Cancer and Prostatic Diseases (2008) 11, 148–152 & 2008 Nature Publishing Group All rights reserved 1365-7852/08 $30.00 www.nature.com/pcan ORIGINAL ARTICLE

PSA reduction (after antibiotics) permits to avoid or postpone in selected patients

V Serretta1, A Catanese1, G Daricello1, R Liotta1, R Allegro2, A Martorana3, F Aragona3 and D Melloni1 1Department of Internal Medicine, Cardiovascular and Nephro-Urological Diseases, Institute of , University of Palermo, Palermo, Italy; 2Department of Statistics, GSTU (Gruppo Studi Tumori Urologici) Foundation, Palermo, Italy and 3Department of Human Pathology, University of Palermo, Palermo, Italy

Microscopic foci of may induce prostate-specific antigen (PSA) increase. PSA reduction after antibiotics might identify those patients in whom biopsy can be avoided. Ninety-nine patients received ciprofloxacin for 3 weeks, of whom 59 showed PSA reduction. Histology detected small foci of prostatitis in 65% of cases. Carcinoma was found in 40 and 20.3% of patients with unchanged or decreased PSA, respectively (P ¼ 0.03). No cancer was detected if PSA decreased below 4 ng/ml or more than 70%. Biopsy can be postponed, with a low risk of missing a cancer, if PSA decreases more than 70% or below 4 ng/ml. and Prostatic Diseases (2008) 11, 148–152; doi:10.1038/sj.pcan.4500996; published online 17 July 2007

Keywords: prostate cancer; prostatitis; antibiotics; PSA;

