A new diagnostic strategy for prostate cancer


Editor's Introduction

Urine TMPRSS2:ERG fusion transcript stratifies prostate cancer risk in men with elevated serum PSA

annotated by

During their lifetime, 18% of all American men will get prostate cancer, a slow-growing cancer which rarely causes symptoms. Because both surgery and radiotherapy commonly cause harmful side effects, it is important that only patients with aggressive cancers receive these treatments. Currently, the diagnosis of prostate cancer is based on blood samples to detect prostate specific antigen (PSA), followed by biopsy and microscopy of tissue. However, these techniques are not very accurate and over-diagnosis of prostate cancer is common. Many men receive treatment even though their cancer will never cause disease, exposing the men to the associated side effects of treatment. The need for more accurate screening and diagnostic methods is high. In this paper, the authors report a novel test based on changes in the chromosome seen in most prostate cancers.

Paper Details

Original title
Urine TMPRSS2:ERG fusion transcript stratifies prostate cancer risk in men with elevated serum PSA
Scott Tomlins
Original publication date
Vol. 3, Issue 94, p. 94ra72
Issue name
Science Translational Medicine


More than 1,000,000 men undergo prostate biopsy each year in the United States, most for "elevated" serum prostate-specific antigen (PSA). Given the lack of specificity and unclear mortality benefit of PSA testing, methods to individualize management of elevated PSA are needed. Greater than 50% of PSA-screened prostate cancers harbor fusions between the transmembrane protease, serine 2 (TMPRSS2) and v-ets erythroblastosis virus E26 oncogene homolog (avian) (ERG) genes. Here, we report a clinical-grade, transcription-mediated amplification assay to risk stratify and detect prostate cancer noninvasively in urine. The TMPRSS2:ERG fusion transcript was quantitatively measured in prospectively collected whole urine from 1312 men at multiple centers. Urine TMPRSS2:ERGwas associated with indicators of clinically significant cancer at biopsy and prostatectomy, including tumor size, high Gleason score at prostatectomy, and upgrading of Gleason grade at prostatectomy. TMPRSS2:ERG, in combination with urine prostate cancer antigen 3 (PCA3), improved the performance of the multivariate Prostate Cancer Prevention Trial risk calculator in predicting cancer on biopsy. In the biopsy cohorts, men in the highest and lowest of threeTMPRSS2:ERG+PCA3 score groups had markedly different rates of cancer, clinically significant cancer by Epstein criteria, and high-grade cancer on biopsy. Our results demonstrate that urineTMPRSS2:ERG, in combination with urine PCA3, enhances the utility of serum PSA for predicting prostate cancer risk and clinically relevant cancer on biopsy.


Although the use of serum prostate-specific antigen (PSA) to screen for prostate cancer is widespread clinically(1), PSA has several limitations as an early detection biomarker. PSA is highly specific for tissue of prostatic origin, but is not cancer-specific. Moreover, serum levels are frequently elevated in benign conditions. Currently, whereas most men undergo needle biopsy when levels of PSA are more than 4.0 ng/ml, less than half of these biopsies result in a diagnosis of prostate cancer(23). PSA also has sensitivity limitations, as shown by the Prostate Cancer Prevention Trial (PCPT), which demonstrated that 15% of men with PSA of 0 to 4.0 ng/ml have prostate cancer, of which 15% have high Gleason grade disease (45). Despite the development of multivariate models multivariate models, such as the PCPT risk calculator that incorporates PSA and other clinical factors in an attempt to provide an individual risk estimate (6), men are commonly referred for biopsy in the United States on the basis of their serum PSA concentration alone.

Moreover, screening with PSA has probably led to the overdiagnosisof prostate cancer—an estimated 23 to 44% of all screening-detected cancers would never have caused symptoms (7)—and overtreatment. Two randomized trials evaluating the effect of PSA screening on prostate cancer mortality showed that during the first decade of follow-up, PSA screening has a modest effect on prostate cancer mortality, with substantial risks of negative biopsy and overdiagnosis and overtreatment of indolent cancer (cancer that would not cause symptoms in a lifetime) (38). Together, these results highlight the limitations of the current serum PSA–based paradigm of prostate cancer early detection. Recognizing these limitations, several groups, including the United States Preventative Task Force, the American Cancer Society, and the American Urological Association, have advocated for individualized decision making between a patient and his physician regarding PSA screening and/or proceeding to biopsy (9). Biomarkers to assist this process, however, are lacking.

Several modifications of serum PSA, including free PSA, rate of PSA change (PSA velocity), various PSA isoforms, and related proteins, have been proposed as prostate cancer biomarkers that can be used to help PSA-screened men make more informed decisions about proceeding to biopsy (10). These strategies, however, rely on surrogate biomarkers that are tissue-specific and not intrinsically cancer-specific. An alternative approach is to develop clinically robust assays for cancer-specific biomarkers that have been identified through genomic and transcriptomic studies (11).

Recently, chromosomal rearrangementswere identified in prostate cancer that fuse the 5′ untranslated region of the androgen-regulated gene transmembrane protease, serine 2 (TMPRSS2) with v-ets erythroblastosis virus E26 oncogene homolog (avian) (ERG) or ets variant 1 (ETV1)ERG and ETV1 are both members of the erythroblastosis virus E26 transformation-specific (ETS)transcription factor family (12). Subsequent studies confirmed ETS gene fusions in about 50% of PSA-screened prostate cancers (13). Fusions between TMPRSS2 and ERG, which result in a truncated ERG protein product, represent about 90% of all ETS gene fusions (13). Fusion of TMPRSS2 and ERG loci at the chromosomal level [as detected byfluorescence in situ hybridization (FISH)] and subsequent overexpression of the TMPRSS2:ERG transcript and truncated ERG protein product are essentially 100% specific for the presence of prostate cancer in tissue-based studies (13–15). Additionally, multiple studies have demonstrated that TMPRSS2:ERG fusions are only detectable in about 15% of high-grade prostatic intraepithelial neoplasia (PIN) lesions, invariably adjacent to fusion-positive cancer (1618). In vitro and in vivo functional studies have also demonstrated a functional role for TMPRSS2:ERG fusions in prostate cancer oncogenesis(13171920). Together, TMPRSS2:ERG gene fusions are highly specific biomarkers that define a distinct molecular subtype of prostate cancer.

The protein product of the TMPRSS2:ERG fusion is neither chimeric nor known to be secreted, which precludes the possibility of antibody-based detection in serum (as for PSA). However, a clinical-grade, urine-based assay for the noncoding transcript prostate cancer antigen 3 (PCA3) [a prostate-specific gene overexpressed in greater than 95% of prostate cancers (21)] has been developed and has proven useful as an adjunct to serum PSA for prostate cancer detection (2223). In addition, research-grade, reverse transcription–polymerase chain reaction (RT-PCR)–based assays have shown that TMPRSS2:ERG mRNA is indeed detectable in urine (2428).

To translate these findings to clinical practice, we have developed a clinical-grade, transcription-mediated amplification (TMA) assay for quantifying TMPRSS2:ERG mRNA, which is normalized to the amount of PSAmRNA by TMA (which controls for the abundance of prostate cells and prostate mRNA) to generate a TMPRSS2:ERG “score.” The assay is based on the same technology as the PCA3 assay. We tested prospectively collected, post–digital rectal exam (DRE) urine from men presenting for biopsy and/or prostatectomy. We then correlated urine TMPRSS2:ERG levels with clinicopathologic features, including indicators of clinically significant cancer. We also measured PCA3 in the same urine specimens and report the combined performance of TMPRSS2:ERG and PCA3 for prostate cancer risk stratification of PSA-screened men. This work represents an initial step in using a panel of cancer-specific biomarkers for early detection of prostate cancer.

