Prognostic significance of prostate cancer originating from the transition zone

Abstract

Purpose:

Transition zone (TZ) cancers are reported to have better biochemical relapse-free survival (bRFS) after radical prostatectomy (RP) than cancers from the peripheral zone (PZ). To understand the influence of tumor location, we compared bRFS for TZ and PZ cancers stratified for risk using known clinical and pathological prognostic factors.

Patients and Methods:

The surgical pathology and outcomes of 494 patients were reviewed. Cancers originating from the TZ and PZ were identified from step sectioning of surgical specimens and tumor mapping. Univariate and multivariate analyses of bRFS after RP were compared.

Results:

TZ cancers were present in 89 (18%) patients. On univariate analysis, most factors predicted bRFS, although cancer location did not: 5-year bRFS was 85% for TZ vs. 77% for PZ (P = 0.12). However, on multivariate analysis, all factors except SV involvement were significant, including TZ cancer location (P = 0.04, HR = 1.88 [1.02–3.47]). Interestingly, TZ location was correlated with improved 5-year bRFS for cancers > 2 cc (81% for TZ vs. 65% for PZ, P = 0.017), for preop PSA >10 (80% for TZ vs. 59% for PZ, P = 0.027), and for PSAV > 2 (85% for TZ vs. 66% for PZ, P = 0.08). However, TZ cancers showed no difference in outcome for small volumes, low preop PSA, low PSAV, or high Gleason grade.

Conclusions:

TZ cancers that are large, with high preop PSA, low Gleason scores, and high PSAV show better outcomes than their PZ counterparts. However, high-grade cancer tumor location had no apparent influence on outcome. Tumor location could be considered in subsets for optimal prognostication.

1. Introduction

Roughly 1 out of every 5 prostate cancers arises from the transition zone (TZ). The TZ is separated from the rest of the prostate gland by compressed fibromuscular tissue and is located in the anterior and central part of the gland. The majority of TZ cancers are located in the mid- and apical prostate [1]. Due to their location, most TZ cancers are nonpalpable and are less often detected by standard biopsy schemes; consequently they are larger when diagnosed compared with peripheral zone (PZ) cancers. Despite larger volumes, several studies have demonstrated that TZ cancers display a more favorable biologic behavior compared with those arising from the PZ [1,2–6]. These studies have reported superior biochemical relapse-free survival (bRFS) after radical prostatectomy (RP) in matched pair analysis [1]. Whether the difference in outcomes after RP between TZ and PZ cancers is simply the consequence of anatomical location within the gland or due to biological differences between cancers in these locations is not known.

Until recently, TZ cancers would only be identified if one performed detailed tumor mapping of the prostate gland after RP. However, as extended biopsy schemes have become the recommended standard [7,8], more TZ cancers appear to be identified, mainly because of the 2 to 4 additional biopsy cores that are directed anteriorly. Although biopsies alone will have limited reliability in identifying predominant TZ cancers, advances in imaging technologies such as high-resolution MRI promise to some day allow for the precise location and extent of tumor to be known. Therefore, the question of whether tumor location (PZ vs. TZ) contributes to prognosis is becoming increasingly important in counseling patients for therapy and in selecting management options. In this study, we compare TZ and PZ prostate cancers in terms of clinical and pathologic differences, in terms of the reliability of biopsy Gleason grading, and in terms bRFS after RP.

2. Patients and methods

2.1. Patient population

The surgical pathology and outcomes of 494 patients who underwent RP between 1989 and 2000 at Stanford were reviewed. This retrospective study was part of an IRB approved research program at our institution, and only those patients with a signed consent for retrospective chart review were included. All patients were previously untreated and diagnosed with sextant biopsies. Step sectioning of surgical specimens and tumor mapping have been previously described [1]. Briefly, it consists of 3 mm serial transverse sectioning and mapping of tumor and identification of the prostate epithelial boundary defining the transition zone on the H and E slides. Computerized planimetry was then used to reconstruct in 3 dimensions the total cancer volume and proportional distribution within each zone. The primary zone of cancer origin (TZ or PZ) was then assigned according to that zone containing the majority of the cancer volume [1]. The primary and secondary pathologic Gleason grades were assigned according to the volumetric proportional amount of that grade present. All pathology was performed by a single pathologist (John E. McNeal, MD). Clinical factors include clinical stage according to the AJCC 4th edition [9], preoperative PSA, and biopsy Gleason grade [10]. Pathologic factors include pathological Gleason grade, involvement status of seminal vesicle, capsule, margin and lymph node, total cancer volume, and prostate weight. Median follow-up time was 6 years. The clinical and pathological characteristics of the study patients are summarized in Table 1.

Table 1.