Introduction biopsy. The aim of our study was to investigate the possibility of reducing the number of prostate biopsies in Disruption of the natural anatomic and physiologic patients showing PSA decrease or normalization after barriers between the prostatic milieu and the blood- antibiotic therapy. This approach could be useful in stream is an important factor determining increased patients for whom it is necessary to postpone biopsy and serum prostate-specific antigen (PSA) levels. Inflamma- in patients, with previous negative biopsies, willing to tion alters prostatic duct integrity causing PSA leakage avoid biopsy until further PSA increase. A secondary from the acini and ductal lumina. In 1989, Dalton was the aim was to identify the level of PSA reduction that must first to report total PSA elevation in .1 be reached to avoid biopsy, limiting as much as possible Many experimental and clinical studies suggest a the risk of missing a cancer. correlation between acute and chronic prostatitis and increased serum PSA levels.2–6 Subclinical inflammation of the prostate could elevate serum PSA in asymptomatic Patients and methods patients without clinically detectable prostate cancer. 7 Ninety-nine asymptomatic Caucasian men requiring Gerstenbluth et al., in 2002, reported chronic prostatitis urological consultation for PSA between 4 and 10 ng/ as a common finding in radical speci- dl, with negative for cancer and mens. In the majority of cases, prostatitis is an incidental candidate to biopsy, were entered in our study. Exclusion pathological finding that causes no clinical symptoms. criteria were abnormal mid-void urine, urinary tract These patients are categorized by the National Institutes infection, urinary symptoms, clinical signs or documen- of Health (NIH) into category IV prostatitis (asympto- ted history of prostatitis. The patients should have at matic inflammatory prostatitis) (Supplementary Infor- least two previous elevated PSA detections, showing no mation). It has been suggested that repeating the decrease greater than 10%, to confirm the possibility of measurement of PSA in symptomatic men can help 8,9 spontaneous PSA reduction. All patients were submitted avoid unnecessary prostatic biopsy. Although prosta- to physical examination and rectal exam before biopsy. titis may cause PSA elevation, asymptomatic patients are Prostate gland enlargement was a common finding, but not routinely screened for this disease before transrectal there was no single objective finding indicative of prostate cancer or prostatitis. Since our patients were asymptomatic, they did not undergo prostate Correspondence: Professor V Serretta, Department of Internal and laboratory-specific analyses to rule out prostatitis. Medicine, Cardiovascular and Nephro-Urological Diseases, Institute Written informed consent was obtained from all patients. of Urology, University of Palermo, Via Pietro Calandra 12, int 25 90148 Basal total-PSA (t-PSA) and free-PSA (f-PSA) determina- Tommaso Natale, Palermo, Italy. E-mail: [email protected] tions were repeated in all patients at study entry. Received 23 March 2007; revised 4 May 2007; accepted 5 June 2007; Eighteen patients (18.2%), with repeated PSA values published online 17 July 2007 between 10.2 and 20 ng/dl, were accepted in the study as PSA reduction after antibiotics V Serretta et al 149 they needed to temporarily postpone the biopsy. These Table 1 The characteristics of the 99 patients at entry patients were informed of the increased risk of prostate tumour. Ciprofloxacin (500 mg b.i.d.) was given orally for Patients’ characteristics Mean s.d. Median Range 3 weeks. t-PSA and f-PSA determinations were repeated Age (years) 65 6.7 65 48–83 2 weeks after therapy. Independently from the value PSA (ng/ml) 8.1 3.2 7.3 4.1–20.0 of the repeated PSA, 2–4 weeks after therapy cessation, PSA 4–10 (ng/ml), 81 patients 6.9 1.8 6.7 4.1–10.0 all patients were submitted to transrectal ultrasound- PSA 10, 2–20 (ng/ml), 18 patients 13.7 2.6 13.1 10.2–20.0 guided biopsy (biplanar mode, B&K sonograph). Free PSA 1.3 1.4 1.0 0.1–7.4 f-PSA/t -PSA ratio 0.15 0.10 0.14 0.01–0.49 Prostate volume was estimated assuming an ellipsoid Prostate volume (cm3) 37.8 15.5 37.0 19.0–82.0 shape. It was evaluated only at the time of the trans- rectal biopsy, and no attempt to study volume variations, Abbreviation: f-PSA/t-PSA ratio, free/total-PSA ratio. presumably of small entity, caused by antibiotic therapy was done. Ultrasound-guided transrectal biopsy, using an 18-G needle fitted to an automatic biopsy gun, was Table 2 Effect of antibiotic therapy on t-PSA and f-PSA/t-PSA performed. The number of cores varied from 12 to 21 ratio according to patients’ age, prostate volume, number of PSA after antibiotic therapy Mean s.d. Median Range previous biopsies and presence of suspicious area detected by ultrasound. In particular, 12 prostate cores PSA (ng/ml) 7.03 2.79 6.60 1.00–19.27 were obtained in 61 patients, 14–16 cores in 25 patients Free PSA 1.19 0.75 1.02 0.09–3.57 owing to suspicious area at ultrasound and 21 cores f-PSA/t-PSA ratio 0.17 0.09 0.16 0.01–0.38 in 13 patients with previous negative biopsy or with PSA reduction (ng/ml) 2.87 2.93 2.00 0.18–13.80 3 PSA reduction (%) 28.07 21.36 22.22 2.37–91.39 prostate volume between 60 and 82 cm . Local anaes- PSA ratio f/t variation 0.06 0.14 0.06 À0.38 to 0.34 thesia with transrectal infiltration of 5 ml of 20% mepivacaine solution, through a 20-G needle at the Cancer detection rates Patients % Cancer % junction between each seminal vesicle and the prostate, was performed. Routine haematoxylin–eosin-stained Unchanged/increased PSA 40 40.4 16 40 slides were obtained and were routinely examined to Decreased PSA 59 59.6 12 20.3 PSA reduction 470% 5 5.1 00 avoid any deviation from the usual clinical practice. PSA reduction 450% 13 13.1 2 2 In addition, a pathological systematic analysis of the 4pbasal PSAp10 ng/ml 6 6.1 00 specimens was performed to score the inflammation 10obasal PSAp20 ng/ml 7 7.1 2 2 that was graded using a four-point scale according to PSAo4 ng/ml 8 8.1 00 6 the system formulated by Irani: grade 0, no inflamma- Abbreviation: f-PSA/t-PSA ratio, free/total-PSA ratio. tory cells; grade 1, scattered inflammatory cell infiltrate Bold values indicate no prostate cancer detected. within the stroma without lymphoid nodules; grade 2, non-confluent lymphoid nodules; and grade 3, confluent infiltrate of mononuclear inflammatory cells. Fifty-nine patients (59.6%) showed t-PSA reduction Patients with PSA reduction and negative biopsy will after therapy. In eight of them (8.1%), PSA decreased be followed with PSA determinations and digital rectal below 4 ng/ml. The mean and median PSA reduction exam (DRE) at 6-month intervals for the first year and rates were 28 and 22%, respectively. The reduction of then yearly. t-PSA had a widespread variation ranging between 2.4 and 91.4%. The median absolute reduction was 2.0 ng/ml, up to 13.8 ng/ml. PSA ratio remained almost Statistics unchanged, with a median value of 0.16 and a median Statistical analysis was performed by regression analysis. rate of change of 6%, as f-PSA varied according to t-PSA. The univariate associations between tumour absence At statistical analysis, comparing the baseline character- at biopsy with patients’ age, basal t-PSA and f-PSA, istics of patients with and without t-PSA reduction after t-PSA and f-PSA reduction after therapy were examined antibiotics, the former had a significantly higher baseline and differences assessed by w2 and Fisher’s exact tests. PSA value (median value 9.0 vs 5.0 ng/ml; Po0.0001). Statistical comparisons of the groups were performed On the other hand, no difference emerged in terms of using the Mann–Whitney U test and the Kruskal–Wallis age, prostate volume, f-PSA/t-PSA ratio, numbers of non-parametric analyses of variance. Multivariate logis- cores and previous negative biopsies. Routine histology tic regression analysis was performed and significant revealed benign prostate hyperplasia, prostate carcinoma ORs with 95% confidence interval (CI) were calculated. and chronic prostatitis in 65 (65.7%), 28 (28.3%) and 6 (6%) patients, respectively (Table 3). Conversely, the systematic grading of the inflammation showed grade 0 in 35 patients and small microscopic foci of prostatitis Results in 60 patients, grade 1, 2 and 3 in 52, 6 and 5 patients, respectively. No correlation was found between basal The characteristics of the 99 patients entered in our study PSA values and routine histology (P ¼ 0.49) or inflam- are reported in Table 1. Median prostate volume was mation score. Median PSA basal value in patients 37 cm3 and median t-PSA value 7.3 ng/ml. Ciprofloxacin with grade 0–1 and 2–3 was 6.8 (range: 4.1–17.0) and was well tolerated and treatment interruption for toxicity 8.7 ng/ml (range: 5.0–15.2), respectively (P ¼ 0.44). Simi- was never required. larly, no difference emerged between the PSA reduction The effect of the antibiotic therapy on t-PSA and rate after antibiotics and the grade of inflammation. f-PSA/t-PSA ratio is shown in Table 2. Median PSA reduction was 10.2 and 14.9% in patients

