PROSTATE CANCER is the leading cancer among men in

Prognostic Model for Predicting Survival in Men With Hormone-Refractory Metastatic Prostate Cancer By Susan Halabi, Eric J. Small, Philip W. Kantoff, ...
6 downloads 2 Views 217KB Size
Prognostic Model for Predicting Survival in Men With Hormone-Refractory Metastatic Prostate Cancer By Susan Halabi, Eric J. Small, Philip W. Kantoff, Michael W. Kattan, Ellen B. Kaplan, Nancy A. Dawson, Ellis G. Levine, Brent A. Blumenstein, and Nicholas J. Vogelzang Purpose: To develop and validate a model that can be used to predict the overall survival probability among metastatic hormone-refractory prostate cancer patients (HRPC). Patients and Methods: Data from six Cancer and Leukemia Group B protocols that enrolled 1,101 patients with metastatic hormone-refractory adenocarcinoma of the prostate during the study period from 1991 to 2001 were pooled. The proportional hazards model was used to develop a multivariable model on the basis of pretreatment factors and to construct a prognostic model. The area under the receiver operating characteristic curve (ROC) was calculated as a measure of predictive discrimination. Calibration of the model predictions was assessed by comparing the predicted probability with the actual survival probability. An independent data set was used to validate the fitted model. Results: The final model included the following factors: lactate dehydrogenase, prostate-specific antigen, alkaline

phosphatase, Gleason sum, Eastern Cooperative Oncology Group performance status, hemoglobin, and the presence of visceral disease. The area under the ROC curve was 0.68. Patients were classified into one of four risk groups. We observed a good agreement between the observed and predicted survival probabilities for the four risk groups. The observed median survival durations were 7.5 (95% confidence interval [CI], 6.2 to 10.9), 13.4 (95% CI, 9.7 to 26.3), 18.9 (95% CI, 16.2 to 26.3), and 27.2 (95% CI, 21.9 to 42.8) months for the first, second, third, and fourth risk groups, respectively. The corresponding median predicted survival times were 8.8, 13.4, 17.4, and 22.80 for the four risk groups. Conclusion: This model could be used to predict individual survival probabilities and to stratify metastatic HRPC patients in randomized phase III trials. J Clin Oncol 21:1232-1237. © 2003 by American Society of Clinical Oncology.

ROSTATE CANCER is the leading cancer among men in the United States, accounting for 31% of all male malignancies.1 The American Cancer Society estimated that 189,000 men in the United States would be diagnosed with prostate cancer and that 30,200 men would die from this cancer during 2002.1 It is estimated that one in six men will be diagnosed with prostate cancer sometime during his lifetime and that one in 30 men will die of this disease.2 Several prognostic models have been developed to predict different outcomes in various patient populations, from untreated clinically localized cancer to patients in other clinical states. The vast majority of models, however, are based on predicting outcomes: those that were pathologic,3,4 prostate-specific antigen (PSA) recurrence,5-10 or disease recurrence in patients with clinically localized prostate cancer.

Few studies have identified prognostic models that are predictive of survival in men with hormone-refractory prostate cancer (HRPC). In an analysis of 85 patients with metastatic hormoneresistant prostate cancer, Berry et al11 identified factors predictive of short survival duration. These included age (⬎ 65 years); severe bone pain; poor performance status; presence of soft tissue metastases; anemia; and elevated levels of lactate dehydrogenase (LDH), acid phosphatase, alkaline phosphatase, and prolactin. In an analysis of 1,020 patients, Emrich et al12 identified factors that were predictive of survival in order of importance: previous hormone response, anorexia, elevated acid phosphatase, pain, elevated alkaline phosphatase, obstructive symptoms, tumor grade, performance status, anemia, and age at diagnosis. Kantoff et al13 identified prognostic factors on the basis of 242 metastatic HRPC patients. These factors were alkaline phosphatase, LDH, baseline PSA, and hemoglobin. Other factors identified in other studies were greater than 50% decline in PSA,14-16 changes in PSA after therapy,17,18 weight loss,19 extent of bone metastasis,19,20 pretreatment serum testosterone level,20 and any decline in PSA.21 Biologic markers such as plasma and urine vascular endothelial growth factor and reverse transcriptase polymerase chain reaction for PSA have been identified as statistically significant predictors of overall survival in androgen-independent cancer patients.22-25 Recently, colleagues at Memorial Sloan-Kettering Cancer Center (New York, NY) developed and validated a pretreatment nomogram.26 The objective of this study was to develop a pretreatment prognostic model that could be used to predict survival probability among men with HRCP. Data from six Cancer and Leukemia Group B (CALGB) studies were used to examine the relationship between baseline factors and overall survival. Furthermore, independent data sets were used to validate the prognostic model.

