New Thoughts on the PSA Test

In an article in Science they discuss some recent work on looking more deeply into the PSA test. We also recently did an analysis of PSA in a time  series basis but the key was understanding certain growth constants.

The Science article states:

In cancer research, one never-ending quest is the hunt for biomarkers, proteins in the blood that reveal tumors deep in the body. The goal is to find a reliable predictor that enables early diagnosis and better treatment of disease. Most biomarkers identified so far have been disappointing, but millions of men are tested for one of them: PSA, or prostate-specific antigen, which cells in the prostate gland churn out at higher levels when cancer is present.

But some men with high PSA levels don't have the disease, leading to unnecessary biopsies, and some with low PSA levels do. PSA testing is routine in some places, yet large clinical trials in the United States and Europe are questioning how big an impact it has on reducing cancer deaths. This week, a group in Iceland suggests a new way to assess PSA results: Acknowledge that under normal circumstances some men produce more (and some produce less) PSA, and individualize healthy and high levels for each man tested by focusing on genetic variants that affect PSA levels. 

However these researchers found certain sections of DNA, SNPs, which allowed for significantly improved classification. They state:

The researchers on the PSA study, led by geneticists Julius Gudmundsson and Kari Stefansson of deCODE Genetics in Reykjavik, began by scanning the genomes of about 16,000 men to identify places in the genome that modify the PSA level in blood. They found six, including three that were previously identified. Then the group compared more than 5000 men with prostate cancer and 41,000 without. They assessed whether variants at the six sites modified PSA and prostate cancer risk.

Yet they also look back at what we consider an erroneous result (see our analysis on PSA Testing), the NEJM European study, where they state:

Last year in The New England Journal of Medicine, the U.S. study reported no difference in mortality between the group that had regular PSA screening and the one that did not. The European team, led by Schröder, found that those who got PSA screening were 20% more likely to survive prostate cancer. Schröder attributes the contrasting results to different study designs; both groups are continuing to follow the men. 

We still consider their results flawed because of two factors. First they held to a PSA of 4.0 which we now know is in error and second the testing was sporadic and not periodic, as well as having a duration too long between samples. We also argued that they failed to look at the data time series, namely the velocity issues, and the % Free PSA, a critical element.

Now using the SNP data we can use the model we developed which is a birth-death model for cells and then uses maximum likelihood estimates to get the most powerful test. Using the SNP data from this study it is possible to improve the tightness of fit and better represent the actual B-D constants. Worth a thought.

The analysis of PSA data will as a necessity include a better understanding of the cell dynamics for both benign and cancerous cells. That will go to the heart of any worthwhile measurement. Static measurements will always have a fatal flaw as has been discussed by Dougherty.