A machine learning approach to predict progression on active surveillance for prostate cancer

PURPOSE

Robust prediction of progression on active surveillance (AS) for prostate cancer can allow for risk-adapted protocols. To date, models predicting progression on AS have invariably used traditional statistical approaches. We sought to evaluate whether a machine learning (ML) approach could improve prediction of progression on AS.

PATIENTS AND METHODS

We performed a retrospective cohort study of patients diagnosed with very-low or low-risk prostate cancer between 1997 and 2016 and managed with AS at our institution. In the training set, we trained a traditional logistic regression (T-LR) classifier, and alternate ML classifiers (support vector machine, random forest, a fully connected artificial neural network, and ML-LR) to predict grade-progression. We evaluated model performance in the test set. The primary performance metric was the F1 score.

RESULTS

Our cohort included 790 patients. With a median follow-up of 6.29 years, 234 developed grade-progression. In descending order, the F1 scores were: support vector machine 0.586 (95% CI 0.579 - 0.591), ML-LR 0.522 (95% CI 0.513 - 0.526), artificial neural network 0.392 (95% CI 0.379 - 0.396), random forest 0.376 (95% CI 0.364 - 0.380), and T-LR 0.182 (95% CI 0.151 - 0.185). All alternate ML models had a significantly higher F1 score than the T-LR model (all p <0.001).

CONCLUSION

In our study, ML methods significantly outperformed T-LR in predicting progression on AS for prostate cancer. While our specific models require further validation, we anticipate that a ML approach will help produce robust prediction models that will facilitate individualized risk-stratification in prostate cancer AS.

Comprehensive genomic profiling of urine DNA for urothelial carcinoma detection and risk prediction

450

Background: Clinical diagnosis and risk stratification of patients with urothelial carcinoma (UC) remains a challenge, with high rates of recurrence and disease progression following treatment. Urinary comprehensive genomic profiling (uCGP) has significant potential to aid in both diagnosis and prognostication of non-muscle-invasive and muscle-invasive disease. Methods: uCGP was performed on urine specimens collected at 9 centers across the US from 577 subjects prior to cystoscopy. 152 subjects were UC tumor positive (de novo and recurrence), 191 had a history of UC but negative by surveillance cystoscopy at time of collection, and 234 were urology control subjects undergoing cystoscopy without evidence of UC. Urine DNA was sequenced and comprehensively profiled across 60 genes for 6 classes of mutations using the CLIA-validated UroAmplitude test. Disease detection and molecular grade (high grade vs. low grade) algorithms were trained (n=345) and validated (n=232) in independent cohorts. Results: Among UC tumor positives, grade distribution was 53% high grade, 41% low grade, and 6% unknown. Stage distribution was Tis (5%), Ta (57%), T1 (16%), ≥T2 (15%), Tx (7%). 99% of tumor positive patients had one or more mutation identified. Interestingly, 69% of UC surveillance negative and 49% of urology controls also had at least one high impact mutation. The prevalence of mutations among controls necessitates machine learning algorithms to classify disease status. In validation, de novo tumor diagnosis demonstrated sensitivity of 93.8% and specificity of 89.4% and a NPV of 98.8% in urology controls. Recurrent tumors were detected with a PPV of 73.5%, sensitivity of 62.5% and specificity of 89.0% in patients with a history of UC. Molecular grading predicted high-grade with a PPV of 90.9% and a specificity of 96.7% compared to pathology. Urinary TP53 mutations were enriched in ≥T2 tumors relative to Ta (OR=14.8 [95%CI 4.6-47.5], P=0.00001). Copy number alterations were also associated with increased risk of muscle invasion, metastasis, and enriched for CIS relative to Ta tumors (≥T2: OR=6.4 [95%CI 1.8-22.9], P=0.019; CIS: OR=10.5 [95%CI 1.9-58.9], P=0.04). Conclusions: We developed and validated a uCGP test that provides robust noninvasive detection of UC across a diverse group of patients and clinical contexts, including non-muscle-invasive and muscle-invasive UC. Mutations with actionable or prognostic value are found in most subjects. These data suggest that uCGP classifies tumor presence with better performance than traditional urinary biomarkers. Importantly, uCGP identifies genomic markers of muscle invasion, metastasis, and CIS. With longer term follow-up, uCGP mutational profiles may reveal important prognostic information regarding risk of disease recurrence and progression. Additional studies are underway to further support the generalizability of these findings.