P

From the Department of Biostatistics and Bioinformatics and CALGB Statistical Center, Duke University Medical Center, Durham, NC; Urologic Oncology Program, University of California at San Francisco, San Francisco, CA; The Lank Center for Genitourinary Oncology, Department of Adult Oncology, Dana-Farber Cancer Institute, Boston MA; Memorial Sloan-Kettering Cancer Center, New York, and Department of Medicine, Roswell Park Cancer Institute, Buffalo, NY; University of Maryland, Baltimore, MD; and Section of Hematology and Oncology, Department of Medicine, University of Chicago Medical Center, Chicago, IL. Submitted June 17, 2002; accepted December 23, 2002. Supported in part by grants from the National Cancer Institute (CA31946 and CA 36601), National Institutes of Health, Department of Health and Human Services, Bethesda, MD, to the Cancer and Leukemia Group B (R. Schilsky). The contents of this article are solely the responsibility of the authors and do not necessarily represent the official views of the National Cancer Institute. Address reprint requests to Susan Halabi, PhD, Duke University Medical Center, Box 3958, Durham, NC 27710; email: [email protected]. © 2003 by American Society of Clinical Oncology. 0732-183X/03/2107-1232/$20.00

1232

Journal of Clinical Oncology, Vol 21, No 7 (April 1), 2003: pp 1232-1237 DOI: 10.1200/JCO.2003.06.100

Downloaded from ascopubs.org by 37.44.207.71 on January 27, 2017 from 037.044.207.071 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.

1233

PREDICTING SURVIVAL IN HRPC PATIENTS

PATIENTS AND METHODS

Table 1.

Baseline Characteristics of the Learning and Validation Samples Learning Sample (N ⫽ 760)

Study Population Data on 1,101 men from six clinical trials conducted by the CALGB between 1992 to 1998 were pooled. All patients signed informed consent forms before study registration. These studies were CALGB 9181, 9182, 9480, 9680, 9780, and 9583.13,16,27-30 The patient population consisted of men with prostate cancer who had progressive metastatic disease and for whom both androgen ablation and antiandrogen withdrawal had failed. CALGB 9181 was a randomized study of low-dose versus high-dose megestrol. One hundred forty-nine patients were randomly assigned to treatment on this study. CALGB 9182 was a phase III study in which 242 patients were randomly assigned to receive either hydrocortisone with mitoxantrone or hydrocortisone alone. CALGB 9480 was a phase III study of 390 patients randomly assigned to receive three different fixed doses of suramin. CALGB 9583 was a phase III study in which 260 patients were randomly assigned with equal probability to receive antiandrogen withdrawal alone followed by ketoconazole at progression or antiandrogen withdrawal plus simultaneous ketoconazole and hydrocortisone. CALGB 9680 was a randomized phase II trial of high-dose mitoxantrone/granulocyte-macrophage colony-stimulating factor and low-dose steroids. Twenty-one patients without pelvic irradiation received 21 mg/m2 mitoxantrone every 3 weeks (arm 1), whereas 24 patients who had pelvic irradiation received 17 mg/m2 (arm 2). CALGB 9780 was a phase II study in which 46 patients were treated with docetaxel, estramustine, and low-dose hydrocortisone. For all studies, eligible patients had progressive adenocarcinoma of the prostate after androgen ablation, an Eastern Cooperative Oncology Group (ECOG) performance status of 0 to 2, and adequate hematologic, renal, and hepatic function. Additional details have been published elsewhere.13,16,21,27-30

