Citations

3 Background: There is a clear need to develop a clinically relevant molecular subtyping approach for prostate cancer. We hypothesized that prostate cancer can be subtyped based on luminal versus basal lineage. Methods: We applied the PAM50 classifier, which is used clinically to identify luminal and basal cancers in breast cancer, to subtype a total of 3,782 prostate cancer samples using a high-density microarray platform run in a CLIA-certified laboratory. We examined the associations of these subtypes and clinical outcomes. Results: We demonstrate that PAM50 segregates prostate cancer into three reproducible subtypes in both retrospective cohorts and on prospective validation: luminal A (33.3%-34.3%), luminal B (28.5%-32.6%), and basal (34.1%-37.1%). Luminal B prostate cancers exhibited the worst clinical prognoses, followed by basal and luminal A subtypes (10-year biochemical recurrence-free survival: 29/39/41%; distant metastasis-free survival: 53/73/73%; prostate cancer-specific survival: 78/86/89%; overall survival: 69/80/82% respectively) on both univariable and multivariable analyses accounting for standard clinicopathologic prognostic factors. Known luminal lineage markers, such as NKX3.1 and KRT18, and the basal lineage CD49f signature, were enriched in luminal- and basal-like cancers respectively, demonstrating the connection between these subtypes and established prostate cancer biology. While both luminal-like subtypes were associated with increased AR expression and signaling, only luminal B prostate cancers were significantly associated with post-operative response to androgen deprivation therapy (ADT) in a subset analysis matching patients based on clinicopathologic variables (interaction p = 0.006, luminal B 10-year metastasis: 33% (treated) vs. 55% (untreated), non-luminal B: 37% (treated) vs. 21% (untreated)). Conclusions: These findings contribute novel insight into the biology of prostate cancer, and provide translatable clinical tools for personalizing post-operative ADT for patients with prostate cancer. Similar to breast cancer, these findings suggest that luminal/basal subtyping may be useful in treatment selection in prostate cancer.

142 Background: Despite documented oncological benefit, postoperative adjuvant radiotherapy (aRT) utilization in prostate cancer (PCa) patients is still limited in the US. We aimed to develop and internally validate a risk stratification tool incorporating the Decipher score, along with routinely available clinicopathologic features, to identify patients who would benefit the most from aRT. Methods: Our cohort included a total of 512 PCa patients treated with RP at one of four US academic centers between 1990-2010. All patients had ≥ pT3a disease, positive margins, and/or pathologic lymph node invasion (LNI). Multivariable Cox regression analysis (MVA) tested the relationship between available predictors (including Decipher score) and clinical recurrence (CR), which were then used to develop a novel risk stratification tool. Our study adhered to the TRIPOD guidelines for development of prognostic models. Results: Overall, 21.9% patients received aRT. Median follow-up in censored patients was 8.3 years. The 10-year CR rate was 4.9% vs. 17.4% in patients treated with aRT vs. initial observation (p < 0.001). Pathological T3b/T4 stage, Gleason score 8-10, LNI and Decipher score > 0.6 were independent predictors of CR (all p < 0.01) Cumulative number of risk factors was 0, 1, 2, and 3-4 in respectively 46.5, 28.9, 17.2, and 7.4% of patients. Adjuvant RT was associated with decreased CR rate in patients with ≥ 2 risk factors (10-year CR rate 10.1% in aRT vs. 42.1% in initial observation, p = 0.008), but not in those with < 2 risk factors (p = 0.23). Conclusions: Utilizing the novel model to indicate aRT might reduce overtreatment, decrease unnecessary side effects, and reduce risk of CR in the subset of patients (~25% of all patients with aggressive pathological disease) who really benefit from this therapy.

133 Background: The genomic classifier, Decipher, has been validated to predict risk of metastasis after radical prostatectomy (RP). However, the cohort size and event rate in the previous studies did not allow for a thorough investigation into performance within individual clinicopathologic or treatment subgroups. In this study, we present the first meta-analysis of the performance of the 22-marker genomic classifier in men with prostate cancer (PCa) post-RP. Methods: MEDLINE, EMBASE, and the Decipher genomic resource information database were searched for published reports of men with PCa treated by RP between 2010 and 2016 where the benefit of the Decipher genomic classifier test was assessed. The primary end point was the ability of Decipher to independently improve prognostication of regional or distant metastasis over routine clinicopathologic factors. Meta-analysis was performed with random-effects modeling, and extent of heterogeneity between studies was determined with the I2 test. Results: Five studies (975 total patients, and 855 with individual patient genomic and clinicopathologic data) were eligible for analysis. The median follow-up was 8 years. All patients had clinical high-risk disease, yet 60.9%, 22.6%, and 16.5% of patients were classified as low, intermediate, and high-risk, respectively by Decipher and had 10-year cumulative incidence rates of metastases of 5.5%, 15.0% and 26.7% (p < 0.001), respectively. Adjusting for standard clinicopathologic variables, on multivariable analysis Decipher remained a statistically significant predictor of metastasis (hazard ratio [HR] 1.30 per 0.1 unit, 95% confidence interval [CI] 1.14-1.47, p < 0.001), and the summary HR for metastasis of Decipher across the 5 studies was 1.52 (95% CI 1.39-1.67) per 0.1 unit. Conclusions: The genomic classifier test, Decipher, has the ability to independently improve prognostication of men post-RP, as well as within nearly all clinicopathologic and treatment subgroups. Strong consideration should be given to incorporating the use of genomic testing in clinical decision making and clinical trials to better individualize treatment.