Multi-institutional matched comparison of radical cystectomy to trimodality therapy for muscle-invasive bladder cancer

433

Background: Prior randomized controlled trials (RCT) comparing bladder preservation to radical cystectomy (RC) for muscle invasive bladder cancer (MIBC) closed early due to lack of accrual. Given that no future RCTs are foreseen, and in the absence of level 1 data, we aimed to provide the best evidence possible on outcomes of matched cohorts comparing trimodality therapy (TMT, maximal transurethral resection of bladder tumor followed by concurrent chemoradiation) to RC in order to guide management. Methods: This retrospective analysis included 703 patients with MIBC clinical stage T2-T3/4aN0M0 MIBC urothelial carcinoma of the bladder, 421 RC and 282 TMT who would have been eligible for both TMT or RC, treated at the Massachusetts General Hospital, Boston; Princess Margaret Cancer Centre, Toronto; and University of Southern California, Los Angeles between 2005-2017. To compare homogeneous cohorts, all patients included in this analysis had solitary tumors < 7 cm, no or unilateral hydronephrosis, and no multifocal carcinoma in situ. Treatment propensity scores were estimated using logistic regression, and patients were matched 3:1 with replacement. Covariates included age, sex, clinical T stage, hydronephrosis, (neo)adjuvant chemotherapy, body mass index, smoking history, and ECOG status. Overall survival (OS) was estimated with adjusted Cox models; cancer-specific survival (CSS), distant failure-free survival, pelvic nodal failure-free survival and metastasis-free survival (combined distant and pelvic nodal failure) were estimated with adjusted competing risk models. Our primary endpoint of interest was metastasis-free survival. The analysis was performed as intent-to-treat. Results: The 3:1 matched cohort comprised of 1,116 patients (834 RC vs 282 TMT). After matching, age (71.3 vs 71.6), cT2 clinical stage (88 vs 90%), presence of hydronephrosis (12 vs 10%), and use of (neo)adjuvant chemotherapy (60 vs 65%) were similar between RC and TMT cohorts. Salvage cystectomy was performed in 38 patients (13%) treated by TMT. At 5 years, metastasis-free (73 vs 78%, p = 0.07), distant failure-free (78 vs 82%, p = 0.14), and pelvic nodal failure-free (96 vs 94%, p = 0.33) survival were not statistically different between RC and TMT, whereas CSS and OS favored TMT (78 vs 85%, p = 0.02; 70 vs 78%, p < 0.001). Outcomes for RC and TMT were not different among centers. Final pT stage in the RC patients was: pT0 14%, pT1 7%, pT2 29%, pT3/4 42% and N+ 24%. Peri RC mortality was 2.1% and median number of nodes removed was 40. NMIBC recurrence occurred in 57/278 (20.5%) TMT patients. Conclusions: This large multi-institutional contemporary study provides the best evidence to date, in the absence of randomized trials, supporting TMT for select patients with MIBC. Oncologic outcomes seem to be equivalent between TMT and RC, affirming the position that TMT should be offered as an effective alternative.

Integrative clinical and molecular characterization of translocation renal cell carcinoma

Translocation renal cell carcinoma (tRCC) is a poorly characterized subtype of kidney cancer driven by MiT/TFE gene fusions. Here, we define the landmarks of tRCC through an integrative analysis of 152 patients with tRCC identified across genomic, clinical trial, and retrospective cohorts. Most tRCCs harbor few somatic alterations apart from MiT/TFE fusions and homozygous deletions at chromosome 9p21.3 (19.2% of cases). Transcriptionally, tRCCs display a heightened NRF2-driven antioxidant response that is associated with resistance to targeted therapies. Consistently, we find that outcomes for patients with tRCC treated with vascular endothelial growth factor receptor inhibitors (VEGFR-TKIs) are worse than those treated with immune checkpoint inhibitors (ICI). Using multiparametric immunofluorescence, we find that the tumors are infiltrated with CD8⁺ T cells, though the T cells harbor an exhaustion immunophenotype distinct from that of clear cell RCC. Our findings comprehensively define the clinical and molecular features of tRCC and may inspire new therapeutic hypotheses.

Predicting survival after radical prostatectomy: Variation of machine learning performance by race

BACKGROUND

Robust prediction of survival can facilitate clinical decision-making and patient counselling. Non-Caucasian males are underrepresented in most prostate cancer databases. We evaluated the variation in performance of a machine learning (ML) algorithm trained to predict survival after radical prostatectomy in race subgroups.

METHODS

We used the National Cancer Database (NCDB) to identify patients undergoing radical prostatectomy between 2004 and 2016. We grouped patients by race into Caucasian, African-American, or non-Caucasian, non-African-American (NCNAA) subgroups. We trained an Extreme Gradient Boosting (XGBoost) classifier to predict 5-year survival in different training samples: naturally race-imbalanced, race-specific, and synthetically race-balanced. We evaluated performance in the test sets.

RESULTS

A total of 68,630 patients met inclusion criteria. Of these, 57,635 (84%) were Caucasian, 8173 (12%) were African-American, and 2822 (4%) were NCNAA. For the classifier trained in the naturally race-imbalanced sample, the F1 scores were 0.514 (95% confidence interval: 0.513-0.511), 0.511 (0.511-0.512), 0.545 (0.541-0.548), and 0.378 (0.378-0.389) in the race-imbalanced, Caucasian, African-American, and NCNAA test samples, respectively. For all race subgroups, the F1 scores of classifiers trained in the race-specific or synthetically race-balanced samples demonstrated similar performance compared to training in the naturally race-imbalanced sample.