TMPRSS2:ERG assay accuracy

To assess the accuracy of the TMPRSS2:ERG TMA assay, we compared it to FISH for ERG rearrangement on 208 needle biopsy cores from 122 patients in the University of Michigan Health System (UMHS) biopsy cohort. On the basis of sufficient PSA expression (PSA copies/ml >3000), 206 of 208 (99%) cores were informative for further TMPRSS2:ERG evaluation. As shown in table S1, 40 of 79 (51%) informative cancerous cores were positive for ERG rearrangement by FISH; all informative benign cores evaluated by FISH (n = 72) were negative. By TMA, 41 of 79 (52%) cancerous and 5 of 127 (4%) benign cores were positive for TMPRSS2:ERG, with an overall concordance between FISH and TMA in informative cores of 92%. These results further support the cancer specificity of TMPRSS2:ERG and demonstrate the accuracy of the TMA assay.

Urine TMPRSS2:ERG in men undergoing prostatectomy

To assess associations between urine TMPRSS2:ERG and clinicopathological parameters in prostate cancer, we first examined a cohort of 218 men presenting for radical prostatectomy at UMHS. This cohort was selected because prostatectomy provides a more complete assessment of tumor pathology than biopsy, and offers the opportunity to compare biopsy and prostatectomy pathology (see fig. S1 for flow chart of all urine samples in the study). Of 218 men, 187 (86%) had informative urine, based on sufficient urine PSA expression (average PSAcopies/ml >15,000), for further TMPRSS2:ERG score evaluation. Clinicopathological characteristics for informative men are shown in Table 1. Consistent with the notion that biopsy often undersamples disease burden, 35 of 70 (50%) informative men with biopsy Gleason 6 disease were upgraded to Gleason 7 at prostatectomy, similar to previous reports (29).


Table 1. Clinicopathological data, including parameters at preceding biopsy and associations with TMPRSS2:ERG score for informative prostatectomy (RRP) patients (n = 187). The number of patients with data for each parameter is given. (Top) Median with interquartile range (IQR) for each parameter, Spearman's rho (rs), and P value (Spearman rank correlation) for the correlation of each parameter with urine TMPRSS2:ERG score. (Bottom) Number (with percentage) for each parameter, median TMPRSS2:ERG score with 95% confidence interval (CI), and P values from Wilcoxon rank-sum tests. pT, pathologic tumor.

Table 1

This table presents data from men who had their prostate surgically removed (academic prostatectomy cohort). By including these patients in the study, the authors were able to correlate the TMPRSS2:ERG score with the correct Gleason grading of the cancer and tumor extent.

What is a bioposy?

A biopsy involves removing a sample of cells or tissue from your body.  The cells/tissue are then sent to a laboratory for further testing.


Gleason grading

When the cells/tissue from a biopsy of the prostate are analyzed by a pathologist, they can be assigned a Gleason grade which correlates to the microscopic tumor patterns of their biopsy specimen. In this case, grade does not refer to the percentage you get right on a test, rather it refers to the complexity of the pattern of cells the pathologist sees under the microscope.


How to interpret the Table

When only using biopsies for grading, there is always a risk of getting a sample that is too small and not representative of the whole tumor. In order to appreciate this table, it is essential to have a basic understanding of statistical tests and P values.

If two parameters correlate, e.g., one is elevated when the other is similarly so, this can be shown using a test like Spearman's correlation test.

AP value < 0.05 is regarded as statistically significant, meaning that the observed relationship is unlikely to be due to chance. As you can see, parameters such as number of biopsies with cancer, cancer involvement in biopsy core, high clinical/tumor stage (a cancer grading system), and Gleason score are all associated with higher TMPRSS2:ERG score.

In contrast, parameters like age, race, prostate weight (although almost significant), and time to prostatectomy did not correlate to TMPRSS2:ERG score.

Interestingly, PSA value did not correlate to TMPRSS2:ERG score.

Questions for this Table

Prostate cancer is more common in the United States, Europe, the Caribbean, and Australia, compared to the rest of the world. Rates of prostate cancer can vary up to 30 fold in different countries. Why do you think that is? And why was there no difference observed in race in this study?

An ideal early detection biomarker not only would distinguish patients with and without cancer but would also be associated with clinically significant cancer. At prostatectomy, significant cancer is most commonly defined on the basis of tumor size, high Gleason score (>6), and non–organ-confined disease (pathologic tumor >2). Similarly, models, which commonly include tumor size and Gleason score, have been developed for defining significant cancer in men with positive biopsies (30). In prostatectomy patients, TMPRSS2:ERG score was positively associated with markers of tumor volume, including number of positive cores and maximum percentage of tumor involvement of a single core in the preceding biopsy (Table 1). TMPRSS2:ERG score was also positively associated with maximum tumor dimension at prostatectomy, but was not associated with prostate weight, serum PSA, or PSA density (PSAD) at prostatectomy (Table 1).

TMPRSS2:ERG score was significantly higher in men with high prostatectomy Gleason score (>6 versus 6) and was associated with upgrading at prostatectomy from biopsy (6 to >6 versus 6 to 6) (Table 1). Finally, consistent with associations with indicators of significant pathology at biopsy and prostatectomy, TMPRSS2:ERG score was significantly higher in men with significant versus insignificant cancer at prostatectomy and preceding biopsy.


Table 2. Associations between urine TMPRSS2:ERG score and clinicopathological parameters among informative academic biopsy patients (n = 606).The number of patients with data for each parameter is given. (Top) Spearman’s rho (rs) and P value (Spearman rank correlation) for the correlation of each parameter with urine TMPRSS2:ERG score. (Bottom) Median TMPRSS2:ERG score with 95% CI and P values from Wilcoxon rank-sum tests.

Table 2

The men included in this table have been referred for needle biopsy at three academic centers (the academic biopsy cohort).

Out of the 606 men, 44% were diagnosed with prostate cancer, and these men had higher TMPRSS2:ERG score compared to the men who did not have cancer. Same as in Table 1, the TMPRSS2:ERG score in this group was associated with number and percentage of biopsy core with cancer. It did, however, increase with age.

Important though, after grading the tumors that were found, there was no correlation observed between Gleason score, clinical/tumor stage, and TMPRSS2:ERG score.

Questions for this Table

When taking prostate biopsies, what is the purpose of taking more than one sample/core?

How many cores are usually obtained?

How many do you think are sufficient?

How much of the prostate does a single core sample?

Urine TMPRSS2:ERG in an academic biopsy cohort

Of the 623 men in the academic biopsy cohort [combined UMHS, University of Laval (UL), and the Veterans Administration San Diego (VASD) Medical Center cohorts], 606 (97%) had informative urine based on sufficient urine PSA expression (average PSA copies/ml >15,000) for TMPRSS2:ERG score evaluation. Prebiopsy clinicopathological characteristics for these men are shown in table S2. On biopsy, 269 of the 606 (44%) informative men were diagnosed with prostate cancer. TMPRSS2:ERG was higher in men diagnosed with cancer compared to those with noncancerous diagnoses (Table 2). TMPRSS2:ERG score was positively associated with direct markers of tumor volume, including maximum percentage of tumor involvement of a single core and number of positive cores (Table 2). TMPRSS2:ERG score was significantly higher in men with significant versus insignificant cancer on biopsy (Table 2). TMPRSS2:ERG score was correlated with increasing age (Table 2) and was higher in men at initial biopsy compared with repeat biopsy, but was not significantly correlated or associated with other prebiopsy clinicopathological parameters, including indicators of prostate size (Table 2).

