Forty percent of prostate cancers remain idle and never affect a person’s health. For this reason, physicians often choose to monitor rather than treat the cancer. To help physicians and patients decide on the best approach for low-risk prostate cancer, a Johns Hopkins team developed an application that predicts the likely risks of the cancer.
“It’s one of the coolest models I have seen in the 30 years I have been at Johns Hopkins,” says urologist Ballentine Carter. “I was surprised by what it could forecast.”
As one of the first projects out of the Johns Hopkins Individualized Health Initiative (Hopkins inHealth), the application synthesizes an individual’s demographic, clinical, biomarker and biopsy data into graphs that quantify the person’s disease risks.
Biostaticians Yates Coley, Mufaddal Mamawala and Scott Zeger devised the algorithms that produce the predictions using 20 years of data collected from nearly 1,300 patients with low-risk prostate cancer.
The color-coded graphs show three things: the probability that the cancer will become severe, the likely results of future diagnostic blood tests and the likelihood that the next biopsy will find evidence of a more dangerous tumor. Such information is valuable when making decisions about medical treatment.
For example, if it is unlikely that someone’s low-risk cancer will be reclassified as more severe, then the physician and patient may delay future biopsies. Alternatively, if the model predicts that the marker in future blood tests will increase, the clinician may recommend a biopsy.
The team is now conducting a study on patients’ experience with the tool, and the Technology Innovation Center is incorporating the application into Epic. The goal is to bring data from Epic into the application and provide results to any clinician seeing a patient with prostate cancer.