Statistical Analysis The main end point was survival duration. Survival duration was defined as the time between randomization (or registration for the nonrandomized studies) and death. Patients were censored if they were known to be alive or they were lost to follow-up. The Kaplan-Meier product-limit estimator was used to estimate the survival distribution.31 The proportional hazards model was used to assess the prognostic significance of baseline factors in univariable and multivariable analyses.32 The proportional hazards model was used to develop a multivariable model and to validate the prognostic model. The goal was to use two thirds (n ⫽ 760) of the 1,101 patients for the learning set and one third (n ⫽ 341) of the patients for the validation set. Data for the learning set were used from three protocols (CALGB 9181, 9182, and 9480) to develop the model (nomogram). PSA, LDH, and alkaline phosphatase levels had heavily right-skewed distributions and were modeled using the log transformation. Martingale and Schoenfeld residuals were used to check the adequacy of the linearity and the proportional hazards assumptions. 33 We assessed the predictive performance of the final model (nomogram) by internal validation using the bootstrap resampling technique.33,34 For each of the 200 bootstrap samples, the model was refitted and then tested on the original sample to obtain a bias-corrected estimate of predictive accuracy (ie, when the model is applied to an independent sample of patients).33 The area under the receiver operating characteristic curve (ROC) was calculated as a measure of predictive discrimination in the original sample and the bootstrapped validation samples. An index of 0.5 indicates no discrimination ability, whereas a value of 1 indicates perfect discrimination.33 Calibration of the model predictions was evaluated by comparing the predicted probability at 12 and 24 months with the Kaplan-Meier survival probability. In addition, data from three protocols (CALGB 9583, 9680, and 9780) were used to validate the final model. For each patient, we calculated a risk score of death using the parameter estimates from the final model that was developed from the learning sample. The final model was used to obtain individual predicted probability of survival for each patient on the basis of his covariates, and data were categorized on the basis of the quartile of the predicted probability.35 The statistical analyses for model (nomogram) development and validation were done using the S-plus software (Statistical Sciences, Seattle, WA) and, as well, software that is available electronically in the public domain.36

Demographics Median age, years Interquartile range Race, % white Metastases* % Bone % Lymph node involvement % Lung % Liver Gleason sum % 2-4 % 5-7 % 8-10 Median years since diagnosis Interquartile range Performance status %0 %1 %2 % Measurable disease % Visceral disease % Prior therapy† (any) % Surgical castration % LHRH analog % Estrogen % Progesterone agent % Antiandrogen % Other Laboratory data Hemoglobin, g/dL Interquartile range PSA, ng/mL1 Interquartile range Alkaline phosphatase, IU/L1 Interquartile range LDH, U/L1 Interquartile range

Validation Sample (N ⫽ 341)

71 65 to 75 84

72 65 to 77 81

92 32 8 7

85 34 6 6

10 46 44 3 2 to 6

6 49 45 4 2 to 7

41 46 13 32 13 99 52 55 6 5 77 17

54 38 8 39 11 99 30 73 5 3 93 10

12.4 11.1 to 13.5 126 46 to 336 172 105 to 355 225 173 to 409

12.4 11.4 to 13.4 77 23 to 227 125 91 to 237 216 178 to 437

Abbreviations: LHRH, luteinizing hormone-releasing hormone; PSA, prostatespecific antigen; LDH, lactate dehydrogenase. *Patients may have more than one metastasis. †Patients may have more than one type of prior therapy.

RESULTS

Learning Data Table 1 displays the baseline characteristics of the 760 patients in the learning sample and 341 patients in the validation samples. The median age of the learning sample was 71 years, and 84% of these patients were white. Eighty-seven percent of the patients had an ECOG performance status of 0 or 1. Eighty-two percent of patients had bone metastases and 31% had lymph node involvement. Median baseline PSA was 126 ng/mL, median hemoglobin was 12 g/dL, and median alkaline phosphatase was 171 U/L. Median survival duration among these patients was 13 months (95% confidence interval [CI], 12 to 14), and median length of follow-up among surviving patients was 37 months (Fig 1). Univariable Analysis Figure 2 presents the shape of each baseline predictor on the log hazard of death. PSA, LDH, alkaline phosphatase, performance status, Gleason sum, and hemoglobin were strongly related with log hazard of death. Table 2 presents the univariable survival analysis of the baseline factors. Statistically significant factors of survival were

Downloaded from ascopubs.org by 37.44.207.71 on January 27, 2017 from 037.044.207.071 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.

1234

HALABI ET AL Table 2.

Univariable Analysis of Predictors on the Basis of 760 Patients in the Learning Sample

Factor

Fig 1.

