Predicting the clinical course of prostate cancer

Abstract

Risk stratification in prostate cancer remains a significant clinical challenge. A study in this issue of the JCI describes an exciting application of high-throughput functional genomic technology to further refine our understanding of treatment failure risk in prostate cancer patients .

Since Walsh and Donker first described the pelvic anatomy that allowed for the development of the nerve-sparing anatomic radical prostatectomy in 1982, the morbidity associated with the surgical treatment of clinically localized prostate cancer has decreased substantially (1). The subsequent advent of prostate-specific antigen (PSA) screening has led to a substantial stage migration in newly diagnosed adenocarcinoma of the prostate, and this has resulted in a higher likelihood of surgical cure. Despite these therapeutic advances, our ability to accurately predict the risk of treatment failure for an individual patient with prostate cancer remains limited. The current tools we utilize to guide critical decisions, such as whether or how aggressively to treat prostate cancer, are based on serum PSA levels, biopsy Gleason score, and clinical stage. Despite the incorporation of powerful multifactorial nomograms into our decision process, the ability to predict individual patient outcome remains limited (2, 3).

Novel prognostic indicators

In this issue of the JCI, a report by Glinsky et al. attempts to advance our understanding and ability to stratify the risk of treatment failure for patients with localized prostate cancer undergoing radical prostatectomy (4). Relying on the enormous power of recent advances in functional human genomics, Glinsky et al. have analyzed the gene expression profiles of human prostate cancer samples in a total of 100 different tumors. The authors screened 21 patients for 12,625 gene transcripts to identify genes that would predict relapse-free survival following radical prostatectomy using the Affymetrix GeneChip system. Subsequently, using a smaller set of highly discriminating five-gene clusters, they were able to predict clinical outcome in a larger validation set of 79 patients.

This study represents the largest clinical series of genomic classification in prostate cancer published to date using a high-throughput functional human genomic technique. The virtually unlimited power of microarray technology to discriminate minute genetic differences in clinical specimens has the potential to revolutionize the power of clinicians to accurately identify high-risk groups of patients (5). However, several potential drawbacks in the current study should be noted. First, this study was performed using the postoperative radical prostatectomy specimen as the tissue source. This is likely due to the greater ease of tissue acquisition in this setting. In order to best assist in clinical decision making, predictive information should be available prior to definitive treatment. This would require DNA analysis of transrectal ultrasound–guided biopsy material. This approach is subject to the unavoidable risk of biopsy sampling error due to the known heterogeneity of multifocal prostate cancer. This heterogeneity is currently manifested by assessments of variability of Gleason Score within a given tumor. It remains to be determined whether distinct foci of cancer within a given tumor will have similar or different genetic profiles. The postoperative risk stratification utilized in the current study is only able to assist in choosing patients in need of early adjuvant therapy or closer follow-up.

Statistical analysis

In addition, although the authors present a detailed analysis of their data that suggests that these discriminating gene clusters can predict clinical outcome of prostate cancer beyond the traditional factors such as stage, Gleason score, and PSA level, their univariate approach to this analysis is flawed (4). By separating patient groups into high Gleason score or high PSA or high stage and then allowing the cluster analysis to discriminate outcome, they have limited the power of these traditional risk factors, which should be considered simultaneously in each case. A potentially more powerful approach to this problem would be to add gene cluster analysis data in a multivariate model which already includes stage, Gleason score, and PSA level to determine if the predictive ability of gene cluster analysis is in fact additive to current approaches (6). It remains to be determined how accurate and individually predictive such an approach will be. Further stratification of a given patient’s risk of treatment failure is important, but the ultimate goal must remain accurate and reproducible individual prediction. Only then will such approaches accomplish more than altering informed consent in selection of therapy.

Despite these limitations the presented data should be heralded as a significant advance in our ability to leverage the awesome power of functional genomics in a clinically useful fashion. It is this translational approach to risk stratification that is most likely to lead to progress in our understanding of each individual prostate cancer patient’s risk of treatment failure and allow us to intelligently counsel patients regarding what today remains a complex and often confusing array of treatment options.

Footnotes

See the related article beginning on page 913.

Nonstandard abbreviations used: prostate-specific antigen (PSA).

Conflict of interest: The authors have declared that no conflict of interest exists.

References

1. Walsh, PC, Donker, PJ. Impotence following radical prostatectomy: insight into etiology and prevention. J. Urol. 1982. 128:492-497.
   View this article via: PubMed Google Scholar
2. Kattan, MW, Eastham, JA, Stapleton, AM, Wheeler, TM, Scardino, PT. A preoperative nomogram for disease recurrence following radical prostatectomy for prostate cancer. J. Natl. Cancer Inst. 1998. 90:766-771.
   View this article via: PubMed CrossRef Google Scholar
3. D’Amico, AV, et al. Pretreatment nomogram for prostate-specific antigen recurrence after radical prostatectomy or external-beam radiation therapy for clinically localized prostate cancer. J. Clin. Oncol. 1999. 17:168-172.
   View this article via: PubMed Google Scholar
4. Glinsky, GV, Glinskii, AB, Stephenson, AJ, Hoffman, RM, Gerald, WL. Gene expression profiling predicts clinical outcome of prostate cancer. J. Clin. Invest. 2004. 113:913-923. doi:10.1172/JCI200420032.
   View this article via: JCI PubMed Google Scholar
5. Glinsky, GV, Ivanova, YA, Glinskii, AB. Malignancy-associated regions of transcriptional activation: gene expression profiling identifies common chromosomal regions of recurrent transcriptional activation in human prostate, breast, ovarian and colon cancers. Neoplasia. 2003. 5:221-228.
6. Kattan, MW, Wheeler, TM, Scardino, PT. Postoperative nomogram for disease recurrence after radical prostatectomy for prostate cancer. J. Clin. Oncol. 1999. 17:1499-1507.
   View this article via: PubMed Google Scholar