- © 2005 by American Society of Clinical Oncology
Identifying Patients at Risk for Significant Versus Clinically Insignificant Postoperative Prostate-Specific Antigen Failure
- From the Department of Radiation Oncology, Brigham and Women's Hospital and Dana-Farber Cancer Institute, Harvard Medical School, Boston, MA; Department of Statistics, University of Connecticut, Storrs, CT; Department of Psychiatry, Washington University School of Medicine, St Louis, MO; and Department of Urology, Northwestern Feinberg School of Medicine, Chicago, IL
- Address reprint requests to Anthony V. D'Amico, MD, PhD, Brigham and Women's Hospital, 75 Francis St, Department of Radiation Oncology L-2 Level, Boston, MA 02115; e-mail: adamico{at}lroc.harvard.edu
Abstract
Purpose We evaluated whether men at risk for significant versus clinically insignificant prostate-specific antigen (PSA) failure after radical prostatectomy could be identified using information available at diagnosis.
Patients and Methods A prospective prostate cancer screening study that enrolled, diagnosed, and treated 1,011 men with radical prostatectomy at Barnes-Jewish Hospital (St Louis, MO) from January 1, 1989, to June 1, 2002, for localized prostate cancer formed the study cohort. Preoperative predictors of a postoperative PSA doubling time (DT) of less than 3 months and more than 12 months or no PSA failure were identified using logistic regression.
Results A preoperative PSA velocity more than 2.0 ng/mL/yr (P = .001) and biopsy Gleason score 7 (P = .006) or 8 to 10 (P = .003) were significantly associated with having a postoperative PSA DT less than 3 months. A PSA level less than 10 ng/mL (P = .005), a nonpalpable cancer (P = .001) with a Gleason score ≤ 6 (P = .0002), and a preoperative PSA increase that did not exceed 0.5 ng/mL/yr (P = .03) were significantly associated with a postoperative PSA DT of at least 12 months or no PSA failure. Most men with these preoperative characteristics and a postoperative PSA DT of 12 months or more had a persistent postoperative PSA level of at least 0.2 ng/mL that did not exceed 0.25 ng/mL after a median follow-up of 3.6 years.
Conclusion A postoperative PSA DT less than 3 months is associated with a preoperative PSA velocity more than 2.0 ng/mL/yr and high-grade disease. Select men with a postoperative PSA DT more than 12 months may not require salvage radiation therapy.
INTRODUCTION
A patient can be counseled regarding his risk of prostate-specific antigen (PSA) failure and eventual prostate cancer-specific mortality (PCSM) after radical prostatectomy (RP) using the PSA level, Gleason score, and T category at diagnosis.1 However as PSA-based screening has become more widely practiced, the absolute level of PSA has become less useful as a predictive factor because most patients present with a PSA level less than 10 ng/mL.2 Despite this shift toward lower PSA levels, the change in the PSA level during the year before diagnosis has been suggested to be significantly associated with PCSM.3
In a group of 1,095 men serially screened with PSA and diagnosed with prostate cancer at a median PSA level of 4.3 ng/mL, a preoperative PSA increase of more than 2 ng/mL during the year before diagnosis was associated with a significantly shorter time to death from prostate cancer and death from any cause when compared with patients whose PSA velocity was 2 ng/mL/yr or less. In addition, patients who experience a rapid compared with a protracted increase in the serum PSA level after RP have been shown to have a significantly shorter time to metastases4 and death as a result of prostate cancer.4,5
Given that a PSA doubling time (DT) of less than 3 months leads to a shorter overall survival because of increased rates of PCSM, whereas PCSM for patients with a PSA DT of at least 12 months is rare during the decade after treatment, the management of these patients at the time of PSA failure should differ. Specifically, patients who experience the life-shortening type of PSA failure should be selected for studies examining the possible benefit of combined-modality therapies (eg, chemotherapy plus hormonal therapy) compared with practiced approaches (eg, hormonal therapy) at the time of PSA failure. Moreover, identifying patients at risk for a short PSA DT at presentation affords the patient the opportunity to enter a clinical trial at diagnosis that compares new treatment approaches with the current standard. At the other extreme, some patients with a PSA DT of more than 12 months may benefit from postoperative radiation therapy, whereas others may fare just as well with observation, given that their protracted increase in PSA may be from residual benign prostate tissue. This distinction may be enhanced using information available at diagnosis. Therefore, this study was performed to investigate whether men at risk for significant versus clinically insignificant PSA failure after RP could be identified using information available at diagnosis.