Overall survival in the learning and validation samples.

performance status, Gleason sum, alkaline phosphatase, LDH, hemoglobin, and PSA. Multivariable Analysis In multivariable analysis, statistically significant prognostic factors of overall survival were performance status, Gleason sum on the original prostatectomy or prostate biopsy specimen, LDH, PSA, alkaline phosphatase, and hemoglobin level. The strongest prognostic factor was performance status followed by Gleason sum, LDH, alkaline phosphatase, PSA, and hemoglobin levels (Table 3).

Age, years ⱕ 71 ⬎ 71 Race White Nonwhite Performance status 0 1 2 Gleason sum 2-4 5-7 8-10 Disease measurability Measurable Evaluable Visceral disease No Yes Hemoglobin* ⱕ 12.4 ⬎ 12.4 Alkaline phosphatase* ⱕ 172 ⬎ 172 PSA* ⱕ 126 ⬎ 126 LDH* ⱕ 225 ⬎ 225

Survival, Months

No. of Patients

No. of Events

Median

Range

379 381

356 358

13 13

12 to 14 11 to 15

.8623

641 119

603 111

13 14

12 to 14 12 to 16

.4503

315 350 95

290 331 93

18 11 4

16 to 19 10 to 12 4 to 6

⬍ .0001

73 335 315

63 313 306

18 14 11

13 to 24 12 to 16 10 to 12

⬍ .0001

243 515

230 482

12 14

9 to 13 13 to 15

.2953

658 102

617 97

14 8

12 to 15 6 to 12

.0289

387 373

370 344

10 17

9 to 12 15 to 18

⬍ .0001

382 378

343 371

18 10

16 to 19 9 to 11

⬍ .0001

381 379

346 368

16 11

14 to 17 10 to 12

⬍ .0001

381 379

345 369

17 10

16 to 19 8 to 11

⬍ .0001

P

Abbreviations: PSA, prostate-specific antigen; LDH, lactate dehydrogenase. *The factors were dichotomized on the basis of the median value of the factor. The P values were significant on the basis of the proportional hazard model when the variable was treated as a continuous variable.

Figure 3 presents a nomogram constructed on the basis of the fitted proportional hazards model (Table 3). This nomogram can be used to estimate the median and 12- and 24-month probability of survival. The nomogram is employed by determining a patient’s position on each predictor scale. Prognostic points are located on the top axis of each scale. The points for each Table 3.

Multivariable Model Predicting Overall Survival Duration

Factors

Performance status 0 1 2 Gleason sum ⬍8 8-10 Log(LDH) Log(alkaline phosphatase) Log(PSA) Visceral disease No Yes Hemoglobin Fig 2.

Plots showing the relationship of predictors with hazard of death.

Parameter Estimate

HR

95% CI

1.00 1.48 2.19

Referent 1.31 to 1.67 1.94 to 2.47

1.00 1.40 1.37 1.23 1.10

Referent 1.20 to 1.62 1.21 to 1.55 1.12 to 1.36 1.05 to 1.15

1.00 1.17 0.92

Referent 0.95 to 1.46 0.87 to 0.97

⬍ .0001

0.335

0.312 0.211 0.093 0.161

⫺0.082

P

⬍ .0001

0.392

⬍ .0001 ⬍ .0001 ⬍ .0001 .147

⬍ .0001

Abbreviations: HR, hazard ratio; CI, confidence interval; LDH, lactate dehydrogenase; PSA, prostate-specific antigen.

Downloaded from ascopubs.org by 37.44.207.71 on January 27, 2017 from 037.044.207.071 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.

1235

PREDICTING SURVIVAL IN HRPC PATIENTS

Fig 3.

Pretreatment nomogram predicting probability of survival.

predictor are summed and plotted at the total points axis (the fourth line from the bottom). Assuming the patient is alive, a vertical line drawn from the total points axis directly straight down to the 12-month (or 24-month) survival probability will indicate the patient’s probability of survival for 12 months (or 24 months). We evaluated the nomogram (Fig 3) for its discriminative ability or its ability to separate patients with different outcomes.33 The area under the ROC curve using the 760 patients was 0.69 in the learning sample. We also evaluated the calibration (internal validation) of the nomogram. The predictions from the nomogram were close to actual probability of survival (data not shown). Validation Sample Patients in the validation data set had an improved survival compared with patients in the learning samples (Table 1). The median survival duration for 341 patients was 17 months (95%