Prostate Cancer and Prostatic Diseases PSA reduction after antibiotics V Serretta et al 150 Table 3 Statistical analysis: PSA variation in relation to histological exam and to baseline characteristics of the study population Characteristics Histological exam P* Subgroups P*

Prostate cancer No prostate cancer PSA reduction PSA increase

patients, n (%) 28 (28.3) 71 (71.7) 59 (59.6) 40 (40.4) Age, median (range) 66.5 (53–83) 65 (48–80) 0.10 65 (54–83) 65 (48–79) 0.24 Basal PSA, median (range) 7.2 (4.4–20.0) 7.3 (4.1–17.5) 0.49 9.0 (4.3–20.0) 5.9 (4.1–14.9) o0.0001 Prostate volume, median (range) 25.6 (19–65) 40 (20–82) 0.02 33 (20–82) 41 (19–82) 0.16 f-PSA/t-PSA ratio, median (range) 0.12 (0.05–0.49) 0.14 (0.01–0.29) 0.39 0.19 (0.02–0.42) 0.12 (0.01–0.29) 0.08 PSA % variation, median (range) 2.5 (À66.6–71.2) À11.2 (À91.4–54.5) 0.02 — — —

Abbreviation: f-PSA/t-PSA ratio, free/total-PSA ratio. *Mann–Whitney U test (one-tail).