Comparison of clinical and pathologic characteristics for patients with prostate cancer located in the peripheral zone (PZ) or transition zone (TZ)

Factor	PZ (n = 405)	TZ (n = 89)	P value
Median age (years)	58	63.5	0.65
Clinical T-stage			<0.0001
T1c	51% (207)	78% (69)
T2	49% (198)	22% (20)
Pre-op PSA (ng/ml)			<0.0001
Median	7.4	10.8
IQR	4.9–10.5	6.3–17.1
≤10	73% (294)	47% (42)
>10–20	22% (89)	34% (30)
>20	5% (22)	19% (17)
Biopsy GS			0.12
≤+3	54% (216)	63% (56)
3+4	29% (116)	18% (16)
4+3	7.5% (30)	11% (10)
≥4+4	9.5% (38)	8% (7)
Pathologic GS			0.016
≤+3	18% (69)	24% (21)
3+ 4	63% (254)	46% (41)
4+ 3	20% (80)	28% (25)
≥4+4	0.5% (2)	2% (2)
Seminal vesicle positive	9% (36)	2% (2)	0.032
Surgical margins positive	16% (66)	29% (26)	0.0053
Capsule involved	36% (145)	15% (13)	<0.0001
Cancer volume (cc)			0.001
Median	2.3	4.4
IQR	1.2–4.7	2.2–7.3
LN positive	5% (21)	7% (6)	0.57
Prostate Weight (gm)			0.64
Median	47	47
IQR	38–60	40–61
Year of surgery			0.31
1989–92	33% (134)	26% (23)
1993–96	43% (172)	44% (39)
1997–2000	24% (98)	30% (27)

PZ = peripheral zone; TZ = transition zone; PSA = prostate specific antigen; GS = Gleason sum; IQR = interquartile range (25th–75th percentile).

2.2. Statistical analysis

Biochemical failure was defined as a persistently detectable or rising postop PSA of 0.1 ng/ml or higher. Follow-up PSAs were taken at 3-month intervals for the first year postop, at 6-month intervals for the following 2 years, and yearly thereafter. Factors at diagnosis are compared as both continuous and categorical variables. Differences between TZ and PZ cancers are tested for significance with the Wilcoxon statistic for PSA as a continuous variable, with the t-test for cancer volume and prostate weight as continuous variables, and χ² for all categorical variables. bRFS after RP were compared with univariate Kaplan Meier [11] analyses and with proportional hazards multivariate analyses. The annual PSAV (the rate of change in PSA per year) has been shown to be an independent factor for relapse when comparing PSAV ≤ 2 vs. > 2 ng/ml/y [12,13]. In our study, it was available for 101 patients (75 PZ patients and 26 TZ patients). It was calculated by simple linear regression from 2 or more PSA measurements available prior to the positive biopsy diagnosis. The median time interval between PSA measurements was 12 months.

3. Results

TZ cancers were present in 89 of 494 (18%) patients. Biochemical failure occurred in 94 PZ patients (23%) and in 14 TZ patients (16%) with a median time to failure of 1.5 years (IQR 0.3 to 3.8 years). A full comparison of clinical and pathological features between TZ and PZ cancers is summarized in Table 1. Compared with PZ cancers, TZ cancers were more likely to be nonpalpable (T1c), have higher preop PSA, and larger cancers, but less capsular and seminal vesicle involvement (all P < 0.05). There was no significant difference in rate of lymph node involvement, prostate weight, or frequency of cancer location over the years spanned by this study.

All factors were significant on univariate analysis in predicting for bRFS except for cancer location, where the 5-year bRFS was 85% for TZ vs. 77% for PZ (trend P = 0.12). Kaplan Meier analysis demonstrated that TZ cancers have a trend for improved bRFS compared with PZ cancers for all patients despite having significantly worse clinical and pathologic features (Fig. 1). To evaluate the effect of preoperative PSA level on outcome for TZ and PZ cancers, we analyzed outcomes for patients with preop PSA ≤ 10 vs. > 10 ng/ml (Fig. 2). In general, TZ cancers appeared to have higher bRFS than PZ cancers, although the difference was statistically significant only for patients with preop PSA > 10, where the 5-year bRFS was 80% for TZ vs. 59% for PZ (P = 0.027). In Fig. 3, outcomes were evaluated in patients with tumor volumes ≤ 2 cc and > 2 cc. Once again, tumor location appeared to affect outcome only in the high risk group, in that TZ cancers had higher bRFS than PZ cancers only for cancers > 2 cc (5-year bRFS was 81% for TZ vs. 65% for PZ cancers, P = 0.017). To determine whether tumor location influenced outcome in patients with other high risk features, we evaluated outcomes for tumors with pGS ≤ 3+4 vs. ≥ 4+3 Fig. (4). Intriguingly, TZ cancers showed higher bRFS only for patients with pGS ≤ 3+4 (P = 0.056). When we compared outcomes for patients with a PSAV ≤ 2 vs. > 2 ng/ml/y, shown in Fig. 5, outcomes appeared to be better for TZ cancers and this relationship approached statistical significance for PSAV > 2 ng/ml/y (P = 0.08).

