Mr. Miller is employed as a statistical consultant for Dr. Robert W. Veltri at the James Buchanan Brady Urological Institute of the Johns Hopkins Hospital.
Natural history of disease progression in patients who fail to achieve an undetectable prostate-specific antigen level after undergoing radical prostatectomy
Version of Record online: 6 OCT 2004
Copyright © 2004 American Cancer Society
Volume 101, Issue 11, pages 2549–2556, 1 December 2004
How to Cite
Rogers, C. G., Khan, M. A., Craig Miller, M., Veltri, R. W. and Partin, A. W. (2004), Natural history of disease progression in patients who fail to achieve an undetectable prostate-specific antigen level after undergoing radical prostatectomy. Cancer, 101: 2549–2556. doi: 10.1002/cncr.20637
- Issue online: 16 NOV 2004
- Version of Record online: 6 OCT 2004
- Manuscript Accepted: 2 AUG 2004
- Manuscript Revised: 12 JUL 2004
- Manuscript Received: 4 MAY 2004
- National Institute of Health/National Cancer Institute SPORE. Grant Number: P50CA58236
- Prostate Cancer Foundation (CapCURE)
- prostate carcinoma;
- prostate-specific antigen (PSA);
- prostate-specific antigen kinetics;
- disease progression;
To the authors' knowledge, the natural history of disease progression to distant metastasis is unknown in men who fail to achieve an undetectable prostate-specific antigen (PSA) level after radical retropubic prostatectomy (RRP),. The authors assessed the clinical outcome of men with a persistently detectable PSA level after RRP for clinically localized prostate carcinoma.
Between 1989 and 2002, 160 men failed to achieve an undetectable PSA level (≥ 0.1 ng/mL) after undergoing RRP for clinically localized prostate carcinoma. No patient received adjuvant therapy before documented metastasis. The Kaplan–Meier method was used to estimate distant metastasis-free survival. Univariate and multivariate Cox proportional hazards regression was used to assess the ability of clinical and pathologic variables to predict distant metastasis-free survival.
The probability of distant metastasis-free survival at 3 years, 5 years, and 10 years was reported to be 68%, 49%, and 22%, respectively. Seventy-five men (47%) developed distant metastases after RRP (median time to metastases of 5.0 years; range, 0.5–13 years). The combination of RRP Gleason score, seminal vesicle status, and lymph node status resulted in 3 risk groups for the prediction of distant metastasis-free survival (hazards ratio [HR] = 1.6; P < 0.01). The slope of PSA changes approximately 3–12 months after RRP at a cutoff value ≥ 0.05 was found to be even more predictive of distant metastasis-free survival (HR = 2.9; P < 0.01).
Many patients remained free of metastatic disease for an extended period despite failing to achieve an undetectable PSA level after undergoing RRP for clinically localized prostate carcinoma. However, other patients experienced rapid disease progression to distant metastasis. The authors defined clinical (PSA slope) and pathologic (Gleason score) prognostic variables to help identify those patients with a higher risk of developing distant metastasis after undergoing RRP. Cancer 2004. © 2004 American Cancer Society.
Despite advancements in the early detection of adenocarcinoma of the prostate (prostate carcinoma) and improvements in surgical technique, approximately 25% of patients will experience biochemical disease recurrence over a 10-year period after undergoing radical retropubic prostatectomy (RRP).1–4 Over time, the majority of these men will eventually develop distant metastases and/or will die of prostate carcinoma.5 Several pretreatment and posttreatment clinical and pathologic variables, including serum prostate-specific antigen (PSA) level, Gleason score, surgical confinement status, seminal vesicle (SV) status, and pelvic lymph node (LN) invasion, have been found to indicate the probability of biochemical disease recurrence after RRP.2–4, 6–8 However, in men who have a persistently detectable PSA level without ever achieving an undetectable PSA level after undergoing RRP, the natural history of disease progression has not been well described.
The goals of the current study were to attempt to predict the time to distant metastasis and to stratify the patients into risk groups using pretreatment clinical and biopsy pathology data as well as postoperative pathology and PSA kinetics.