with grade 0–1 and 2–3 inflammation, respectively Table 4 Predictive factors of tumour absence at biopsy (P ¼ 0.52). Nevertheless, prostate carcinoma was found Predictive factors Univariate analysis Multivariate analysis in 16 (40%) out of 40 cases with increased or unchanged w2 test (P-value) adjusted OR (95% CI) PSA and in 12 (20.3%) out of 59 patients showing PSA reduction (P ¼ 0.02). In nine patients with PSA lowered t-PSA (o10 vs P ¼ 0.429 0.90 (0.75–1.08) to less than 70% of the basal value and/or decreased X10 ng/ml) below 4 ng/ml (eight patients), no cancer was detected. Patients’age P ¼ 0.396 0.95 (0.89–1.02) X Excluding the 18 patients with initial PSA values above (o65 vs 65 years) t-PSA after therapy P ¼ 0.101 0.90 (0.75–1.08) 10 ng/ml, out of the remaining 81 patients, no cancer was (o4vsX4 ng/ml) detected in six patients with PSA reduction to less PSA reduction (ng/ml) P ¼ 0.471 0.648 (0.453–0.929) than 50% of the basal value and in three more patients (o2vsX2 ng/ml) with PSA decreased below 4 ng/ml (Table 4). Consi- t-PSA reduction (%) P ¼ 0.033 2.611 (1.075–6.516) dering age, number of biopsy cores, prostate volume, 10 1.163 (1.009–1.356) baseline PSA, f-PSA/t-PSA ratio, absolute PSA variation 20 1.352 (1.018–1.840) 30 1.572 (1.027–2.495) and PSA reduction rate, only prostate volume and PSA 40 1.827 (1.036–3.384) reduction rate showed a correlation with the absence of 50 2.124 (1.045–4.590) prostate cancer. However, prostate volume had no 60 2.470 (1.054–6.225) statistical significance at multivariate analysis, as it was 70 2.872 (1.064–8.443) equally distributed between patients with or without 80 3.339 (1.073–11.451) PSA reduction. On the other hand, PSA reduction rate 90 3.882 (1.083–15.530) showed an independent strong correlation with the Abbreviations: CI, confidence interval; t-PSA, total PSA. absence of prostate cancer at biopsy, ORs varying from 1.2 to 3.9 (median 2.6). Particularly, for PSA reduction rates over 50 and 70%, PSA sensitivity was 75 and 100% In our study, the routine histological exam revealed and negative predictive value 77 and 100%, respectively. chronic prostatitis in a small number of specimens. The lack of correlation was probably due to the fact that our patients were asymptomatic and that histology was obtained after antibiotics. On the other hand, when an Discussion extensive pathological examination was performed to score the inflammation, thus also detecting very small The present study analyses the effect of antibiotic foci of chronic prostatitis, 60% of patients showed foci therapy on f-PSA and t-PSA to investigate if a relevant of inflammation, even after antibiotics. In accordance PSA reduction induced by antibiotics could be related to with the findings of Kwak et al.,12 we did not find a decreased cancer detection rate at biopsy. In particular, any statistically significant correlation between the PSA we investigated whether antibiotics lower PSA to normal values and the extent of inflammation in biopsy speci- levels, thus permitting postponement of prostate biopsy mens. Apparently, this observation does not support the until further PSA increase. It has recently been suggested correlation between antibiotics and PSA reduction. that, in managing men with elevated PSA, the decision Nevertheless, patients with PSA reduction after anti- for biopsy based on one single value may be inappropri- biotics had a baseline t-PSA significantly higher than that ate. Prostate biopsy could, in fact, be avoided or of the other patients in spite of the lower incidence of postponed in men with normal DRE who have a normal carcinoma. Although the exact implication of this finding repeated PSA, eventually after a course of antibiotics.8,9 is not completely clear, the higher PSA values could be Minimizing negative prostate biopsies is an important due to the fact that small foci of inflammation could alter goal. Nadler et al.10 demonstrated the presence of chronic prostate integrity more extensively than small foci of inflammatory cells in 64.3% of men undergoing trans- carcinoma. rectal ultrasound biopsy for elevated PSA. Brett et al.11 In our experience, PSA decreased after antibiotics found, in 300 randomly selected men, category IV in almost 60% of patients, returning within normal limits prostatitis prevalence of 32.2% and serum PSA levels in 8% of them. Thus, PSA normalization was far significantly higher (P ¼ 0.0004) than in men without less common than that reported by other authors in prostatitis. patients with increased PSA and laboratory signs of