Fig. 1.

Kaplan Meier curves comparing the bRFS between TZ and PZ cancers. Symbols represent censoring events.

Fig. 2.

Kaplan Meier curves comparing the bRFS between TZ and PZ cancers. Top panel is for patients with preop PSA ≤ 10 ng/ml and lower panel for those with preop PSA > 10 ng/ml. Symbols represent censoring events.

Fig. 3.

Kaplan Meier curves comparing the bRFS between TZ and PZ cancers. Top panel is for patients with cancer volumes ≤ 2 cc and lower panel is for patients with cancer volumes > 2 cc. Symbols represent censoring events.

Fig. 4.

Kaplan Meier curves comparing the bRFS between TZ and PZ cancers. Top panel is for patients with pGS ≤ 3+4 and lower panel is for patients with pGS ≥ 4+3. Symbols represent censoring events.

Fig. 5.

Kaplan Meier curves comparing the bRFS between TZ and PZ cancers. Top panel is for patients with PSAV ≤ 2 ng/ml/y and lower panel is for patients with PSAV > 2 ng/ml/y. Symbols represent censoring events.

On multivariate analysis, all factors except for SV involvement were significantly correlated with bRFS, including TZ cancer location (P = 0.04, HR = 1.88 [1.02–3.47]). The univariate and multivariate analyses are summarized in Table 2.

Table 2.

Univariate and multivariate analysis of factors associated with biochemical relapse-free survival (bRFS)

Factor	5-year bRFS (%)	Univariate P value	Multivariate P value	Hazard ratio (95% CI)
Preop PSA (ng/ml)
≤10	84.8	<0.0001	0.022	0.44 (0.22–0.89)
10–20	72.0		0.073	0.53 (0.26–1.06)
>20	41.4		a	a
Path Gleason grade
≤3+3	98.9	<0.0001	0.0009	0.034 (0.005–0.25)
3+4	81.5		<0.0001	0.39 (0.25–0.59)
≥4+3	48.7		a	a
Seminal vesicle involvement
Positive	20.4		a	a
Negative	82.2	<0.0001	0.78	0.92(0.51–1.65)
Margin status
Positive	49.1		a	a
Negative	83.8	<0.0001	<0.0001	0.42 (0.27–0.65)
Capsular involvement
Positive	50.7		a	a
Negative	90.5	<0.0001	0.0001	0.39 (0.25–0.63)
Lymph node involvement
Positive	21.1		a	a
Negative	81.3	<0.0001	0.042	0.47 (0.23–0.97)
Cancer volume
≤2cc	92.4		0.0076	0.44 (0.24–0.8)
>2cc	69.0	<0.0001	a	a
Cancer location
PZ	76.9	0.12	0.044	1.88 (1.02–3.47)
TZ	85.1		a	a
Year of surgery
1989–92	76.4	0.27	0.59	1.18 (0.65–2.16)
1993–96	77.2		0.99	0.99 (0/56–1.78)
1997–2000	81.3		a	a

PZ = peripheral zone; TZ = transition zone; PSA = prostate specific antigen; GS = Gleason sum.

^aReference factor.

Gleason grading of prostate needle biopsies has been shown to differ significantly compared with radical prostatectomy grading. It is unknown whether tumor location (PZ vs. TZ) influences the frequency or extent of grade changes between biopsy and prostatectomy grade. Since grade is an important determinant of prognosis, we compared the reliability of the biopsy grade to accurately predict the final pathologic grade in PZ and TZ cancers. For this comparison, we made a distinction between four Gleason score (GS) groups: 3+3 or lower, 3+4, 4+3, and 4+4 or higher. We defined a clinically significant upgrade as those cancers that were GS 3+3 on biopsy and GS 3+4 or higher on final surgical grading, or GS 3+4 on biopsy and 4+3 or higher on surgical GS, or GS 4+3 on biopsy and 4+4 or higher on surgical GS. We defined downgrading as the converse relationships. Overall, there was an exact GS match in 39% of patients, an upgrade in 46% and a downgrade in 15%. There was no difference in the biopsy to surgical grade change between TZ and PZ cancers (P = 0.54). These results are summarized in Table 3.

Table 3.

Comparison of clinically significant biopsy-to-surgical grade change between peripheral zone (PZ) and transition zone (TZ) cancers

Biopsy-to-surgical grade changea	PZ	TZ
Upgrade	46% (184)	46.1% (41)
Same grade	38.5% (154)	42.7% (38)
Downgrade	15.5% (62)	11.2% (10)
χ2 P = 0.54

^aUpgrade defined as biopsy grade 3+3 to surgical grade 3+4 or higher, or biopsy grade 3+4 to surgical grade 4+3 or higher. Downgrade defined as the converse grade changes.