Urine TMPRSS2:ERG in a community biopsy cohort

To demonstrate the utility of urine TMPRSS2:ERG outside of academic medical centers, we also evaluated a community biopsy cohort. In this cohort, owing to ongoing assay optimization, we used a second-generation TMPRSS2:ERG TMA assay (which uses TMPRSS2:ERG primers different from that of the initial assay). Of 471 men in the community biopsy cohort, 463 (98%) had informative urine based on sufficient urine PSA expression (average PSA copies/ml >10,000) for TMPRSS2:ERG score evaluation, using the second-generation assay. Prebiopsy clinicopathological characteristics for these men are shown in table S3.

On biopsy, 204 of the 463 (44%) informative men in the community biopsy cohort were diagnosed with prostate cancer. TMPRSS2:ERG score was higher in informative men diagnosed with cancer compared to those with noncancerous diagnoses (Table 3). In informative men with cancer, TMPRSS2:ERG score was positively associated with maximum tumor involvement of a single core, percentage of total involvement, and number of cores positive (Table 3), consistent with findings in the prostatectomy and academic biopsy cohorts. Finally, TMPRSS2:ERG score was higher in men with Gleason score >6 versus 6 (Table 3). All associations between clinicopathological parameters in patients with benign and cancerous biopsies and TMPRSS2:ERG and PCA3 scores are given in Table 3.


Table 3. Associations between urine TMPRSS2:ERG score and clinicopathological parameters among informative community biopsy patients (n = 463). The number of patients with data for each parameter is given. (Top) Spearman’s rho (rs) and P value (Spearman rank correlation) for the correlation of each parameter with urine TMPRSS2:ERG score. (Bottom) MedianTMPRSS2:ERG score with 95% CI and P values from Wilcoxon rank-sum tests.

Table 3

Among men that had been referred to a community hospital (community cohort), 44% were diagnosed with cancer after biopsy.

Also in this group, men with cancer who had a higher number of positive biopsy cores and the percentage of total involvement had a higher TMPRSS2:ERG score.

In this group, however, TMPRSS2:ERG score also correlated to Gleason score, i.e., patients with more aggressive tumors had a higher TMPRSS2:ERG score.

Questions for this Table

Why do you think there is a difference in correlation to Gleason score in this group compared with the groups presented in Table 1 and 2?

TMPRSS2:ERG score for individualizing cancer risk in men with elevated serum PSA undergoing biopsy

On the basis of associations with the presence of cancer and significant pathology, we explored several clinically applicable models for using urine TMPRSS2:ERG to individualize prostate cancer risk in PSA-screened men presenting for biopsy. We first compared receiver operating characteristic (ROC) curves for TMPRSS2:ERG score and serum PSA. Among men with informative urine (where urine PSA expression was sufficient for TMPRSS2:ERG score evaluation) with measured serum PSA, TMPRSS2:ERG had significantly increased area under the curve (AUC; the probability that a classifier will rank a randomly chosen positive instance higher than a randomly chosen negative one) compared to serum PSA in the academic biopsy cohort (n = 606; 0.71 versus 0.61) (fig. S2A). In the community cohort, although TMPRSS2:ERG had increased AUC compared to serum PSA (particularly at high specificity), this difference was not significant (n = 460; 0.65 versus 0.59) (fig. S2B).

Combined TMPRSS2:ERG and PCA3 scores for predicting the presence of prostate cancer on biopsy

Although nearly 100% specific for the presence of prostate cancer in tissue-based studies, TMPRSS2:ERG gene fusions occur in only ~50% of PSA-screened prostate cancers. Previous studies using research-grade assays have demonstrated that measuring both PCA3 and TMPRSS2:ERG in urine outperforms either marker alone for predicting the presence of prostate cancer on biopsy (2425). Thus, to optimize performance, we analyzed the performance of urine TMPRSS2:ERG in combination with PCA3 (both measured using TMA assays from the same urine sample). PCA3 was available for all evaluable men in both the academic prostatectomy and the biopsy cohorts, and in 459 of 463 (99%) of men in the community biopsy cohort. Among informative men with measured TMPRSS2:ERGPCA3, and serum PSA, the TMPRSS2:ERG+PCA3 score had a significantly increased AUC compared to serum PSA in both the academic (Fig. 1A) and the community biopsy cohorts (Fig. 1B).


Fig. 1.  Urine TMPRSS2:ERG+PCA3 score for the prediction of prostate cancer in men undergoing needle biopsy. (A) ROC curves for serum PSA (red) and TMPRSS2:ERG+PCA3 score (blue) for predicting prostate cancer on informative academic biopsy patients with evaluable serum PSA and PCA3 score (n = 606). (B) As in (A), except for all informative community biopsy patients (n = 456). (C) A model incorporating TMPRSS2:ERG+PCA3 score into the PCPT risk calculator for predicting prostate cancer on biopsy (informative men without a previous diagnosis of cancer who had all required PCPT risk calculator information and PCA3 score) was trained on VASD and UL biopsy patients (n = 202) and tested on the UMHS EDRN biopsy patients (n = 261). ROCs for the PCPT risk calculator alone (red) and the PCPT +TMPRSS2:ERG+PCA3 score (blue) for UMHS testing patients are shown. (D) As in (C), except a model was fitted for all informative community biopsy patients (n = 436). AUC and P values for comparison of the curves (DeLong’s test) are given for all panels.

Figure 1

The authors report a new test, the TMPRSS2:ERG score, but have the ambition of increasing its accuracy by combining the assessment with Prostrate Cancer Antigen 3 (PCA3).

Figure 1 A and B.

In Fig. 1 A and 1 B, they compare this combination with the PSA test in the academic cohort (1A) and the community cohort (1B). What is measured in a ROC curve is the sensitivity (Y axis) and specificity (X axis). The higher these two factors are, the greater the area under the curve is. As you can see, there is significant difference in the two tests (P < 0.001), i.e.,TMPRSS2:ERG + PCA3 is better to predict cancer than PSA.

Further on, a comparison was done between TMPRSS2:ERG score and the PCPT risk calculator, showing the same advantage to the TMPRSS2:ERG score.

Questions for this Table

What would a ROC curve for an ideal diagnostic test look like? Please draw an example.

We next assessed the ability of TMPRSS2:ERG+PCA3 score to increase the AUC of the multivariate PCPT risk calculator for predicting prostate cancer diagnosis on biopsy. Among informative men in the academic and community biopsy cohorts, 463 and 439, respectively, had not been diagnosed with prostate cancer previously and the clinical information required for the PCPT risk calculator was available. The PCPT risk calculator showed significantly increased AUC compared to serum PSA alone in both the academic (n = 463; 0.64 versus 0.60) and the community (n = 439; 0.66 versus 0.61) cohorts (fig. S3). To evaluate the performance of TMPRSS2:ERG in combination with PCA3 in the academic cohort, we used the subset of UL and VASD patients (n = 202) to train a model incorporating TMPRSS2:ERG+PCA3 score into the PCPT risk calculator (see Materials and Methods). We then tested this model on the remaining UMHS patients (n = 261). In the independent UMHS testing set, incorporation of TMPRSS2:ERG+PCA3 score significantly increased the AUC of the PCPT risk calculator (Fig. 1C). To explore the potential clinical benefit of incorporating TMPRSS2:ERG+PCA3 score into the PCPT risk calculator, we performed decision curve analysis (31) on the same modeled data. Incorporation of TMPRSS2:ERG+PCA3 into the PCPT risk calculator resulted in net benefit across all threshold probabilities (the probability of cancer on biopsy at which an individual would choose to undergo biopsy) from ~15 to 90% (fig. S4), thus demonstrating improved clinical outcome.