CI, 14 to 19 months), and the median follow-up for surviving patients was 24 months. We evaluated the nomogram for its discriminative ability and calibration (external validation). The area under the ROC curve using the 341 patients was 0.68 in the validation samples. Figure 4 presents how the predictions from the model at 12 and 24 months compared with the actual survival probability for the 341 patients in our analysis (calibration). Furthermore, patients were grouped into quartiles on the basis of the median of the predicted survival duration. Figure 5 presents the observed survival curves for the four risk groups. The four risk groups have different observed survival probability (P ⬍ .001). The observed median survival durations were 7.5 months (95% CI, 6.2 to 10.9 months), 13.4 months (95% CI, 9.7 to 26.3 months), 18.9 months (95% CI, 16.2 to 26.3 months), and 27.2 months (95% CI, 21.9 to 42.8 months) for the first, second, third, and fourth risk

Downloaded from ascopubs.org by 37.44.207.71 on January 27, 2017 from 037.044.207.071 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.

1236

HALABI ET AL

Fig 4.

Fig 5.

Calibration of the nomogram.

groups, respectively. The corresponding median predicted survival times were 8.8, 13.4, 17.4, and 22.8 months for the four risk groups. DISCUSSION

We developed and validated a prognostic model that can be used to stratify patients with HRPC in future randomized studies. To our knowledge, this study is among the first few studies to develop a pretreatment prognostic model in men with HRPC. A similar nomogram was independently developed by Smaletz et al.26 Similar to other studies,11,12,26,37,38 we found that performance status, Gleason sum, LDH, PSA, and alkaline phosphatase were significant prognostic factors of overall survival. Unlike Smaletz et al,26 Gleason sum was an important variable in our model. Because biologic markers, such as plasma and urine vascular endothelial growth factor and reverse transcriptase polymerase chain reaction for PSA, were not available on all patients, we could not use such data in the model development or in the validation; we plan to do so as data are collected. Our model is reasonably accurate in terms of correctly predicting survival probability. The accuracy of the prediction is, however, higher at 1 year as opposed to 2 years. This may simply reflect difficulty in predicting further into the future or may reflect real differences in overall survival rates between those patients in the learning data set and those in the validation data sets. Additional research should validate this model prospectively. Available therapies for HRPC patients are palliative and have not been shown to prolong survival duration.39,40 However, survival is clearly dependent on pretreatment clinical variables. Identification of prognostic factors is important to classify patients into different strata.41 An understanding of the distribution of patients into these studies could account for results reported in phase II clinical trials.

Observed survival by risk groups.

For phase III trials, the stratification of patients ensures that the treatment groups are balanced with respect to the known or possible factors to avoid the possibility of confounding. Furthermore, the utility of risk stratification may help identify subsets of patients that may have prolonged survival duration. Indeed, it is possible that some treatment will be beneficial for only a subgroup of patients but not for others. The proposed model has several strengths. First, the model incorporated a large number of patients with metastatic HRPC. Second, the data were prospectively collected from multicenter protocols that enrolled and treated patients on carefully controlled and monitored clinical trials that are of high quality. Third, detailed treatment information and outcome data were available from these trials. Limitations of this study include the fact that patients enrolled in these studies were required to have an ECOG performance status of 0 to 2 and were also deemed appropriate for participation in a clinical trial. Thus, the results of the study cannot be generalized to the entire HRPC patient population. Second, the patient population is heterogeneous because we pooled data from six different studies that enrolled patients in the study period between 1992 and 1998. However, this heterogeneity may also be the strength of the data, which increases the general applicability of the derived model. Finally, the model that was developed and validated did not include biologic markers as predictors of overall survival. Future research is needed that incorporates such markers. In summary, the model developed here has been thoroughly validated within the CALGB. The variables included in the prognostic model are routinely collected in clinical practice and can be easily incorporated to derive a prognostic score. This nomogram will be helpful to obtain individual survival probability and to stratify patients in future randomized phase III studies.

APPENDIX

The appendix is available online at www.jco.org.

Downloaded from ascopubs.org by 37.44.207.71 on January 27, 2017 from 037.044.207.071 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.