4. Discussion

We have shown that compared with PZ cancers TZ cancers are generally diagnosed as nonpalpable disease (clinical stage T1c), have higher PSA levels, and have higher tumor volumes. Despite this, they less frequently present with capsular or seminal vesicle involvement after RP. In patients with high risk features, tumor location in the TZ was associated with improved bRFS with one important exception: tumor Gleason grade. Improved outcomes were observed in patients with TZ cancers when stratified for high preop PSA, larger cancer volume, and high PSAV. The strength of these associations likely explains the strong independent association of TZ location with improved outcomes on multivariate analysis after correction for all clinical and pathologic factors. Our findings suggest that accurate characterization of tumor location could contribute to prediction of bRFS after radical prostatectomy. While there is no reliable means of determining tumor location in the pretreatment setting currently, as prostate cancer imaging technologies are developed, predictive algorithms should take into account tumor location. Since all pretreatment predictive algorithms include Gleason grading of the biopsy samples, it is reassuring that significant changes in grading occurred at the same frequency among TZ and PZ cancers.

High tumor grade (GS ≥ 4+3) appeared to confer a poor outcome even for TZ cancers and canceled out the effect of tumor location in bRFS. Previous studies have demonstrated that Gleason grade is the most important determinant of recurrence after therapy and death due to prostate cancer. Studies from our institution have demonstrated that an increase in the percentage of Gleason patterns 4 and 5 is the most powerful predictor of poor outcome after surgery [14]. Our current study suggests that in high grade cancers, location is moot, and that the biology of the cancer is most reflected in the high grade appearance of the tumor. While some have speculated that TZ cancers possess a different biology than PZ cancers, our data suggests that high grade cancers in both locations share biological features, and these tumors could share molecular genetic features.

As with all retrospective studies that span a long time-period, one has to be aware of the potential effects that stage migration might have upon outcomes [15]. In our study that spans from 1989 to 2000, we did not find that year of treatment was significant in either the proportion of TZ cancers diagnosed or in the likelihood of bRFS (see Tables 1 and 2). We also note that a few of the sub-categories of our data contain very small numbers of patients (e.g., TZ cancers with GS ≥ 4+4, or TZ cancers with SV involvement, see Table 1) and represents a limitation of this study.

While previous studies nearly universally concur on the apparent favorable biology of TZ cancers [1,2–5], there is just one dissenting series [16,17]. Our findings shed light on the possible reasons for that discrepancy. In the study showing no effect of location on outcome, 63 TZ patients were matched with 63 PZ patients for pathologic T-stage, pathologic Gleason grade, and margin status, and were shown to possess the same bRFS at 5 years after RP. However, that study failed to match for either preop PSA or cancer volume, and their TZ patients had significantly higher preop PSA and cancer volume (each by a factor of ~2) compared with the matched PZ patients. Furthermore, the paradoxical effect of Gleason grade could account for their findings, since any biases toward high Gleason grade would obscure the differences in outcome between TZ and PZ cancers. Another study that also used a matched-pair analysis of 79 patients in each group [1] but that included all clinical and pathologic factors (including PSA and cancer volume) demonstrated a significant difference in the 5-year bRFS, 71% for TZ cancers vs. 49% for PZ cancers (P = 0.0002).

We note that in our series there were 15% of TZ cancers possessing ECE associated with a positive surgical margin rate of 29%, whereas nearly the converse rates were observed for PZ cancers (36% and 16%, respectively). A consistently similar finding was observed in other series [1,16,17]. We hypothesize that this unexpected finding is likely the result of surgical technique with respect to the anatomical location of the cancer within the prostate gland. With this in mind, one might be more attentive to tumor location and margin location for patients referred for postoperative radiotherapy.

Our study confirms the prevalence of TZ cancers and their clinical and biologic difference with PZ prostate cancers. It also demonstrates the relative importance in identifying these cancers, both within the context of analyzing outcomes after primary therapy and in the context of assessing individual prognostic groups at the time of patient counseling. A recent study showed that differences between TZ and PZ cancers undermines the accuracy of the Partin tables in predicting pathologic stage [6] by overestimating the extent of disease. With the increased potential ability of extended biopsy schemes to detect TZ cancers, it will be worthwhile to identify and study those patients subsequently treated with radiotherapy to see of the same favorable biologic behavior is observed as with RP. It may be also be of value to more routinely identify TZ cancers from prostatectomy patients as it may influence those patients being considered for adjuvant or salvage radiotherapy, or for high risk patients being considered for clinical trials of adjuvant therapies.