Purpose Despite documented oncologic benefit, use of postoperative adjuvant radiotherapy (aRT) in patients with prostate cancer is still limited in the United States. We aimed to develop and internally validate a risk-stratification tool incorporating the Decipher score, along with routinely available clinicopathologic features, to identify patients who would benefit the most from aRT. Patient and Methods Our cohort included 512 patients with prostate cancer treated with radical prostatectomy at one of four US academic centers between 1990 and 2010. All patients had ≥ pT3a disease, positive surgical margins, and/or pathologic lymph node invasion. Multivariable Cox regression analysis tested the relationship between available predictors (including Decipher score) and clinical recurrence (CR), which were then used to develop a novel risk-stratification tool. Our study adhered to the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis guidelines for development of prognostic models. Results Overall, 21.9% of patients received aRT. Median follow-up in censored patients was 8.3 years. The 10-year CR rate was 4.9% vs. 17.4% in patients treated with aRT versus initial observation ( P < .001). Pathologic T3b/T4 stage, Gleason score 8-10, lymph node invasion, and Decipher score > 0.6 were independent predictors of CR (all P < .01). The cumulative number of risk factors was 0, 1, 2, and 3 to 4 in 46.5%, 28.9%, 17.2%, and 7.4% of patients, respectively. aRT was associated with decreased CR rate in patients with two or more risk factors (10-year CR rate 10.1% in aRT v 42.1% in initial observation; P = .012), but not in those with fewer than two risk factors ( P = .18). Conclusion Using the new model to indicate aRT might reduce overtreatment, decrease unnecessary adverse effects, and reduce risk of CR in the subset of patients (approximately 25% of all patients with aggressive pathologic disease in our cohort) who benefit from this therapy.

90 Background: Biochemical recurrence (BCR) may be an indicator of metastatic prostate cancer; however, it isn't specific enough to differentiate patients who will experience rapid onset of metastases from those who do not. Such lack of specificity leads to overtreatment of men who will likely never experience metastatic disease in their lifetime. We hypothesize that genomic biomarkers, detectable in primary tumor tissue, may better identify rapidly progressing metastatic prostate cancer. Methods: FFPE prostatectomy specimens (n = 309) from the Mayo Clinic tumor registry were profiled using a high-density expression array. Patients were retrospectively classified into three outcome groups: NED (no evidence of disease recurrence); BCR-only (patients that experienced BCR but no metastatic progression); and MET (metastatic disease confirmed by positive CT or bone scans) and were matched on several criteria for analysis . Uni- and multi-variable analyses between outcome groups were used to analyze clinical and genomic variables (adjusting for adjuvant hormone therapy as a confounding factor). Results: Median follow-up was 16.8 years and MET patients were diagnosed a mean of 3.1 years following BCR. Univariable and multivariable analysis of clinical variables failed to show statistically significant differences between NED and BCR-only outcome groups. Differential expression analysis between NED and BCR-only groups showed no significant differences after false discovery correction. Conversely, 6,277 and 52,020 biomarkers were differentially expressed between MET and the BCR-only and NED groups, respectively. Conclusions: Considering the clinical and genomic results together, patients who experienced rapid metastatic disease were significantly different from both the NED and BCR-only groups. However, no significant differences were identified between the BCR-only and NED outcome groups. These results imply that BCR is not a good surrogate for men who are likely to experience rapid onset of metastatic disease and lethal prostate cancer.

