The percentage of prostate needle biopsy cores with carcinoma from the more involved side of the biopsy as a predictor of prostate specific antigen recurrence after radical prostatectomy

Results from the Shared Equal Access Regional Cancer Hospital (SEARCH) database

Abstract

BACKGROUND

The authors previously found that, although the total percentage of prostate needle biopsy cores with carcinoma was a significant predictor of prostate specific antigen (PSA) failure among men undergoing radical prostatectomy (RP), there was a trend toward a lower risk of recurrence in patients with positive bilateral biopsies, suggesting that high-volume, unilateral disease was a worse predictor of outcome than an equivalent number of positive cores distributed over two lobes. In the current study, the authors sought to compare the total percentage of cores with carcinoma directly with the percentage of cores from the more involved or dominant side of the prostate with carcinoma for their ability to predict outcome among men who underwent RP.

METHODS

A retrospective survey of 535 patients from the Shared Equal Access Regional Cancer Hospital database who underwent RP at 4 different equal-access medical centers between 1988 and 2002 was undertaken. The total percentage of cores positive was compared with the percentage of cores positive from the dominant and nondominant sides for their ability to predict biochemical recurrence after RP. The best predictor then was compared with the standard clinical variables PSA, biopsy Gleason score, and clinical stage in terms of ability to predict time to PSA recurrence after RP using multivariate analysis.

RESULTS

The adverse pathologic features of positive surgical margins and extracapsular extension were significantly more likely to be ipsilateral to the dominant side on the prostate biopsy. The percentage of cores positive from the dominant side provided slightly better prediction (concordance index [C] = 0.636) for PSA failure than the total percentage of cores positive (C = 0.596) and markedly better than the percentage of cores from the nondominant side (C = 0.509). Cutoff points for percentage of cores positive from the dominant side were identified (< 34%, 34–67%, and > 67%) that provided significant risk stratification for PSA failure (P < 0.001). On multivariate analysis, the percentage of cores positive from the dominant side was the strongest independent predictor of PSA recurrence (P < 0.001). Biopsy Gleason score (P = 0.017) also was a significant, independent predictor of recurrence. There was a trend, which did not reach statistical significance, toward an association between greater PSA values and biochemical failure (P = 0.052). Combining the PSA level, biopsy Gleason score, and percentage of cores positive from the dominant side of the prostate resulted in a model that provided a high degree of prediction for PSA failure (C = 0.671).

CONCLUSIONS

The percentage of cores positive from the dominant side of the prostate was a slightly better predictor of PSA recurrence than was the total percentage of cores positive. Using the percentage of cores from the dominant side along with the PSA level and the biopsy Gleason score provided significant risk stratification for PSA failure. Cancer 2003. Published 2003 by the American Cancer Society.

Prostate needle biopsy (PNBx) tumor volume is an important predictor of outcome for men undergoing treatment for prostate carcinoma.1–4 This is often and most easily quantified by determining the percentage of biopsy cores with carcinoma. We recently found that among 1094 men who underwent radical prostatectomy (RP), the percentage of positive biopsy cores was the strongest independent predictor of adverse pathologic findings and prostate specific antigen (PSA) recurrence.5 It is noteworthy that, in multivariate analysis, after controlling for the total percentage of cores positive, there was a trend toward a lower risk of PSA recurrence for patients who had bilateral positive biopsies, suggesting that high-volume unilateral disease was a worse predictor of outcome than was an equivalent number of positive cores distributed over two lobes. We hypothesized that the percentage of cores positive from the dominant side may be a better surrogate for tumor volume and thus may be correlated more strongly with disease-free survival. However, multivariate analysis has limitations for comparing two strongly correlated variables, such as the percentage of total cores positive and the percentage of cores positive from the dominant side. Thus, in the current study, we used alternate statistical approaches to compare these two variables. Because both biopsy tumor volume measurements are equally easy to calculate, whichever variable provided the best risk assessment would justify its use relative to the other, even if the differences were not statistically significant.

We compared the total percentage of cores positive and the percentage of cores positive from the dominant side and the nondominant side using the Shared Equal Access Regional Cancer Hospital (SEARCH) database of patients who underwent RP.6 The best outcome predictor then was compared with the standard clinical variables clinical stage, biopsy Gleason score, and PSA level to develop a multivariate model for predicting PSA failure. Based on the best predictors in this multivariate model, we assessed the 2-year PSA recurrence risk after RP.