Among informative men in the community biopsy cohort, we were able to perform the PCPT risk calculation and derive a PCA3 score for 436 men. Incorporation of TMPRSS2:ERG+PCA3 score significantly increased the AUC of the PCPT risk calculator (Fig. 1D). Similarly, we also explored the ability of TMPRSS2:ERG+PCA3 score to improve an alternative nomogram (which requires free PSA) for predicting cancer on initial biopsy developed by Chun et al. (32). Although this nomogram had increased AUC compared to serum PSA alone, this difference was not statistically significant (0.71 versus 0.64) (fig. S5A). Nevertheless, incorporation of TMPRSS2:ERG+PCA3 into the nomogram resulted in significantly increased AUC compared to the nomogram alone (0.77 versus 0.71) (fig. S5B). The effect of incorporating TMPRSS2:ERG score alone into the PCPT risk calculator is shown in fig. S2, C and D.

TMPRSS2:ERG+PCA3 score groups stratify prostate cancer risk on biopsy

We next evaluated whether stratifying patients on the basis of TMPRSS2:ERG and PCA3 scores could individualize prostate cancer risk in men with elevated serum PSA. Because the TMPRSS2:ERG assays used in the academic biopsy and prostatectomy (initial assay) and community biopsy (second generation) cohorts were different, direct comparison of TMPRSS2:ERG scores was not possible (PCA3 scores were directly comparable across all cohorts). We therefore applied a quartile-based approach using TMPRSS2:ERG and PCA3 scores in our prostatectomy cohort (where pathology was better defined than in the biopsy cohort) to identify cutoffs that binned men into three distinct TMPRSS2:ERG+PCA3 score groups (lowest, intermediate, or highest) in each biopsy cohort. As shown in Table 4, 363 (34%), 346 (32%), and 356 (33%) of 1065 men in the combined biopsy cohorts were in the lowest, intermediate, and highest TMPRSS2:ERG+PCA3 score groups, respectively. Biopsy resulted in a cancer diagnosis in 21%, 43%, and 69% of men in the lowest, intermediate, and highest groups, respectively. Moreover, 7%, 20%, and 40% of men in the respective lowest, intermediate, and highest groups were diagnosed with Gleason score >6 cancer (Table 4). Of the 966 men with required clinical information for determining Epstein criteria for significance of cancer on biopsy (see Methods), 15%, 33%, and 61% of men in the lowest, intermediate, and highest groups, respectively, had Epstein criteria–defined significant cancer (table S4). Results for the individual biopsy cohorts are also shown in Table 4 and table S4.


Table 4. Risk of cancer (and high-grade cancer) on biopsy based on stratification byTMPRSS2:ERG+PCA3 score. Number (n) of patients in groups defined byTMPRSS2:ERG+PCA3 scores and the number with cancer or high-grade cancer (Gleason >6) in each group are given. Risk ratios with 95% CI between highest and lowestTMPRSS2:ERG+PCA3 groups, and Fisher’s exact test P values are shown.

Table 4

Based on TMPRSS2:ERG score, the authors divided the patients into three groups (lowest, intermediate, highest).

As you can see in this table, patients in the highest group had the highest probability to have cancer, and more men in this group had a high Gleason score (>6). The opposite was true for the lowest group.

Risk ratio levels describe the difference in the risk of getting cancer and/or high Gleason score if you are in the highest versus the lowest group.

A 95% confidence interval (CI) indicates that the true value is expected to be with the confidence interval values 95 out of 100 times. The authors want to increase the relevance of their study by showing that this is true in many patients' groups, i.e., seen in both academic and community hospitals.

Questions for this Table

Why did the authors compare groups instead of comparing actual TMPRSS2:ERG + PCA3 scores?

TMPRSS2:ERG+PCA3 score groups compared to PCPT risk calculator

These results demonstrate that men stratified by TMPRSS2:ERG+PCA3 scores have markedly different risks of cancer, high-grade cancer, and clinically significant cancer on biopsy. To be useful clinically, however, TMPRSS2:ERG+PCA3 score–based stratification should add to currently used clinicopathological information. Thus, we compared PCPT risk calculator and PCPT high-grade risk calculator scores for men in the lowest, intermediate, and highest TMPRSS2:ERG+PCA3 score groups from the biopsy cohorts. As shown in Fig. 2, calculated PCPT risks of cancer (or Gleason score >6 cancer) differed markedly from the actual risks observed in the lowest and highest TMPRSS2:ERG+PCA3 score groups of men. For example, academic patients in the lowest TMPRSS2:ERG+PCA3 score group had a 20% actual risk of prostate cancer on biopsy, but of men evaluable for PCPT risk score calculation, 95% had PCPT risk scores greater than 20%; similarly, 90% of PCPT-evaluable men in the highest TMPRSS2:ERG+PCA3 score group had PCPT risk scores less than their actual risk of cancer on biopsy (72%) (Fig. 2A). Results were similar with regard to high-grade cancer (Gleason score >6) risk in the academic cohort (Fig. 2C), and in the community biopsy cohort for cancer (Fig. 2B) and high-grade cancer risk (Fig. 2D).


Fig. 2.  Comparison of Prostate Cancer Prevention Trial (PCPT) risk calculations for cancer (or high-grade cancer) risk on biopsy to actual rates in men stratified byTMPRSS2:ERG+PCA3 scores. (A to D) Men in the academic (A and C) or community (B and D) biopsy cohorts were assigned to lowest, intermediate, or highest TMPRSS2:ERG+PCA3 score groups and actual risks of cancer, or high-grade cancer, were determined for each group (green line). Calculated PCPT risk scores for cancer (A and B), or high-grade cancer (C and D), are plotted for all men with PCPT calculator required clinical information. The median of PCPT risk scores for men in each TMPRSS2:ERG+PCA3 score group is indicated by the blue line. Men with noncancerous and cancerous biopsies (A and B), or noncancerous/low-grade cancer and high-grade cancer on biopsy (C and D), are indicated by black and red points, respectively.

Figure 2

In fig. 2, the three groups based on TMPRSS2:ERG + PCA3 score levels are compared to the PCPT risk calculator.

PCPT did not correlate to the actual risk of getting cancer (2A and 2B), or getting a cancer with high grade (2C and 2D).

If you compare the blue and green lines, you can see that they do not match at all; i.e., PCPT does not measure what clinicians might think they are. The TMPRSS2:ERG + PCA3 score better corresponds to disease presence and severity of disease.

Questions for this Table

What kind of critique do you think this study will get? Is this a test we can start using right away? If not, why not? What questions would you like to ask the scientists in order to evaluate the study?

We report the development of a quantitative, urine-based TMA assay for TMPRSS2:ERG, which is associated with indicators of significant cancer at biopsy and prostatectomy. We demonstrate the utility of this assay alone and in combination with PCA3 expression in urine for individual risk stratification of serum PSA–screened men presenting for prostate biopsy in two multicenter cohorts. Development, validation, and clinical implementation of early detection biomarkers for prostate cancer, particularly specific markers of “aggressive” cancer, are complicated by several aspects of prostate cancer biology and current clinical management. First, the gold standard for detecting prostate cancer based on a biomarker, needle biopsy, is not image-guided and is limited by substantial undersampling. This has been demonstrated by ex vivo studies on prostates at autopsy, which showed that extended (≥10 cores) and saturation biopsy schemes miss about 40% of all cancers (including ~20% of significant cancers) detectable when the entire prostate is pathologically examined (3334). Second, given the high prevalence of indolent prostate cancer [~30 to 40% among men aged 50 to 80 years (35)], a 100% specific and sensitive test for prostate cancer is unlikely to be achieved and clinically undesirable. Third, localized prostate cancer is often multifocal—with heterogeneous TMPRSS2:ERG status between individual foci (13)—and individual foci cannot currently be followed definitively by imaging or biopsy. Therefore, little is known about the natural history of “aggressive” prostate cancer; furthermore, it is unclear whether an ideal biomarker would need to detect tiny foci of high-grade aggressive cancers (which are rarely detected with current PSA screening and biopsy paradigms). Alternatively, if aggressive cancers progress from low-grade cancers, the ideal biomarker would identify foci of low-grade cancer that might require acquisition of additional aberrations, such as those that might only develop in disseminated cells years after supposedly curative treatment. Finally, defining cutoffs or algorithms for reporting and subsequent clinical decision making will be a balance between sensitivity and specificity, which is complicated in prostate cancer by the above factors.