40 Background: Gleason score (GS) is the most widely used grading system of cell differentiation in prostate cancer and one of the best pathological predictors of disease progression. Patients with high GS (> 7) are the most likely to experience metastasis, whereas patients with low GS (< 7) are expected to have favorable long-term outcomes. In addition to the subjective nature of GS assessment, patients with GS 7 represent a heterogeneous group in terms of patient outcomes. In this study a biomarker signature is developed and validated which could improve the prediction of high risk disease among GS 7 radical prostatectomy (RP) patients. Methods: Patient specimens from the Mayo Clinic RP Registry (n = 764) were used for feature selection and training of a 392-feature K-nearest neighbor (KNN, k = 11) classifier. These 392 genomic features were identified by assessing differential expression between patients with GS < 7 and those with GS > 7, using a Bonferroni adjusted T-Test with a p-value threshold of 0.05. The classifier was subsequently validated in two independent patient cohorts (Memorial Sloan Kettering [MSKCC] and German Cancer Research Center [DKFZ]) and compared to a state of the art biomarker signature (Penney et al. 2011). Results: In the MSKCC dataset (GSE21034) our model segregated GS < 7 from GS > 7 patients (n = 56) with an area under the receiver operating characteristic curve (AUC) of 0.97, comparable to the AUC of 0.94 obtained by Penney et al. in their independent validation set (n = 45). Strong performance was observed by our model when discriminating between primary Gleason grade (pGG) 3 and pGG 4 & 5 patients in the MSKCC (n = 130) and DKFZ prostate cancer (GSE29079, n = 47) datasets, achieving AUCs of 0.75 and 0.87, respectively. In the challenging GS 7 subset, our model segregated GS 3 + 4 and 4 + 3 patients in the DKFZ dataset (n = 64) with an AUC of 0.81 outperforming the Penney et al. signature (AUC = 0.60). Conclusions: A biomarker signature was developed which discriminates between low and high GS patients, outperforming a previously reported signature. Further validation of this biomarker signature in additional post RP patients, as well as, pretreatment biopsy specimens is warranted.

36 Background: Following radical prostatectomy (RP), 30-40% of patients have adverse pathology and are deemed at high risk for metastatic progression. The objective of this study was to validate the ability of Decipher, a genomic classifier (GC), to improve prediction of metastatic disease progression compared with clinical variables in order to better identify candidates for therapy intensification. Methods: A previously developed 22-feature GC model was validated in a prospectively designed case-cohort study of a clinically high-risk population (i.e., with one or more adverse pathological features) of 1,010 RP patients treated at Mayo Clinic between 2000-2006. A random sample of 20% of the cohort was subjected to microarray analysis and GC scores were generated for 219 patients. The primary endpoint, the c-index for predicting metastatic disease progression (i.e., positive bone or CT scans) was evaluated in a blinded analysis. Cox modeling and decision curve analyses were used to compare the performance of GC to individual clinical variables and prediction models. Results: GC had a c-index 0.79 (95% CI 0.71-0.86) that was significantly better than any single clinical variable. Cumulative incidence curves in the cohort showed that 72% of patients had low GC scores with only 3% and 6% incidence of metastatic disease at 5 and 10 years post RP. In contrast, for the 28% of patients with high GC scores, the cumulative incidence was 17% and 25% at 5 and 10 years post RP. Decision curve analysis showed that the GC model had higher overall net benefit compared to clinical variables over a wide range of ‘decision-to-treat’ thresholds for risk of metastasis. In multivariable modeling with clinicopathologic variables, GC remained the only significant independent predictor of metastasis (HR=1.51, for each 0.1 unit increment, p<0.001). Conclusions: GC can better predict metastatic disease progression compared with clinical variables and may select among patients with adverse pathology a majority that is in fact at low risk for metastasis.

9 Background: Prostate cancer molecular subtypes based on ETS gene fusions and SPINK1 were originally identified through outlier gene expression profiling analysis. Such molecular subtypes may have utility in disease stratification and clonality assessment, complementing available purely prognostic tests. Hence, we determined the analytical validity of molecular subtyping in a large sample of PCa treated with radical prostatectomy. Methods: We analyzed Affymetrix Human Exon 1.0ST GeneChip expression profiles for 1,577 patients from 8 radical prostatectomy (RP) cohorts. Multi-feature random forest classifiers and outlier analysis were used to define microarray-based molecular subtypes. Results: A random forest (RF) classifier was trained and validated to predict ERG fusion status using a subset with known ERG rearrangement status defined by FISH, achieving >95% sensitivity and specificity in the validation subset. Less frequent rearrangements involving other ETS genes or SPINK1 over-expression were predicted based on gene expression outlier analysis. Across cohorts, 45%, 9% 8% and 38% of PCa were classified as ERG+, ETS+, SPINK+, and Triple Negative, respectively. Global gene expression analysis shows that the four subtypes could be collapsed into three entities (ERG+, ETS+ and SPINK+/Triple Negative) based on expression patterns and clinical characteristics similarity. A series of multivariable analyses further revealed, ERG+ to be associated with lower pre PSA and Gleason scores but more likely to have EPE and occur in patients with European American ancestry compared to the ETS+, SPINK+/Triple Negative tumors (p<0.001). In contrast, patients with ETS+ were more likely to have SVI compared to both ERG+ and SPINK/Triple Negative (p=0.01), while SPINK+/Triple Negative had higher Gleason scores and were more likely to occur in African Americans (p<0.001). Conclusions: The Decipher platform can accurately determine ERG rearrangement status and PCa molecular subtypes. Inclusion of molecular subtyping, such as m-ERG status, may enable additional precision medicine opportunities in prognostic tests