1237

PREDICTING SURVIVAL IN HRPC PATIENTS

REFERENCES 1. American Cancer Society: Cancer Facts & Figures 2002. American Cancer Society, Atlanta, GA, 2002 2. Ries LAG, Eisner MP, Kosary CL, et al: SEER Cancer Statistics Review: 1973-1998. Bethesda, MD, National Cancer Institute, NIH publication 00-2789, 2000 3. Partin AW, Yoo J, Carter HB, et al: The use of prostate specific antigen, clinical stage and Gleason score to predict pathological stage in men with localized prostate cancer. J Urol 150:110-114, 1993 4. Partin AW, Kattan MW, Subong EN, et al: Combination of prostatespecific antigen, clinical stage, and Gleason score to predict pathological stage of localized prostate cancer: A multi-institutional update. J Am Med Assoc 277:1445-1451, 1997 5. Kattan MW, Stapleton AM, Wheeler TM, et al: Evaluation of a nomogram used to predict the pathologic stage of clinically localized prostate carcinoma. Cancer 79:528-537, 1997 6. D’Amico AV, Whittington R, Schultz D, et al: Outcome based staging for clinically localized adenocarcinoma of the prostate. J Urol 158:14221426, 1997 7. D’Amico AV. Whittington R, Malkowicz SB, et al: The combination of preoperative prostate specific antigen and postoperative pathological findings to predict prostate specific antigen outcome in clinically localized prostate cancer. J Urol 160:2096-2101, 1998 8. D’Amico AV, Whittington R, Malkowicz SB, et al: Pretreatment nomogram for prostate-specific antigen recurrence after radical prostatectomy or external-beam radiation therapy for clinically localized prostate cancer. J Clin Oncol 17:168-172, 1999 9. Bauer JJ, Connelly RR, Sesterhenn IA, et al: Biostatistical modeling using traditional variables and genetic biomarkers for predicting the risk of prostate carcinoma recurrence after radical prostatectomy. Cancer 79:952962, 1997 10. Bauer JJ, Connelly RR, Seterhenn IA, et al: Biostatistical modeling using traditional preoperative and pathological prognostic variables in the selection of men at high risk for disease recurrence after radical prostatectomy for prostate cancer. J Urol 159:929-933, 1998 11. Berry WR, Laszlo J, Cox E, et al: Prognostic factors in metastatic and hormonally unresponsive carcinoma of the prostate. Cancer 44:763-775, 1979 12. Emrich LJ, Priore RL, Murphy GP, et al: Prognostic factors in patients with advanced stage prostate cancer. Cancer Res 45:5173-5179, 1985 13. Kantoff PW, Halabi S, Conaway MR, et al: Hydrocortisone with or without mitoxantrone in men with hormone refractory prostate cancer: The results of CALGB 9182. J Clin Oncol 17:2506-2513, 1999 14. Kelly WK, Scher HI, Mazumdar M, et al: Prostate-specific antigen as a measure of disease outcome in metastatic hormone-refractory prostate cancer. J Clin Oncol 11:607-615, 1993 15. Smith DC, Dunn RL, Strawderman MS, et al: Change in serum prostate-specific antigen as a marker of response to cytotoxic therapy for hormone-refractory prostate cancer. J Clin Oncol 16:1835-1843, 1998 16. Small EJ, Halabi S, Ratain MJ, et al: A randomized study of three different doses of suramin administered with a fixed dosing schedule in patients with advanced prostate cancer: Results of intergroup 0159 (CALGB 9480). J Clin Oncol 20:3369-3375, 2002 17. Sridhara R, Eisenberger MA, Sinibaldi VJ, et al: Evaluation of prostate-specific antigen as a surrogate marker for response of hormonerefractory prostate cancer to suramin therapy. J Clin Oncol 13:2944-2953, 1995 18. Scher HI, Kelly WM, Zhang ZF, et al: Post-therapy serum prostatespecific antigen level and survival in patients with androgen-independent prostate cancer. J Natl Cancer Inst 91:244-251, 1999 19. DeWys WD, Begg CB, Brodovsky H, et al: A comparative clinical trial of adriamycin and 5-fluorouracil in advanced prostatic cancer: Prognostic factors and response. Prostate 4:1-11, 1983 20. Soloway MS, Ishikawa S, van der Zwaag R, et al: Prognostic factors in patients with advanced prostate cancer. Urology 33:53-56, 1989

