Prostate Ca: Novel model predicts ‘true cancer state’

September 1, 2016

Although various calculators are available for predicting biopsy results in men with prostate cancer being managed by active surveillance, a novel model developed by researchers at Johns Hopkins University is unique for its ability to predict a patient’s “true cancer state”; ie, the Gleason score that would be assigned on whole-gland analysis after radical prostatectomy.

San Diego-Although various calculators are available for predicting biopsy results in men with prostate cancer being managed by active surveillance, a novel model developed by researchers at Johns Hopkins University is unique for its ability to predict a patient’s “true cancer state”; ie, the Gleason score that would be assigned on whole-gland analysis after radical prostatectomy. 

Read: AUA '16 PCa take-homes: PSA drop, active surveillance are key themes

Initial evaluation of the model’s performance in a cohort of 191 men in active surveillance who opted to undergo radical prostatectomy demonstrated that it had good accuracy in predicting the probability of having a true Gleason score ≥7. Mean absolute error was almost 0, reported first author Yates Coley, PhD, at the AUA annual meeting in San Diego. A paper discussing the model was more recently published in European Urology (Aug. 11, 2016 A [Epub ahead of print]).

“Active surveillance has been shown to be a safe alternative to curative intervention for men with favorable-risk prostate cancer. Yet overtreatment of localized disease persists, and uncertainty surrounding biopsy Gleason scores is a key factor driving that problem,” said Dr. Coley, postdoctoral fellow in biostatistics at Johns Hopkins School of Public Health, Baltimore.

“Our model aims to address the latter issue by predicting the true Gleason score,” added Dr. Coley, who worked on the study with H. Ballentine Carter, MD, and colleagues.

The Bayesian joint model was fit with data from men in the Johns Hopkins Active Surveillance Cohort. It incorporates patient age and uses all PSA and biopsy Gleason score data obtained at diagnosis and during active surveillance.

“Our methodology reflects the fact that we will have more confidence that an individual is actually harboring indolent disease when he has multiple biopsy Gleason scores of 6 versus a single score from the diagnostic biopsy,” Dr. Coley said.

The model also assumes that PSA values will gradually rise with age, correlate in an individual over time, and suggest the presence of more aggressive prostate cancer if the level is higher or rising quickly.

Next: Tool will be integrated with EHR

 

Tool will be integrated with EHR

The authors are developing an interactive decision support tool based on the model that will be integrated with the electronic health record. All data will be automatically uploaded and updated predictions generated immediately.

Also see: Landmark prostate cancer screening trial criticized

It uses the patient’s age and all PSA values and biopsy Gleason scores from diagnosis onward along with calendar date and time in active surveillance to predict whether the full prostate Gleason score at the time of diagnosis and throughout active surveillance is 6, 3+4, 4+3, or 8 to 10. It also generates a likely PSA trajectory and risk of reclassification.

Dr. Coley cautioned that currently the tool is not considered applicable to men on active surveillance outside of the Johns Hopkins program, but future plans include testing its performance in an external population. In addition, the authors will be studying how use of the decision support tool might impact patient anxiety and overall satisfaction with care.

So that the model reflects the most up-to-date understanding of risk, they are also hoping to continually refine it using emerging information on the predictive value of new variables, including multiparametric magnetic resonance imaging and genomic markers.

More on Prostate Cancer:

Active surveillance initiative raises adoption of strategy

Assay predicts outcomes in men taking mCRPC agent

Data offer new insight on HIFU’s use in low-risk PCa