MATERIALS AND METHODS

Patients

After obtaining Institutional Review Board approval from each institution, data from consecutive patients who underwent RP at the West Los Angeles, Palo Alto, and San Francisco Veterans Administration Medical Centers and the San Diego Naval Medical Center between 1988 and 2002 were combined into the SEARCH database. Patients who had received preoperative androgen-deprivation therapy or radiation therapy were excluded. Patients without data regarding the number of cores or the number of cores positive (n = 384), patients who had < 6 cores obtained (n = 215), and patients who had data missing on side-specific percentage of cores (n = 564) were excluded. Patients with clinical stage T1a/T1b disease also were excluded (n = 5). This resulted in a study population of 535 patients, who were categorized as white (n = 287), black (n = 140), Hispanic (n = 45), Asian (n = 21), or other/unknown race/ethnicity (n = 42).

All PNBx samples were obtained with ultrasound guidance. Most patients (52%) underwent sextant biopsies; 8% of patients had 8 cores obtained, 14% of patients had 10 cores obtained, and 9% of patients had 12 cores obtained. The remaining 17% of patients had between 7 and 18 cores obtained. The percentage of cores positive was calculated by dividing the number of positive cores by the total number of cores. The side (left or right) with the greatest percentage of cores positive was considered the dominant side, and the contralateral side was considered the nondominant side.

The prostatectomy specimens were sectioned in accordance with each institution's protocol.7–10 Recurrence was defined as a single PSA level > 0.2 ng/mL or 2 values at 0.2 ng/mL. Forty-two patients had no follow-up data. These patients were included to evaluate the predictors of adverse pathologic characteristics, but not of biochemical recurrence.

Statistics

Comparisons between the dominant side on biopsy and the side of adverse pathologic features (positive surgical margins, extracapsular extension [ECE], or seminal vesicle invasion) were made using a chi-square test. The biopsy tumor burden measurements were compared further for their ability to predict time to PSA failure using the concordance index,11 which is the probability that given two randomly selected patients, the patient who developed a recurrence first was, in fact, the one predicted to do so. Cutoff points for the percentage of cores positive from the dominant side were identified by combining patients with similar PSA recurrence risks as determined by log-rank analysis. Cutoff points for the total percentage of cores positive were identified previously using the same patient cohort.5 Multivariate analysis of predictors of time to biochemical recurrence was performed using a forward, stepwise Cox proportional hazards model, with P < 0.15 indicating that a variable should be entered into the model at each step. If the generated model contained variables with P > 0.10, then the variable with the highest P value was deleted successively until only variables with P < 0.10 remained. The biopsy Gleason score (< 7, 7, and > 7), the PSA level (< 10 ng/mL, 10–20 ng/mL, and > 20 ng/mL), and the clinical stage (T1c/T2a vs. T2b/T2c vs. T3) were examined as categoric variables using previously defined cutoff points.12, 13 Using the coefficients of the multivariate Cox model along with a bootstrapping technique with 1000 replicates to determine the 95% confidence interval, the 2-year PSA recurrence risk was determined. All clinical variables (age, clinical stage, PSA level, and biopsy Gleason score) and pathologic variables (surgical Gleason score, pathologic stage, ECE, surgical margin status, seminal vesicle invasion, and lymph node involvement) were similar among the centers. Therefore, data from all centers were combined for analysis. All statistical analyses were performed using STATA software (Version 7.0; Stata Corp., College Station, TX).

RESULTS

Table 1 demonstrates the clinical and pathologic characteristics of the study population. If a patient had a positive surgical margin or ECE, then it was significantly more likely to be ipsilateral to the dominant side of the biopsy (Table 2). Among patients with unilateral positive surgical margins, an even stronger correlation with the side of dominant biopsy was found, with 69% of positive margins located ipsilateral to the dominant side of the biopsy. Similarly, among patients with unilateral were positive ECE, 77% ipsilateral to the dominant side of the biopsy. There was no significant correlation between the side of seminal vesicle invasion and the side of dominance on biopsy, although only 10% of patients had seminal vesicle invasion.