CONCLUSIONS

A ML algorithm trained using NCDB data to predict survival after radical prostatectomy demonstrates variation in performance by race, regardless of whether the algorithm is trained in a naturally race-imbalanced, race-specific, or synthetically race-balanced sample. These results emphasize the importance of thoroughly evaluating ML algorithms in race subgroups before clinical deployment to avoid potential disparities in care.

Combination MRI-targeted and systematic prostate biopsy may overestimate gleason grade on final surgical pathology and impact risk stratification

PURPOSE

Gleason grade (GG) on prostate biopsy is important for risk stratification and clinical decision making. Multiparametric MRI (mpMRI) improved detection of clinically significant disease and some studies suggest that MRI-fusion biopsy combined with systematic biopsy results in fewer upgrades on final surgical pathology. However, the downgrade rate is unclear and there is controversy in the literature. The objectives of this study are to assess the concordance of combination biopsy with final surgical pathology, and furthermore, to specifically determine downgrade rates.

MATERIALS AND METHODS

In our institutional mpMRI-ultrasound fusion biopsy database, 173 underwent targeted and systematic biopsy followed by radical prostatectomy (RP). GG on targeted, systematic and combination (targeted and systematic) biopsy were compared with GG on RP. Concordance rates between biopsy types were compared with the McNemar test. Proportion of GG upgrade or downgrade at the time of RP was also evaluated.

RESULTS

Surgical pathology was concordant with 44.5% of systematic biopsies, 46.8% of targeted biopsies, and 56.7% of combination biopsies. Combination biopsy significantly overestimated the final GG on RP compared to systematic biopsy (16.8% vs. 8.7% RR 1.93, 95% CI 1.36-2.75, P < 0.001). Downgrade rate from unfavorable to favorable intermediate-risk disease was 46.2%, and from high-risk to intermediate-risk disease was 45.1%.

CONCLUSIONS

Combination (targeted and systematic) biopsy is associated with the highest concordance rate between biopsy and RP pathology when compared with systematic or targeted biopsy alone. However, targeting MRI lesions and therefore the higher risk components, may at times overestimate the final surgical pathology which can result in overtreatment of what may truly be less aggressive disease.

Transperineal multiparametric magnetic resonance imaging-ultrasound fusion-targeted prostate biopsy combined with standard template improves perineural invasion detection

Perineural invasion (PNI) on biopsy is associated with adverse features in prostate cancer (PCa). Transrectal multiparametric magnetic resonance imaging (MRI)-targeted biopsy (TBx) has shown to detect higher presence of PNI than standard template biopsy (SBx). Transperineal biopsy provides effective cancer detection with lower complications than the transrectal approach. We compared PNI detection efficiency between SBx and TBx through transperineal approach. We identified patients with PCa who underwent transperineal TBx and concomitant standard 20-core template SBx from September 2019 to February 2021. Clinical, MRI imaging and biopsy characteristics were evaluated and compared between TBx and SBx. Two hundred thirty-eight patients with PCa underwent concomitant transperineal SBx and TBx procedures. Combined PNI+ (SBxPNI+ and/or TBxPNI+) was identified in 77 of 238 (32.4%) patients. SBx detected 23.9% PNI-positive patients and TBx detected 19.3% PNI-positive patients of all patients with PCa. Patients with PNI were with significantly different clinicopathological characteristics than patients without PNI. Although significantly more positive PCa cores and higher positive PCa core rate were found in the SBx method, patients with SBxPNI+ only shared similar features as TBxPNI+only patients. Of 176 cases with both SBxPCa and TBxPCa, TBx could detect 19 (15.1%) more PNI cases than SBx while SBx could detect 24 (18.3%) more PNI cases than TBx. Multiparametric MRI fusion-targeted biopsy in combination with template biopsy through transperineal approach achieved PNI detection rate over 30% of PCa cases. The increased PNI detection may improve the model to select active surveillance candidates in clinical practice.

To stage or not to stage: determining the true clinical significance of the biopsy tract through perinephric fat in assessing renal cell carcinoma

AIMS

Perinephric fat invasion (PFI) is a key component of renal cell carcinoma (RCC) staging, but there are limited data pertaining to biopsy tract seeding (BTS) resulting in perirenal tissue involvement [BTS with perinephric fat invasion (BTS-P)].The aim is to correlate clinical outcomes with pathologic stage to determine whether the presence of BTS-P should be considered a criterion to stage RCC as part of the pT3a category in the absence of any other upstaging variables.

