- Email: [email protected]
PERCENT OF CORES POSITIVE FOR CANCER IS A BETTER PREOPERATIVE PREDICTOR OF CANCER RECURRENCE AFTER RADICAL PROSTATECTOMY THAN PROSTATE SPECIFIC ANTIGEN
0022-5347/04/1714-1492/0 THE JOURNAL OF UROLOGY® Copyright © 2004 by AMERICAN UROLOGICAL ASSOCIATION
Vol. 171, 1492–1499, April 2004 Printed in U.S.A.
DOI: 10.1097/01.ju.0000118690.05943.c0
PERCENT OF CORES POSITIVE FOR CANCER IS A BETTER PREOPERATIVE PREDICTOR OF CANCER RECURRENCE AFTER RADICAL PROSTATECTOMY THAN PROSTATE SPECIFIC ANTIGEN IGNACIO F. SAN FRANCISCO, MEREDITH M. REGAN, ARIA F. OLUMI
AND
WILLIAM C. DEWOLF*
From the Division of Urologic Surgery and Biometrics Center (MMR), Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, Massachusetts
ABSTRACT
Purpose: We examined the prognostic significance of clinical and pathological variables on outcome following radical retropubic prostatectomy (RRP) in a cohort of patients in the postprostate specific antigen (PSA) era. Materials and Methods: We reviewed the clinical and pathological data on a cohort of 476 patients who underwent RRP for localized prostate cancer between January 1990 and July 2001 by 1 urologist (WCD). Median age, preoperative PSA and followup were 61 years, 5.8 ng/ml and 49 months, respectively. We used Cox proportional hazard modeling to evaluate the prognostic significance of clinical and pathological variables for cancer recurrence, defined as 2 successive PSA determinations 0.3 ng/ml or greater. Results: Of the 476 patients 53 (11%) had recurrence. Estimated cancer nonprogression probability was 86% (95% CI 83 to 90) and 76% (95% CI 68 to 86) at 5 and 10 years, respectively. Two multivariate analyses were performed. The first analysis, using only preoperative indicators, found that the percent of biopsy cores positive for cancer and biopsy Gleason score were the best predictive indicators of recurrence. The second multivariate analysis, using preoperative and postoperative indicators, found that the percent of biopsy cores positive for cancer, RRP Gleason score and the combined pathological stage/margin status variable were the best predictive indicators of recurrence. PSA was not found to be an important predictor of recurrence on either multivariate analysis. Patients with a percent of biopsy cores in the upper half of the distribution (greater than 28% positive) were at significantly increased risk for recurrence compared with those in the lower half of the distribution (28% or less positive) (HR 3.86, p ⬍0.001). Conclusions: The percent of cores positive for cancer was a better predictor of cancer recurrence than PSA in this post-PSA era RRP series. In addition, surgical Gleason score and pathological stage/surgical margins were also independent predictors of cancer recurrence after RRP. These 3 predictors are displayed in a nomogram-type format to summarize estimated 5 and 10-year recurrence-free probabilities. KEY WORDS: prostate, prostatic neoplasms, prostatectomy, biopsy, recurrence
Since its introduction in 1987, prostate specific antigen (PSA) has been the most commonly used marker for the diagnosis of prostate cancer and the detection of cancer recurrence after radical retropubic prostatectomy (RRP).1 In fact, because of its clinical usefulness, PSA screening has led to stage migration and the earlier detection of prostate cancer.2, 3 It has been shown that preoperative parameters, such as PSA, clinical stage, biopsy Gleason score4, 5 and percent of cancer in the needle biopsy,6, 7 can predict treatment failure. However, only PSA as a preoperative parameter has been shown to be of significant value when combined in multivariate analysis with other postoperative parameters, such as prostatectomy specimen Gleason score, extracapsular extension, surgical margin status, seminal vesicle involvement and lymph node metastases, for predicting recurrence.8, 9 Stage migration data, including those presented in this study, suggest that PSA is becoming less reliable as a preoperative predictor of postoperative recurrence on multivariate
analysis. With this observation in mind we re-investigated the role of preoperative parameters, eg PSA and the percent of cores positive for cancer in the needle biopsy, in combination with postoperative parameters to identify reliable predictors of outcome in patients who have undergone RRP. This analysis is of special interest and it is unique for several reasons. It is a uniform single surgeon series with 92% of the study cohort having followup during the last 2 years of the 11-year study period and 95% having at least 2 followup visits after RRP. All diagnoses and surgery were performed in the post-PSA era and most prostate biopsy results were assessed by a single core embedding technique that has been shown to improve the pathological evaluation of needle cores.10 All of these factors tended to increase the reliability and fidelity of the data. In addition, we propose that specific independent predictors of recurrence as assessed by these data, ie the percent of biopsy cores positive for cancer, RRP Gleason score and pathological stage/margin status variable, can be displayed in a nomogram-type format to summarize recurrence-free probability estimates.
Accepted for publication November 21, 2003. Supported by the Hershey Family Foundation at the Beth Israel Deaconess Medical Center. * Correspondence: 330 Brookline Ave., Division of Urologic SurMATERIALS AND METHODS gery, Beth Israel Deaconess Medical Center, Boston, Massachusetts Patient population. We reviewed the clinical data and pa02215 (telephone: 617-6678903; FAX: 617-6677292; e-mail: wdewolf@ thology reports of 481 consecutive patients with clinically bidmc.harvard.edu). 1492
1493
PERCENT OF CORES WITH CANCER PREDICTS PROSTATE CANCER RECURRENCE
localized prostate cancer who underwent anatomical RRP from January 1, 1990 to July 31, 2001 by 1 urologist (WCD). Patients with neoadjuvant hormonal therapy were excluded from this study. In addition, 2 patients were excluded because of death a few months after the surgery from causes other than prostate cancer and, therefore, they had no followup. Another 3 patients were excluded because most clinical and pathological data were not available. Therefore, the final cohort of our study consisted of 476 patients. Median age in the whole cohort was 61 years (range 35 to 77) and median followup was 49 months (range 1 to 134). Clinical and pathological evaluation. Prostate needle biopsies that were positive for cancer and performed elsewhere were reviewed by pathologists from our institution to confirm the diagnosis of cancer and Gleason score. Ten pathologists were involved in the formulation of Gleason scores and positivity of the prostate cores, and the current study is a retrospective analysis of these cores. All prostatectomy specimens were reviewed at our combined genitourinary pathology conference, which included a single senior genitourinary pathologist.11 Furthermore, all positive or questionable biopsies were confirmed by 1 of the 2 senior pathologists. Prostate biopsies performed at our institution were done with a 6.5 MHz transrectal probe (General Electric, Milwaukee, Wisconsin). Needle biopsies were obtained using a Pro-Mag 2.2 automatic biopsy instrument (Manan Medical Products, Inc., Northbrook, Illinois) with an 18 gauge ⫻ 25 cm cutting biopsy needle.11 Between 3 and 16 cores (median 7) were sampled. The percent of cores positive for cancer was calculated by the number of cores with cancer divided by the total number of cores multiplied by 100. Each prostate biopsy done at our institution was individually embedded and labeled for maximal surface area analysis.10 RRP specimens were submitted for microscopic examination, including the apex and base margins, 1 section from each seminal vesicle and 3 sections from each lobe (apical, mid and base). If cancer was not found in the sections, additional tissue was analyzed. The 1997 TNM staging system from the American Joint Committee of Cancer was used for clinical and pathological staging.12 Clinical staging consisted of digital rectal examination with T2a and T2b defined as any firmness or suspicious area of cancer in the corresponding prostatic anatomy. T2 was not necessarily intended to indicate a hard nodule. Clinical staging also included serum PSA measurement. When PSA was greater than 10 ng/ml, Gleason score was 7 or there was possible clinical T3 disease on rectal examination, abdominal and pelvic computerized tomography, and bone scan were also done. Pathological stage and margin status were combined as a single variable for analysis because combination of these variables are better predictors of recurrence than each alone.13 Positive or negative surgical margins were indicated by ⫹ and ⫺, respectively, following the staging assignment. This combined variable included organ confined disease with negative surgical margins (pT2⫺), organ confined with positive margins (pT2⫹), and extracapsular extension without (pT3a⫺) and with (pT3a⫹) positive surgical margins. In addition, the usual designations for seminal vesicle invasion (pT3b) and lymph node metastases (N1) were used. Followup. The first PSA value was routinely obtained 6 weeks after surgery. Patients were then followed with digital rectal examination and serum PSA every 6 months for 1 year and then once yearly thereafter, although PSA determinations were usually made twice yearly. Cancer recurrence was defined as 2 consecutive serum PSA values of 0.3 ng/ml or higher. Statistical analysis. Clinical, surgical and pathological factors are summarized as the median and IQR (25th, 75th percentiles) or as the number and percent of patients. We used univariate and multivariate Cox proportional hazards regression modeling to assess the prognostic significance of
clinical and pathological factors on time to cancer recurrence. The HR, 95% CI and 2-sided p values are reported. For modeling continuous variables were categorized into a natural categorization if one existed, or if not, by quartiles or at the median. Multivariate analysis used forward stepwise regression to identify a best model using p ⬍0.01 as a criterion to remain in the model. The model was limited to 5 variables at most because there were 53 events. The 5 and 10-year recurrence-free probabilities were estimated by exponentiation of the negative cumulative hazard estimated from the model and 95% CIs were calculated from the log of the negative log of the estimated survival function. The additional explanatory value of PSA as 2 indicator variables of PSA 4 to 10 ng/ml or greater than 10 ng/ml at the termination of model selection was assessed with a 2 df chi-square statistic. For statistical analysis SAS, version 8.0 (SAS Institute, Inc., Cary, North Carolina) software was used. RESULTS
Patient characteristics. Table 1 lists preoperative clinical characteristics of the cohort. Median patients age was 61 years (range 35 to 77) and median pre-RRP PSA was 5.85 TABLE 1. Preoperative clinical and biopsy characteristics of 476 patient cohort Variable Median age at diagnosis (IQR) Median ng/ml pre-RRP PSA (IQR) Median No. cores (IQR) Median No. Ca pos cores (IQR) Median % Ca pos cores (IQR) No. clinical stage (%): T1a T1b T1c T2a T2b No. Gleason score (%): 3–5 6 7 8 9
61 (55, 65) 5.85 (4.17, 9.05) 7 (6, 10) 2 (1, 3) 28 (16, 50) 1 (0.2) 4 (0.8) 140 (29.4) 224 (47.1) 107 (22.5) 10 (2.1) 342 (71.9) 92 (19.3) 22 (4.6) 10 (2.1)
TABLE 2. Surgical and postoperative pathological characteristics of 476 patient cohort Variable No. nerve sparing surgery (%): Unilat Bilat Median gm prostate wt (IQR)* No. pathological stage (%): T2a T2b T3a T3b No. Gleason score (%): 5–6 7 (3⫹4) 7 (4⫹3) 8–9 No. pos margins (%): Apical Posterior Bladder Lat No. seminal vesicle involvement (%) No. lymph node metastases (%) No. pathological stage/margin status (%): pT2⫺ pT2⫹ pT3a⫺ pT3a⫹ pT3b N1 (lymph node pos) * Total of 424 patients.