Table 1. Preoperative Clinical and Pathologic Characteristics of Men Undergoing Radical Prostatectomy

Characteristic	Value
SD: standard deviation.
No. of patients	535
Mean age ± SD (yrs)	62.8 ± 6.8
PSA (ng/mL)
Mean ± SD	10.6 ± 11.2
Median	7.8
Biopsy Gleason score: no. of patients (%)
2–4	39 (7)
5–6	307 (58)
7	128 (24)
8–10	59 (11)
Mean score ± SD	6.2 ± 1.1
Median score	6.0
Percentage of cores positive total
Mean ± SD	35 ± 22
Median	33
Percentage of cores positive from the more involved side
Mean ± SD	56 ± 28
Median	50
Percentage of cores positive from the less involved side
Mean ± SD	15 ± 25
Median	0
Clinical stage: no. of patients (%)
T1c/T2a	383 (74)
T2b/c	131 (25)
T3	3 (1)
Pathologic Gleason score: no. of patients (%)
2–4	14 (3)
5–6	254 (48)
7	193 (36)
8–10	69 (13)
Mean score ± SD	6.5 ± 1.1
Median score	6.0
Pathologic stage: no. of patients (%)
T2	361 (68)
T3	160 (30)
T4	7 (1)
Positive surgical margins: no. of patients (%)	166 (31)
Extracapsular extension: no. of patients (%)	143 (27)
Seminal vesicle invasion: no. of patients (%)	52 (10)
Positive lymph nodes: no. of patients (%)	10 (2)

Table 2. Comparison of the Dominant Side on Biopsy with the Side of Adverse Pathologic Findings after Radical Prostatectomy

Adverse finding	Dominant side on biopsy (%)		P value (chi-square test)
	Left	Right
Incidence of positive surgical margins
Left	34 (25)	25 (13)	0.022
Right	22 (16)	38 (21)	—
Incidence of extracapsular extension
Left	30 (22)	24 (13)	0.041
Right	26 (19)	44 (23)	—
Incidence of seminal vesicle invasion
Left	9 (6)	8 (4)	0.576
Right	6 (4)	8 (4)	—

With a median 27-month follow-up (mean, 34 months) among patients without a biochemical recurrence, 103 patients (21%) developed a biochemical recurrence. The total percentage of cores positive and the percentage of cores positive from the dominant side and the nondominant side were compared for their ability to predict biochemical recurrence. Using the concordance index, a slight improvement in risk assessment was identified for the percentage of cores from the dominant side (concordance index [C] = 0.636) relative to the total percentage of cores positive (C = 0.596), with the percentage of cores from the nondominant side having the lowest predictive ability (C = 0.509).

To examine whether the ability of these tumor volume measurements for predicting PSA failure was improved among men with extended biopsy schemes, data from men who underwent sextant biopsies and from men who had > 6 cores obtained (median, 10 cores obtained) were examined separately. Among men who underwent sextant biopsy, the percentage of cores positive from the dominant side was more predictive of PSA failure compared with the total percentage of cores positive (C = 0.670 vs. 0.612, respectively). Similarly, among men who had > 6 cores obtained, the percentage of cores positive from the dominant side was more predictive of PSA failure than was the total percentage of cores positive (C = 0.651 vs. 0.613, respectively). The ability of the percentage of cores positive from the nondominant side to predict PSA failure among men who underwent sextant biopsies (C = 0.535) or among men who had > 6 cores obtained (C = 0.492) was markedly lower compared with the total percentage or dominant-side percentage of cores positive.

To identify clinically relevant cutoff points for the percentage of cores positive from the dominant side, we separated patients into 6 groups based on the number of cores positive, assuming a double-sextant or 12-core biopsy, corresponding to 6 cores per side (Table 3). Thus, patients with < 17% positive cores from the dominant side (1 of 6 cores positive) were grouped together. Similarly, patients with 17–34% positive cores on the dominant side (2 of 6 cores positive) were grouped together. Groups with similar PSA recurrence risks were combined into a low-risk group (1 or 2 of 6 cores positive; < 34%), an intermediate-risk group (3 or 4 of 6 cores positive; 34–67%), and a high-risk group (> 4 of 6 cores positive; > 67%) that provided significant risk stratification for PSA recurrence (P < 0.001; log-rank test) (Fig. 1).