MATERIALS AND METHODS
Patient Population and Inclusion Criteria
The current study was conducted under Health Insurance Portability and Accountability Act regulations and internal review board approval. Between January 1989 and December 2002, 9056 consecutive men (5019 men with long-term follow-up and 2680 men with postoperative PSA information) underwent RRP with staging pelvic lymphadenectomy for clinically localized prostate carcinoma (classified as T1, T2, or T3a disease) at an urban tertiary referral center. Informed consent was obtained from all patients before surgery. A total of 224 men failed to achieve an undetectable PSA level 3 months after undergoing RRP. A technically detectable serum PSA level was defined as ≥ 0.1 ng/mL. The reasons for exclusion from the current study included receipt of neoadjuvant hormonal therapy (13 patients [6%]), receipt of immediate postoperative adjuvant radiation therapy before the onset of metastasis (19 patients [8%]), and receipt of early adjuvant hormonal therapy before the onset of metastasis (32 patients [14%]). The remaining 160 patients in the final study group (mean age, of 58 years; range, 38–75 years) had a minimum follow-up time of 1 year.
Preoperative Evaluation and Pathologic Characteristics
Patients were staged according to the 1992 American Joint Committee on Cancer staging guidelines,9 including digital rectal examination (DRE) and routine serum PSA assays (Hybritech Tandem-R and E, Beckman Coulter, San Diego, CA, and TOSOH, Tosoh Medics, South San Francisco, CA). Pathologic diagnosis of prostate carcinoma was made preoperatively from digital-guided or transrectal ultrasound-guided prostate biopsies. The Gleason scoring system was used for histologic grading of the prostate needle biopsy and prostatectomy specimens.10 Pathologic evaluation of the surgical specimens was performed as described previously.11
Postoperative follow-up was obtained through routine serum PSA assays and DRE performed at quarterly intervals the first year, semiannually the second year, and annually thereafter. Radionuclide bone scans were performed initially because of biochemical persistence and on a yearly basis thereafter, unless they were performed earlier for symptoms suggestive of distant metastasis. A positive bone scan result or other radiographic or histologic (LN biopsy) evidence was used for the diagnosis of distant metastasis.
No patients in the final cohort received neoadjuvant or adjuvant hormonal therapy or radiation therapy before the onset of metastasis, based on available medical records. Therefore, neither neoadjuvant nor adjuvant hormonal therapy nor radiation therapy was found to have an impact on the time to distant metastasis in the current analysis.
All statistical analyses were performed using the Stata (Version 8.2) software package (Stata Corporation, College Station, TX). The proportions of the various clinical and pathologic variables in patients with and without the development of distant metastases were compared using the Fisher exact test. The mean age and preoperative PSA levels were compared between the two patient groups using the Student t test. The distribution of the biopsy Gleason scores and the RRP Gleason scores in all patients was compared using the Wilcoxon matched-pair signed rank test. For all analyses of distant metastasis-free survival, the time to distant metastasis was determined from the time of the RRP. There were not enough prostate carcinoma-specific deaths to evaluate overall survival effectively. Therefore, only distant metastasis-free survival was assessed in the current study. Kaplan–Meier plots for estimates of distant metastasis-free survival probabilities were generated based on clinical and pathologic variables. The resulting curves in each plot were compared using log-rank tests and the P values were reported. Cox proportional hazards regression was used to determine univariate and multivariate hazards ratios (HR) for the ability of the parameters to predict the time to distant metastasis and/or prostate carcinoma-specific death. All multivariate Cox regressions used backwards stepwise selection at a stringency level of P < 0.05. The changes in PSA level after surgery, represented as slopes, were calculated for patients with ≥ 2 PSA values available between 3–12 months after surgery (n = 119 patients) by plotting the PSA levels after surgery on the y-axis values and the time interval in months after surgery on the x-axis. The slopes of the resulting plots were determined for each evaluable patient. A cutoff value ≥ 0.05 for the post-RRP PSA slopes (which was the median PSA slope value in all 119 patients) was used to group the patients into favorable (slope < 0.05) and unfavorable (slope ≥ 0.05) prognostic groups.
Table 1 provides data concerning patient age, preoperative PSA level, biopsy and prostatectomy Gleason scores, and other pathologic variables for the 85 patients (53%) without distant metastasis compared with the 75 patients (47%) with distant metastasis and/or prostate carcinoma-specific death. The mean follow-up time for all patients was 5.3 years (standard error ± 3.7 years; range, 1–17 years). Of all the routine pathologic variables assessed, clinical stage and biopsy Gleason scores were found to differ significantly (P < 0.05) between the two patient groups (Table 1). In addition, the proportion of patients in the two groups with positive SV and LN status was significantly different (P < 0.01 and P = 0.03, respectively). It is interesting to note that in both groups of patients, the distribution of Gleason score sums at the time of biopsy was significantly different from the distribution of Gleason scores in the RRP specimens (P < 0.001).