MATERIALS AND RESULTS

We identified 304 renal biopsies from patients with subsequent nephrectomies for RCC; 33 of the tumours contained PFI. Each case was reviewed to determine the presence of BTS-P and other forms of invasion [e.g. non-BTS-P PFI, sinus fat invasion (SFI), and/or renal vein invasion (RVI)], and these findings were compared with survival outcomes. Ten (30%) of 33 tumours with PFI showed BTS-P as the only finding, and were otherwise pT1 tumours; six (60%) patients were alive without disease (AWOD) (mean, 77.5 months), three were lost to follow-up (LTF), and one died of other disease (DOOD). Two patients showed true PFI plus BTS-P; one was LTF and one is AWOD at 107 months. Ten (43%) of 23 patients with tumours with true invasion (PFI ± SFI and/or RVI) are AWOD (mean, 97.7 months), eight (35%) died of disease (DOD), four were LTF, and one DOOD. Kaplan-Meier survival curves showed that the cancer-specific survival was significantly worse in patients with true invasion (P = 0.044) than in those with BTS-P as the sole finding.

CONCLUSION

Patients with tumours showing BTS-P only appear to have better outcomes than those with other non-PFI invasion, suggesting that this finding should not be upstaged to pT3a. Additional studies are needed to corroborate the significance of our observations.

Conditional progression-free survival in men on active surveillance for prostate cancer stratified by NCCN risk

222

Background: Active surveillance (AS) is an accepted management strategy for men with very low, low, and select cases of favorable intermediate National Comprehensive Cancer Network (NCCN) risk prostate cancer (PCa). However, how patients’ risk of disease progression evolves over time during AS has not been well defined. Conditional survival measures the probability a patient will continue to survive some number of years, given that they have already survived a certain number without progression. We evaluated our AS cohort to investigate overall and conditional progression free survival on AS, stratified by the NCCN risk groups. Methods: We reviewed our institutional database of 1254 men enrolled in AS for localized PCa from 1996-2016. Our AS protocol includes prostate specific antigen (PSA) and digital rectal exam (DRE) every 4-6 months for 3 years, then annually. Mandatory confirmatory 12 core biopsy is done at 12-18 months. Multiparametric magnetic resonance imagining (mpMRI) or additional systematic or MRI-fusion biopsies are done at the discretion of physician and patient. Overall freedom from pathologic grade progression on follow-up biopsy and treatment free survival were estimated using the Kaplan-Meier method. Survival curves were compared pairwise using the Log-rank test and adjusted for false discovery rates with the Benjamini-Hochberg procedure. Three-year conditional survival estimates were derived for both outcomes from the Kaplan-Meier estimator. Results: Of 1254 men, 521 (41.6%) met criteria for very low, 606 (48.4%) for low, and 125 (10.0%) for favorable intermediate NCCN risk at diagnosis. Median follow-up time was 6.5 years (IQR 4.1-9.4). Median pathologic grade progression free survival in years was significantly longer for very low risk (7.8, 95% CI 6.8-11.2) compared to low risk men (5.6, 95% CI 4.7-6.9), however neither was significantly different from favorable intermediate risk men (5.9). There was no significant difference in treatment free survival between the three risk groups. At diagnosis, the three-year risk for pathologic grade progression (24%, 95% CI 21-27%) and progression to treatment (22%, 95% CI 20-25%) were similar. However, with increasing time of event-free AS, the conditional probability of pathologic grade progression increased, while that of progression to treatment decreased. Conclusions: Our results demonstrate that despite a mild increase in pathologic progression free survival in very low risk men, there was no clear difference in overall treatment free survival between very low, low, and select favorable intermediate NCCN risk men. Further, with increased time spent on AS, despite elevated rates of pathologic progression, patient progression to treatment decreased. This trend may be indicative of changes in goals of care as men with PCa age and should be closely monitored during AS.

The impact of a positive family history on clinical and pathologic outcomes of active surveillance for prostate cancer