64 (13.4) 111 (23.3) 56 (45–69) 156 (32.8) 279 (58.6) 24 (5.0) 17 (3.6) 311 101 33 31 94 68 7 5 25 17 5
(65.3) (21.2) (6.9) (6.5) (19.7) (14.3) (1.5) (1.1) (5.3) (3.6) (1.1)
366 (76.9) 68 (14.3) 5 (1.1) 19 (4.0) 13 (2.7) 5 (1.1)
1494
PERCENT OF CORES WITH CANCER PREDICTS PROSTATE CANCER RECURRENCE
ng/ml (range 0.4 to 63.0). A median of 7 biopsy cores (range 3 to 16) were obtained with a median of 2 cores (28%) positive for cancer. Of the patients 72% had a Gleason score of 6 on needle biopsy. Table 2 lists surgical and pathological characteristics of the cohort. More than 90% of cases were stage T2 and 303 (64%) had a Gleason score of 6. Of the patients 20% had 1 or more positive margins. Therefore, 371 patients (almost 80%) had specimen confined disease, ie pT2 or pT3. Only 5 patients (1.1%) had lymph node metastases. Disease recurrence and survival. The cohort of 476 patients was followed a median of 49 months. Only 53 patients (11.1%) experienced disease recurrence following RRP. The estimated cancer nonprogression probabilities were 86% (95% CI 83 to 90) and 76% (95% CI 68 to 86) at 5 and 10 years, respectively (fig. 1). Fifteen patients have died, of whom only 2 (0.4%) died of prostate cancer. Univariate analysis to predict disease recurrence. Various clinical and pathological factors were evaluated on univariate analysis as independent predictors of disease recurrence (table 3). Preoperative clinical parameters that were associated with an increased hazard of disease recurrence were elevated PSA, the number or percent of biopsy cores positive for cancer above the median and higher Gleason score (each p ⬍0.01). Nerve sparing RRP was marginally associated with an increased hazard of recurrence (p ⫽ 0.05). Combined pathological stage and margin status was statistically significant (p ⬍0.0001). As expected, categories T2⫹, T3a⫹, T3b and N1 had significantly increased hazards of recurrence compared with specimen confined disease (pT2⫺ and pT3a⫺). Multivariate analysis to predict disease recurrence. We performed 2 sets of multivariate analyses, the first involving
only preoperative variables and the second involving combination preoperative, surgical and postoperative variables. 1) Analysis of preoperative clinical variables identified 2 parameters that were statistically significant at the ␣ ⫽ 0.01 level, namely biopsy Gleason score and percent of biopsy cores positive for cancer (table 4). Elevated total PSA was not statistically significant at the ␣ ⫽ 0.01 level and, therefore, it did not remain in the model (chi-square 6.3 (2 df), p ⫽ 0.043). Compared with patients who had a biopsy Gleason score of less than 7, patients with a Gleason score of 7 and 8 –9 had significantly increased hazards of recurrence (HR 3.92 and 12.24, respectively, p ⬍0.0001). Patients with the percent of positive biopsy cores in the upper half of the distribution (median greater than 28% positive cores, ie 2 or more positive cores in a 6 core biopsy, or 4 or more in a 12 core biopsy) had a significantly increased hazard of recurrence compared with those in the lower half of the distribution (median of 28% or less positive cores, ie 1 positive core in a 6 core biopsy, or 3 or fewer in a 12 core biopsy) (HR 3.62, p ⬍0.001). 2) When surgical and postoperative variables were added to the preoperative variables in multivariate analysis, 3 parameters remained significant at the ␣ ⫽ 0.01 level, namely the percent of biopsy cores positive for cancer, RRP Gleason score and the pathological stage/margin status combined variable (table 4). Patients with the percent of biopsy cores in the upper half of the distribution (greater than 28% positive) had a significantly increased hazard of recurrence compared with those in the lower half of the distribution (28% or less positive) (HR 3.86, p ⬍0.001). Compared with patients who had a Gleason score of 5, 6 or 7 (3⫹4), patients with a Gleason score of 7 (4⫹3) and 8 –9 had a significantly increased hazard of recurrence (HR 4.27 and 8.35, respectively, p ⬍0.0001).
FIG. 1. Kaplan-Meier estimate of time to disease recurrence in 476 men with clinically localized prostate cancer who underwent anatomical RRP showing 5 and 10-year recurrence-free probabilities and number at risk.
1495
PERCENT OF CORES WITH CANCER PREDICTS PROSTATE CANCER RECURRENCE
TABLE 3. Univariate analysis of preoperative, surgical and postoperative pathological candidate predictors of disease recurrence Variables
HR (95% CI)
Age at diagnosis vs younger than 60: 60–69 70 or Older Preop ng/ml PSA vs less than 4: 4–10 Greater than 10 Pos cores greater than median vs median or less: 3 or Greater vs 1 or 2 Greater than 28% vs 28% or less Clinical stage vs T1: T2a T2b Gleason score vs 6 or less: 7 8–9 Nerve sparing surgery yes vs no Prostate wt gm quartiles vs quartile 1: Quartile 2 Quartile 3 Quartile 4 Pathological stage vs T2: T3a T3b Gleason score vs 6 or less: 7 (3⫹4) 7 (4⫹3) 8–9 Pos margins yes vs no Seminal vesicle involvement yes vs no Lymph node metastases yes vs no Pathological stage/pos margins vs specimen confined pT2⫺ ⫹ pT3a⫺: pT2⫹ pT3a⫹ pT3b N1 (lymph node pos)
p Value (log rank test) 0.15
0.91 (0.50, 1.65) 1.95 (0.84, 4.51)
⬍0.0001
8.46 (1.15, 62.44) 26.10 (3.55, 192.24) 5.50 (2.92, 10.36) 4.78 (2.32, 9.84) 1.25 (0.63, 2.48) 1.55 (0.72, 3.31) 5.36 (2.83, 10.16) 13.94 (6.76, 28.76) 0.54 (0.29, 1.01) 1.19 (0.56, 2.55) 1.21 (0.56, 2.63) 0.51 (0.20, 1.30) 6.21 (3.05, 12.63) 15.84 (7.62, 32.96) 2.49 11.45 17.11 4.50 12.62 34.25
(1.10, 5.64) (5.35, 24.51) (7.99, 36.65) (2.62, 7.75) (6.19, 25.72) (9.94, 117.97)
⬍0.0001 ⬍0.0001 0.53 ⬍0.0001 0.05 0.29
⬍0.0001 ⬍0.0001
⬍0.0001 ⬍0.0001 ⬍0.0001 ⬍0.0001
3.67 (1.83, 7.36) 13.06 (6.03, 28.28) 18.09 (7.44, 43.95) 79.15 (21.68, 288.99)
TABLE 4. Multivariate analysis of candidate predictors for disease recurrence Variables
HR (95% CI)
p Value*
Preop clinical ⫹ biopsy† % Pos biopsy cores: Lower distribution half (28 or less) Upper distribution half (greater than 28) Biopsy Gleason score: 3–6 7 8–9
Reference 3.62 (1.74, 7.54) Reference 3.92 (2.04, 7.53) 12.24 (5.90, 25.40) Preop, surgical ⫹ postop pathological†
— ⬍0.001 — ⬍0.0001 ⬍0.0001
% Pos biopsy cores: Lower distribution half (28 or less) Reference — Upper distribution half (greater than 28) 3.86 (1.85, 8.04) ⬍0.001 RRP Gleason score: 5–7 (3⫹4) Reference — 7 (4⫹3) 4.27 (2.13, 8.55) ⬍0.0001 8–9 8.35 (4.15, 16.78) ⬍0.0001 Pathological stage/margin status: Specimen confined pT2⫺ ⫹ pT3a⫺ Reference — pT2⫹ 2.39 (1.17, 4.87) 0.017 pT3a⫹, pT3b ⫹ N1 (lymph node pos) 7.94 (3.94, 15.99) ⬍0.0001 * Final model was re-parameterized for presentation, so that each variable represents increased recurrence hazard relative to its reference category, which is the category with the better prognosis, and because of this re-parameterization, p values are not strictly p ⬍0.01. † PSA did not provide additional explanatory value to final model since 2 indicator variables PSA 4 to 10 and greater than 10 ng/ml did not achieve statistical significance at ␣ ⫽ 0.01.