|Variables||No. of patients (%)||P value|
|Persistent PSA only (n = 85) (%)||Persistent PSA with distant metastasis (n = 75) (%)|
|Age (range)||58.3 (41–71 yrs)||59.24 (38–75 yrs)||0.40a|
|Serum PSA (range)||13.6 (3.1–68.5 ng/mL)||16.8 (0.1–71.9 ng/mL)||0.15*|
|Clinical T classification||< 0.01b|
|T1c||37 (43)||11 (15)|
|T2a||26 (31)||25 (33)|
|T2b||19 (22)||22 (29)|
|T2c||3 (4)||98 (112)|
|T3a||0 (0)||89 (12)|
|Biopsy Gleason score||< 0.01b|
|5–6||29 (34)||22 (29)|
|7 (3 + 4)||31 (36)||17 (23)|
|7 (4 + 3)||15 (18)||9 (12)|
|8–9||10 (12)||27 (36)|
|RRP Gleason score||0.27b|
|6||7 (8)||5 (7)|
|7 (3 + 4)||26 (31)||16 (21)|
|7 (4 + 3)||23 (27)||17 (23)|
|8–9||29 (34)||37 (49)|
|SM positive||35 (41)||43 (57)||0.04b|
|LN positive||24 (28)||34 (45)||0.03b|
|SV positive||28 (33)||45 (60)||< 0.01b|
A Kaplan–Meier plot of the time (in years) from RRP to the development of clinically evident distant metastasis for the 160 patients in the current study who failed to achieve an undetectable PSA level is shown in Figure 1. At the time of our report, 75 patients (47%) who failed to achieve an undetectable PSA level after undergoing RRP had evidence of distant metastasis, with subsequent prostate carcinoma-specific deaths reported to occur in 34 (21%) of these patients. The probability of distant metastasis-free survival was 68% at 3 years (95% confidence interval [95% CI], 59–75%), 49% at 5 years (95% CI, 39–58%), 38% at 7 years (95% CI, 27–48%), and 22% at 10 years (95% CI, 11–35%). Figure 2 demonstrates the ability of several clinicopathologic variables to predict distant metastasis-free survival. The ability of the biopsy Gleason score to predict distant metastasis-free survival time using a cutoff value ≤ 7 to stratify the patients into favorable (biopsy Gleason score ≤ 7) and unfavorable (biopsy Gleason score > 7) prognostic groups is illustrated in Figure 2A. Figures 2B–D show various groupings of clinical stage, a cutoff value ≤ 7 for the RRP Gleason score, and the combination of SV and LN status, respectively. The differences in the median distant metastasis-free survival times between the favorable and unfavorable prognostic groups for each of the clinicopathologic variables in Figure 2 were found to be statistically significant (log-rank test, P ≤ 0.01). The median time to distant metastasis for patients in the unfavorable prognostic groups was 2 years earlier than that of the favorable prognostic groups for all clinicopathologic variables except biopsy Gleason score, for which the difference in median distant metastasis-free survival was reported to be 3.5 years.
The results of the univariate Cox regression analysis assessing the ability of clinicopathologic variables to predict the time to distant metastasis are shown in Table 2. The PSA slope 3–12 months after undergoing RRP was found to be the most powerful predictor of distant metastasis-free survival, followed by biopsy Gleason score and clinical stage (HR = 2.9, 2.6, and 2.0, respectively). The Kaplan–Meier plot using the combination of SV status, LN status, and RRP Gleason score to form three prognostic risk groups is presented in Figure 3. Overall, the three prognostic risk groups were significant with regard to the prediction of time-dependent failure (distant metastasis) (HR = 1.56; P < 0.01). Although the difference in the median distant metastasis-free survival times for Groups 2 and 3 was not found to be significantly different (log-rank test, P = 0.10), the difference in the median distant metastasis-free survival time for Groups 1 and 2 was found to approach statistical significance (log-rank test, P = 0.05). Table 3 details the variables included in the stratification of the three risk groups.
