Monitoring PSA levels in prostate cancer patients after radiotherapy can help clinicians tell the difference between a recurrence and a benign “bounce.”
By John Halamka, M.D., M.S., Dwight and Dian Diercks President, Mayo Clinic Platform and Paul Cerrato, M.A., senior research analyst and communications specialist, Mayo Clinic Platform
Following patients treated with radiotherapy for prostate cancer remains a challenge for several reasons. Their sheer number — about 3.5 million in the U.S. alone — can be overwhelming for clinicians who are also trying to manage new patients. In addition, hospitals can lose track of survivors over time. And given the recommendation that these patients should be seen every 3-6 months, it’s no surprise that patients who are lost to follow-up may experience a delayed detection of recurrence.
Radiation oncologists and urologists usually monitor these patients by measuring prostate specific antigen (PSA) over time. The Phoenix definition, or an increase of 2 ng/ML or more above a post-treatment low level after radiotherapy, is considered the standard definition for biochemical cancer recurrence, which requires additional testing and workup. However, benign PSA bounce, another phenomenon, can look like a recurrence, but eventually drops down.
The challenge for clinicians is determining whether a patient’s increased PSA is a recurrence or a bounce, and if additional testing should be pursued or not. This causes uncertainty and anxiety for both patients and healthcare providers and could also lead to unnecessary testing like MRI or PSMA PET. Additionally, uncertainty can lead to delays in diagnosis of true recurrences and a missed opportunity for re-treatment when disease burden is at the lowest.
Figure 1
As Figure 1 illustrates, it can be difficult to distinguish the two events. Mark Waddle, M.D., with the Department of Radiology Oncology at Mayo Clinic, and his colleagues have developed an algorithm that integrates EMR data with the latest AI tools to address the problem. Their Prostate Cancer Control Tower uses large language model (LLM) prompts to create a unified EMR Integrated Dashboard, taking advantage of data from Mayo Clinic Platform. It accepts longitudinal PSA curves and clinical variables extracted from LLMs as inputs. Using XGBOOST as the machine learning model, the team has conducted a retrospective analysis that was able to detect cancer recurrence about six months earlier than usual while maintaining the same true positive and false positive rates as the Phoenix criterion.
Figure 2
Figure 2 shows an example of a patient at two years post-treatment. The Prostate Cancer Control Tower enables clinicians to detect an increase in PSA levels, while biochemical recurrence is detected a three-year mark and clinical recurrence shortly after that. The team hopes that this lead time will result in more located disease and more cures. They are planning a prospective trial to confirm these results.
Several other investigators have also been attempting to differentiate prostate cancer recurrence from a benign bounce. One international team analyzed over 800 patients who had undergone radiotherapy and found that a PSA bounce usually occurs early on after treatment while a real recurrence is characterized by an increase in PSA much later. Specifically, median time to first PSA rise was ~16–17 months for a bounce versus ~34 months for biochemical failure (p<0.0001). Similarly, an independent study found that median time to first PSA rise was 15.1 months for bounce versus 22–30 months for true failure (p=0.001).
Once upon a time, cancer was a death sentence for most patients. As the latest research demonstrates, prostate cancer now joins the list of several other malignancies that have become manageable challenges.
Share this:
