Prostate Cancer and Prostatic Disease (2014) 17, 144–148 & 2014 Macmillan Publishers Limited All rights reserved 1365-7852/14 www.nature.com/pcan

ORIGINAL ARTICLE

The association of tumor volume with mortality following radical prostatectomy JJ Knoedler1, RJ Karnes1, RH Thompson1, LJ Rangel2, EJ Bergstralh2 and SA Boorjian1 BACKGROUND: Data regarding the prognostic significance of tumor volume (TV) in prostate cancer are conflicting. Herein, we evaluated the association of TV with prostate cancer mortality following radical prostatectomy (RP), and assessed the additive prognostic value of TV to an established predictive model. METHODS: We identified 13 687 patients who underwent RP without preoperative therapy between 1987 and 2009. TV was estimated using the prolate ellipsoid formula. Survival was estimated using the Kaplan–Meier method and compared with the logrank test. Cox proportional hazard regression models were used to evaluate the association of TV with mortality. The ability of TV to enhance the performance of an established prognostic model (Mayo Clinic GPSM (Gleason, PSA, seminal vesicle and margin status) score) was assessed using the c-index. RESULTS: Median TV was 1.57 cm3 (interquartile range (IQR) 0.48–4.19). Increasing TV was associated with significantly higher risks of seminal vesicle invasion (hazard ratio (HR) 1.58; Po0.0001), positive surgical margins (HR 1.28; Po0.0001) and lymph node involvement (HR 1.26; Po0.0001). Median postoperative follow-up was 9.4 years (IQR 5.0–14.5). Patient grouping into quartiles according to TV resulted in a significant stratification of outcome, as the 15-year cancer-specific survival by TV quartile was 99%, 98%, 95% and 88%, respectively (Po0.0001). Moreover, on multivariate analysis, greater TV remained associated with significantly increased risks of systemic progression (HR 1.27; Po0.0001), death from prostate cancer (HR 1.29; Po0.0001) and all-cause mortality (HR 1.05; Po0.0001). Meanwhile, addition of TV to the GPSM score increased the c-index for the model’s prediction of prostate cancer mortality from 0.803 to 0.822. CONCLUSIONS: TV is associated with survival following RP, and enhances, although modestly, the performance of an established prediction model. As such, TV warrants continued assessment in risk stratification tools. Prostate Cancer and Prostatic Disease (2014) 17, 144–148; doi:10.1038/pcan.2013.61; published online 28 January 2014 Keywords: tumor volume; radical prostatectomy; mortality; risk stratification

INTRODUCTION The prognostic importance of the extent of cancer noted in radical prostatectomy (RP) specimens has been well described.1 However, the extent of cancer may variously be recorded as tumor stage, surgical margin status and/or tumor volume (TV). Indeed, these parameters have been found to be tightly correlated in series to date.2–8 Nevertheless, although tumor stage and margin status have been included within prostate cancer prediction models,9,10 the importance of even reporting TV has been questioned,4 despite a recommendation from the International Society of Urologic Pathology that ‘y some quantitative estimate of cancer volume should be undertakeny’11 in RP specimens. Notably, among men undergoing RP, data regarding the prognostic significance of TV are conflicting.1–3,5–8,12–23 In particular, the close correlation of TV with tumor stage and margin status8 has often obscured the ability to establish the predictive value for this feature. Moreover, studies to date have largely been limited by small sample sizes, relatively short-term follow up and/or by the use of the outcome measure of biochemical recurrence (BCR). In fact, the natural history of BCR has been illustrated to be heterogeneous, and does not in most cases result in the more clinically relevant events of metastasis or death from prostate cancer.24,25 In addition, whether inclusion

of TV within established predictive variables would improve estimation of prognosis remains to be determined. Herein, then, we evaluated the association of TV with clinicopathologic outcomes including systemic cancer recurrence and mortality following RP, and assessed the additive prognostic value of TV to an established prostate cancer predictive model.