225

Background: Active surveillance (AS) is the preferred management strategy for men with low-risk prostate cancer. However, approximately one in three men on AS experience progression of disease leading to treatment within 5 years, highlighting an urgent unmet need to reliably distinguish indolent from aggressive prostate cancer and improve patient selection criteria for AS. Germline genetic testing for DNA repair gene mutations is now recommended for patients with newly diagnosed prostate cancer and a strong family history of prostate cancer or BRCA1/2-related cancers, as such mutations have been associated with more aggressive forms of the disease. Here, we investigated the impact of family history on AS outcomes, under the hypothesis that men at high genetic risk for prostate cancer are at greater risk for progression to treatment on AS. Methods: We retrospectively reviewed detailed family history data of 958 patients from our institutional database of men enrolled in AS between 1997-2019. Data on family history of prostate cancer and hereditary cancer syndrome ( BRCA1/2-related prostate, breast, ovarian and/or pancreatic cancers) were collected and integrated into a composite family history score incorporating the number of relatives with each cancer weighted by degree of relatedness. A strong family history was defined as a composite score representing > 1 first-degree relative equivalent. The primary outcome was biopsy progression and secondary outcomes were adverse pathologic features at prostatectomy and biochemical recurrence. Statistical analysis was conducted using the Kaplan-Meier method and Cox proportional hazards regression. Results: In univariate analysis, a strong family history suggestive of a hereditary cancer syndrome (HR 1.37 [1.03-1.90], P = 0.033) was associated with a significant increased risk of biopsy progression; however, any family history of prostate cancer (HR 1.10 [0.89-1.35], P = 0.38) and a strong family history of prostate cancer (HR 1.35 [0.92-1.98], P = 0.13) were not significant. In multivariate analysis, a strong family history suggestive of a hereditary cancer syndrome remained a statistically significant predictor of biopsy progression (HR 1.42 [1.03-1.96], P = 0.03), after adjusting for age, percent core involvement on initial biopsy and PSA density. No significant association was found between family history and adverse features on surgical pathology or biochemical recurrence. Conclusions: A positive family history suggestive of a hereditary cancer syndrome is associated with an increased risk of biopsy progression on AS and is an independent predictor of biopsy progression. Men with such a family history may still be safely offered AS but should be counseled about the higher risk of progression. Further work to investigate the underlying genetic factors responsible for this increased risk is warranted.

Resolution of a High Grade and Metastatic BK Polyomavirus-Associated Urothelial Cell Carcinoma Following Radical Allograft Nephroureterectomy and Immune Checkpoint Treatment: A Case Report

BACKGROUND

BK viral infection in the posttransplant setting continues to cause serious morbidity with effects ranging from allograft nephropathy and dysfunction to urothelial malignancy.

RESULTS

In this report, we present a patient that developed BK-associated nephropathy and, 6 years later, locally advanced urothelial malignancy in the renal allograft with nodal, muscle, and extremity involvement. Following radical allograft nephroureterectomy, he was treated with palliative radiation and the immune checkpoint inhibitor atezolizumab. Follow-up imaging at 1 year demonstrated radiographic complete response.

CONCLUSIONS

This report supports the growing body of evidence supporting the association of urothelial malignancy and BK virus infection in renal transplant recipients. Further, it highlights the novel application of immune checkpoint inhibitors in the treatment of advanced posttransplant malignancy, in particular when the allograft is removed and the tumor is possibly of donor origin.

Viral integration in BK polyomavirus-associated urothelial carcinoma in renal transplant recipients: multistage carcinogenesis revealed by next-generation virome capture sequencing

BK polyomavirus (BKPyV)-associated cancer after transplantation has gained increasing attention. However, the role of BKPyV integration on oncogenesis is still unclear. In this study, next-generation virome capture sequencing of primary and metastatic tumors were performed in three patients with BKPyV-associated urothelial carcinoma after renal transplantation. As a result, a total of 332 viral integration sites were identified in the six tumors. Integration of BKPyV in both primary and metastatic tumors followed the mechanism of microhomology-mediated end joining mostly, since microhomologies between human and BKPyV genomes were significantly enriched in flanking regions of 84% of the integration sites. Viral DNA breakpoints were nonrandom and tended to assemble in large T gene, small T gene and viral protein 2 gene. There were three, one and one consensus integration sites between the primary and metastatic tumors, which affected LINC01924, eIF3c, and NEIL2 genes in the three cases respectively. Thus, we concluded that integration of BKPyV was a continuous process occurring in both primary and metastatic tumors, generating heterogenous tumor cell populations. Through this ongoing process, certain cell populations might have gained growth advantage or metastatic potential, as a result of viral integration either affecting the cellular genes where the viral DNA integrated to or altering the expression or function of the viral genes.

Renal biopsy cell cycle proliferation (CCP) score to predict adverse surgical pathology in renal cell carcinoma

747

Background: The role of renal mass biopsy (RMB) in risk stratifying patients with renal cell carcinoma (RCC) is unclear. We sought to determine whether the cell cycle proliferation (CCP) score from RMB can improve risk stratification of localized RCC. Methods: We identified patients with RCC who underwent RMB and subsequent partial/radical nephrectomy from 2000-2014. We used multivariable logistic regression to determine the association of patient-level variables and biopsy CCP score with adverse surgical pathology (Fuhrman grade 3-4, pT stage≥3, papillary type II histology, or evidence of metastasis at surgery). Effect size was estimated with odds ratios (OR) and discriminative performance with AUC. Results: Overall, 94 of 202 patients (46%) had adverse surgical pathology. On multivariable analysis, CCP score >0 was associated with 2.38-fold increased odds of adverse pathology (Table). Relative to the model omitting CCP score (AUC=0.70), the addition of CCP score as a continuous (AUC=0.731) or binary (AUC=0.730) variable yielded increased discriminative performance. Similar associations were observed in an analysis limited to patients with low-grade tumors on biopsy (bCCP: OR 2.44, p=0.024; cCCP: OR 1.57, p=0.11). In both models, increased lesion size on imaging was consistently associated with adverse pathology. Conclusions: Among patients with RCC, biopsy CCP score >0 was independently associated with adverse pathology, suggesting this classifier provides prognostic information beyond conventional pathologic data. Biopsy CCP score could be used to better guide patient-specific management.[Table: see text]