21. Halabi S, Conaway MR, Small EJ, et al: Serum prostate specific antigen as a predictor of survival in prostate cancer patients treated with second-line hormonal therapy (CALGB 9181). Prostate J 3:18-25, 2001 22. Kantoff PW, Halabi S, Farmer DA, et al: Prognostic significance of reverse transcriptase-polymerase chain reaction (RT-PCR) for prostate specific antigen (PSA) in men with hormone refractory prostate cancer (HRPC). J Clin Oncol 19:3025-3028, 2001 23. Halabi S, Small EJ, Hayes DF, et al: The prognostic significance of reverse transcriptase polymerase chain reaction (RT-PCR) for prostate specific antigen (PSA) in prostate cancer patients with one prior hormonal therapy: A nested study within CALGB 9583. J Clin Oncol 21:490-495, 2003 24. Bok RA, Halabi S, Fei DT, et al: Vascular endothelial growth factor and basic fibroblast growth factor urine levels as predictors of outcome in hormone refractory prostate cancer: A CALGB study. Cancer Res 61:25332536, 2001 25. George DJ, Halabi S, Shepard TF, et al: Prognostic significance of plasma vascular endothelial growth factor (VEGF) levels in patients with hormone refractory prostate cancer: A CALGB study. Clin Cancer Res 7:1932-1936, 2001 26. Smaletz O, Scher HI, Small EJ: A nomogram for overall survival of patients with progressive metastatic prostate cancer following castration. J Clin Oncol 20:3972-3982, 2002 27. Dawson NA, Conaway MR, Halabi S, et al: A randomized study comparing standard vs. moderately high dose megestrol acetate in advanced prostate cancer: CALGB 9181. Cancer 88:825-834, 2000 28. Savarese DM, Halabi S, Hars V, et al: Phase II study of docetaxel, estramustine, and low-dose hydrocortisone in men with hormone-refractory prostate cancer: A final report of CALGB 9780. J Clin Oncol 19:2509-2516, 2001 29. Levine EG, Halabi S, Roberts JD, et al: Higher doses of mitoxantrone among men with hormone refractory prostate cancer: A Cancer and Leukemia Group B study. Cancer 94:665-672, 2002 30. Small EJ, Halabi S, Picus J, et al: A prospective randomized trial of antiandrogen withdrawal alone or antiandrogen withdrawal in combination with high dose ketoconazole in androgen independent prostate cancer patients: Results of CALGB 9583. Proc Am Soc Clin Oncol 20:174a-695, 2001 31. Kaplan E, Meier P: Nonparametric estimation from incomplete observations. J Am Stat Assoc 53:457-481, 1958 32. Cox DR: Regression models and life tables. J R Stat Soc B 34:187-220, 1974 33. Harrell FE Jr., Lee KL, Califf RM, et al: Regression modelling strategies for improved prognostic prediction. Stat Med 3:143-152, 1984 34. Efron B, Gong G: A leisurely look at the bootstrap, the jackknife, and cross-validation. Am Stat 37:36-48, 1983 35. Chen SH, George SL: The bootstrap and identification of prognostic factors via Cox’s proportional hazards regression model. Stat Med 4:39-46, 1985 36. Harrell FE: Design: S-plus function for biostatistical/epidemiological modeling, testing, estimation, validation, graphics, prediction, and typesetting by storing enhanced model design attributes in the fit, 1994. http:// [email protected] 37. Petrylak DP, Scher HI, Li Z, et al: Prognostic factors for survival of patients with bidimensionally measurable metastatic hormone-refractory prostatic cancer treated with single-agent chemotherapy. Cancer 70:2870-2878, 1992 38. Pienta KJ, Redman B, Hussain M, et al: Phase II evaluation of oral estramustine and oral etoposide in hormone-refractory adenocarcinoma of the prostate. J Clin Oncol 12:2005-2012, 1994 39. Vogelzang NJ, Crawford ED, Zietman A: Current clinical trial design issues in hormone-refractory prostate carcinoma. Cancer 82:2093-2101, 1998 40. Scher HI, Mazumdar M, Kelly WK: Clinical trials in relapsed prostate cancer: Defining the target. J Natl Cancer Inst 88:1623-1634, 1996 41. George DJ, Kantoff PW: Prognostic indicators in hormone refractory prostate cancer. Urol Clin North Am 26:303-310, 1999

Downloaded from ascopubs.org by 37.44.207.71 on January 27, 2017 from 037.044.207.071 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.

Suggest Documents