James Hebert

Sc.D

University of South Carolina, Cancer Prevention and Control Program

Distinguished Professor, Director

PLCO (Learn more about this study)

PLCO-15

Feb 22, 2013

Prostate Specific Antigen (PSA) Growth Curves: an Innovation to Improve Prostate Cancer Screening

The increased detection of non-significant Prostate Cancer (PrCA) forms the basis of recent controversies on PSA-based screening recommendations. Because PrCA can be aggressive (especially in Blacks and in younger men diagnosed with the disease) the challenge is to improve screening specificity for clinically significant disease (i.e., PrCA associated with a high probability of disability and death and therefore should necessitate medical treatment) while distinguishing non-significant PrCA that has little likelihood of causing significant harm. We aim to analyze repeated measures of PSA and medical treatment data to develop a tool that will improve both sensitivity and specificity of the PSA-based screening test to detect significant PrCA, while differentiating it from non-significant PrCA.

We hypothesize that the use of multiple PSA tests will provide data to distinguish virulent, significant PrCA. If results obtained from the study are consistent with our hypothesis, then a significant portion of the unnecessary prostate biopsies and PrCA treatment may be avoided, thus reducing unwanted expenses while improving quality of life and cost-effectiveness.

We aim to develop growth graphs/curves based on statistical models of repeated PSA measurements. PSA growth curves of men confirmed never to have had PrCA will be compared with those of men with significant and non-significant PrCA. These data will be used to test our hypothesis that these curves will be significantly different, with differences described quantitatively. A clinician (either manually or automatically using electronic health record) could then overlay an individual patient’s observed PSA change over time on our model-generated graphs. Based on the pattern of the overlay in the bedside graphical tool, shared informed decisions may then be made regarding the need for diagnostic workup (i.e., prostate biopsy) and treatment.

This research is highly innovative as it will apply longitudinal data analysis-repeated measures to build on a well-established medical finding concerning the relationship between the pattern of PSA changes over time, to distinguish significant from non-significant PrCA. This tool will be significantly different from traditional measures of PSA change such as PSA velocity (PSA doubling or regression algorithms). PSA velocity or doubling time, which generally is based on two-point measures of PSA and assumes linear PSA change, is used almost exclusively for post-treatment monitoring of disease recurrence. By contrast, this approach will provide a description of the natural history of PrCA that encompasses essential characteristic of disease virulence.

We will follow a nested case-control study design. The cases will be patients with confirmed diagnoses of significant PrCA. Three different control groups will be identified; patients without any prostatic disease, patients with benign hyperplasia (BPH), patients with non-significant PrCA. Using a classic retrospective approach, we will “follow” all patients’ repeated PSA measures over time and describe the trajectories at which those measures changed with time. We will use nonlinear mixed models to model the PSA trajectories (growth curves). We will obtain a description of the mean growth of the cases and the controls over the study time; and the estimated parameters can explain the variability between and within subjects.

The overall goal of this research is to identify and refine a means for differentiating “significant” prostate cancer (i.e., virulent disease with high potential for causing harm in leading to death) from any other condition that could be related to an increased PSA measure in any particular (i.e., one) point in time. This has very important implications, especially for population subgroups that are at higher-than-average risk of virulent prostate cancer (e.g., African-Americans and young men). We have set out these specific aims and support of our overall goal:
1. To establish separate graphical reference ranges for the pattern of PSA change over time (called PSA growth curve) using longitudinal repeated measures of PSA from patients confirmed to: a) be PrCA free, b) have non-significant PrCA, and c) have significant PrCA. The 3 curves will be stratified by age and race.

2. To compare the curves from patients with clinically significant PrCA to those without any PrCA and estimate specificity and sensitivity based on the resulting curves

3. To validate the resulting curves in a different population in which individuals have multiple PSA measures: We will apply the same analysis using enterprise-wide national electronic health record data on more than 5 million Veterans from the Department of Veterans Affairs (VA).

4. To develop a nomogram or other scoring/comparative technique that can be used clinically to aid clinicians and patients in using serial PSA measures to aid in diagnosing “significant” (i.e., virulent) prostate cancer.

Principal Investigator:
James R Hébert Sc.D.
Health Sciences Distinguished Professor
Arnold School of Public Health, University of South Carolina. Director, Statewide Cancer Prevention & Control Program

Co-investigators:
Susan E. Steck Ph.D., M.P.H., R.D. Associate Professor
Department of Epidemiology & Biostatistics
Cancer Prevention and Control Program
Arnold School of Public Health
University of South Carolina

Bo Cai Ph.D.
Associate Professor
Department of Epidemiology & Biostatistics
Arnold School of Public Health
University of South Carolina

Gowtham Rao MD, PhD, MPH
William J.B. Dorn Veterans Affairs Medical Center
Department of Epidemiology & Biostatistics
Arnold School of Public Health
University of South Carolina

Azza Shoiabi Bpharm. MPH.
Department of Epidemiology & Biostatistics
Arnold School of Public Health
University of South Carolina