To increase sensitivity for predicting biopsy outcome, we combined TMPRSS2:ERG with the noncoding transcript PCA3 for individual risk stratification. We envision the first step toward a more specific early detection strategy for prostate cancer to be improved management of men with elevated serum PSA. We have shown that in multiple cohorts, elevated urine TMPRSS2:ERG is associated with the presence of prostate cancer and features of aggressiveness. Moreover, men with extremes of TMPRSS2:ERG+PCA3 have different risks of cancer (and high-grade cancer) on biopsy. Thus, in combination with other clinical parameters, urine TMPRSS2:ERG+PCA3 may be useful initially in guiding the urgency of biopsy after the detection of elevated serum PSA.

Urine TMPRSS2:ERG+PCA3 score may have additional utility for common scenarios encountered in the early diagnosis of prostate cancer. For example, men in the highest TMPRSS2:ERG+PCA3 score group with a negative biopsy might benefit from close follow-up, because their chances of having cancer are high (table S5). Similarly, men in the highest TMPRSS2:ERG+PCA3 score group enrolling in active surveillance may wish to consider more extensive biopsy, because their risk of having significant disease that was undersampled on initial biopsy is high. Inclusion of other ETS fusions or aberrantly expressed genes that define additional molecular prostate cancer subtypes, such as SPINK1 (25), into a multiplexed panel, may improve accuracy. Last, in light of increasing evidence that ETS fusion–positive and ETS fusion–negative cancers are distinct molecular subtypes with different biological behavior and response to therapy (1336), the clinically robust assay described here might have potential prognostic utility.

A limitation of this study is that most patients in the study (>85%) were Caucasian, and additional studies will be required to determine whether our findings extend to all men. Furthermore, because men in this study were PSA-screened (and elected to undergo biopsy), studies in non–PSA-screened cohorts will be required to determine the potential utility of TMPRSS2:ERG+PCA3 in that clinical context.

In summary, we report a novel TMA-based urine assay for TMPRSS2:ERG, a highly specific tissue biomarker for prostate cancer that also has a documented role in driving prostate cancer tumorigenesis. We demonstrate that urine TMPRSS2:ERG score is associated with the presence of cancer, tumor volume, and clinically significant cancer in prostatectomy and biopsy patients. The combination of TMPRSS2:ERG and PCA3 improves on the multivariate PCPT risk calculator, suggesting utility for risk stratification beyond currently measured clinicopathological parameters. Last, we demonstrate that stratification by TMPRSS2:ERG+PCA3 scores identifies groups with markedly different risks of cancer, high-grade cancer, and clinically significant cancer on biopsy. Given the uncertainty currently surrounding the utilization of serum PSA for the early detection of prostate cancer, urine TMPRSS2:ERG in combination with PCA3 score may provide an opportunity to help men and their physicians make more informed decisions about early detection, biopsy, and management of prostate cancer in the context of elevated serum PSA.


Patient cohorts

Post-DRE urine was prospectively collected with informed consent from 623 men referred for needle biopsy at three academic centers: the UMHS (n = 317); the UL, Quebec, Canada (n = 213); and the VASD Medical Center (n = 93). Prospective post-DRE urine was also collected with informed consent from 471 men presenting for needle biopsy at seven community clinics throughout the United States. Specimens from UL and VASD were collected from July 2006 to October 2008. Specimens from UMHS were collected from March 2008 to June 2009 and are part of an Early Detection Research Network (EDRN) biopsy cohort. Specimens from the community clinics were collected from August 2007 to June 2008. Post-DRE urine was prospectively collected with informed consent from 218 men before radical prostatectomy at UMHS, including 48 men who had their diagnostic biopsy performed at UMHS (and are also included in that biopsy cohort) and 170 men who presented for prostatectomy. See fig. S1 for a flow chart of all urine specimens.

Clinicopathological information was determined at each academic institution, needle biopsy pathology was reviewed at UMHS for all prostatectomy patients, and a central pathologist reviewed needle biopsy pathology from the community clinics. Men with previous treatment for prostate cancer (prostatectomy, radiation, hormone therapy, and chemotherapy), men with surgical treatment of the prostate within 6 months of urine collection (or previous biopsy within 6 weeks), men taking 5α-reductase inhibitors or testosterone within 3 months of urine collection, or men with prostatitis at the time of urine collection were excluded. For the prostatectomy cohort, men whose urine was collected >6 months before prostatectomy were excluded.

TMA of TMPRSS2:ERG in urine specimens

Urine specimens were obtained immediately after DRE (involving three sweeps of each lateral prostate lobe), refrigerated, and processed within 4 hours by mixing with an equal volume of urine transport medium (detergent-based stabilization buffer; PROGENSA PCA3 Urine Specimen Kit, Gen-Probe Inc.) and stored above −70°C until analysis. Samples not processed according to protocol were excluded. Amounts of TMPRSS2:ERGa and PSA mRNA (the latter to control for the abundance of prostate cells and prostate mRNA in the urine) were determined by TMA, an isothermal nucleic acid amplification method (23). Briefly, in separate procedures, TMPRSS2:ERGa and PSA mRNAs were captured from processed urine specimens through hybridization to oligonucleotide complements coupled to magnetic microparticles. The targets were amplified by TMA. Products were detected with chemiluminescent DNA probes with the hybridization protection assay. Aliquots of the same specimen were assessed for PCA3 expression by TMA with the PROGENSA PCA3 assay (Gen-Probe Inc.) according to the manufacturer’s instructions (2223). Urine samples were tested by TMA for TMPRSS2:ERG and PSA in triplicate and duplicate, respectively. TMPRSS2:ERG and PSA mRNA copy levels were interpolated from calibration curves derived by assaying in vitro transcripts. For each specimen, the TMPRSS2:ERG score was calculated as follows: 100,000 × (average urine TMPRSS2:ERGa copies/ml)/(average urine PSA copies/ml).PCA3 scores were calculated as 1000 × (average urine PCA3 copies/ml)/(average urine PSA copies/ml).

Initial and second-generation TMPRSS2:ERG TMA assays

Academic biopsy and prostatectomy specimens were assayed with a TMPRSS2:ERG TMA assay using TMPRSS2:ERGa primers that span the junction between TMPRSS2 (NM_005656.3, exon 1) and ERG (NM_004449.4, exon 4), producing an amplification product of 86 nucleotides. PSA primers were the same as those used in the PROGENSA PCA3 assay (23). Samples with average PSA copies/ml >15,000 were considered informative.

Owing to ongoing assay optimization, community biopsy specimens were assayed with a second-generation assay using TMPRSS2:ERGa primers that also span the junction between TMPRSS2 and ERG, but produce an amplification product of 117 nucleotides; PSA primers were the same as those used in the PROGENSA PCA3 assay (23). Samples with average PSA copies/ml >10,000 were considered informative.

Tissue specimens

Formalin-fixed biopsy cores (n = 153; 74 benign, 79 cancerous) from 105 patients in the UMHS needle biopsy cohort with ERG rearrangement status determined by FISH were used for comparison with TMA for TMPRSS2:ERG expression. Because all benign cores were negative by FISH for ERG rearrangement, TMA for TMPRSS2:ERG expression was performed on an additional 55 benign cores (presumed negative for ERG rearrangement for analysis), resulting in a total of 208 cores (129 benign and 79 cancerous) from 122 patients evaluated by TMA.