Detection of clinically significant prostate cancer after negative fusion and systematic biopsy

288

Background: MRI/Ultrasound fusion biopsy of the prostate has enhanced the detection of clinically significant prostate cancer (csPCa). Detection of csPCa is greatest when fusion and systematic biopsies are combined. However, the finding of a negative fusion and negative systematic biopsy in patients with suspicious lesion on imaging raises the question of either falsely positive imaging or a false negative biopsy. Methods: We retrospectively reviewed our database of patients undergoing MRI/transrectal US-guided fusion biopsy. All images were graded according to the Prostate Imaging Reporting and Data System version (PIRADS) 2.0. Patients underwent targeted biopsy followed by systematic 12-core double sextant biopsy within the same session. csPCa was defined as Grade Group (GG) ≥2 PCa. Patients with no prostate cancer (PCa) found on biopsies were followed. MRI studies with PIRADS v2 score ≤ 2 were considered to have no MRI evidence of PCa. Results: A total of 400 patients had at least one PIRADS ≥3 lesion and underwent fusion/systematic biopsy. Of these, 113 (28.3%) patients had no evidence of PCa on either fusion or systematic biopsy. Median follow-up was 32.5 months. 44 (39%) patients underwent repeat MRI and of these, 24 (54%) had no evidence of PCa on repeat MRI. PIRADS lesion disappearance was associated with lower PSA Density (PSAd) (0.12 vs 0.20; P = 0.0319) and decreased progression to repeat biopsy (8.33% vs 95%; P < 0.0001). Patients who had a repeat biopsy had a greater PSAd ( 0.21 vs 0.12; P = 0.0054). Of 113 patients with negative initial biopsy, 23 (20.4 %) underwent repeat biopsy: 16 (14.2 %) had PCa and 11 (9.7%) had csPCa. Thus, 48% of patients who underwent repeat biopsy had csPCa. Among patients with a PCa on repeat biopsy, cancer was sampled by MRI targeted cores in 80% of patients. Conclusions: Despite a negative initial fusion/systematic biopsy, at least 10% of patients were subsequently diagnosed with clinically significant PCa. The combination of elevated PSAd and the persistence of a suspicious lesion on repeat MRI appears selective for previously missed PCa. However, after negative fusion biopsy, repeat MRI yields a high rate of PIRADS lesion disappearance in patients with low PSAd.

Using high-resolution magic angle spinning magnetic resonance spectroscopy to characterize the metabolomic profile of renal cell carcinoma

710

Background: Renal cell carcinoma (RCC) is a metabolic disease, with the various subtypes exhibiting aberrations in different metabolic pathways. Metabolomics may offer greater sensitivity for revealing disease biology than evaluation of tissue morphology. In this study, we investigate the metabolomic profile of RCC using high resolution magic angle spinning (HRMAS) magnetic resonance spectroscopy (MRS). Methods: Tissue samples were obtained from radical or partial nephrectomy specimens that were fresh frozen & stored at -80ºC. Tissue HRMAS MRS was performed by a Bruker AVANCE spectrometer. Metabolomic profiles of RCC & adjacent benign renal tissue were compared, and false discovery rates (FDR) were used to account for multiple testing. Regions of interest (ROI) with FDR < 0.05 were selected as potential predictors of malignancy. The Wilcoxon rank sum test was used to compare median MRS relative intensities for the candidate predictors. Logistic regression was used to determine odds ratios for risk of malignancy based on abundance of each metabolite. Results: There were 38 RCC (16 clear cell, 11 papillary, 11 chromophobe) & 13 adjacent normal tissue specimens (matched pairs). Metabolic predictors of malignancy based on FDR include histidine, phenylalanine, phosphocholine, serine, phosphocreatine, creatine, glycerophosphocholine, valine, glycine, myo-inositol, scylla-inositol, taurine, glutamine, spermine, acetoacetate & lactate. Higher levels of spermine, histidine & phenylalanine at 3.15-3.13 ppm were associated with a decreased risk of RCC (OR 4x10−5, 95% CI 7.42x10−8, 0.02), while 2.84-2.82 ppm increased the risk of malignant pathology (OR 7158.67, 95% CI 6.3, 8.3x106), and the specific metabolites characterizing this region remain to be identified. Tumor stage did not appear to affect the metabolomics of malignant tumors, suggesting that metabolites are more dependent on histologic subtype. Conclusions: HRMAS-MRS identified many metabolites that may predict RCC. We demonstrated that those in the 3.14-3.13 ppm ROI were present in lower levels in RCC, while higher levels of metabolites in the 2.84-2.82 ppm ROI substantially increased the risk of RCC.