Prostate Cancer and Prostatic Diseases PSA reduction after antibiotics V Serretta et al prostatitis.4,5,13 After antibiotic treatment, Potts et al.5 in selected patients needing to postpone biopsy and in 151 documented PSA normalization in 42% of 51 patients those patients with previous negative biopsies reluctant and Brett et al.11 in 41% of 86 patients with an increased to undergo a second or a third procedure. In patients PSA and diagnosis of prostatitis. showing PSA decreasing to less than 50% of initial value A major bias of our study is that it is not a randomized or normalizing after antibiotics, PSA could be monitored one. In some patients, serum PSA would have dropped postponing biopsy until further increase. On the other without antibiotics owing to the spontaneous variation in hand, our results are also valid for patients showing serum PSA. Thus, our results could be due to a lower spontaneous PSA variations. incidence of prostate cancer in patients with greater PSA variability. We tried to minimize the risk of PSA variability, including only patients with persistent Conclusions elevated PSA values in at least two determinations before study entry. On the other hand, it must be After antibiotic treatment, the resultant PSA reduction emphasized that our results can be also applied to rate was strongly related to the absence of cancer at patients showing similar spontaneous PSA reduction. biopsy. PSA decreases in 60% and returns within normal Another limitation is that we could not establish the limits in 8% of patients. Eleven per cent of patients with proportion of men, albeit small, who could have a PSA between 4 and 10 ng/ml will show, after antibiotics, clinically detectable chronic prostatitis. Our patients a decrease of PSA value to less than 50% of the initial were not screened for prostatitis as they were asympto- value and/or below 4 ng/ml. The risk of missing a matic. cancer in these patients is low and PSA can be monitored Our results are very similar to those obtained by until further rise. Obviously, antibiotics cannot be Hochreiter et al.8 in 35 patients with NIH category IV administered in all patients to avoid biopsy in a small prostatitis and elevated PSA. They showed, after anti- number of them, but could be given in selected cases bacterial therapy, PSA reduction in 63% of patients, with requiring postponement of prostate biopsy and in PSA returning to normal values in 9% of them, thus patients with previous negative biopsies reluctant to avoiding prostate biopsy. repeat the procedure. In our experience, the f-PSA/t-PSA ratio was not a discriminator between cancer and chronic inflammation, even after a course of antibacterial therapy, as f-PSA Acknowledgements appeared to vary according to t-PSA. Many other studies investigated the correlation between PSA, f-PSA/t-PSA We thank the GSTU (‘Gruppo Studio Tumori Urologici’) ratio and inflammatory disorders of the prostate14–16 Foundation for the kind support for the statistical with contradictory results. Scattoni et al.,16 compared the analysis. pathology results of prostatectomy with the preoperative f-PSA/t-PSA ratio and found a decreased ratio in the presence of histological signs of acute prostatitis. Lorente References et al.17 found significant variations in t-PSA and f-PSA/ t-PSA ratios in patients with benign prostatic hyper- 1 Dalton DL. Elevated serum prostate-specific antigen due to plasia (BPH)-associated prostatitis and not in patients acute bacterial prostatitis. Urology 1989; 33: 465–469. with BPH or cancer. On the contrary, Ornstein et al.18 did 2 Neal Jr DE, Clejan S, Sarma D, Moon TD. Prostate specific not find any change in f-PSA/t-PSA ratio in prostatitis. antigen and prostatitis. I. Effect of prostatitis on serum PSA in 19 the human and nonhuman primate. Prostate 1992; 20: 105–111. Yavascaoglu et al., in accordance with our results, 3 Game` X, Vincendeau S, Palascak R, Milcent S, Fournier R, demonstrated t- and f-PSA reduction but no variation in Houlgatte A. Total and free serum prostate specific antigen f-PSA/t-PSA ratio after antibacterial therapy in patients levels during the first month of acute prostatitis. Eur Urol 2003; with PSA levels between 4 and 10 ng/ml. 43: 702–705. Our study shows a significantly lower cancer detection 4 Hasui Y, Marutsuka K, Asada Y, Ide H, Nishi S, Osada Y. rate in patients with decreased PSA after antibiotic Relationship between serum prostate-specific antigen and therapy, demonstrating a correlation between PSA reduc- histological prostatitis in patients with benign prostatic hyper- tion and negative biopsy, OR varying from 1.2 to 3.9 for plasia. Prostate 1994; 25: 91–96. reduction percentages between 10 and 90%. Out of 99 5 Potts JM. Prospective identification of National Institutes of patients, considering PSA reduction greater than 70% Health category IV prostatitis in men with elevated prostate specific antigen. JUrol2000; 164: 1550–1553. and/or below 4 ng/ml, 9% of prostate biopsies could be 6 Irani J, Levillain P, Goujon JM, Bon D, Dore B, Aubert J. avoided or postponed until further PSA increase, Inflammation in benign prostatic hyperplasia: correlation with missing no prostate tumour. Excluding patients with prostate specific antigen value. JUrol1997; 157: 1301–1303. PSA value greater than 10 ng/ml, a PSA reduction rate of 7 Gerstenbluth RE, Seftel AD, Mac Lennan GT, Rao RN, Corty EW, 50% can be adopted and 11% of biopsies avoided until Ferguson K et al. Distribution of chronic prostatitis in radical further PSA increase. It must be, however, emphasized prostatectomy specimens with upregulation of bcl-2 in areas of that a long follow-up is needed to determine if any of inflammation. JUrol2002; 167: 2267–2270. these men will have prostate cancer in the near future, 8 Hochreiter W, Wolfensberger P, Danuser H, Studer UE. and larger studies are needed to identify the optimal PSA Antibiotic treatment of asymptomatic inflammatory prostatitis in patients with elevated psa: can biopsies be avoided? Eur Urol reduction level to postpone biopsy. 2004; 3 (Suppl 2): 204; Abstract 806. Obviously, a 3-week course of antibiotic treatment is 9 Singh R, Cahill D, Popert R, O’Brien TS. Repeating the an inappropriate and certainly not a cost-effective measurement of prostate-specific antigen in symptomatic approach to reduce the number of prostate biopsies in men can avoid unnecessary prostatic biopsy. BJU Int 2003; 92: unselected cases. However, this approach can be adopted 932–935.