Eight 10-μm sections from each evaluated block were cut; the first and last sections were evaluated by hematoxylin and eosin to ensure homogeneity and presence of cancerous or benign tissue. FISH for ERG rearrangement was performed on the seventh section as previously described (37). For TMA testing, the remaining sections were placed into a single transport tube containing 3 ml of specimen transport media (buffered detergent solution, part number 101768, Gen-Probe Inc.), heated to 60°C in a water bath for 30 min, and placed on ice for 5 min. Paraffin was removed with a sterile swab. The transfer tube was then capped, inverted, and stored at ≤−70°C until testing. The initial TMA assay was then performed on 300 μl of sample added to 100 μl of specimen transport medium, and assays were performed in singlicate for TMPRSS2:ERG andPSA. Specimens with PSA copies/ml >3000 were considered informative and those with TMPRSS2:ERG score >50 were considered positive.

Statistical analysis

Statistical analyses were performed with R, version 2.10.1 (R Foundation for Statistical Computing, http://www.R-project.org). Two-tailed tests were used for all comparisons and P values <0.05 were considered statistically significant.

For biopsy cohorts, Gleason score was assigned from the highest scoring single core and clinically significant cancers were defined by the Epstein criteria (3839) [any clinical stage >T1c, PSAD (serum PSA/prostate volume on ultrasound) ≥0.15 ng/ml per cubic centimeter, Gleason score >6, three or more cores positive, or >50% greatest single core involvement as significant]. Eight evaluable men in the academic biopsy cohort with a previous diagnosis of cancer (on active surveillance) with Gleason score 6 disease who had a negative biopsy in the current study were considered to have clinically insignificant cancer. PCPT risk factor scores and PCPT high-grade risk factor scores (6) were calculated as described (http://deb.uthscsa.edu/URORiskCalc/Pages/figure.jsp#) for all biopsy patients without a previous prostate cancer diagnosis who had all required clinicopathological information (DRE, first-degree family history of prostate cancer, PSA, and history of previous negative biopsy; or DRE, PSA, history of previous negative biopsy, African-American race, and age). Chun et al. nomogram (32) risk percentages were calculated with the online calculator (http://www.nomogram.org/Prostate/pros_calc.php) for all community biopsy patients presenting for initial biopsy who had all required clinicopathological information (age, DRE status, serum PSA, percentage free PSA, prostate volume by ultrasound, and number of biopsy cores).

Significant cancers at prostatectomy were defined by the Epstein criteria as those with maximum tumor dimension >1.0 cm (equivalent to tumor volume >0.5 cm3, because tumor volume is not routinely estimated at UMHS), Gleason score >6, and non–organ-confined disease (seminal vesicle involvement, positive lymph nodes, or extraprostatic extension) (3940). In the prostatectomy cohort, for determining significance at previous biopsy, the above Epstein biopsy criteria were used, except PSAD was calculated with prostate weight at prostatectomy (38), because ultrasound volume at biopsy was not always available.

Associations between cancer status (or TMPRSS2:ERG and PCA3 scores) and clinicopathological variables were assessed with Fisher’s exact test, Wilcoxon rank-sum test, or Spearman’s rho, depending on variable type (continuous or categorical). Confidence intervals on the median were computed with standard distribution-free methods (41). Diagnostic potential was quantified with sensitivity, specificity, likelihood ratio, and AUC. Comparisons between AUCs for different markers on the same collection of patients were performed with a nonparametric approach that accounts for the correlation between markers (42).

The ability of various combinations of biomarkers (TMPRSS2:ERGPCA3, PCPT risk percentage, Chun et al. nomogram, and/or serum PSA) to predict biopsy status was assessed through the use of multivariate logistic regression models. TMPRSS2:ERG, serum PSA, and PCA3 were logarithmically transformed [with the transformation log2(1 + x)] to avoid inappropriate influence of outlying values on prediction. For analysis of PCPT risk percentage in the academic cohort, we divided informative biopsy patients without a previous prostate cancer diagnosis who had all required clinicopathological information into training (VASD and UL patients) and testing (UMHS patients) sets, fit the models of interest to the training set, and used the estimated coefficients of these models to compute the predicted probability of positive biopsy for patients in the test set. The statistical significance of the difference in AUCs was then assessed (42). In all other cases, when training and testing sets were not explicitly constructed, the predictive ability of multivariate combinations of biomarkers was assessed with 10-fold cross-validation. Decision curve analysis to explore potential clinical net benefit was performed as described (31).

To assess for cancer risk in men with extremes of TMPRSS2:ERG and PCA3 scores, we first computed quartiles of TMPRSS2:ERG and PCA3 scores for the men in the prostatectomy cohort. We then used these quartile cutoffs to divide men into three combined TMPRSS2:ERG+PCA3 score groups (lowest, intermediate, and highest), as follows. Men were classified into the highest TMPRSS2:ERG+PCA3 group if at least one of the two individual biomarker scores was in the highest quartile, and into the lowest group if one biomarker score was in the lowest quartile and the other no higher than the second quartile; all other men were classified as intermediate risks. These prostatectomy cohort–derived cutoffs were then used to classify men in the academic biopsy cohort, because the two cohorts used the same TMPRSS2:ERG and PCA3 assays. To classify men in the community biopsy cohort, we used the same PCA3 score cutoffs, but because prostatectomy quartiles for TMPRSS2:ERG scores could not be directly used in this cohort (given the different assays used), we used sample quartiles of TMPRSS2:ERG determined from the community biopsy cohort to determine cutoffs. Rates of cancer, high-grade cancer (Gleason score >6), and clinically significant cancer (by Epstein criteria) in each TMPRSS2:ERG+PCA3 score group were then calculated. For determining clinically significant cancer rates, all VASD patients were excluded from the academic cohort (because the number of positive cores and maximum percentage involvement of individual cores was not recorded), and patients with cancer on biopsy where Epstein significance could not be determined owing to missing clinical information were excluded from both the academic and the community biopsy cohorts.

References and Notes

  1. Berkowitz, D. U. Ekwueme, Use of the prostate-specific antigen test among U.S. men: Findings from the 2005 National Health Interview Survey. Cancer Epidemiol. Biomarkers Prev. 17, 636–644 (2008).

  2. R. L. Grubb III, P. F. Pinsky, R. T. Greenlee, G. Izmirlian, A. B. Miller, T. P. Hickey, T. L. Riley, J. E. Mabie, D. L. Levin, D. Chia, B. S. Kramer, D. J. Reding, T. R. Church, L. A. Yokochi, P. A. Kvale, J. L. Weissfeld, D. A. Urban, S. S. Buys, E. P. Gelmann, L. R. Ragard, E. D. Crawford, P. C. Prorok, J. K. Gohagan, C. D. Berg, G. L. Andriole, Prostate cancer screening in the Prostate, Lung, Colorectal and Ovarian cancer screening trial: Update on findings from the initial four rounds of screening in a randomized trial. BJU Int. 102, 1524–1530 (2008).

  3. F. H. Schröder, J. Hugosson, M. J. Roobol, T. L. Tammela, S. Ciatto, V. Nelen, M. Kwiatkowski, M. Lujan, H. Lilja, M. Zappa, L. J. Denis, F. Recker, A. Berenguer, L. Määttänen, C. H. Bangma, G. Aus, A. Villers, X. Rebillard, T. van der Kwast, B. G. Blijenberg, S. M. Moss, H. J. de Koning, A. Auvinen; ERSPC Investigators, Screening and prostate-cancer mortality in a randomized European study. N. Engl. J. Med. 360, 1320–1328 (2009).

  4. M. S. Lucia, A. K. Darke, P. J. Goodman, F. G. La Rosa, H. L. Parnes, L. G. Ford, C. A. Coltman Jr., I. M. Thompson, Pathologic characteristics of cancers detected in The Prostate Cancer Prevention Trial: Implications for prostate cancer detection and chemoprevention. Cancer Prev. Res. 1, 167–173 (2008).

