Karnes, R. Jeffery

In medical research, the Brier score (BS) and the area under the receiver operating characteristic (ROC) curves (AUC) are two common metrics used to evaluate prediction models of a binary outcome, such as using biomarkers to predict the risk of developing a disease in the future. The assessment of an existing prediction models using data with missing covariate values is challenging. In this article, we propose inverse probability weighted (IPW) and augmented inverse probability weighted (AIPW) estimates of AUC and BS to handle the missing data. An alternative approach uses multiple imputation (MI), which requires a model for the distribution of the missing variable. We evaluated the performance of IPW and AIPW in comparison with MI in simulation studies under missing completely at random, missing at random, and missing not at random scenarios. When there are missing observations in the data, MI and IPW can be used to obtain unbiased estimates of BS and AUC if the imputation model for the missing variable or the model for the missingness is correctly specified. MI is more efficient than IPW. Our simulation results suggest that AIPW can be more efficient than IPW, and also achieves double robustness from miss‐specification of either the missingness model or the imputation model. The outcome variable should be included in the model for the missing variable under all scenarios, while it only needs to be included in missingness model if the missingness depends on the outcome. We illustrate these methods using an example from prostate cancer.

Objective To describe the natural history, reconstructive solutions, and functional outcomes of those men undergoing pubectomy and urinary reconstruction after prostate cancer treatment. Patients and Methods This study retrospectively identified 25 patients with a diagnosis of urosymphyseal fistula (UF) following prostate cancer therapy who were treated with urinary reconstruction with pubectomy. This study describes the natural history, reconstructive solutions, and functional outcomes of this cohort. Results All 25 patients had a history of pelvic radiotherapy for prostate cancer. The median (interquartile range [IQR]) time from prostate cancer treatment to diagnosis of UF was 11 (6, 16.5) years. The vast majority of men (24/25; 96%) presented with debilitating groin pain during ambulation. Posterior urethral stenosis was common (20/25; 80%), with 60% having repetitive endoscopic treatments. Culture of pubic bone specimens demonstrated active infection in 80%. Discordance between preoperative urine and intraoperative bone cultures was common, 21/22 (95.5%). After surgery, major 90‐day complications (Clavien–Dindo Grade III and IV) occurred in eight (32%) patients. Pain was significantly improved, with resolution of pain (24/25; 96%) and restoration of function, the median (IQR) preoperative Eastern Cooperative Oncology Group Performance Status (ECOG PS) was 3 (2, 3) vs median postoperative ECOG PS score of 0 (0, 1). Conclusion Endoscopic urethral manipulation after radiation for prostate cancer is a risk factor for UF. Conservative management will not provide symptom resolution. Fistula decompression, bone resection, and urinary reconstruction effectively treats chronic infection, improves pain and ECOG PS scores.

AbstractBackgroundTo assess whether a prespecified statistical model based on the four kallikrein markers measured in blood—total, free, and intact prostate‐specific antigen (PSA), together with human kallikrein‐related peptidase 2 (hK2)—or any individual marker measured in pretreatment serum were associated with biochemical recurrence‐free (BCR) or metastasis‐free survival after radical prostatectomy (RP) in a subgroup of men with very high‐risk disease.MethodsWe identified 106 men treated at Mayo Clinic from 2004 to 2008 with pathological Gleason grade group 4 to 5 or seminal vesicle invasion at RP. Univariable and multivariable Cox models were used to test the association between standard predictors (Kattan nomogram and GPSM [Gleason, PSA, seminal vesicle and margin status] score), kallikrein panel, and individual kallikrein markers with the outcomes.ResultsBCR and metastasis occurred in 67 and 30 patients, respectively. The median follow‐up for patients who did not develop a BCR was 10.3 years (interquartile range = 8.2‐11.8). In this high‐risk group, neither Kattan risk, GPSM score, or the kallikrein panel model was associated with either outcome. However, after adjusting for Kattan risk and GPSM score, separately, preoperative intact PSA was associated with both outcomes while hK2 was associated with metastasis‐free survival.ConclusionsConventional risk prediction tools were poor discriminators for risk of adverse outcomes after RP (Kattan risk and GPSM risk) in patients with very high‐risk disease. Further studies are needed to define the role of individual kallikrein marker forms in the blood to predict adverse prostate cancer outcomes after RP in this high‐risk setting.

Abstract Purpose: The heterogeneity of androgen receptor (AR)-activity (AR-A) is well-characterized in heavily treated metastatic castration-resistant prostate cancer (mCRPC). However, the diversity and clinical implications of AR-A in treatment-naïve primary prostate cancer is largely unknown. We sought to characterize AR-A in localized prostate cancer and understand its molecular and clinical implications. Experimental Design: Genome-wide expression profiles from prostatectomy or biopsy samples from 19,470 patients were used, all with independent pathology review. This was comprised of prospective discovery (n = 5,239) and validation (n = 12,728) cohorts, six retrospective institutional cohorts with long-term clinical outcomes data (n = 1,170), and The Cancer Genome Atlas (n = 333). Results: A low AR-active subclass was identified, which comprised 9%–11% of each cohort, and was characterized by increased immune signaling, neuroendocrine expression, and decreased DNA repair. These tumors were predominantly ERG and basal subtype. Low AR-active tumors had significantly more rapid development of recurrence or metastatic disease across cohorts, which was maintained on multivariable analysis [HR, 2.61; 95% confidence interval (CI), 1.22–5.60; P = 0.014]. Low AR-active tumors were predicted to be more sensitive to PARP inhibition, platinum chemotherapy, and radiotherapy, and less sensitive to docetaxel and androgen-deprivation therapy. This was validated clinically, in that low AR-active tumors were less sensitive to androgen-deprivation therapy (OR, 0.41; 95% CI, 0.21–0.80; P = 0.008). Conclusions: Leveraging large-scale transcriptomic data allowed the identification of an aggressive subtype of treatment-naïve primary prostate cancer that harbors molecular features more analogous to mCRPC. This suggests that a preexisting subgroup of patients may have tumors that are predisposed to fail multiple current standard-of-care therapies and warrant dedicated therapeutic investigation.

