Stanford researchers develop model to find blood biomarkers that estimate tumour size

In any battle, size matters. Identifying cancer in its infancy, before it spreads, will likely increase the odds of defeating it. However, no method currently exists to reliably translate the results of common blood-screening tests meant to suss out malignancy into the size of the tumour challenging both patient and doctor.

Now researchers at the Stanford University School of Medicine have begun to map out a way to correlate the levels of so-called blood biomarkers with cancer volume. The effort will guide the development of new tests to facilitate early detection of the disease.

“Early cancer detection is a very challenging but important goal for the cancer field,” said Sanjiv Sam Gambhir, MD, PhD, professor of radiology and senior author of the research published in the Aug. 18 issue of PLoS Medicine. “This modelling work enables a very deep understanding of the problems that will have to be solved for blood-based cancer biomarkers to be successful in this effort.”

Gambhir is also head of the Molecular Imaging Program at Stanford. He and radiologist Amelie Lutz, MD, developed the mathematical model using two common blood biomarkers: PSA, or prostate-specific antigen, which is often elevated in prostate cancer, and CA125, which serves as a marker for follow-up therapy in ovarian cancer. They found that the minimum tumour sizes predicted by their calculations roughly matched what is found in clinical practice, indicating that they are on the right track.

“We’re pretty happy that we came up with rather realistic tumour sizes,” said Lutz. “Although this is a very basic model, it should give researchers a tool to use when deciding if a particular secreted protein would be a good biomarker.”

Good biomarkers are eagerly sought in the quest for early cancer diagnosis. A number of things happen when a cell turns cancerous. It begins to divide when it shouldn’t, and it may start to make and secrete proteins into the bloodstream that non-cancerous cells around it do not. Ideally, physicians screening for cancer would be able to administer a simple blood test to identify even minute amounts of these proteins, or biomarkers, very early in the development of the disease and to initiate life-saving therapy.

In reality, however, cancers rarely provide such a foolproof “tell.” Because many biomarkers, including CA125 and PSA, are secreted by both healthy and cancerous cells, physicians and researchers assessing their levels in the bloodstream must make educated guesses as to where to assign the cut-off between the high end of normal and the low end of worrisome. Misjudging this number can lead either to unacceptably high numbers of false-positive results or to overlooking already dangerous disease.

Lutz used published data from cells grown in culture dishes to estimate how much CA125 or PSA is secreted into the surrounding cell growth media. The team then devised a mathematical model to translate the amount secreted by the cells into levels that could be expected to appear in the bloodstream for specific tumour volumes. If the marker is tumour-specific, the minimum detectable amount of biomarker in the bloodstream marked the smallest tumour that could likely be detected with this method. If the marker is secreted by both normal and cancerous cells, the threshold amount for suspicion is necessarily higher and the minimum detectable tumour size is larger.

By its very nature, the technique required many simplifications of what is a very complex biological process. But it worked surprisingly well. For example, because PSA is secreted by both normal and cancerous cells, the smallest prostate tumour predicted by the method is about the size of a small pea. If it were secreted only by cancer cells, PSA could possibly be used to detect tumours much smaller: about the size of the period at the end of this sentence.

“We’re making assumptions about how tumours behave in the human body,” said Lutz.
“Although there’s not always a linear relationship between tumour size and secretion levels, our model identifies some important characteristics of a good biomarker. We also try to point out additional physiological parameters and studies needed to develop more precise models.”

With this approach, a protein that is highly secreted only by cancer cells would warrant further investigation. Also, a panel, or group, of good cancer-specific biomarkers is likely to be more clinically useful than just one protein, since not all tumours may express the same markers.

“It would be helpful to have a panel of biomarkers that complement each other,” said Lutz. “A lot of effort is going in to finding the perfect panel, and we hope that this work will help researchers home in on specific candidates.”

Lutz and Gambhir’s Stanford colleagues on the research include Juergen Willmann, MD, assistant professor of radiology; Frank Cochran, PhD, postdoctoral scholar; and Pritha Ray, PhD, research associate.

(Source: PLoS Medicine: Stanford University: September 2008)