Using high-resolution magic angle spinning magnetic resonance spectroscopy to characterize the metabolomic profile of fat-poor angiomyolipoma and renal cell carcinoma

711

Background: Fat-poor angiomyolipoma (AML) can be difficult to differentiate from renal cell carcinoma (RCC) radiographically and may lead to biopsy or unnecessary intervention. In vivoplatforms with the ability to identify tumor histology based on metabolic profiles may avoid unnecessary procedures & their complications. The metabolomics of AML have not been characterized, & research into this area may provide targetable molecules for large AMLs. In this study, we investigate the metabolomic profile of AMLs compared to clear cell RCC (ccRCC) using high resolution magic angle spinning (HRMAS) magnetic resonance spectroscopy (MRS). Methods: Tissue samples were obtained from radical or partial nephrectomy specimens that were fresh frozen & stored at -80ºC. Tissue HRMAS MRS was performed by a Bruker AVANCE spectrometer. Metabolomic profiles of RCC & adjacent benign renal tissue were compared, and false discovery rates (FDR) accounted for multiple testing. Regions of interest (ROI) with FDR <0.05 were considered potential predictors of ccRCC rather than AML. The Wilcoxon rank sum test was used to compare median MRS relative intensities for candidate predictors. Logistic regression was used to determine odds ratios for risk of malignancy based on abundance of each metabolite. Results: There were 16 ccRCC samples & 7 AML specimens. Candidate predictors of malignancy rather than AML based on FDR p-values include histidine, phenylalanine, phosphocholine, serine, alanine, glutamate, glutathione, glycerophosphocholine, & glutamine. While an abundance of these metabolites is associated with higher risk of malignancy, the odds ratio was particularly high in the 3.5-3.49 ppm spectral region (OR 2.99x106, 95% CI 3.27, 2.73x1012, p=0.033)in ccRCC samples. Conclusions: HRMAS MRS identified metabolites that may help differentiate fat-poor AML from ccRCC. In particular, metabolites in the 3.5-3.49 ppm spectral region increased the risk of harboring RCC. Our findings may contribute to future in vivostudies to help identify which patients require intervention for malignancy & which may be observed for benign AML without requiring biopsy.

Concordance of systematic and fusion biopsy with surgical pathology

93

Background: Multiparametric MRI is increasingly used in prostate cancer detection. Previous studies have shown that detection rate of clinically significant cancer is higher in MRI targeted biopsy than systematic biopsy. However, the concordance between the Gleason score on fusion biopsy and radical prostatectomy is less well known. The objective of this study is to look for predictors of histopathologic concordance between biopsy (fusion and systematic) and radical prostatectomy. Methods: We used an institutional database of men who underwent mpMRI-ultrasound fusion targeted and systematic biopsy followed by radical prostatectomy. Gleason score on targeted, systematic and combination (targeted + systematic) biopsy were compared with Gleason score on radical prostatectomy, and concordance was recorded. The McNemar test was used to compare concordance and upgrade rates. Predictors of concordance and upgrade such as age, prostate volume, PSA, PSA density, Gleason score on biopsy, number of targets reported on mpMRI, and PI-RADS score were evaluated with Fisher’s exact test and logistic regression. Results: Surgical pathology was concordant with 47.4% of systematic biopsies, 52.0% of targeted biopsies and 58.4% of combination biopsies. There was no significant difference in concordance rates between systematic and targeted biopsy (P = 0.37). However, combination biopsy was superior to both systematic (RR 1.23, 95% CI 1.08-1.40, P = 0.03) and targeted biopsy (RR 1.12, 95% CI 1.02 – 1.24, P = 0.03) in predicting concordance with surgical pathology. Risk of upgrade to a higher Gleason score on surgical pathology was significantly lower with combination biopsy compared to systematic (RR 0.57, 95% CI 0.46-0.69, P < 0.001) or targeted biopsy alone (RR 0.72, 95% CI 0.61-0.84, P = 0.001). Upgrade rates were 43.9% for systematic biopsy, 34.7% for targeted, and 24.9% for combination. Lastly, we found no significant predictors of concordance or upgrade. Conclusions: Combination biopsy is associated with the highest concordance rate between biopsy and radical prostatectomy when compared with systematic or targeted biopsy alone. Performing targeted biopsy alone will underestimate tumour aggressiveness on surgical pathology.