Table 3. Log-Rank Analysis of Cutoff Points for the Percentage of Positive Biopsy Cores to Predict Biochemical Recurrence after Radical Prostatectomy

Percentage of cores positive	Corresponding no. of cores positive on the more involved side assuming a 12-core biopsy	No. of patients	Observed events	Expected events	Ratio of observed to expected events
Initial categorization
< 17%	1	31	2	5.52	0.36
17–34%	2	174	20	37.59	0.53
> 34–50%	3	55	13	10.01	1.30
> 50–67%	4	118	24	26.51	0.91
> 67–84%	5	25	8	4.63	1.73
> 84	6	95	37	19.75	1.87
Consolidated categorization
< 34%	1–2	205	22	43.10	0.51
34–67%	3–4	173	37	36.52	1.01
> 67%	5–6	120	45	24.38	1.85

Figure 1. Six-year Kaplan–Meier estimates of prostate specific antigen (PSA) failure-free survival of patients treated with radical prostatectomy, stratified by the percentage of positive biopsy cores from the more involved side of the biopsy (pairwise P values: < 34% vs. 34–67%, P = 0.027; < 34% vs. > 67%, P < 0.001; 34–67% vs. > 67%, P = 0.007).

Cutoff points for the percentage of cores positive from the dominant side and the percentage of total cores positive were compared with cutoff points for the standard clinical variables PSA level, biopsy Gleason score, and clinical stage in terms of their ability to predict PSA failure (Table 4). On multivariate analysis, the percentage of cores positive from the dominant side was the strongest predictor of biochemical recurrence (P < 0.001). Biopsy Gleason score (P = 0.017) also was a significant independent predictor of recurrence. There was a trend, which did not reach statistical significance, toward an association between greater PSA values and biochemical failure (P = 0.052). Combining the PSA level, biopsy Gleason score, and percentage of cores positive from the dominant side resulted in a model that provided a high degree of prediction for PSA failure (C = 0.671). Using the coefficients of the multivariate Cox model along with a bootstrapping technique to determine the 95% confidence interval, the 2-year PSA recurrence risk stratified by PSA level, biopsy Gleason score, and percentage of cores positive from the dominant side was calculated and is presented in tabular form (Table 5).

Table 4. Cox Proportional Hazards Analysis of Factors that Predicted the Time to Biochemical Recurrence after Radical Prostatectomy

Factor	HR	95% CI	P value
HR: hazard ratio; 95% CI: 95% confidence interval; PSA: prostate specific antigen.
Univariate analysis
Percentage of cores from the more involved side	1.89	1.47–2.42	< 0.001
Total percentage of cores involved	1.66	1.32–2.09	< 0.001
Biopsy Gleason score	1.71	1.33–2.20	< 0.001
Serum PSA level	1.65	1.28–2.12	< 0.001
Clinical stage	1.60	1.09–2.36	0.016
Multivariate analysis
Percentage of cores from the more involved side	1.65	1.26–2.17	< 0.001
Biopsy Gleason score	1.40	1.06–1.84	0.017
Serum PSA level	1.31	1.00–1.72	0.052

Table 5. Two-Year Risk of Prostate Specific Antigen Recurrence (with 95% Confidence Interval) after Radical Prostatectomy

PSA level	Biopsy Gleason score
	2–6		7		8–10
	Risk	95% CI	Risk	95% CI	Risk	95% CI
HR: hazard ratio; 95% CI: 95% confidence interval; PSA: prostate specific antigen.
PSA < 10 ng/mL
Percentage of cores from dominant side
< 34%	7	4–10	9	6–12	13	5–21
34–67%	11	8–14	15	10–20	21	9–33
> 67%	17	8–26	23	10–36	31	11–51
PSA 10–20 ng/mL
Percentage of cores from dominant side
< 34%	9	7–11	12	8–16	17	7–27
34–67%	13	9–17	19	13–25	26	13–39
> 67%	21	11–31	28	15–41	38	17–59
PSA > 20 ng/mL
Percentage of cores from dominant side
< 34%	11	5–17	15	6–24	21	7–35
34–67%	17	9–25	23	12–34	32	14–50
> 67%	26	11–41	35	18–52	46	23–69

DISCUSSION

Although PSA levels and biopsy Gleason score are important outcome predictors among men undergoing therapy for prostate carcinoma, recent efforts have focused on incorporating PNBx tumor volume as an adjunct to improve pretreatment risk stratification.1–4 One of the most easily obtainable measurements of biopsy tumor volume is the total percentage of cores positive, which predicted outcome significantly after RP, as reported previously by both our group and others.1–4 However, we recently found that the percentage of cores positive from the dominant side of the prostate may be an even stronger predictor of recurrence.5 To evaluate this hypothesis further, we directly compared the total percentage of cores positive with the percentage of cores positive from the dominant side and found that the percentage of cores from the dominant side provided slightly better risk assessment than did the total percentage of cores positive. Because both measurements are equally easy to calculate, the slight improvement in risk assessment justifies the routine clinical use of the percentage of cores positive from the dominant side, rather than the total percentage of cores positive, to predict PSA failure. However, to maintain this important information regarding the dominant side, biopsy cores need to be submitted at least in separate bottles from the right and left sides.