  5. I. M. Thompson, D. K. Pauler, P. J. Goodman, C. M. Tangen, M. S. Lucia, H. L. Parnes, L. M. Minasian, L. G. Ford, S. M. Lippman, E. D. Crawford, J. J. Crowley, C. A. Coltman Jr., Prevalence of prostate cancer among men with a prostate-specific antigen level < or =4.0 ng per milliliter. N. Engl. J. Med. 350, 2239–2246 (2004).

  6. I. M. Thompson, D. P. Ankerst, C. Chi, P. J. Goodman, C. M. Tangen, M. S. Lucia, Z. Feng, H. L. Parnes, C. A. Coltman Jr., Assessing prostate cancer risk: Results from the Prostate Cancer Prevention Trial. J. Natl. Cancer Inst. 98, 529–534 (2006).

  7. G. Draisma, R. Etzioni, A. Tsodikov, A. Mariotto, E. Wever, R. Gulati, E. Feuer, H. de Koning, Lead time and overdiagnosis in prostate-specific antigen screening: Importance of methods and context. J. Natl. Cancer Inst. 101, 374–383 (2009).

  8. G. L. Andriole, E. D. Crawford, R. L. Grubb III, S. S. Buys, D. Chia, T. R. Church, M. N. Fouad, E. P. Gelmann, P. A. Kvale, D. J. Reding, J. L. Weissfeld, L. A. Yokochi, B. O’Brien, J. D. Clapp, J. M. Rathmell, T. L. Riley, R. B. Hayes, B. S. Kramer, G. Izmirlian, A. B. Miller, P. F. Pinsky, P. C. Prorok, J. K. Gohagan, C. D. Berg; PLCO Project Team, Mortality results from a randomized prostate-cancer screening trial. N. Engl. J. Med. 360, 1310–1319 (2009).

  9.  S. A. Strope, G. L. Andriole, Prostate cancer screening: Current status and future perspectives. Nat. Rev. Urol. 7, 487–493 (2010).

  10. F. H. Jansen, M. Roobol, G. Jenster, F. H. Schröder, C. H. Bangma, Screening for prostate cancer in 2008 II: The importance of molecular subforms of prostate-specific antigen and tissue kallikreins. Eur. Urol. 55, 563–574 (2009).

  11. R. J. Leary, I. Kinde, F. Diehl, K. Schmidt, C. Clouser, C. Duncan, A. Antipova, C. Lee, K. McKernan, F. M. De La Vega, K. W. Kinzler, B. Vogelstein, L. A. Diaz Jr., V. E. Velculescu, Development of personalized tumor biomarkers using massively parallel sequencing. Sci. Transl. Med. 2, 20ra14 (2010).

  12. S. A. Tomlins, D. R. Rhodes, S. Perner, S. M. Dhanasekaran, R. Mehra, X. W. Sun, S. Varambally, X. Cao, J. Tchinda, R. Kuefer, C. Lee, J. E. Montie, R. B. Shah, K. J. Pienta, M. A. Rubin, A. M. Chinnaiyan, Recurrent fusion of TMPRSS2 and ETS transcription factor genes in prostate cancer. Science 310, 644–648 (2005).

  13. S. A. Tomlins, A. Bjartell, A. M. Chinnaiyan, G. Jenster, R. K. Nam, M. A. Rubin, J. A. Schalken, ETS gene fusions in prostate cancer: From discovery to daily clinical practice. Eur. Urol. 56, 275–286 (2009).

  14. B. Furusato, S. H. Tan, D. Young, A. Dobi, C. Sun, A. A. Mohamed, R. Thangapazham, Y. Chen, G. McMaster, T. Sreenath, G. Petrovics, D. G. McLeod, S. Srivastava, I. A. Sesterhenn, ERG oncoprotein expression in prostate cancer: Clonal progression of ERG-positive tumor cells and potential for ERG-based stratification. Prostate Cancer Prostatic Dis. 13, 228–237 (2010).

  15. K. Park, S. A. Tomlins, K. M. Mudaliar, Y. L. Chiu, R. Esgueva, R. Mehra, K. Suleman, S. Varambally, J. C. Brenner, T. MacDonald, A. Srivastava, A. K. Tewari, U. Sathyanarayana, D. Nagy, G. Pestano, L. P. Kunju, F. Demichelis, A. M. Chinnaiyan, M. A. Rubin, Antibody-based detection of ERG rearrangement-positive prostate cancer. Neoplasia 12, 590–598 (2010).

  16. B. Han, R. Mehra, R. J. Lonigro, L. Wang, K. Suleman, A. Menon, N. Palanisamy, S. A. Tomlins, A. M. Chinnaiyan, R. B. Shah, Fluorescence in situ hybridization study shows association of PTEN deletion with ERG rearrangement during prostate cancer progression. Mod. Pathol. 22, 1083–1093 (2009).

  17. B. S. Carver, J. Tran, A. Gopalan, Z. Chen, S. Shaikh, A. Carracedo, A. Alimonti, C. Nardella, S. Varmeh, P. T. Scardino, C. Cordon-Cardo, W. Gerald, P. P. Pandolfi, Aberrant ERG expression cooperates with loss of PTEN to promote cancer progression in the prostate. Nat. Genet. 41, 619–624 (2009).

  18. J. M. Mosquera, S. Perner, E. M. Genega, M. Sanda, M. D. Hofer, K. D. Mertz, P. L. Paris, J. Simko, T. A. Bismar, G. Ayala, R. B. Shah, M. Loda, M. A. Rubin, Characterization of TMPRSS2-ERG fusion high-grade prostatic intraepithelial neoplasia and potential clinical implications. Clin. Cancer Res. 14, 3380–3385 (2008).

  19. Y. Zong, L. Xin, A. S. Goldstein, D. A. Lawson, M. A. Teitell, O. N. Witte, ETS family transcription factors collaborate with alternative signaling pathways to induce carcinoma from adult murine prostate cells. Proc. Natl. Acad. Sci. U.S.A. 106, 12465–12470 (2009).

  20. J. C. King, J. Xu, J. Wongvipat, H. Hieronymus, B. S. Carver, D. H. Leung, B. S. Taylor, C. Sander, R. D. Cardiff, S. S. Couto, W. L. Gerald, C. L. Sawyers, Cooperativity of TMPRSS2-ERG with PI3- kinase pathway activation in prostate oncogenesis. Nat. Genet. 41, 524–526 (2009).

  21. M. J. Bussemakers, A. van Bokhoven, G. W. Verhaegh, F. P. Smit, H. F. Karthaus, J. A. Schalken, F. M. Debruyne, N. Ru, W. B. Isaacs, DD3: A new prostate-specific gene, highly overexpressed in prostate cancer. Cancer Res. 59, 5975–5979 (1999).

  22. L. Deras, S. M. Aubin, A. Blase, J. R. Day, S. Koo, A. W. Partin, W. J. Ellis, L. S. Marks, Y. Fradet, H. Rittenhouse, J. Groskopf, PCA3: A molecular urine assay for predicting prostate biopsy outcome. J. Urol. 179, 1587–1592 (2008).

  23. J. Groskopf, S. M. Aubin, I. L. Deras, A. Blase, S. Bodrug, C. Clark, S. Brentano, J. Mathis, J. Pham, T. Meyer, M. Cass, P. Hodge, M. L. Macairan, L. S. Marks, H. Rittenhouse, APTIMA PCA3 molecular urine test: Development of a method to aid in the diagnosis of prostate cancer. Clin. Chem. 52, 1089–1095 (2006).