|Parameters||Categories||Distant metastatic risk from RRPa|
|Positive||Negative||HR||P value||Chi-square test|
|Clinical T classification at diagnosis||T1b/T1c vs. T2a/T2b vs. T2c/T3a||1.96||<0.01||12.47|
|Biopsy Gleason score||8–9||5–7||2.56||<0.01||13.34|
|SV and LN status combined||SV+ and LN+||SV-/LN-, SV+/LN-, or SV-/LN+||1.91||0.01||6.73|
|RRP Gleason score||8–9||6–7||1.83||0.01||6.62|
|RRP Gleason score||8–9 vs. 7 (4 + 3) vs. 6–7 (3 + 4)||1.46||< 0.01||7.83|
|PSA slope after RRPb||Slope ≥ 0.05||Slope < 0.05||2.91||< 0.01||13.78|
|Group no.||No. of Distant Metastases (%)||Definition|
|1||49 (35)||Gleason score 6 and 7 (3 + 4) and SV-/LN- or SV+/LN- or SV-/LN+|
|2||40 (42)||Gleason score 7 (4 + 3)|
|3||71 (58)||Gleason score 8–9|
|Gleason score 6 and 7 (3 + 4) and SV+/LN+|
The PSA slope 3–12 months after undergoing RRP was calculated for 119 of 160 patients who had ≥ 2 PSA values available after undergoing RRP. A Kaplan–Meier plot for the post-RRP PSA slope using a cutoff value ≥ 0.05 to categorize the patients into favorable and unfavorable prognostic groups is presented in Figure 4. The median time to distant metastasis for patients with a post-RRP PSA slope < 0.05 was dramatically different compared with patients with a post-RRP PSA slope ≥ 0.05 (13.0 years vs. 3.5 years, respectively; log-rank test, P < 0.001, HR = 2.9).
Multivariate Cox regression analysis was performed using RRP pathology variables including prebiopsy PSA levels, PSA density, prostate mass, and post-RRP PSA slope. The only variables that were found to be significant on multivariate analysis were the PSA slope after RRP (HR = 2.7; P < 0.01) and RRP Gleason score (HR = 1.8; P = 0.04).
Pound et al.5 showed that the natural history of disease progression to distant metastases and death after biochemical disease recurrence is variable but can be prolonged. To our knowledge, there are no reliable data concerning the timing and natural history of disease progression to distant metastasis for men who fail to achieve an undetectable PSA level after undergoing RRP.
What does a persistently detectable PSA level after RRP mean? Two potential explanations for a PSA level that does not become undetectable within 6 weeks after a successful RRP include the presence of systemic, micrometastatic disease that went undetected preoperatively, and the presence of residual benign prostate tissue left behind at surgery (i.e., overpreservation of the bladder neck, incomplete prostatectomy with residual capsular tissue at the apex, and capsular incisions exposing benign tissue).12 Men with a PSA level in the early postoperative period that is detectable but stable often do not have disease progression.13 Benign glandular tissue at the margins of surgical resections is a possible explanation for stable postoperative PSA levels ≤ 0.3 ng/mL.14 The misinterpretation of persistent, low levels of serum PSA caused by the presence of unresected benign glands as systemic disease could lead potentially to unnecessary, aggressive adjuvant treatment in these patients. If PSA fails to reach undetectable levels after surgery but remains stable at levels < 0.4 ng/mL for ≤ 3 years after surgery, it is reasonable to consider retained benign tissue as a potential source of detectable PSA.13 However, if a patient's detectable PSA level increases above 0.4 ng/mL within 3 years after surgery or never decreases below 0.4 ng/mL, the presence of systemic disease should be strongly considered. One possible method to monitor this possibility of residual micrometastatic disease might be the use of technology that detects circulating prostate cells, preferably cancer cells, in the blood.15, 16 More recently, a new commercially available technique already has been quite successful in breast carcinoma cases and could prove of value in patients with prostate carcinoma.17, 18
The current study analysis demonstrates that some men who fail to achieve an undetectable PSA level will experience slow disease progression and remain free of distant metastasis for an extended length of time. In the current study, 38% of patients had no evidence of metastases for ≥ 7 years. The median time from surgery for all patients to disease progression to distant metastases was 5 years. A significant number of men who fail to achieve an undetectable PSA level will experience rapid disease progression. In the current study, 32% of the patients were reported to develop metastases within 3 years. It is important to direct efforts at the early identification of those patients who will develop disease progression more rapidly. PSA kinetics, such as PSA velocity, PSA density, and PSA doubling time, have proven valuable in estimating the aggressiveness and location of disease in patients with biochemical disease recurrence.5, 19–22 It is interesting to note that PSA density was not found to be predictive of subsequent distant metastasis in the current study. However, we observed that the kinetics of change in PSA level from 3–12 months after RRP (represented by a slope) was strongly predictive of earlier time to distant metastasis in the current analysis (Fig. 4). Our analysis demonstrates that it might be clinically useful to monitor these treatment failures at more frequent intervals (3 months) for the first 18 months after undergoing RRP to calculate PSA slopes earlier in the disease process.