MATERIALS AND METHODS After institutional review board approval was obtained, we identified 18 916 patients who underwent RP between 1987 and 2009 at our institution. Surgical procedures were performed by different surgeons using standard techniques. The extent of pelvic lymph node dissection varied with individual surgeon and over the time period of the study. Patients who underwent preoperative radiation therapy or androgen deprivation therapy (n ¼ 1944), foreign patients without follow-up (n ¼ 789), patients missing the variable of TV (n ¼ 2286) as well as patients who did not authorize use of their records (n ¼ 210) were excluded. Thus, the final cohort for analysis contained 13 687 patients. Nearly all of the patients (12 271/13 687; 89.7%) were treated with open retropubic approach to prostatectomy, with 1416 (10.3%) having undergone robotic-assisted RP. Tumors were classified according to the 2010 AJCC 7th edition TNM staging system.26 TV was estimated using the prolate ellipsoid formula.4,27 TV was assessed from the index, or largest

1 Department of Urology, Mayo Clinic, Rochester, MN, USA and 2Department of Health Sciences Research, Mayo Clinic, Rochester, MN, USA. Correspondence: Dr SA Boorjian, Department of Urology, Mayo Clinic, 200 First Street Southwest, Rochester, MN 55905, USA. E-mail: [email protected] Received 17 September 2013; revised 21 November 2013; accepted 17 December 2013; published online 28 January 2014

Tumor volume predicts death after prostatectomy JJ Knoedler et al

145 Table 1.

Patient clinicopathologic demographics 5

No. (%)

Clinical stage (n ¼ 13 593) T1 T2 T3/4

6069 (44.3) 6820 (49.8) 704 (5.2)

Biopsy Gleason score (n ¼ 11 631) 6 7 8–10

7939 (68.3) 2987 (25.7) 705 (6.1)

Pathologic tumor stage (n ¼ 13 683) pT2 pT3a pT3b pT4 Pathological Gleason score (n ¼ 13 462) 6 7 8–10 Positive surgical margin pN þ

10 258 1913 1464 48

(75) (14.0) (10.7) (0.4)

7847 4622 993 4007 639

(58.3) (34.3) (7.4) (29.3) (4.7)

0

-5

-10 Year of surgery

Figure 1.

tumor, when multifocal lesions were present within the prostate. Adjuvant therapy, including androgen deprivation and radiation therapy, was defined as treatment received within 90 days of RP, and was administered at the discretion of the treating physician. Salvage therapy, defined as secondary treatment beyond 90 days from RP, was similarly recommended according to individual physician discretion. Medical androgen deprivation therapy was generally intended to be lifelong. However, given the retrospective nature of this study, it is uncertain whether some patients discontinued treatment at a later date. Postoperative assessments including physical examination and serum PSA testing were conducted quarterly to semi-annually for the first 2 years after surgery, semi-annually to annually for the next 3 years and annually thereafter. BCR was defined here as a PSA X0.4 ng ml  1.28,29 Local recurrence was defined as cancer demonstrated on biopsy of the prostatic bed or by receipt of salvage radiation therapy to the prostatic bed without evidence of systemic recurrence. Systemic progression involved demonstrable metastatic deposits on imaging or pathologic confirmation of prostate cancer on biopsy of tissue outside the prostatic fossa. Vital status was identified from death certificates or physician correspondence. Statistical analyses were performed using the SAS software package (SAS Institute, Cary, NC). The association of clinical variables with TV was explored using Spearman’s correlation and univariate logistic regression as appropriate. Survival was estimated using the Kaplan–Meier method and compared with the log-rank test. Patients were censored at last follow-up or death if the end point of interest had not been attainted. Cox proportional hazard regression models were used to evaluate the association of clinicopathologic variables, including TV, with outcome. Because of extreme skewness, TV and preoperative PSA were analyzed on a log-2 scale, where the reported hazard ratio (HR) is that associated with a doubling. The ability of TV to enhance the performance of the GPSM (Gleason, PSA, seminal vesicle and margin status) score, a prognostic model developed at our institution9 and demonstrated to stratify patients’ risk of BCR9,10 and cancer-specific survival,10 was assessed using the c-index. All statistical tests were two sided, with a P-value of o0.05 considered statistically significant.

RESULTS Patient clinicopathologic demographics are provided in Table 1. Median patient age at RP was 63 years (interquartile range (IQR) 58–68), and median body mass index was 27.6 kg m  2 (IQR 25.4– 30.2). Median preoperative PSA was 6.3 ng ml  1 (IQR 4.4–9.8), and median prostate volume was 32.4 cm3 (IQR 24.8–44.1). Median TV for the overall cohort was 1.57 cm3 (IQR 0.48–4.19). Interestingly, & 2014 Macmillan Publishers Limited

Log2 (Tumor volume estimated)

Variable

1987-1995

1996-2002

2003-2009

Time trends in tumor volume at radical prostatectomy.

