tennessee; RESULTS Patrick Guiñan,Cook County Hospital; Robert C. Ranigan, University of Kentucky Medical Center; Hugh Fisher, Albany Medical College Hospital; and Douglas E. Objective Response. The results of the univariate analyses Johnson, M. D. Anderson Hospital. 4 The abbreviation used is: NPCP, National Prostatic Cancer Project. of the prognostic importance of the 13 factors with respect to Received 1/22/85; revised 4/25/85; accepted 7/11/85. objective response to treatment are summarized in Table 3. With CANCER RESEARCH VOL. 45 OCTOBER 1985 5173 Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 1985 American Association for Cancer Research. PROGNOSTIC FACTORS IN PROSTATE CANCER Table 1 Patient population summary NPCP treatment entered used protocol100TreatmentregimensCyclophosphamide responseFailureYractivated1973Yrclosed1975No.on protocol125No.in analyses57 (6f 5-Fluorouracil Standard treatmentHormone 200 Estramustine phosphate Failure 1974 1976 125 59(6) Streptozotocin Standard treatment 300 imidazole-carboxamide Failure 1975 1977 165 78(8) Procarbazine Cyclophosphamide 400 Prednimustine Failure 1976 1977 135 63(6) Prednimustine + estramustine 500 Diethylstilbestrol or orchiectomy Newly 1976 1980 301 131 (13) Diethylstilbestrol + cyclophos- diagnosed phamide Estramustine phosphate 600 Diethylstilbestrol Stable 1976 1982 188 100(10) Diethylstilbestrol + Cyclophos phamide Diethylstilbestrol + estramustine phosphate 700 Hydroxyurea Failure 1977 1979 125 66(6) Methyl-chloroethyl-cyclohexy- nitrosourea Cyclophosphamide 800 Estramustine phosphate Failure 1977 1979 121 56(5) Vincristine Estramustine phosphate + vin- cristine 1100 Estramustine phosphate Failure 1979 1981 189 130(13) Methotrexate cis-Platinum 1200 Estramustine phosphate Failure 1979 1981 149 96(9) c/s-Platinum Estramustine phosphate + c/s- platinum 1300 Diethylstilbestrol or orchiectomy Newly 1980 1983 317 184(18) Estramustine phosphate diagnosed Diethylstilbestrol or orchiectomy + 5-fluorouracil + Cyclophos phamide Totals 1940 1020(53) " Numbers in parentheses, percentage used in analyses. the exception of tumor grade and age of the patient, each of the categories of the factors in the model are listed in Table 5. As variables considered was found to be of prognostic importance. expected, the most important prognostic factor is previous hor Patients who are stable on hormones and patients who have mone response status (Table 4). After adjusting for the effect of failed previous hormone therapy have a poorer prognosis (i.e., a this factor, the next most important prognostic factor is whether smaller probability of a treatment success, as defined above) or not a patient requires analgesics. The remaining factors which than newly diagnosed patients, as evidenced by the negative were included in the multivariate model, in order of importance, coefficients for the "stable" and "failure" categories shown in are pain, elevated acid phosphatase, and anemia. Table 3. Prognosis improves with improved