Current Management of Small Renal Masses, Including Patient Selection, Renal Tumor Biopsy, Active Surveillance, and Thermal Ablation

Renal cancer represents 2% to 3% of all cancers, and its incidence is rising. The increased use of ultrasonography and cross-sectional imaging has resulted in the clinical dilemma of incidentally detected small renal masses (SRMs). SRMs represent a heterogeneous group of tumors that span the full spectrum of metastatic potential, including benign, indolent, and more aggressive tumors. Currently, no composite model or biomarker exists that accurately predicts the diagnosis of kidney cancer before treatment selection, and the use of renal mass biopsy remains controversial. The management of SRMs has changed dramatically over the last two decades as our understanding of tumor biology and competing risks of mortality in this population has improved. In this review, we critically assess published consensus guidelines and recent literature on the diagnosis and management of SRMs, with a focus on patient treatment selection and use of renal mass biopsy, active surveillance, and thermal ablation. Finally, we highlight important opportunities for leveraging recent research discoveries to identify patients with SRMs at high risk for renal cell carcinoma-related mortality and minimize overtreatment and patient morbidity.

Comparative Effectiveness of Bladder-preserving Tri-modality Therapy Versus Radical Cystectomy for Muscle-invasive Bladder Cancer

INTRODUCTION

There are limited randomized data comparing radical cystectomy (RC) with bladder-sparing tri-modality therapy (TMT) in the treatment of muscle-invasive bladder cancer (MIBC). Both strategies are thought to have similar survival outcomes with different morbidity profiles. We compare the effectiveness of TMT and RC using decision-analytic modeling and the endpoint of quality-adjusted life years (QALYs).

PATIENTS AND METHODS

Using a Markov model, we simulated the lifetime outcomes after TMT versus RC ± neoadjuvant chemotherapy for 67-year-old patients with clinical stage T2-T4aN0M0 MIBC. Model probabilities and utilities were extracted from the literature. The incremental effectiveness was reported in QALYs and sensitivity analyses were performed.

RESULTS

For all patients with MIBC, although the model showed identical survival, TMT was the most effective strategy with an incremental gain of 0.59 QALYs over RC (7.83 vs. 7.24 QALYs, respectively). When limiting the model to favorable, contemporary cohorts in both the TMT and RC strategies, TMT remained more effective with an incremental gain of 1.61 QALYs (9.37 vs. 7.76 QALYs, respectively). One-way sensitivity analyses demonstrated the model was sensitive to the quality of life parameters (ie, the utilities) for RC and TMT. When testing the 95% confidence interval of the RC utility parameter the model demonstrated an incremental gain with TMT from -0.54 to 4.23 QALYs. Probabilistic sensitivity analysis demonstrated that TMT was more effective than RC for 63% of model iterations.

CONCLUSIONS

This modeling study found that treatment of MIBC with organ-sparing TMT in appropriately-selected patients may result in a gain of QALYs relative to RC.

The prognostic impact of a negative confirmatory biopsy in men on active surveillance for prostate cancer

76

Background: Active surveillance (AS) is increasingly used in managing low-risk and favorable intermediate-risk prostate cancer. To mitigate the risk of unsampled higher risk disease, most institutional AS protocols call for a confirmatory prostate biopsy within 12-18 months following initial diagnostic biopsy. Here, we investigate whether the results of confirmatory biopsy impact the outcomes of men on AS. Methods: We retrospectively reviewed our institutional database of men enrolled in AS between 1997-2014 with a minimum follow-up of 6 months (n = 974). Biopsies with any prostate cancer were considered positive. Biopsies containing only benign prostatic tissue, prostatic intraepithelial neoplasia (PIN), or atypical small acinar proliferation (ASAP) were considered negative. Statistical analysis was conducted using the Kaplan-Meier method and Cox proportional hazards regression. Results: At diagnosis, median age was 67 years (IQR 62-72) and median PSA was 5.1 ng/mL (IQR 3.9-6.8). The vast majority of patients had Gleason ≤6 (97%) and clinical stage T1 (92%) disease. With a median follow-up of 4.8 years, 702 (72%) patients underwent a confirmatory biopsy. 67% of confirmatory biopsies were positive for prostate cancer; 33% were negative (167 benign, 40 PIN, and 22 ASAP). Of the 702 patients, 33% progressed to treatment, with pathologic progression the most common reason (77%). Univariate predictors of progression to treatment included initial clinical stage ( P= 0.04), involvement of > 20% of any core on diagnostic biopsy ( P < 0.01), PSA density ≥0.15 ( P < 0.001), and confirmatory biopsy status ( P < 10-14). In multivariate analysis, a negative confirmatory biopsy remained the strongest predictor of progression to treatment (HR 0.12 [95%CI 0.06-0.24], P < 10-8). Confirmatory biopsy status was not associated with risk of adverse pathology on RP, metastasis-free survival, disease-specific survival, or overall survival. Conclusions: A negative confirmatory biopsy is associated with a significantly lower rate of progression to treatment among men on AS. This may serve as a useful tool for prognostication and help determine the need for further repeat biopsies for men on AS.