For example, of patients with a Gleason score of less than 7 or 3⫹4 only 5.6% had recurrence (12 of 311 with less than 7 and 11 of 101 with 3⫹4), while 45% (15 of 33) with a Gleason score of 4⫹3 and 48% (15 of 31) with greater than 7 had recurrence. Also, patients with positive margins without evidence of extracapsular extension (pT2⫹) had an increased hazard of recurrence compared with patients with specimen confined disease (pT2⫺ or pT3a⫺) (HR 2.39, p ⫽ 0.017). As expected, patients with nonorgan confined disease/positive surgical margins (pT3a⫹), seminal vesicle invasion (pT3b) or lymph node metastasis (N1) had a significantly increased hazard of recurrence (HR 7.94, p ⬍0.0001). Figures 2 to 4 show univariate KaplanMeier curves for these 3 factors.
Table 5 lists 5 and 10-year recurrence-free probabilities with the 95% CI estimated from the model in a nomogramtype format. The reference group with a low RRP Gleason score, specimen confined disease and low percent of positive biopsy cores were estimated to have 5 and 10-year recurrence-free probabilities of above 0.98 (95% CI 0.96, 0.99) and 0.96 (95% CI 0.90, 0.98), respectively. For example, if the patient had a high percent of positive biopsy cores (greater than 28%) and the other 2 factors were held constant, the estimated 5-year recurrence-free probability decreased from 0.98 to 0.93. In the worst case scenario a patient with an RRP Gleason score of 8 –9, nonorgan, nonspecimen confined disease and a high percent of positive biopsy cores was esti-
1496
PERCENT OF CORES WITH CANCER PREDICTS PROSTATE CANCER RECURRENCE
FIG. 2. Kaplan-Meier estimate of time to disease recurrence in 476 men who underwent RRP by whether percent of positive biopsy cores was above (greater than 28%) or below median.
mated to have a 5-year recurrence-free probability of 0.01 (95% CI 0.00, 0.08). DISCUSSION
We defined multiple preoperative and postoperative parameters that are candidate clinical guidelines to help determine the risks of disease recurrence after RRP. We performed 2 multivariate analyses. The first analysis involved preoperative parameters determined to be important for defining the risk of recurrence. Only the percent of positive biopsy cores and Gleason score were found to be significant at the ␣ ⫽ 0.01 level, while PSA was not (table 4). The second multivariate analysis combined preoperative and postoperative parameters thought to be important for predicting recurrence. The results indicated that the percent of cores positive for cancer, RRP Gleason score and pathological stage/margin status were the best independent predictors of recurrence. Again, preoperative PSA did not achieve statistical significance. The major contribution of this study is that it shows that the percent of cores positive for cancer in needle biopsies are a better predictor of biochemical recurrence after RRP than PSA. Likewise, other series have found a role for the percent of positive cores to help define tumor stage and volume,14 and outcome after RRP.6, 7 D’Amico et al found that the percent of positive cores was predictive of outcome only when preoperative parameters were included on multivariate analysis.6 Finally, Freedland et al reported that the percent of biopsy tissue with cancer, rather than the percent of cores with cancer, was predictive of biochemical failure after RRP.7 In contrast, our study shows that the percent of positive cores,
when combined with postoperative parameters, is highly predictive of relapse (tables 4 and 5). We believe that the value of the percent of positive cores as a predictor of cancer recurrence is related to at least 3 factors. 1) It has been demonstrated that the percent of cores positive for cancer, as determined by the equation, [(number of cores positive for cancer/ total number of cores) ⫻ 100] is the expression of real tumor volume.14 2) the extended needle biopsy technique improves cancer detection15 and cancer grading accuracy11 3) According to a previous study, a higher rate of cancer can be detected when cores are embedded individually,16 a technique that we have used for RRP specimen analysis.10 In our statistical analysis of biopsy results we implemented a priori a standard statistical approach and created groups by distribution quartiles. It was done because there are no generally accepted cutoffs for grouping subjects according to positive biopsy cores, as there are for PSA, for example. It led to a cutoff above the median value of 28% positive cores. In light of our results a careful examination and determination of an optimal cutoff or cutoffs for determining risk according to positive biopsy cores should be done. Such an analysis, for example creating groups by distribution deciles, could not be performed in the current study because the small number of recurrences in the cohort limited the proportional hazards model to 5 covariates at most. Longer followup with more data points would answer these concerns. Why was preoperative PSA not an important preoperative parameter for cancer recurrence in contrast to the percent of biopsy cores positive for cancer? 1) Multivariate analyses showed the most important parameters in determining re-
PERCENT OF CORES WITH CANCER PREDICTS PROSTATE CANCER RECURRENCE
1497
FIG. 3. Kaplan-Meier estimate of time to disease recurrence in 476 men who underwent RRP by surgical Gleason score
currence but did not exclude the importance of preoperative PSA. If PSA were added to the final model of preoperative variables, PSA would add explanatory value (p ⫽ 0.043), although it did not meet the modeling criteria. However, if PSA were added to the final model including preoperative and postoperative variables, it would not add explanatory value (p ⫽ 0.12). 2) This cohort of patients was only from the post-PSA era, which means that it was a homogeneous group with few patients with PSA greater than 20 ng/ml and 80% with PSA less than 10 ng/ml. Therefore, because of the homogeneity of PSA in most of the cohort, PSA should not be expected to be as good a determinant of recurrence compared with other parameters. As a consequence of a better selection of patients for RRP in the post-PSA era, only 2 patients died of prostate cancer in this series. The independent predictive role of RRP specimen Gleason score in the outcome of patients after RRP is not a new finding.9, 17–19 On univariate analysis the Gleason score groups 4⫹3 and greater than 7 were at significantly higher risk for recurrence than the Gleason score 5– 6 group (table 3 and fig. 3). However, the risk of recurrence in the Gleason score 3⫹4 group was only marginally increased. On multivariate analysis it was no longer important to distinguish the Gleason 3⫹4 group from the Gleason score 5– 6 group (table 4). Therefore, according to our model and data we can see a markedly different outcome in patients with primary Gleason grade 3 and primary Gleason grade 4, as represented by the Gleason scores 3⫹4 and 4⫹3 (table 5), similar to the findings of others.20
The significance of pathological stages and positive surgical margins as independent predictors of recurrence is well known and has been described in several publications.9, 17–19 In this study we wished to combine the 2 variables to differentiate organ confined, nonorgan confined, specimen confined, and nonspecimen confined disease, similar to Paulson.13 We found that patients with specimen confined cancer (T2⫺ and T3⫺) had better outcomes than patients with organ confined cancer with positive margins (T2⫹) (table 5 and fig. 4). Patients with T3a⫺ disease had better than expected results (fig. 4), possibly because of the low number of patients in this category. According to our data surgical margins are important for the patient prognosis after RRP, even in patients with organ confined disease. It is especially true in patients who have Gleason scores 4⫹3 and greater than 7 with a high percent of positive cores (table 5 and fig. 4). In contrast, none of the 196 patients who were low risk in all 3 parameters (ie Gleason score 5– 6 or 3⫹4, T2⫺, T3⫺ and 28% or less of positive cores for cancer) have had recurrence to date after RRP.