The risk of developing distant metastatic disease when PSA persists in the detectable range after RRP is performed was shown to correlate with biopsy and RRP Gleason scores. Men with RRP Gleason scores < 8 had a 62% chance of remaining free of metastasis 5 years after surgery compared with a 30% probability in men with higher grade tumors (Gleason score of 8–10). The risk of developing metastatic disease when PSA persists after RRP also was shown to correlate with clinical stage. The reason for presenting two different models of outcome was to compare postoperative pathology (Fig. 3) and PSA kinetic data (Fig. 4) in predicting distant metastasis-free survival. Also, only a subset of patients had at least two PSA measurements available for calculation of PSA slope, whereas postoperative pathology reports were available for all patients. However, all variables were analyzed simultaneously using multivariate analysis and both RRP Gleason score and post-RRP PSA slope were found to be independent prognostic variables. We believe that both models are informative and provide important clinical information related to patient risk. However, it appears that PSA slope after undergoing RRP was a better predictor of distant metastasis-free survival time in this cohort of patients.
Pound et al.5 described the natural history of disease progression to distant metastasis after biochemical disease recurrence. They demonstrated that time to biochemical disease progression (≤ 2 years), prostatectomy Gleason score (> 7), and PSA doubling time (≤ 10 months) was predictive of the probability as well as the time to the development of metastatic disease. One potential limitation of the study was that two to three predictive variables were time dependent. The concept of immediate adjuvant therapy for patients who are most likely to harbor aggressive disease and are at increased risk of developing metastatic disease recently has gained interest and has been shown to provide some benefit.23 The ability to classify patients with a persistently detectable PSA into risk groups for the development of distant metastasis could be useful in helping physicians and patients decide who should be treated early with hormonal therapy or investigational protocols and who should undergo surveillance. The PSA slope 3–12 months after RRP was found to be a significant predictor in the current study, but this variable is time dependent. It would be ideal to be able to predict the likelihood of developing distant metastasis based entirely on information available at the time of surgery or soon thereafter. A pathologic-based tool referred to as quantitative nuclear grading, which utilizes DNA labeling image-based technology that can be applied to paraffin-embedded, archived tissue sections, has been shown to increase prognostic ability beyond what is possible with clinical and pathologic variables.24 This information should help physicians and patients to make educated treatment decisions in the early postoperative period when designing and enrolling patients in investigational protocols. In addition, the detection of circulating prostate carcinoma cells in the blood is another technology that may facilitate the prediction of patients likely to develop distant metastases.15–18
The limitations of the current study included the retrospective design and limited follow-up in some patients. Follow-up was limited for 47 of the patients in this study (30%) who had only 1–2 years of follow-up after undergoing RRP. Only six of these patients had developed distant metastasis at the last date of follow-up. The poor correlation between biopsy and prostatectomy Gleason scores in the patient sample in the current study made it difficult to use this critical pretreatment pathology variable to predict distant metastasis. However, the phenomenon of histologic upgrading from biopsy to prostatectomy specimen, which most likely accounts for the lack of correlation in the current study, is to our knowledge well established.23, 25 The findings of the current study will require external validation.
In the current study, we report on the natural history of disease progression to distant metastasis in men with a failure to achieve an undetectable PSA level after undergoing RRP for clinically localized disease. Many patients remained free of metastatic disease for an extended period without other forms of therapy despite failing to achieve an undetectable PSA level, whereas another group of patients experienced rapid disease progression. Postoperative pathology provides some guidance in the identification of patients who do not achieve an undetectable PSA level after RRP who are at a higher risk for eventual distant disease metastasis. However, the post-RRP PSA slope, as calculated using PSA levels 3–12 months after surgery, was found the best predictor of time to distant metastasis in this cohort of men. The latter finding would indicate that men at high risk for prostate carcinoma progression based on preoperative and postoperative pathology should be monitored more frequently in the first 3–18 months after undergoing RRP to allow kinetic analysis of their PSA levels. Finally, it would appear that a Partin-like table, nomogram, or computer program could be constructed with the inputs derived from the current study to manage men at high risk for disease progression.
- 9BeahrsOH, HensonDE, HutterRVP, MyersMH editors. AJCC cancer staging manual. 4th ed. Philadelphia: J.B. Lippincott Co., 1992: 181–186.
- 10The Veterans Adminstrative Cooperative Urological Research Group. Histological grading and clinical staging of prostatic carcinoma. In: TannenbaumM, editor. Urologic pathology: the prostate. Philadelphia: Lea & Febiger, 1977: 171–198.,