  24. D. Hessels, F. P. Smit, G. W. Verhaegh, J. A. Witjes, E. B. Cornel, J. A. Schalken, Detection of TMPRSS2-ERG fusion transcripts and prostate cancer antigen 3 in urinary sediments may improve diagnosis of prostate cancer. Clin. Cancer Res. 13, 5103–5108 (2007).

  25. B. Laxman, D. S. Morris, J. Yu, J. Siddiqui, J. Cao, R. Mehra, R. J. Lonigro, A. Tsodikov, J. T. Wei, S. A. Tomlins, A. M. Chinnaiyan, A first-generation multiplex biomarker analysis of urine for the early detection of prostate cancer. Cancer Res. 68, 645–649 (2008).

  26. B. Laxman, S. A. Tomlins, R. Mehra, D. S. Morris, L. Wang, B. E. Helgeson, R. B. Shah, M. A. Rubin, J. T. Wei, A. M. Chinnaiyan, Noninvasive detection of TMPRSS2:ERG fusion transcripts in the urine of men with prostate cancer. Neoplasia 8, 885–888 (2006).

  27. J. Nilsson, J. Skog, A. Nordstrand, V. Baranov, L. Mincheva-Nilsson, X. O. Breakefield, A. Widmark, Prostate cancer-derived urine exosomes: A novel approach to biomarkers for prostate cancer. Br. J. Cancer 100, 1603–1607 (2009).

  28. K. Rostad, O. J. Hellwinkel, S. A. Haukaas, O. J. Halvorsen, A. M. Oyan, A. Haese, L. Budaus, H. Albrecht, L. A. Akslen, T. Schlomm, K. H. Kalland, TMPRSS2:ERG fusion transcripts in urine from prostate cancer patients correlate with a less favorable prognosis. APMIS 117, 575–582 (2009).

  29. F. Dong, J. S. Jones, A. J. Stephenson, C. Magi-Galluzzi, A. M. Reuther, E. A. Klein, Prostate cancer volume at biopsy predicts clinically significant upgrading. J. Urol. 179, 896–900 (2008).

  30. S. F. Shariat, P. I. Karakiewicz, C. G. Roehrborn, M. W. Kattan, An updated catalog of prostate cancer predictive tools. Cancer 113, 3075–3099 (2008).

  31. A. J. Vickers, E. B. Elkin, Decision curve analysis: A novel method for evaluating prediction models. Med. Decis. Making 26, 565–574 (2006).

  32. F. K. Chun, A. Briganti, M. Graefen, F. Montorsi, C. Porter, V. Scattoni, A. Gallina, J. Walz, A. Haese, T. Steuber, A. Erbersdobler, T. Schlomm, S. A. Ahyai, E. Currlin, L. Valiquette, H. Heinzer, P. Rigatti, H. Huland, P. I. Karakiewicz, Development and external validation of an extended 10-core biopsy nomogram. Eur. Urol. 52, 436–444 (2007).

  33. G. P. Haas, N. B. Delongchamps, R. F. Jones, V. Chandan, A. M. Serio, A. J. Vickers, M. Jumbelic, G. Threatte, R. Korets, H. Lilja, G. de la Roza, Needle biopsies on autopsy prostates: Sensitivity of cancer detection based on true prevalence. J. Natl. Cancer Inst. 99, 1484–1489 (2007).

  34. N. B. Delongchamps, G. de la Roza, R. Jones, M. Jumbelic, G. P. Haas, Saturation biopsies on autopsied prostates for detecting and characterizing prostate cancer. BJU Int. 103, 49–54 (2009).

  35. C. Gosselaar, M. J. Roobol, F. H. Schröder, Prevalence and characteristics of screen-detected prostate carcinomas at low prostate-specific antigen levels: Aggressive or insignificant? BJU Int. 95, 231–237 (2005).

  36. G. Attard, A. H. Reid, T. A. Yap, F. Raynaud, M. Dowsett, S. Settatree, M. Barrett, C. Parker, V. Martins, E. Folkerd, J. Clark, C. S. Cooper, S. B. Kaye, D. Dearnaley, G. Lee, J. S. de Bono, Phase I clinical trial of a selective inhibitor of CYP17, abiraterone acetate, confirms that castration-resistant prostate cancer commonly remains hormone driven. J. Clin. Oncol. 26, 4563–4571 (2008).

  37. J. M. Mosquera, R. Mehra, M. M. Regan, S. Perner, E. M. Genega, G. Bueti, R. B. Shah, S. Gaston, S. A. Tomlins, J. T. Wei, M. C. Kearney, L. A. Johnson, J. M. Tang, A. M. Chinnaiyan, M. A. Rubin, M. G. Sanda, Prevalence of TMPRSS2-ERG fusion prostate cancer among men undergoing prostate biopsy in the nited States. Clin. Cancer Res. 15, 4706–4711 (2009).

  38. P. J. Bastian, L. A. Mangold, J. I. Epstein, A. W. Partin, Characteristics of insignificant clinical T1c prostate tumors. A contemporary analysis. Cancer 101, 2001–2005 (2004).

  39. J. I. Epstein, P. C. Walsh, M. Carmichael, C. B. Brendler, Pathologic and clinical findings to predict tumor extent of nonpalpable (stage T1c) prostate cancer. JAMA 271, 368–374 (1994).

  40. W. Siu, R. L. Dunn, R. B. Shah, J. T. Wei, Use of extended pattern technique for initial prostate biopsy. J. Urol. 174, 505–509 (2005).

  41. E. L. Lehmann, H. J. M. D’Abrera, Nonparametrics: Statistical Methods Based on Ranks (HoldenDay, San Francisco, CA, 1975), p. xvi.

  42. E. R. DeLong, D. M. DeLong, D. L. Clarke-Pearson, Comparing the areas under two or more correlated receiver operating characteristic curves: A nonparametric approach. Biometrics 44, 837–845 (1988).

  43. Funding: Supported in part by Gen-Probe Inc., the Early Detection Research Network (U01 CA111275 and U01 CA113913), NIH SPORE (P50 CA69568), and R01 CA132874. S.A.T. is supported by a Young Investigator Award from the Prostate Cancer Foundation. A.M.C. is supported by the Burroughs Wellcome Fund, the Doris Duke Foundation, and the Prostate Cancer Foundation. Author contributions: S.A.T. and A.M.C. had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. S.A.T., S.M.J.A., A.B., Y.P., M.G.S., M.A.R., D.R.R., H.R., J.G., J.T.W., and A.M.C. designed the study. S.M.J.A., S. Miick, J.R.D., S.W., P.H., J.M., H.R., and J.G. developed and performed the TMA assays. J.L.S., L.S.-M., D.W., B. Hollenbeck, K.S., J.S., Y.F., J.B.A., S. Meyers, and J.T.W. obtained, evaluated, and processed specimens. R.V. maintained clinical information. B. Han, L.W., and N.P. performed FISH assays. S.A.T. and R.J.L. analyzed the data. S.A.T. wrote the first draft of the manuscript, which was reviewed by all authors. Competing interests: The University of Michigan and Brigham and Women’s Hospital have been issued a patent on the detection of ETS gene fusions in prostate cancer, on which S.A.T., M.A.R., D.R.R., and A.M.C. are listed as coinventors. The diagnostic field of use has been licensed to Gen-Probe. A.M.C. and J.T.W. have served as consultants for Gen-Probe. Gen-Probe has provided material support for a clinical trial evaluating PCA3, on which J.T.W. is the primary investigator. Y.F. is cofounder of Diagnocure, which licensed diagnostic rights to the PCA3 gene to Gen-Probe. J.S. and M.A.R. have received honoraria from Gen-Probe. M.G.S. has received research funding from Beckman-Coulter and Source-MDx. S.M.J.A., S. Miick, S.W., P.H., J.M., J.R.D., A.B., Y.P., H.R., and J.G. are employees of Gen-Probe. The remaining authors declare no conflicts of interest