CONCLUSIONS
The percent of cores positive for cancer was a better predictor of cancer recurrence than PSA in this post-PSA era RRP series. In addition, surgical Gleason score and pathological stage/surgical margins were also independent predictors of cancer recurrence after RRP. These 3 predictors are displayed in a nomogram-type format to summarize esti-
1498
PERCENT OF CORES WITH CANCER PREDICTS PROSTATE CANCER RECURRENCE
FIG. 4. Kaplan-Meier estimate of time to disease recurrence in 476 men who underwent RRP by pathological stage/margin status category TABLE 5. Nomogram-type summary of estimated 5 and 10-year recurrence-free probability No. Pts
No. Recurrences
196 143 20 33 10 10 7 10 1 7 1 7 8 7 4 3 1 8
0 9 0 5 3 6 0 5 0 4 1 5 3 2 2 3 0 5
RRP Gleason Score 5, 5, 5, 5, 5, 5,
6, 7 (3⫹4) 6, 7 (3⫹4) 6, 7 (3⫹4) 6, 7 (3⫹4) 6, 7 (3⫹4) 6, 7 (3⫹4) 7 (4⫹3) 7 (4⫹3) 7 (4⫹3) 7 (4⫹3) 7 (4⫹3) 7 (4⫹3) 8, 9 8, 9 8, 9 8, 9 8, 9 8, 9
Pathological Stage/ Margin Status pT2⫺, pT3a⫺ pT2⫺, pT3a⫺ pT2⫹ pT2⫹ pT3a⫹, pT3b, pT3a⫹, pT3b, pT2⫺, pT3a⫺ pT2⫺, pT3a⫺ pT2⫹ pT2⫹ pT3a⫹, pT3b, pT3a⫹, pT3b, pT2⫺, pT3a⫺ pT2⫺, pT3a⫺ pT2⫹ pT2⫹ pT3a⫹, pT3b, pT3a⫹, pT3b,
N1 N1
N1 N1
N1 N1
mated 5 and 10-year recurrence-free probabilities. Longer followup with more data points and verification of these data would allow to use this nomogram in general clinical prac-
% Pos Biopsy Cores 28 or Less Greater than 28 or Less Greater than 28 or Less Greater than 28 or Less Greater than 28 or Less Greater than 28 or Less Greater than 28 or Less Greater than 28 or Less Greater than 28 or Less Greater than
28 28 28 28 28 28 28 28 28
Estimated Recurrence-Free Probability (95% CI) 5-Yr
10-Yr
0.98 (0.96, 0.99) 0.93 (0.88, 0.96) 0.96 (0.89, 0.98) 0.84 (0.71, 0.91) 0.86 (0.69, 0.94) 0.55 (0.31, 0.74) 0.92 (0.81, 0.97) 0.73 (0.52, 0.86) 0.82 (0.60, 0.93) 0.47 (0.22, 0.68) 0.52 (0.19, 0.77) 0.08 (0.01, 0.26) 0.85 (0.68, 0.94) 0.54 (0.27, 0.74) 0.68 (0.37, 0.86) 0.23 (0.04, 0.50) 0.28 (0.05, 0.58) 0.01 (0.00, 0.08)
0.96 (0.90, 0.98) 0.84 (0.72, 0.91) 0.90 (0.75, 0.96) 0.66 (0.40, 0.83) 0.70 (0.37, 0.88) 0.25 (0.04, 0.55) 0.83 (0.58, 0.93) 0.48 (0.19, 0.72) 0.63 (0.26, 0.86) 0.17 (0.02, 0.47) 0.22 (0.01, 0.60) ⬍0.01 (0.00, 0.08) 0.69 (0.38, 0.86) 0.24 (0.04, 0.53) 0.41 (0.08, 0.73) 0.03 (0.00, 0.25) 0.05 (0.00, 0.33) ⬍0.01 (0.00, 0.01)
tice. The low rates of cancer recurrence in our series may have been primarily related to the stage migration of prostate cancer associated with PSA screening.
PERCENT OF CORES WITH CANCER PREDICTS PROSTATE CANCER RECURRENCE
Drs. Glenn Bubley and Yoo-Joung Ko provided a critical reading of the manuscript. 10. REFERENCES
1. Stamey, T. A., Yang, N., Hay, A. R., McNeal, J. E., Freiha, F. S. and Redwine, E.: Prostate-specific antigen as a serum marker for adenocarcinoma of the prostate. N Engl J Med, 317: 909, 1987 2. Catalona, W. J., Smith, D. S., Ratliff, T. L. and Basler, J. W.: Detection of organ-confined prostate cancer is increased through prostate-specific antigen-based screening. JAMA, 270: 948, 1993 3. Humphrey, P. A., Keetch, D. W., Smith, D. S., Shepherd, D. L. and Catalona, W. J.: Prospective characterization of pathological features of prostatic carcinomas detected via serum prostate specific antigen based screening. J Urol, 155: 816, 1996 4. Kattan, M. W., Eastham, J. A., Stapleton, A. M., Wheeler, T. M. and Scardino, P. T.: A preoperative nomogram for disease recurrence following radical prostatectomy for prostate cancer. J Natl Cancer Inst, 90: 766, 1998 5. Partin, A. W., Kattan, M. W., Subong, E. N., Walsh, P. C., Wojno, K. J., Oesterling, J. E. et al: Combination of prostate-specific antigen, clinical stage, and Gleason score to predict pathological stage of localized prostate cancer. A multi-institutional update. JAMA, 277: 1445, 1997 6. D’Amico, A. V., Whittington, R., Malkowicz, S. B., Schultz, D., Fondurulia, J., Chen, M. H. et al: Clinical utility of the percentage of positive prostate biopsies in defining biochemical outcome after radical prostatectomy for patients with clinically localized prostate cancer. J Clin Oncol, 18: 1164, 2000 7. Freedland, S. J., Csathy, G. S., Dorey, F. and Aronson, W. J.: Percent prostate needle biopsy tissue with cancer is more predictive of biochemical failure or adverse pathology after radical prostatectomy than prostate specific antigen or Gleason score. J Urol, 167: 516, 2002 8. Kattan, M. W., Wheeler, T. M. and Scardino, P. T.: Postoperative nomogram for disease recurrence after radical prostatectomy for prostate cancer. J Clin Oncol, 17: 1499, 1999 9. Hull, G. W., Rabbani, F., Abbas, F., Wheeler, T. M., Kattan, M. W. and Scardino, P. T.: Cancer control with radical pros-
11.
12.
13. 14.
15. 16.
17.
18.
19. 20.
1499
tatectomy alone in 1,000 consecutive patients. J Urol, 167: 528, 2002 Kao, J., Upton, M., Zhang, P. and Rosen, S.: Individual prostate biopsy core embedding facilitates maximal tissue representation. J Urol, 168: 496, 2002 San Francisco, I. F., DeWolf, W. C., Rosen, S., Upton, M. and Olumi, A. F.: Extended prostate needle biopsy improves concordance of Gleason grading between prostate needle biopsy and radical prostatectomy. J Urol, 169: 136, 2003 Fleming, I. D., Cooper, J. S., Hensen, D. E., Hutter, R. V. P., Kennedy, B. J., Murphy, G. P. et al: Prostate. In: AJCC Cancer Staging Manual, 5th ed. Philadelphia: Lippincott-Raven Publishers, vol. 34, pp. 219 –224, 1997 Paulson, D. F.: Impact of radical prostatectomy in the management of clinically localized disease. J Urol, 152: 1826, 1994 Sebo, T. J., Bock, B. J., Cheville, J. C., Lohse, C., Wollan, P. and Zincke, H.: The percent of cores positive for cancer in prostate needle biopsy specimens is strongly predictive of tumor stage and volume at radical prostatectomy. J Urol, 163: 174, 2000 Eskew, L., Bare, R. L. and McCullough, D. L.: Systematic 5 region prostate biopsy is superior to sextant method for diagnosing carcinoma of the prostate. J Urol, 157: 199, 1997 Lane, R. B., Jr., Lane, C. G., Mangold, K. A., Johnson, M. H. and Allsbrook, W. C., Jr.: Needle biopsies of the prostate: what constitutes adequate histologic sampling? Arch Pathol Lab Med, 122: 833, 1998 Zincke, H., Oesterling, J. E., Blute, M. L., Bergstralh, E. J., Myers, R. P. and Barrett, D. M.: Long-term (15 years) results after radical prostatectomy for clinically localized (stage T2c or lower) prostate cancer. J Urol, 152: 1850, 1994 Pound, C. R., Partin, A. W., Epstein, J. I. and Walsh, P. C.: Prostate-specific antigen after anatomic radical retropubic prostatectomy. Patterns of recurrence and cancer control. Urol Clin North Am, 24: 395, 1997 Catalona, W. J. and Smith, D. S.: Cancer recurrence and survival rates after anatomic radical retropubic prostatectomy for prostate cancer: intermediate-term results. J Urol, 160: 2428, 1998 Stamey, T. A., McNeal, J. E., Yemoto, C. M., Sigal, B. M. and Johnstone, I. M.: Biological determinants of cancer progression in men with prostate cancer. JAMA, 281: 1395, 1999
Vol. 171, 1492–1499, April 2004 Printed in U.S.A.