8 Background: A number of biomarkers related to cell cycle, angiogenesis or apoptosis have been found to be associated with patient outcome in tissue samples from men treated with first line radiation therapy in RTOG clinical trials using immunohistochemical staining and analysis. In a prior study, four biomarkers (Ki-67, MDM2, p16 and Cox-2) and clinical covariates were included in a model of distant metastasis (DM; Pollack et al, Clin Cancer Res, 2014 [epub ahead of print]) risk. The current study tested the hypothesis that these genes are prognostic for DM in men treated primarily with total prostatectomy using RNA expression profiling. Methods: RNA fromprostatectomy samples from Cleveland Clinic (CC, n=182); Mayo Clinic (MC)-I (n = 545) and II (n=235); Memorial Sloan Kettering Cancer Center (MSKCC, n=131); Erasmus Medical Center (EMC, n=48) and Thomas Jefferson University (TJU, n=130) were profiled using 1.4 million RNA features. A Cox proportional hazards model was built on the MC-I training set to combine the 4 biomarkers into a prognostic risk score (4BMSig). 4BMSig was subsequently evaluated for its prognostic significance separately and in combination with clinical risk factors (biopsy Gleason Score, cT-category and Preop-PSA) for DM. Results: 4BMSig was found to discriminate DM patients significantly for the MC-II (AUC = 0.66, p < 0.001), CCF (AUC = 0.68, p < 0.001), and MSKCC (AUC = 0.71, p = 0.04) datasets, and achieved borderline significance for EMC (AUC = 0.70, p = 0.06). 4BMSig did not discriminate DM in the TJU dataset (only 10 DM events). Pooled multivariable analysis (n = 726) with clinical covariates revealed that 4BMSig is a strong independent prognostic covariate for DM (p < 0.001) and prostate cancer specific mortality (p = 0.005). Conclusions: The four genes identified previously as being associated with DM in radiotherapy patients were incorporated herein into 4BMSig, which was found to have potential as a pretreatment prognostic DM risk assessment tool for men treated with prostatectomy. Further validation would consist of testing 4BMSig from RNA in diagnostic tissue from prostate cancer patients prior to prostatectomy.

7 Background: Improved prognostic biomarkers are needed for localized prostate cancer. We undertook an unbiased large-scale analysis of genes associated with metastatic progression and validated the prognostic ability of the top candidate gene. Methods: Prostate cancer samples from prostatectomy patients were analyzed for gene expression using a clinical-grade, high-density Affymetrix GeneChip platform, encompassing >1 million genomic loci and assessed in a CLIA-certified laboratory. Nomination of prognostic candidate genes was performed on a discovery cohort (n=545) and validated on 3 independent cohorts (n=463). Multivariate analyses were performed for the primary endpoint of metastasis. The top prostate-specific gene was further evaluated in 208 additional tumor samples with a novel RNA in-situ hybridization (ISH) assay and in urine samples from 230 patients using PCR. Results: Of all known genes, the long noncoding RNA SChLAP1 ranked first for elevated expression in patients with metastatic progression by receiver-operator-curve analyses. Validation in three independent cohorts confirmed the prognostic value of SChLAP1. On multivariate modeling, SChLAP1 expression independently predicted metastasis within 10 years (odds ratio (OR) = 2.45, 95% confidence interval (CI) 1.70 – 3.53), death within 10 years (OR = 1.93, 95% CI 1.31 – 2.85), and biochemical recurrence within 5 years (OR = 1.76, 95% CI 1.28 – 2.41) with ORs comparable to Gleason score. Evaluation of SChLAP1 expression by RNA ISH confirmed a significant association with disease recurrence (OR = 1.99, 95% CI 1.06 – 3.73). Evaluation of urine SChLAP1 levels demonstrated increased expression in patients at higher risk for disease progression. Conclusions: We perform the largest high-throughput, unbiased study of prostate cancer prognostic biomarkers to date and discover SChLAP1 as a top gene predictive of metastatic progression. We validate SChLAP1 extensively with a clinical-grade assay. We show feasibility of a RNA ISH assay and a non-invasive urine test for SChLAP1. Our results, spanning 1,446 patients from 6 independent patient cohorts, suggest that SChLAP1 represents a very promising biomarker for aggressive clinical course.