we noted a significant decrease in TV over time (Figure 1), as the median TV decreased from 2.62 cm3 among men treated between 1987 and 1995 to 0.97 cm3 among men who underwent RP between 2003 and 2009 (Po0.0001). Next, we evaluated the independent association of TV with adverse pathologic features at RP. We found a statistically significant relationship between TV and pathologic tumor stage (Spearman’s correlation 0.51; Po0.0001) as well as preoperative PSA (Spearman’s correlation 0.36; Po0.0001). Moreover, we noted on multivariate analysis (Table 2) that increased TV was associated with significantly increased risks of seminal vesicle invasion (HR 1.58; Po0.0001), positive surgical margins (HR 1.28; Po0.0001) and positive lymph nodes (HR 1.26; Po0.0001). Similarly, the established prognostic features of PSA, Gleason score and pathologic tumor stage were also associated with these pathologic end points. Median follow-up after RP was 9.4 years (IQR 5.0–14.5), during which time 990 patients were diagnosed with local recurrence, 795 developed systemic progression and 3138 died, including 430 patients who died of prostate cancer. Meanwhile, a total of 591 patients received adjuvant radiation therapy, 1568 adjuvant androgen deprivation therapy, 1627 salvage radiation therapy and 1911 salvage androgen deprivation therapy. We found that TV significantly stratified patient outcomes following surgery, such that when patients were grouped into quartiles according to TV, the 15-year cancer-specific survival was 99%, 98%, 95% and 88%, respectively, from the lowest to the highest TV quartile (Po0.001; Figure 2). In addition, to investigate whether TV would remain associated with outcome after established variables were controlled for, we evaluated the prognostic significance of TV in various subsets of patients. Specifically, we separately tested the association of TV with cancer-specific survival among patients with pathologically organ-confined disease (Figure 3) as well as among patients with nonorgan-confined disease (Figure 4). In both cohorts, we found that TV significantly stratified survival following surgery (Po0.001 for both). In particular, for patients with nonorgan-confined tumors, the 15-year cancer-specific survival according to TV quartile was 95%, 94%, 92% and 83%, respectively, from the lowest to the highest TV quartile. On multivariable analysis, we found that a twofold increase in TV was associated with significantly increased risks of BCR (HR 1.11; Po0.0001; data not shown) and local recurrence (HR 1.16; Po0.0001; data not shown), as well as systemic progression (HR 1.27; Po0.0001), death from prostate cancer (HR 1.29; Po0.0001) and all-cause mortality (HR 1.05; Po0.0001; Table 3). Given the noted temporal trend in TV over the time period of our study, we Prostate Cancer and Prostatic Disease (2014), 144 – 148

Tumor volume predicts death after prostatectomy JJ Knoedler et al

146 Table 2.

Multivariate analysis of the association of tumor volume with adverse pathologic features Positive surgical margin

Year of surgery Patient age Log2(preoperative PSA) Clinical stage Biopsy Gleason score Pathologic Gleason score Pathologic T stagea Log2(tumor volume)

Seminal vesicle invasion

Positive lymph nodes

OR

95% CI

P-value

OR

95% CI

P-value

OR

95% CI

P-value

0.91 1.00 1.30 1.02 0.89 1.36 1.23 1.28

0.90–0.92 1.0–1.01 1.24–1.38 0.95–1.08 0.81–0.98 1.24–1.50 1.15–1.30 1.25–1.32

o0.0001 0.21 o0.0001 0.62 0.02 o0.0001 o0.0001 o0.0001

0.96 1.00 1.42 1.38 1.36 2.47 — 1.58

0.95–0.98 0.99–1.01 1.31–1.54 1.26–1.51 1.19–1.56 2.13–2.87 — 1.50–1.66

o0.0001 0.68 o0.0001 o0.0001 o0.0001 o0.0001 — o0.0001

0.97 0.97 1.36 1.31 1.16 2.17 2.12 1.26

0.95–1.0 0.96–0.99 1.22–1.52 1.15–1.49 0.95–1.42 1.73–2.72 1.84–2.45 1.16–1.37

0.03 0.001 0.01 o0.0001 0.15 o0.0001 o0.0001 o0.0001

Abbreviations: CI, confidence interval; OR, odds ratio. Pathological stage T3b corresponds to seminal vesicle invasion, and is therefore excluded.

a

100 Cancer Specific Survival (%)

100 Cancer Specific Survival (%)

90 Q0–Q1 Q1–Q2 Q2–Q3 Q3–Q4

80 70 60 50 40 30 20 10

P