Prostate Cancer and Prostatic Diseases PSA reduction after antibiotics V Serretta et al 152 10 Nadler RB, Humphrey PA, Smith DS, Catalona WJ, Ratliff TL. distinguish patients with prostate cancer from those with Effect of inflammation and benign prostatic hyperplasia on chronic inflammation of the prostate. JUrol1998; 159: 1595–1598. elevated serum prostate specific antigen levels. JUrol1995; 154: 16 Scattoni V, Raber M, Montorsi F, Da Pozzo L, Brausi M, Calori G. 407–409. Percent of free serum prostate-specific-antigen and histological 11 Brett S, Carver BS, Bozeman CB, Williams BJ, Venable DD. The findings in patients undergoing open prostatectomy for benign prevalence of men with National Institutes of Health category IV prostatic hyperplasia. Eur Urol 1999; 36: 621–630. prostatitis and association with serum prostate specific antigen. 17 Lorente JA, Arango O, Bielsa O, Cortadellas R, Gelabert-Mas A. JUrol2003; 169: 589–591. Effect of antibiotic treatment on serum PSA and percent free PSA 12 Kwak C, Ku JH, Kim T, Park DW, Choi KY, Lee E et al. Effect of levels in patients with biochemical criteria for prostate biopsy subclinical prostatic inflammation on serum PSA levels in men and previous lower urinary tract infections. Int J Biol Markers with clinically undetectable prostate cancer. Urology 2003; 62: 2002; 17: 84–89. 854–859. 18 Ornstein DK, Smith DS, Humprey PA, Catalona WJ. The effect of 13 Bozeman CB, Carver BS, Eastham JA, Venable DD. Treatment of prostate volume, age, total prostate specific antigen level and chronic prostatitis lowers serum prostate specific antigen. JUrol acute inflammation on the percentage of free serum prostate 2002; 167: 1723–1725. specific antigen levels in men without clinically detectable 14 Stancik I, Luftenegger W, Klimpfinger M, Muller MM, Hoeltl W. prostate cancer. JUrol1998; 159: 1234–1237. Effect of NIH-IV prostatitis on free and free-to-total PSA. Eur 19 Yavascaoglu I, Vuruskan H, Kordan Y, Cahskan Z, Oktay B. Urol 2004; 46: 760–764. Effect of chronic inflammation of the prostate on ratio of 15 Jung K, Meyer A, Lein M, Rudolph B, Schnorr D, Loening SA. free-to-total prostate specific antigen in serum. Eur Urol 2004; Ratio of free-to-total prostate specific antigen in serum cannot 3 (Suppl 2): 204; Abstract 805.

Supplementary Information accompanies the paper on the Prostate Cancer and Prostatic Diseases website (http:// www.nature.com/pcan)

Prostate Cancer and Prostatic Diseases