PATIENTS AND METHODS
Patient Selection, Treatment, and Staging
Pretreatment and follow-up information were compiled on 1,804 men who participated in a prospective prostate cancer screening study6 and were treated with radical prostatectomy at Barnes-Jewish Hospital (St Louis, MO) from January 1, 1989, to June 1, 2002, for clinical categories T1c (nonpalpable) and T2 (palpable) prostate cancer. Twelve patients were found to have lymph node metastases at the time of final pathologic evaluation, despite the absence of noted lymph node involvement intraoperatively. These 12 patients had undetectable PSA levels at the first postoperative evaluation, were not treated with adjuvant hormonal therapy, and therefore were included in the study. A total of 689 men who had only a single measurement of PSA preoperatively were excluded from the study, as were 20 men who had received adjuvant radiation therapy. Of the remaining 1,095 men, 84 did not have sufficient preoperative and postoperative PSA data on which to calculate either a preoperative PSA velocity or postoperative PSA DT, respectively, and they were excluded. The remaining 1,011 men comprised the study cohort. No patient received adjuvant hormonal therapy.
Each man provided written informed consent before study entry; the Institutional Review Board approved the study. The median age of the men at the time of initial therapy was 65.4 years (range, 43.2 to 83.5 years). Seventy-two percent of patients were diagnosed on the basis of their PSA and 95% had a PSA of 10 ng/mL or less. The median PSA was 4.3 ng/mL (range, 0.3 to 57.9 ng/mL). Table 1 lists the clinical characteristics of the men before treatment. The preoperative staging has been described previously.3
Clinical Characteristics of the 1,011 Men Before Treatment
Follow-Up
The median follow-up was 4.3 years (range, 0.5 to 12.0 years) and follow-up started on the day of RP and concluded on September 1, 2003, or the date of death, whichever was sooner; no patient was lost to follow-up. Before PSA-defined recurrence, as specified by two consecutive detectable PSA values (> 0.2 ng/dL), patients generally had a serum PSA measurement every 6 months and an annual digital rectal examination. After PSA recurrence, PSA was measured at a median of 4 months (range, 1 to 12 months). At the time of PSA recurrence, biopsy of the anastomosis was not routinely performed. Overall, there were 341 instances of PSA failure.
Statistical Methods
Calculation of the PSA velocity and PSA DT.
Using the PSA value closest in time to diagnosis (median, 1 month [range, 0.25 to 2 months]) and all prior PSA values that were within 1 year of diagnosis, the PSA velocity during the year before diagnosis was calculated using a linear regression analysis.7 The PSA DT was calculated assuming first-order kinetics and using a minimum of two PSA values, each separated in time by a minimum of 3 months and each having a value of more than 0.2 ng/mL. If a patient had a single increase in his PSA value after an undetectable PSA (< 0.2 ng/mL), a PSA DT could not be calculated, and these patients were excluded from the analysis.
PSA velocity, PSA DT, and mortality analyses.
A logistic regression analysis8 was performed to evaluate whether the preoperative PSA level, biopsy Gleason score,9 clinical T category, and preoperative PSA velocity predicted a postoperative PSA DT less than 3 months or either ≥ 12 months or no PSA failure. The PSA level was evaluated as a categoric variable using cut points of ≤ 10 versus greater than 10 ng/mL. Biopsy Gleason score, clinical T category, and the preoperative PSA velocity were categorized using ≤ 6, 7, or 8 to 10; T1c versus T2; and 2.0 or less versus greater than 2.0 ng/mL/yr, respectively. A Cox regression analysis10 was used to test whether the postoperative PSA DT was a predictor of prostate cancer–specific and all-cause mortality after PSA failure. For the purpose of the Cox regression multivariable analyses, the PSA DT was considered as a categoric variable (< 3 v 3 months or more as baseline). Time zero was the date of PSA failure defined as the date of RP if the PSA never decreased to less than 0.2 ng/mL or the date of the first increase to more than 0.2 ng/mL.
