Prostate specific antigen (PSA) as a long-term marker of risk for prostate cancer

In late 2007, based on information derived from the Malmö Preventive Medicine Study in Malmö, Sweden, Hans Lilja and his colleagues proposed that:

  • There may be a predictive connection between the PSA levels of men at 44-50 years of age and their long-term risk for prostate cancer.

Since Dr. Lilja is one of the original developers of commercial forms of the PSA test, we would be wise to consider his suggestions seriously!

The results of their study led Lilja and his colleagues to make three statements in their original paper:

  • “[S]creening men at age 44 to 50 years … may have clinical utility.”
  • “The primary goal for such testing would not be detection of cancer,” but assessment of risk for subsequent follow up.
  • “[A]ny recommendations to undergo biopsy on the basis of our findings would be premature.”

The remainder of this page will deal with the details of this important study, but for many people, the “key learning” is given in the three bullet points above.

In their paper, which is available in full on line, Lilja et al. explain how, between 1974 and 1986, the Malmö Preventive Medicine Study invited all men born between 1926 and 1949,* and who were living in Malmö, to receive a baseline physical examination and have a blood sample drawn. A total of 21,277 men aged between 33 and 50 years of age participated in the project. This was 74 percent of the male population of Malmö.

Based on data in the Swedish Cancer Registry,  498 of these 21,277 men (2.3 percent) had been diagnosed with prostate cancer by December 31, 1999. The Swedish Cancer Registry is one of the most detailed cancer registries in the world. In 1978 it is estimated to have included over 95 percent of all prostate cancers diagnosed in Sweden that year. By 1999 the accuracy of the registry had improved to about 99 percent.

Lilja et al. were therefore able to go back and carry out PSA tests of various types on the blood samples of not only 462 of the 498 men who were subsequently diagnosed with prostate cancer, but also the blood samples of 1,222 “matched controls.” These matched controls are men who did not have any subsequent diagnosis of prostate cancer but who also participated in the original Malmö Preventive Medicine Study, and who could be closely matched to the men who were diagnosed with prostate cancer. The details are available in the original paper.

For not only the prostate cancer patients, but for the matched controls too, Lilja et al. measured:

  • Their total PSA level (tPSA)
  • Their free PSA level (fPSA) and
  • Their complex PSA (cPSA) level

You will need to understand that in most clinical situations a doctor is only measuring a patient’s total PSA level, so when we refer to “PSA” under normal circumstances we are referring to the total PSA or tPSA.

Now there are a lot of sophisticated details in the study described by Lilja et al., particularly when it comes to the statistics, but the results are really fairly straightforward if the assumptions made in the study are correct. (There’s always at least one “if” like this!)

What they found is as follows:

  • First, regardless of the type of PSA measured, the median value of the PSA level was noticeably lower in the men who did not get cancer than it was in the men who did get cancer. In fact it was about twice as high in the men who did get cancer.
  • Second, there was a statistically significant correlation between PSA levels at baseline (based on blood samples drawn between between 1974 and 1986) and risk for prostate cancer, again for all types of PSA measured.
  • Third, and most importantly, even small rises in the total PSA levels correlated with a marked increase in risk for prostate cancer later on, as you will see if you click on the table on the right.

What does this all mean?

In truth, we can’t be entirely sure … yet. Other studies that have been tracking long-term risk for prostate cancer (such as the Baltimore Longitudinal Study of Aging) have also suggested that there is a correlation between early PSA data and risk for prostate cancer. All that we can really do at this time is summarize some of the conclusions drawn by Lilja and his colleagues, as follows:

  • “PSA measured in middle age predicts a diagnosis of prostate cancer up to 25 years later.”
  • “Cancer risk was markedly lower” … for men with total PSA levels of 0.5 ng/ml or less than it was for the population as a whole.
  • “PSA levels of 2-3 ng/mL … were associated with an increase in” risk for prostate cancer of 19 times the average population risk.
  • “[A]ny recommendations to undergo biopsy on the basis of our findings would be premature.”

This study will undoubtedly lead to much greater focus on the potential value of baseline PSA testing in early middle age as a predictor of risk for prostate cancer. Ulmert et al. have, for example, already re-analyzed data from the Malmö Preventive Medicine Study to see if PSA velocity (the speed at which PSA levels increase over time) is also a long-term predictor of risk for prostate cancer. Apparently it was not, based on the available data, but this result may not be definitive.

*The study excluded men born in 1943, 1945, and 1947. Reasons for these exclusions are not given.

Content on this page last reviewed and updated March 17, 2014.