Abstract Introduction: Multivariable models that predict biochemical recurrence (BCR) can indicate patients at higher risk for developing clinical progression after prostatectomy. However, most patients that experience BCR do not metastasize or die from prostate cancer, making BCR an unreliable surrogate endpoint. In this study we compared the clinical and genomic characteristics of prostatectomy patients in order to better delineate the relationship between BCR and aggressive prostate cancer. Methods: Postoperative patients were retrospectively classified into three outcome groups: NED (no evidence of disease recurrence); BCR-only (patients that experienced BCR but didn't have progressive disease as evidenced by bone or CT scans); and CP (clinical progression with BCR plus positive CT or bone scans). Patient specimens were from the Mayo Clinic Radical Prostatectomy Registry, selected using a nested case-control design with 14 years median follow-up. RNA expression levels from FFPE tumor specimens were measured with 1.4 million feature whole-transcriptome oligonucleotide microarrays on 106 matching patients from each outcome group (total 318 patients). We analyzed the clinical covariates of the outcome groups in a pairwise manner (NED vs. BCR-only; NED vs. CP; CP vs. BCR-only), using a conditional logistic model in univariable and multivariable analyses. We performed feature reduction for the genomic data set using Kruskal-Wallis tests and regularized logistic regression. Principal Component Analysis (PCA) was used to quantify the variance of the genomic data with respect to the patient outcome groups. Results: Our initial hypothesis was that all three outcome groups would have different clinical and genomic characteristics. Univariable analysis of clinical variables failed to show statistically significant differences between NED and BCR-only groups. Multivariable analysis of the clinical variables also failed to show statistically significant differences between NED and BCR-only groups. Among the clinical variables only pathological Gleason score was statistically significant when comparing the BCR-only vs. CP and NED vs. CP groups. Genomic analysis found 363 coding and non-coding features that passed thresholds for differential expression between the three outcome groups. PCA established that the primary source of variance in the genomic data stemmed from the separation of CP from NED and BCR-only patients; whereas only a small amount of variance separates NED and BCR-only groups. The 363 features were further compared in the three outcome groups using Wilcoxon tests, adjusting for false-discovery using Benjamini-Hochberg. At a cutoff of p < 0.001 there were no features differentially expressed between NED and BCR-only, while 52 and 34 statistically significant features were found in the NED vs. CP and CP vs. BCR-only comparisons, respectively. Conclusion: Considering the clinical and genomic results together, we find that: 1) Patients that developed clinical progression were significantly different from those that did not, irrespective of whether they experienced BCR or not; 2) the NED and BCR-only groups were statistically very similar to each other. We therefore conclude that BCR may not be the most appropriate endpoint for predictive models that attempt to stratify patients at risk for progression. This abstract is also presented as Poster B44. Citation Format: Anamaria Crisan, Elai Davicion, R. Jeffery Karnes, Robert B. Jenkins, Mercedeh H. Ghadessi, Christine Buerki, Eric J. Bergstrahl, Nicholas Erho, Ismael A. Vergara, Karla V. Ballman, Peter C. Black, Timothy J. Triche. Biochemical recurrence is not a definitive surrogate endpoint for development of clinically useful predictive models for post-prostatectomy patients [abstract]. In: Proceedings of the AACR Special Conference on Advances in Prostate Cancer Research; 2012 Feb 6-9; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2012;72(4 Suppl):Abstract nr PR4.

AbstractPurpose: Conventional therapies to treat prostate cancer (CaP) of androgen-dependent phenotype (ADPC) and castration-resistant phenotype (CRPC) are deficient in outcome which has necessitated a need to identify those agents that could target AR for both disease types. We provide mechanism-based evidence that lupeol (Lup-20(29)-en-3b-ol) is a potent inhibitor of androgen receptor (AR) in vitro and in vivo.Experimental Design: Normal prostate epithelial cell (RWPE-1), LAPC4 (wild functional AR/ADPC), LNCaP (mutant functional/AR/ADPC), and C4-2b (mutant functional/AR/CRPC) cells were used to test the anti-AR activity of lupeol. Cells grown under androgen-rich environment and treated with lupeol were tested for proliferation, AR transcriptional activity, AR competitive ligand binding, AR–DNA binding, and AR–ARE/target gene binding. Furthermore, in silico molecular modeling for lupeol–AR binding was done. Athymic mice bearing C4-2b and LNCaP cell–originated tumors were treated intraperitoneally with lupeol (40 mg/kg; 3 times/wk) and tumor growth and surrogate biomarkers were evaluated. To assess bioavailability, lupeol serum levels were measured.Results: Lupeol significantly inhibited R1881 (androgen analogue) induced (i) transcriptional activity of AR and (ii) expression of PSA. Lupeol (i) competed antagonistically with androgen for AR, (ii) blocked the binding of AR to AR-responsive genes including PSA, TIPARP, SGK, and IL-6, and (iii) inhibited the recruitment of RNA Pol II to target genes. Lupeol sensitized CRPC cells to antihormone therapy. High-performance liquid chromatography analysis showed that lupeol is bioavailable to mice. Lupeol inhibited the tumorigenicity of both ADPC and CRPC cells in animals. Serum and tumor tissues exhibited reduced PSA levels.Conclusion: Lupeol, an effective AR inhibitor, could be developed as a potential agent to treat human CaP. Clin Cancer Res; 17(16); 5379–91. ©2011 AACR.