DOI: 10.1097/01.ju.0000118690.05943.c0
PERCENT OF CORES POSITIVE FOR CANCER IS A BETTER PREOPERATIVE PREDICTOR OF CANCER RECURRENCE AFTER RADICAL PROSTATECTOMY THAN PROSTATE SPECIFIC ANTIGEN IGNACIO F. SAN FRANCISCO, MEREDITH M. REGAN, ARIA F. OLUMI
AND
WILLIAM C. DEWOLF*
From the Division of Urologic Surgery and Biometrics Center (MMR), Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, Massachusetts
ABSTRACT
Purpose: We examined the prognostic significance of clinical and pathological variables on outcome following radical retropubic prostatectomy (RRP) in a cohort of patients in the postprostate specific antigen (PSA) era. Materials and Methods: We reviewed the clinical and pathological data on a cohort of 476 patients who underwent RRP for localized prostate cancer between January 1990 and July 2001 by 1 urologist (WCD). Median age, preoperative PSA and followup were 61 years, 5.8 ng/ml and 49 months, respectively. We used Cox proportional hazard modeling to evaluate the prognostic significance of clinical and pathological variables for cancer recurrence, defined as 2 successive PSA determinations 0.3 ng/ml or greater. Results: Of the 476 patients 53 (11%) had recurrence. Estimated cancer nonprogression probability was 86% (95% CI 83 to 90) and 76% (95% CI 68 to 86) at 5 and 10 years, respectively. Two multivariate analyses were performed. The first analysis, using only preoperative indicators, found that the percent of biopsy cores positive for cancer and biopsy Gleason score were the best predictive indicators of recurrence. The second multivariate analysis, using preoperative and postoperative indicators, found that the percent of biopsy cores positive for cancer, RRP Gleason score and the combined pathological stage/margin status variable were the best predictive indicators of recurrence. PSA was not found to be an important predictor of recurrence on either multivariate analysis. Patients with a percent of biopsy cores in the upper half of the distribution (greater than 28% positive) were at significantly increased risk for recurrence compared with those in the lower half of the distribution (28% or less positive) (HR 3.86, p ⬍0.001). Conclusions: The percent of cores positive for cancer was a better predictor of cancer recurrence than PSA in this post-PSA era RRP series. In addition, surgical Gleason score and pathological stage/surgical margins were also independent predictors of cancer recurrence after RRP. These 3 predictors are displayed in a nomogram-type format to summarize estimated 5 and 10-year recurrence-free probabilities. KEY WORDS: prostate, prostatic neoplasms, prostatectomy, biopsy, recurrence
Since its introduction in 1987, prostate specific antigen (PSA) has been the most commonly used marker for the diagnosis of prostate cancer and the detection of cancer recurrence after radical retropubic prostatectomy (RRP).1 In fact, because of its clinical usefulness, PSA screening has led to stage migration and the earlier detection of prostate cancer.2, 3 It has been shown that preoperative parameters, such as PSA, clinical stage, biopsy Gleason score4, 5 and percent of cancer in the needle biopsy,6, 7 can predict treatment failure. However, only PSA as a preoperative parameter has been shown to be of significant value when combined in multivariate analysis with other postoperative parameters, such as prostatectomy specimen Gleason score, extracapsular extension, surgical margin status, seminal vesicle involvement and lymph node metastases, for predicting recurrence.8, 9 Stage migration data, including those presented in this study, suggest that PSA is becoming less reliable as a preoperative predictor of postoperative recurrence on multivariate
analysis. With this observation in mind we re-investigated the role of preoperative parameters, eg PSA and the percent of cores positive for cancer in the needle biopsy, in combination with postoperative parameters to identify reliable predictors of outcome in patients who have undergone RRP. This analysis is of special interest and it is unique for several reasons. It is a uniform single surgeon series with 92% of the study cohort having followup during the last 2 years of the 11-year study period and 95% having at least 2 followup visits after RRP. All diagnoses and surgery were performed in the post-PSA era and most prostate biopsy results were assessed by a single core embedding technique that has been shown to improve the pathological evaluation of needle cores.10 All of these factors tended to increase the reliability and fidelity of the data. In addition, we propose that specific independent predictors of recurrence as assessed by these data, ie the percent of biopsy cores positive for cancer, RRP Gleason score and pathological stage/margin status variable, can be displayed in a nomogram-type format to summarize recurrence-free probability estimates.
Accepted for publication November 21, 2003. Supported by the Hershey Family Foundation at the Beth Israel Deaconess Medical Center. * Correspondence: 330 Brookline Ave., Division of Urologic SurMATERIALS AND METHODS gery, Beth Israel Deaconess Medical Center, Boston, Massachusetts Patient population. We reviewed the clinical data and pa02215 (telephone: 617-6678903; FAX: 617-6677292; e-mail: wdewolf@ thology reports of 481 consecutive patients with clinically bidmc.harvard.edu). 1492
1493
PERCENT OF CORES WITH CANCER PREDICTS PROSTATE CANCER RECURRENCE
localized prostate cancer who underwent anatomical RRP from January 1, 1990 to July 31, 2001 by 1 urologist (WCD). Patients with neoadjuvant hormonal therapy were excluded from this study. In addition, 2 patients were excluded because of death a few months after the surgery from causes other than prostate cancer and, therefore, they had no followup. Another 3 patients were excluded because most clinical and pathological data were not available. Therefore, the final cohort of our study consisted of 476 patients. Median age in the whole cohort was 61 years (range 35 to 77) and median followup was 49 months (range 1 to 134). Clinical and pathological evaluation. Prostate needle biopsies that were positive for cancer and performed elsewhere were reviewed by pathologists from our institution to confirm the diagnosis of cancer and Gleason score. Ten pathologists were involved in the formulation of Gleason scores and positivity of the prostate cores, and the current study is a retrospective analysis of these cores. All prostatectomy specimens were reviewed at our combined genitourinary pathology conference, which included a single senior genitourinary pathologist.11 Furthermore, all positive or questionable biopsies were confirmed by 1 of the 2 senior pathologists. Prostate biopsies performed at our institution were done with a 6.5 MHz transrectal probe (General Electric, Milwaukee, Wisconsin). Needle biopsies were obtained using a Pro-Mag 2.2 automatic biopsy instrument (Manan Medical Products, Inc., Northbrook, Illinois) with an 18 gauge ⫻ 25 cm cutting biopsy needle.11 Between 3 and 16 cores (median 7) were sampled. The percent of cores positive for cancer was calculated by the number of cores with cancer divided by the total number of cores multiplied by 100. Each prostate biopsy done at our institution was individually embedded and labeled for maximal surface area analysis.10 RRP specimens were submitted for microscopic examination, including the apex and base margins, 1 section from each seminal vesicle and 3 sections from each lobe (apical, mid and base). If cancer was not found in the sections, additional tissue was analyzed. The 1997 TNM staging system from the American Joint Committee of Cancer was used for clinical and pathological staging.12 Clinical staging consisted of digital rectal examination with T2a and T2b defined as any firmness or suspicious area of cancer in the corresponding prostatic anatomy. T2 was not necessarily intended to indicate a hard nodule. Clinical staging also included serum PSA measurement. When PSA was greater than 10 ng/ml, Gleason score was 7 or there was possible clinical T3 disease on rectal examination, abdominal and pelvic computerized tomography, and bone scan were also done. Pathological stage and margin status were combined as a single variable for analysis because combination of these variables are better predictors of recurrence than each alone.13 Positive or negative surgical margins were indicated by ⫹ and ⫺, respectively, following the staging assignment. This combined variable included organ confined disease with negative surgical margins (pT2⫺), organ confined with positive margins (pT2⫹), and extracapsular extension without (pT3a⫺) and with (pT3a⫹) positive surgical margins. In addition, the usual designations for seminal vesicle invasion (pT3b) and lymph node metastases (N1) were used. Followup. The first PSA value was routinely obtained 6 weeks after surgery. Patients were then followed with digital rectal examination and serum PSA every 6 months for 1 year and then once yearly thereafter, although PSA determinations were usually made twice yearly. Cancer recurrence was defined as 2 consecutive serum PSA values of 0.3 ng/ml or higher. Statistical analysis. Clinical, surgical and pathological factors are summarized as the median and IQR (25th, 75th percentiles) or as the number and percent of patients. We used univariate and multivariate Cox proportional hazards regression modeling to assess the prognostic significance of
clinical and pathological factors on time to cancer recurrence. The HR, 95% CI and 2-sided p values are reported. For modeling continuous variables were categorized into a natural categorization if one existed, or if not, by quartiles or at the median. Multivariate analysis used forward stepwise regression to identify a best model using p ⬍0.01 as a criterion to remain in the model. The model was limited to 5 variables at most because there were 53 events. The 5 and 10-year recurrence-free probabilities were estimated by exponentiation of the negative cumulative hazard estimated from the model and 95% CIs were calculated from the log of the negative log of the estimated survival function. The additional explanatory value of PSA as 2 indicator variables of PSA 4 to 10 ng/ml or greater than 10 ng/ml at the termination of model selection was assessed with a 2 df chi-square statistic. For statistical analysis SAS, version 8.0 (SAS Institute, Inc., Cary, North Carolina) software was used. RESULTS
Patient characteristics. Table 1 lists preoperative clinical characteristics of the cohort. Median patients age was 61 years (range 35 to 77) and median pre-RRP PSA was 5.85 TABLE 1. Preoperative clinical and biopsy characteristics of 476 patient cohort Variable Median age at diagnosis (IQR) Median ng/ml pre-RRP PSA (IQR) Median No. cores (IQR) Median No. Ca pos cores (IQR) Median % Ca pos cores (IQR) No. clinical stage (%): T1a T1b T1c T2a T2b No. Gleason score (%): 3–5 6 7 8 9
61 (55, 65) 5.85 (4.17, 9.05) 7 (6, 10) 2 (1, 3) 28 (16, 50) 1 (0.2) 4 (0.8) 140 (29.4) 224 (47.1) 107 (22.5) 10 (2.1) 342 (71.9) 92 (19.3) 22 (4.6) 10 (2.1)
TABLE 2. Surgical and postoperative pathological characteristics of 476 patient cohort Variable No. nerve sparing surgery (%): Unilat Bilat Median gm prostate wt (IQR)* No. pathological stage (%): T2a T2b T3a T3b No. Gleason score (%): 5–6 7 (3⫹4) 7 (4⫹3) 8–9 No. pos margins (%): Apical Posterior Bladder Lat No. seminal vesicle involvement (%) No. lymph node metastases (%) No. pathological stage/margin status (%): pT2⫺ pT2⫹ pT3a⫺ pT3a⫹ pT3b N1 (lymph node pos) * Total of 424 patients.