For all categoric variables, the cut points selected were made before examining the data and were based on clinically relevant previously established stratifications.1,3,5 For all Cox regression analyses, the assumptions of the proportional hazards model were tested and met, and all statistical tests were two sided. Odds ratios and 95% CIs were calculated for all significant predictors of a PSA DT less than 3 months and PSA DT of more than 12 months or no PSA failure. Hazard ratios (HRs) and 95% CIs were calculated for the significant predictors of time to prostate cancer–specific and all-cause mortality after radical prostatectomy. A χ2 metric was used to compare the proportions of prostatectomy T categories, Gleason scores, and margin status among select patients with a postoperative PSA DT of ≥ 12 months and those who did not experience PSA failure. Illustrations of time to prostate cancer–specific and all-cause mortality after radical prostatectomy were made using cumulative incidence11 plots of PCSM and 1− Kaplan and Meier12 plots of survival (or all cause mortality), respectively. No deaths occurred before PSA failure.
RESULTS
Predictors of Clinically Significant PSA Failure and Death
A PSA velocity more than 2.0 ng/mL/yr (P = .001) and biopsy Gleason score 7 (P = .006) or 8 to 10 (P = .003) were significantly associated with a postoperative PSA DT less than 3 months (Table 2). As compared with a PSA DT of ≥ 3 months, a PSA DT less than 3 months was associated with a significantly shorter time to death from prostate cancer (HR, 5.1; 95% CI, 1.6 to 16.4; P = .006) and any cause (HR, 4.0; 95% CI, 1.5 to 11.0; P = .007). As shown in Figure 1, the cumulative incidence estimates of PCSM and 1− Kaplan and Meier estimates of survival or all-cause mortality were significantly higher for patients with a postoperative PSA DT less than 3 months compared with ≥ 3 months. Specifically, at 7 years after PSA failure, the estimates of PCSM and all-cause mortality were 32% (95% CI, 5% to 60%) and 41% (95% CI, 12% to 71%), and 4% (95% CI, 1% to 7%) and 8% (95% CI, 4% to 12%) for patients with a postoperative PSA DT less than 3 months compared with ≥ 3 months, respectively.
Cumulative incidence estimate of prostate cancer–specific mortality and Kaplan and Meier estimate of all-cause mortality stratified by the postoperative prostate-specific antigen doubling time (PSA DT). Log-rank P value for prostate cancer–specific mortality = .002; log-rank P value for all-cause mortality = .003.
Univariable and Multivariable Logistic Regression Analyses to Determine the OR of Experiencing a Postoperative PSA DT < 3 Months or Either a PSA DT > 12 Months or No PSA Failure According to the Clinical Findings at Diagnosis
Predictors of Clinically Insignificant PSA Failure
As listed in Table 2, a postoperative PSA DT of greater than 12 months or no PSA failure was significantly associated with a PSA level of 10 ng/mL or less (P = .005), a nonpalpable cancer (P = .001) with Gleason score of ≤ 6 (P = .0002), and a preoperative PSA increase that did not exceed 0.5 ng/mL/yr (P = .03). For the 40 men with these characteristics and a postoperative PSA DT of greater than 12 months, the distributions of prostatectomy T category (P = .41), Gleason score (P = .69), and margin status (P = .80) were extremely favorable and not significantly different when compared with those of the 89 men with the same preoperative characteristics who did not experience PSA failure (Table 3). Moreover, although 31 of these 40 patients had a persistent PSA level of at least 0.2 ng/mL postoperatively, after a median follow-up of 3.6 years, 24 of these 31 men had PSA levels that had not exceeded 0.25 ng/mL. For these 24 men, the prostate gland volume exceeded 30 mL in 90%, and the median volume was 48 mL (range, 25 to 81 mL).