The percentage of cores positive from the nondominant side was a poor predictor of advanced pathology and PSA recurrence. Indeed, the concordance index for predicting biochemical recurrence was only slightly better than random chance (C = 0.5). There is a possibility that this was due to positive biopsies on the nondominant side caused by secondary, smaller tumors, which do not contribute to adverse pathology or PSA failures. In turn, this suggests that adverse pathology and PSA failure are determined largely by the dominant tumor bulk, as reflected by the percentage of cores positive from the dominant side. Thus, a patient who has 80% of cores positive on one side and 0% positive on the other side would be at greater risk for ECE and PSA failure compared with a patient who has 40% of cores positive on both sides. In agreement with the current results, several previous studies found that as the number or percentage of cores with carcinoma on a single side increased, the risk of ECE on that side increased.8, 14–16 Whether this information, along with standard clinical information (such as PSA level, biopsy Gleason score, and clinical stage), can be used to identify patients who are at a sufficiently high risk of ECE to argue against a nerve-sparing prostatectomy remains to be determined.

Recent studies found that extended biopsy schemes result in an increased detection rate for prostate carcinoma and, specifically, that biopsy locations outside of the standard sextant location are more likely to be positive.17, 18 Therefore, we examined data separately for men who underwent sextant biopsies and men who had > 6 cores obtained. We found that the ability of the percentage of cores positive from the dominant side to predict PSA failure was similar, whether it was examined among men who underwent sextant biopsy or men who underwent an extended biopsy scheme, and was superior to the total percentage of cores for both subsets.

Several previous studies have developed tables for predicting the 2-year PSA recurrence risk using biopsy Gleason score, PSA level, and biopsy tumor volume measurements.1, 19 D'Amico et al.1 developed a nomogram to predict PSA recurrence using the total percentage of cores positive, whereas Nelson et al.19 used the greatest percentage of a biopsy core involved by carcinoma. However, no prior study used the percentage of cores positive from the dominant side of the prostate to predict PSA recurrence. It remains to be determined whether the model developed by either of those prior studies1, 19 or the current study provides the greatest risk stratification for an independent, separate cohort of patients.

Recent studies have questioned the significance of the percentage of cores positive for predicting PSA recurrence relative to more quantified biopsy tumor volume measurements, such as the greatest percentage of a biopsy core involved by carcinoma19, 21 or the total percentage of tissue with carcinoma.4 However, other studies found that the percentage of cores positive was a stronger predictor of PSA recurrence than was the percentage of biopsy tissue with carcinoma among intermediate-risk patients.22, 23 Moreover, not all pathologists report the percentage of tissue with carcinoma in their reports. Thus, Table 5 provides significant stratification for PSA recurrence using readily available clinical information that can be used by both patients and clinicians. Whether a similar table that included more quantified biopsy tumor volume measurements would result in greater stratification remains to be determined. Moreover, to our knowledge, no similar table has been published for patients treated specifically at equal-access medical centers.

In the current study, the highest-risk group identified had only a 46% risk of PSA recurrence within 2 years after surgery. The inability to identify an even-higher-risk group for PSA failure outlines the limitations of using standard clinical variables for predicting PSA recurrence. Indeed, D'Amico et al.20 previously demonstrated that biopsy Gleason score, PSA level, and percentage of cores positive explained < 50% of the variation in PSA recurrence data. Thus, it is clear that additional predictors are needed to improve preoperative risk stratification. Various methods hold the promise of accomplishing this goal, including recent improvements in imaging and the use of molecular markers.20, 24

The current study was limited by its retrospective nature and the fact that follow-up was short. Because all centers in the SEARCH database are equal-access medical centers, the current findings may not apply to patients at non-equal-access medical centers.

In conclusion, because the percentage of cores positive from the dominant side and the total percentage of cores positive are equally easy to calculate, the slight improvement in risk assessment provided by considering the percentage of cores positive from the dominant side justifies doing so in models for predicting PSA recurrence rather than using the total percentage of cores positive. We have described in tabular form the 2-year PSA recurrence risk using a combination of the clinical variables that were important predictors of biochemical recurrence in multivariate analysis: PSA, biopsy Gleason score, and the percentage of cores positive from the dominant side.