64 (13.4) 111 (23.3) 56 (45–69) 156 (32.8) 279 (58.6) 24 (5.0) 17 (3.6) 311 101 33 31 94 68 7 5 25 17 5
(65.3) (21.2) (6.9) (6.5) (19.7) (14.3) (1.5) (1.1) (5.3) (3.6) (1.1)
366 (76.9) 68 (14.3) 5 (1.1) 19 (4.0) 13 (2.7) 5 (1.1)
1494
PERCENT OF CORES WITH CANCER PREDICTS PROSTATE CANCER RECURRENCE
ng/ml (range 0.4 to 63.0). A median of 7 biopsy cores (range 3 to 16) were obtained with a median of 2 cores (28%) positive for cancer. Of the patients 72% had a Gleason score of 6 on needle biopsy. Table 2 lists surgical and pathological characteristics of the cohort. More than 90% of cases were stage T2 and 303 (64%) had a Gleason score of 6. Of the patients 20% had 1 or more positive margins. Therefore, 371 patients (almost 80%) had specimen confined disease, ie pT2 or pT3. Only 5 patients (1.1%) had lymph node metastases. Disease recurrence and survival. The cohort of 476 patients was followed a median of 49 months. Only 53 patients (11.1%) experienced disease recurrence following RRP. The estimated cancer nonprogression probabilities were 86% (95% CI 83 to 90) and 76% (95% CI 68 to 86) at 5 and 10 years, respectively (fig. 1). Fifteen patients have died, of whom only 2 (0.4%) died of prostate cancer. Univariate analysis to predict disease recurrence. Various clinical and pathological factors were evaluated on univariate analysis as independent predictors of disease recurrence (table 3). Preoperative clinical parameters that were associated with an increased hazard of disease recurrence were elevated PSA, the number or percent of biopsy cores positive for cancer above the median and higher Gleason score (each p ⬍0.01). Nerve sparing RRP was marginally associated with an increased hazard of recurrence (p ⫽ 0.05). Combined pathological stage and margin status was statistically significant (p ⬍0.0001). As expected, categories T2⫹, T3a⫹, T3b and N1 had significantly increased hazards of recurrence compared with specimen confined disease (pT2⫺ and pT3a⫺). Multivariate analysis to predict disease recurrence. We performed 2 sets of multivariate analyses, the first involving
only preoperative variables and the second involving combination preoperative, surgical and postoperative variables. 1) Analysis of preoperative clinical variables identified 2 parameters that were statistically significant at the ␣ ⫽ 0.01 level, namely biopsy Gleason score and percent of biopsy cores positive for cancer (table 4). Elevated total PSA was not statistically significant at the ␣ ⫽ 0.01 level and, therefore, it did not remain in the model (chi-square 6.3 (2 df), p ⫽ 0.043). Compared with patients who had a biopsy Gleason score of less than 7, patients with a Gleason score of 7 and 8 –9 had significantly increased hazards of recurrence (HR 3.92 and 12.24, respectively, p ⬍0.0001). Patients with the percent of positive biopsy cores in the upper half of the distribution (median greater than 28% positive cores, ie 2 or more positive cores in a 6 core biopsy, or 4 or more in a 12 core biopsy) had a significantly increased hazard of recurrence compared with those in the lower half of the distribution (median of 28% or less positive cores, ie 1 positive core in a 6 core biopsy, or 3 or fewer in a 12 core biopsy) (HR 3.62, p ⬍0.001). 2) When surgical and postoperative variables were added to the preoperative variables in multivariate analysis, 3 parameters remained significant at the ␣ ⫽ 0.01 level, namely the percent of biopsy cores positive for cancer, RRP Gleason score and the pathological stage/margin status combined variable (table 4). Patients with the percent of biopsy cores in the upper half of the distribution (greater than 28% positive) had a significantly increased hazard of recurrence compared with those in the lower half of the distribution (28% or less positive) (HR 3.86, p ⬍0.001). Compared with patients who had a Gleason score of 5, 6 or 7 (3⫹4), patients with a Gleason score of 7 (4⫹3) and 8 –9 had a significantly increased hazard of recurrence (HR 4.27 and 8.35, respectively, p ⬍0.0001).
FIG. 1. Kaplan-Meier estimate of time to disease recurrence in 476 men with clinically localized prostate cancer who underwent anatomical RRP showing 5 and 10-year recurrence-free probabilities and number at risk.
1495
PERCENT OF CORES WITH CANCER PREDICTS PROSTATE CANCER RECURRENCE
TABLE 3. Univariate analysis of preoperative, surgical and postoperative pathological candidate predictors of disease recurrence Variables
HR (95% CI)
Age at diagnosis vs younger than 60: 60–69 70 or Older Preop ng/ml PSA vs less than 4: 4–10 Greater than 10 Pos cores greater than median vs median or less: 3 or Greater vs 1 or 2 Greater than 28% vs 28% or less Clinical stage vs T1: T2a T2b Gleason score vs 6 or less: 7 8–9 Nerve sparing surgery yes vs no Prostate wt gm quartiles vs quartile 1: Quartile 2 Quartile 3 Quartile 4 Pathological stage vs T2: T3a T3b Gleason score vs 6 or less: 7 (3⫹4) 7 (4⫹3) 8–9 Pos margins yes vs no Seminal vesicle involvement yes vs no Lymph node metastases yes vs no Pathological stage/pos margins vs specimen confined pT2⫺ ⫹ pT3a⫺: pT2⫹ pT3a⫹ pT3b N1 (lymph node pos)
p Value (log rank test) 0.15
0.91 (0.50, 1.65) 1.95 (0.84, 4.51)
⬍0.0001
8.46 (1.15, 62.44) 26.10 (3.55, 192.24) 5.50 (2.92, 10.36) 4.78 (2.32, 9.84) 1.25 (0.63, 2.48) 1.55 (0.72, 3.31) 5.36 (2.83, 10.16) 13.94 (6.76, 28.76) 0.54 (0.29, 1.01) 1.19 (0.56, 2.55) 1.21 (0.56, 2.63) 0.51 (0.20, 1.30) 6.21 (3.05, 12.63) 15.84 (7.62, 32.96) 2.49 11.45 17.11 4.50 12.62 34.25
(1.10, 5.64) (5.35, 24.51) (7.99, 36.65) (2.62, 7.75) (6.19, 25.72) (9.94, 117.97)
⬍0.0001 ⬍0.0001 0.53 ⬍0.0001 0.05 0.29
⬍0.0001 ⬍0.0001
⬍0.0001 ⬍0.0001 ⬍0.0001 ⬍0.0001
3.67 (1.83, 7.36) 13.06 (6.03, 28.28) 18.09 (7.44, 43.95) 79.15 (21.68, 288.99)
TABLE 4. Multivariate analysis of candidate predictors for disease recurrence Variables
HR (95% CI)
p Value*
Preop clinical ⫹ biopsy† % Pos biopsy cores: Lower distribution half (28 or less) Upper distribution half (greater than 28) Biopsy Gleason score: 3–6 7 8–9
Reference 3.62 (1.74, 7.54) Reference 3.92 (2.04, 7.53) 12.24 (5.90, 25.40) Preop, surgical ⫹ postop pathological†
— ⬍0.001 — ⬍0.0001 ⬍0.0001
% Pos biopsy cores: Lower distribution half (28 or less) Reference — Upper distribution half (greater than 28) 3.86 (1.85, 8.04) ⬍0.001 RRP Gleason score: 5–7 (3⫹4) Reference — 7 (4⫹3) 4.27 (2.13, 8.55) ⬍0.0001 8–9 8.35 (4.15, 16.78) ⬍0.0001 Pathological stage/margin status: Specimen confined pT2⫺ ⫹ pT3a⫺ Reference — pT2⫹ 2.39 (1.17, 4.87) 0.017 pT3a⫹, pT3b ⫹ N1 (lymph node pos) 7.94 (3.94, 15.99) ⬍0.0001 * Final model was re-parameterized for presentation, so that each variable represents increased recurrence hazard relative to its reference category, which is the category with the better prognosis, and because of this re-parameterization, p values are not strictly p ⬍0.01. † PSA did not provide additional explanatory value to final model since 2 indicator variables PSA 4 to 10 and greater than 10 ng/ml did not achieve statistical significance at ␣ ⫽ 0.01.