Distribution of Pathologic Characteristics at Radical Prostatectomy Among Patients With the Significant Preoperative Predictors of a Postoperative PSA DT > 12 Months or No PSA Failure, and Stratified by These Categories
DISCUSSION
It is now known that a significant minority of men who experience PSA failure after RP will develop metastases and subsequently die as a result of prostate cancer, and that nearly all of these men can be identified using the postoperative PSA DT.5 It has also been noted that some patients can have residual benign prostate tissue after prostatectomy.13,14 Although a study15 describing how the postoperative PSA DT, PSA level, and pathologic findings at RP can be used to assess who is likely to benefit from salvage postoperative radiation therapy, identification of men in whom a persistent but low postoperative PSA level and/or a protracted increase in postoperative PSA is due to residual prostate cancer compared with benign prostate tissue remains a challenge. This distinction is important, however, when trying to decide whether postoperative radiation therapy should be administered.
In this study, a postoperative PSA DT less than 3 months was associated with a preoperative PSA velocity of more than 2.0 ng/mL/yr, high-grade disease at diagnosis, and a shorter survival due to the increase in PCSM, which is consistent with prior investigations.4,5,15 In addition, a subset of patients with a postoperative PSA DT of at least 12 months and a preoperative PSA change that did not exceed 0.5 ng/mL during the year before diagnosis were found to have prostatectomy findings that were extremely favorable and not significantly different from those of men who did not sustain PSA failure.
Prior studies have suggested that changes in PSA on the order of 0.5 ng/mL/yr could be consistent with benign prostatic hyperplasia.16 Given that the subgroup of men with a PSA DT of more than 12 months had extremely favorable characteristics at RP that were not significantly different when compared with men who did not experience PSA failure, and that the nearly all of them had a persistent PSA level of ≤ 0.2 ng/mL that did not exceed 0.25 ng/mL after more than 3.5 years of median follow-up, it is possible that these men's protracted postoperative PSA increase may be due to residual benign prostate tissue.
The clinical significance of the findings in this study is two-fold. First, patients who experience a PSA DT of less than 3 months have shorter overall survival because of an increase in PCSM and are ideal candidates for a randomized study of hormonal therapy with or without docetaxel (a chemotherapeutic agent that has been shown to prolong survival in patients with hormone-refractory metastatic prostate cancer).17,18 Second, for men with a long postoperative PSA DT reflected in persistent but nearly stable PSA levels, a preoperative PSA velocity of 0.5 ng/mL/yr or less, a PSA level of less than 10 ng/mL, and a nonpalpable biopsy Gleason score of ≤ 6 cancer at diagnosis, postoperative radiation may not be necessary. Therefore, PSA failure is heterogeneous biologic state that can range from clinically insignificant benign proliferation of prostatic epithelial cells to incurable micrometastatic recurrence. Based on the results of this study, patients at risk for these distinct states of postoperative PSA failure can be identified based on information available at diagnosis, including the preoperative PSA velocity.
The potential limitation of this study is that we are not able to distinguish pathologically or radiographically (using biopsy or imaging, respectively) the presence of residual benign prostate versus an indolent microscopic cancer in the surgical bed in the subgroup of men with a long PSA DT. Specifically, biopsy of the surgical anastomosis in these patients has been shown to be subject to a large sampling error, and therefore is not recommended.19 In addition, residual microscopic disease is well below the level of current radiographic resolution, which has been shown to be 0.5 mL using endorectal magnetic resonance imaging with spectroscopy20; therefore, the role of imaging is also limited.
Finally, delivery of salvage radiation therapy to the surgical bed at lower PSA levels15,19,21 (generally less than 2.0 ng/mL) has been shown to lead to better PSA failure-free survival. Therefore, if a period of PSA monitoring before initiation of salvage radiation therapy was used to document whether a persistent postoperative serum PSA level that was stable or slowly increasing required treatment, a reasonable cut point at which radiation therapy should be considered would be at 2.0 ng/mL. Such an algorithm should avoid unnecessary treatment for those men who never reach a PSA level of 2.0 ng/mL, and provide treatment in those who may benefit at a time when a loss of cancer control would not be expected.
In conclusion, a postoperative PSA DT less than 3 months is associated with a preoperative PSA velocity more than 2.0 ng/mL/yr, high-grade disease, and a shorter survival. Select men with a PSA DT more than 12 months may not require salvage radiation therapy.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
Footnotes
-
Authors' disclosures of potential conflicts of interest are found at the end of this article.
- Received December 9, 2004.
- Accepted March 29, 2005.