For example, of patients with a Gleason score of less than 7 or 3⫹4 only 5.6% had recurrence (12 of 311 with less than 7 and 11 of 101 with 3⫹4), while 45% (15 of 33) with a Gleason score of 4⫹3 and 48% (15 of 31) with greater than 7 had recurrence. Also, patients with positive margins without evidence of extracapsular extension (pT2⫹) had an increased hazard of recurrence compared with patients with specimen confined disease (pT2⫺ or pT3a⫺) (HR 2.39, p ⫽ 0.017). As expected, patients with nonorgan confined disease/positive surgical margins (pT3a⫹), seminal vesicle invasion (pT3b) or lymph node metastasis (N1) had a significantly increased hazard of recurrence (HR 7.94, p ⬍0.0001). Figures 2 to 4 show univariate KaplanMeier curves for these 3 factors.
Table 5 lists 5 and 10-year recurrence-free probabilities with the 95% CI estimated from the model in a nomogramtype format. The reference group with a low RRP Gleason score, specimen confined disease and low percent of positive biopsy cores were estimated to have 5 and 10-year recurrence-free probabilities of above 0.98 (95% CI 0.96, 0.99) and 0.96 (95% CI 0.90, 0.98), respectively. For example, if the patient had a high percent of positive biopsy cores (greater than 28%) and the other 2 factors were held constant, the estimated 5-year recurrence-free probability decreased from 0.98 to 0.93. In the worst case scenario a patient with an RRP Gleason score of 8 –9, nonorgan, nonspecimen confined disease and a high percent of positive biopsy cores was esti-
1496
PERCENT OF CORES WITH CANCER PREDICTS PROSTATE CANCER RECURRENCE
FIG. 2. Kaplan-Meier estimate of time to disease recurrence in 476 men who underwent RRP by whether percent of positive biopsy cores was above (greater than 28%) or below median.
mated to have a 5-year recurrence-free probability of 0.01 (95% CI 0.00, 0.08). DISCUSSION
We defined multiple preoperative and postoperative parameters that are candidate clinical guidelines to help determine the risks of disease recurrence after RRP. We performed 2 multivariate analyses. The first analysis involved preoperative parameters determined to be important for defining the risk of recurrence. Only the percent of positive biopsy cores and Gleason score were found to be significant at the ␣ ⫽ 0.01 level, while PSA was not (table 4). The second multivariate analysis combined preoperative and postoperative parameters thought to be important for predicting recurrence. The results indicated that the percent of cores positive for cancer, RRP Gleason score and pathological stage/margin status were the best independent predictors of recurrence. Again, preoperative PSA did not achieve statistical significance. The major contribution of this study is that it shows that the percent of cores positive for cancer in needle biopsies are a better predictor of biochemical recurrence after RRP than PSA. Likewise, other series have found a role for the percent of positive cores to help define tumor stage and volume,14 and outcome after RRP.6, 7 D’Amico et al found that the percent of positive cores was predictive of outcome only when preoperative parameters were included on multivariate analysis.6 Finally, Freedland et al reported that the percent of biopsy tissue with cancer, rather than the percent of cores with cancer, was predictive of biochemical failure after RRP.7 In contrast, our study shows that the percent of positive cores,
when combined with postoperative parameters, is highly predictive of relapse (tables 4 and 5). We believe that the value of the percent of positive cores as a predictor of cancer recurrence is related to at least 3 factors. 1) It has been demonstrated that the percent of cores positive for cancer, as determined by the equation, [(number of cores positive for cancer/ total number of cores) ⫻ 100] is the expression of real tumor volume.14 2) the extended needle biopsy technique improves cancer detection15 and cancer grading accuracy11 3) According to a previous study, a higher rate of cancer can be detected when cores are embedded individually,16 a technique that we have used for RRP specimen analysis.10 In our statistical analysis of biopsy results we implemented a priori a standard statistical approach and created groups by distribution quartiles. It was done because there are no generally accepted cutoffs for grouping subjects according to positive biopsy cores, as there are for PSA, for example. It led to a cutoff above the median value of 28% positive cores. In light of our results a careful examination and determination of an optimal cutoff or cutoffs for determining risk according to positive biopsy cores should be done. Such an analysis, for example creating groups by distribution deciles, could not be performed in the current study because the small number of recurrences in the cohort limited the proportional hazards model to 5 covariates at most. Longer followup with more data points would answer these concerns. Why was preoperative PSA not an important preoperative parameter for cancer recurrence in contrast to the percent of biopsy cores positive for cancer? 1) Multivariate analyses showed the most important parameters in determining re-
PERCENT OF CORES WITH CANCER PREDICTS PROSTATE CANCER RECURRENCE
1497
FIG. 3. Kaplan-Meier estimate of time to disease recurrence in 476 men who underwent RRP by surgical Gleason score
currence but did not exclude the importance of preoperative PSA. If PSA were added to the final model of preoperative variables, PSA would add explanatory value (p ⫽ 0.043), although it did not meet the modeling criteria. However, if PSA were added to the final model including preoperative and postoperative variables, it would not add explanatory value (p ⫽ 0.12). 2) This cohort of patients was only from the post-PSA era, which means that it was a homogeneous group with few patients with PSA greater than 20 ng/ml and 80% with PSA less than 10 ng/ml. Therefore, because of the homogeneity of PSA in most of the cohort, PSA should not be expected to be as good a determinant of recurrence compared with other parameters. As a consequence of a better selection of patients for RRP in the post-PSA era, only 2 patients died of prostate cancer in this series. The independent predictive role of RRP specimen Gleason score in the outcome of patients after RRP is not a new finding.9, 17–19 On univariate analysis the Gleason score groups 4⫹3 and greater than 7 were at significantly higher risk for recurrence than the Gleason score 5– 6 group (table 3 and fig. 3). However, the risk of recurrence in the Gleason score 3⫹4 group was only marginally increased. On multivariate analysis it was no longer important to distinguish the Gleason 3⫹4 group from the Gleason score 5– 6 group (table 4). Therefore, according to our model and data we can see a markedly different outcome in patients with primary Gleason grade 3 and primary Gleason grade 4, as represented by the Gleason scores 3⫹4 and 4⫹3 (table 5), similar to the findings of others.20
The significance of pathological stages and positive surgical margins as independent predictors of recurrence is well known and has been described in several publications.9, 17–19 In this study we wished to combine the 2 variables to differentiate organ confined, nonorgan confined, specimen confined, and nonspecimen confined disease, similar to Paulson.13 We found that patients with specimen confined cancer (T2⫺ and T3⫺) had better outcomes than patients with organ confined cancer with positive margins (T2⫹) (table 5 and fig. 4). Patients with T3a⫺ disease had better than expected results (fig. 4), possibly because of the low number of patients in this category. According to our data surgical margins are important for the patient prognosis after RRP, even in patients with organ confined disease. It is especially true in patients who have Gleason scores 4⫹3 and greater than 7 with a high percent of positive cores (table 5 and fig. 4). In contrast, none of the 196 patients who were low risk in all 3 parameters (ie Gleason score 5– 6 or 3⫹4, T2⫺, T3⫺ and 28% or less of positive cores for cancer) have had recurrence to date after RRP.
CONCLUSIONS
The percent of cores positive for cancer was a better predictor of cancer recurrence than PSA in this post-PSA era RRP series. In addition, surgical Gleason score and pathological stage/surgical margins were also independent predictors of cancer recurrence after RRP. These 3 predictors are displayed in a nomogram-type format to summarize esti-
1498
PERCENT OF CORES WITH CANCER PREDICTS PROSTATE CANCER RECURRENCE
FIG. 4. Kaplan-Meier estimate of time to disease recurrence in 476 men who underwent RRP by pathological stage/margin status category TABLE 5. Nomogram-type summary of estimated 5 and 10-year recurrence-free probability No. Pts
No. Recurrences
196 143 20 33 10 10 7 10 1 7 1 7 8 7 4 3 1 8
0 9 0 5 3 6 0 5 0 4 1 5 3 2 2 3 0 5
RRP Gleason Score 5, 5, 5, 5, 5, 5,
6, 7 (3⫹4) 6, 7 (3⫹4) 6, 7 (3⫹4) 6, 7 (3⫹4) 6, 7 (3⫹4) 6, 7 (3⫹4) 7 (4⫹3) 7 (4⫹3) 7 (4⫹3) 7 (4⫹3) 7 (4⫹3) 7 (4⫹3) 8, 9 8, 9 8, 9 8, 9 8, 9 8, 9
Pathological Stage/ Margin Status pT2⫺, pT3a⫺ pT2⫺, pT3a⫺ pT2⫹ pT2⫹ pT3a⫹, pT3b, pT3a⫹, pT3b, pT2⫺, pT3a⫺ pT2⫺, pT3a⫺ pT2⫹ pT2⫹ pT3a⫹, pT3b, pT3a⫹, pT3b, pT2⫺, pT3a⫺ pT2⫺, pT3a⫺ pT2⫹ pT2⫹ pT3a⫹, pT3b, pT3a⫹, pT3b,
N1 N1
N1 N1
N1 N1
mated 5 and 10-year recurrence-free probabilities. Longer followup with more data points and verification of these data would allow to use this nomogram in general clinical prac-
% Pos Biopsy Cores 28 or Less Greater than 28 or Less Greater than 28 or Less Greater than 28 or Less Greater than 28 or Less Greater than 28 or Less Greater than 28 or Less Greater than 28 or Less Greater than 28 or Less Greater than
28 28 28 28 28 28 28 28 28
Estimated Recurrence-Free Probability (95% CI) 5-Yr
10-Yr
0.98 (0.96, 0.99) 0.93 (0.88, 0.96) 0.96 (0.89, 0.98) 0.84 (0.71, 0.91) 0.86 (0.69, 0.94) 0.55 (0.31, 0.74) 0.92 (0.81, 0.97) 0.73 (0.52, 0.86) 0.82 (0.60, 0.93) 0.47 (0.22, 0.68) 0.52 (0.19, 0.77) 0.08 (0.01, 0.26) 0.85 (0.68, 0.94) 0.54 (0.27, 0.74) 0.68 (0.37, 0.86) 0.23 (0.04, 0.50) 0.28 (0.05, 0.58) 0.01 (0.00, 0.08)
0.96 (0.90, 0.98) 0.84 (0.72, 0.91) 0.90 (0.75, 0.96) 0.66 (0.40, 0.83) 0.70 (0.37, 0.88) 0.25 (0.04, 0.55) 0.83 (0.58, 0.93) 0.48 (0.19, 0.72) 0.63 (0.26, 0.86) 0.17 (0.02, 0.47) 0.22 (0.01, 0.60) ⬍0.01 (0.00, 0.08) 0.69 (0.38, 0.86) 0.24 (0.04, 0.53) 0.41 (0.08, 0.73) 0.03 (0.00, 0.25) 0.05 (0.00, 0.33) ⬍0.01 (0.00, 0.01)
tice. The low rates of cancer recurrence in our series may have been primarily related to the stage migration of prostate cancer associated with PSA screening.
PERCENT OF CORES WITH CANCER PREDICTS PROSTATE CANCER RECURRENCE
Drs. Glenn Bubley and Yoo-Joung Ko provided a critical reading of the manuscript. 10. REFERENCES
1. Stamey, T. A., Yang, N., Hay, A. R., McNeal, J. E., Freiha, F. S. and Redwine, E.: Prostate-specific antigen as a serum marker for adenocarcinoma of the prostate. N Engl J Med, 317: 909, 1987 2. Catalona, W. J., Smith, D. S., Ratliff, T. L. and Basler, J. W.: Detection of organ-confined prostate cancer is increased through prostate-specific antigen-based screening. JAMA, 270: 948, 1993 3. Humphrey, P. A., Keetch, D. W., Smith, D. S., Shepherd, D. L. and Catalona, W. J.: Prospective characterization of pathological features of prostatic carcinomas detected via serum prostate specific antigen based screening. J Urol, 155: 816, 1996 4. Kattan, M. W., Eastham, J. A., Stapleton, A. M., Wheeler, T. M. and Scardino, P. T.: A preoperative nomogram for disease recurrence following radical prostatectomy for prostate cancer. J Natl Cancer Inst, 90: 766, 1998 5. Partin, A. W., Kattan, M. W., Subong, E. N., Walsh, P. C., Wojno, K. J., Oesterling, J. E. et al: Combination of prostate-specific antigen, clinical stage, and Gleason score to predict pathological stage of localized prostate cancer. A multi-institutional update. JAMA, 277: 1445, 1997 6. D’Amico, A. V., Whittington, R., Malkowicz, S. B., Schultz, D., Fondurulia, J., Chen, M. H. et al: Clinical utility of the percentage of positive prostate biopsies in defining biochemical outcome after radical prostatectomy for patients with clinically localized prostate cancer. J Clin Oncol, 18: 1164, 2000 7. Freedland, S. J., Csathy, G. S., Dorey, F. and Aronson, W. J.: Percent prostate needle biopsy tissue with cancer is more predictive of biochemical failure or adverse pathology after radical prostatectomy than prostate specific antigen or Gleason score. J Urol, 167: 516, 2002 8. Kattan, M. W., Wheeler, T. M. and Scardino, P. T.: Postoperative nomogram for disease recurrence after radical prostatectomy for prostate cancer. J Clin Oncol, 17: 1499, 1999 9. Hull, G. W., Rabbani, F., Abbas, F., Wheeler, T. M., Kattan, M. W. and Scardino, P. T.: Cancer control with radical pros-
11.
12.
13. 14.
15. 16.
17.
18.
19. 20.
1499
tatectomy alone in 1,000 consecutive patients. J Urol, 167: 528, 2002 Kao, J., Upton, M., Zhang, P. and Rosen, S.: Individual prostate biopsy core embedding facilitates maximal tissue representation. J Urol, 168: 496, 2002 San Francisco, I. F., DeWolf, W. C., Rosen, S., Upton, M. and Olumi, A. F.: Extended prostate needle biopsy improves concordance of Gleason grading between prostate needle biopsy and radical prostatectomy. J Urol, 169: 136, 2003 Fleming, I. D., Cooper, J. S., Hensen, D. E., Hutter, R. V. P., Kennedy, B. J., Murphy, G. P. et al: Prostate. In: AJCC Cancer Staging Manual, 5th ed. Philadelphia: Lippincott-Raven Publishers, vol. 34, pp. 219 –224, 1997 Paulson, D. F.: Impact of radical prostatectomy in the management of clinically localized disease. J Urol, 152: 1826, 1994 Sebo, T. J., Bock, B. J., Cheville, J. C., Lohse, C., Wollan, P. and Zincke, H.: The percent of cores positive for cancer in prostate needle biopsy specimens is strongly predictive of tumor stage and volume at radical prostatectomy. J Urol, 163: 174, 2000 Eskew, L., Bare, R. L. and McCullough, D. L.: Systematic 5 region prostate biopsy is superior to sextant method for diagnosing carcinoma of the prostate. J Urol, 157: 199, 1997 Lane, R. B., Jr., Lane, C. G., Mangold, K. A., Johnson, M. H. and Allsbrook, W. C., Jr.: Needle biopsies of the prostate: what constitutes adequate histologic sampling? Arch Pathol Lab Med, 122: 833, 1998 Zincke, H., Oesterling, J. E., Blute, M. L., Bergstralh, E. J., Myers, R. P. and Barrett, D. M.: Long-term (15 years) results after radical prostatectomy for clinically localized (stage T2c or lower) prostate cancer. J Urol, 152: 1850, 1994 Pound, C. R., Partin, A. W., Epstein, J. I. and Walsh, P. C.: Prostate-specific antigen after anatomic radical retropubic prostatectomy. Patterns of recurrence and cancer control. Urol Clin North Am, 24: 395, 1997 Catalona, W. J. and Smith, D. S.: Cancer recurrence and survival rates after anatomic radical retropubic prostatectomy for prostate cancer: intermediate-term results. J Urol, 160: 2428, 1998 Stamey, T. A., McNeal, J. E., Yemoto, C. M., Sigal, B. M. and Johnstone, I. M.: Biological determinants of cancer progression in men with prostate cancer. JAMA, 281: 1395, 1999