Raising the Bar on Tumor Marker Prognostic Studies
During the month of August, five leading peer-reviewed biomedical research journals are publishing the same report in their pages and on their Web sites. The report's centerpiece is a single page that resembles a checklist that college freshmen might receive from their professors on how to write a term paper.
In this case, the checklist ticks off 20 recommendations on how to write a different type of paper: a report on a tumor marker prognostic study - investigations of biological markers that may predict patients' clinical courses following a definitive treatment such as chemotherapy or surgery - to be submitted for publication in a peer-reviewed journal.
Developed by an international committee of researchers led by NCI and the European Organization for Research and Treatment of Cancer, the guidelines are intended to facilitate the reporting of the types of data and other information that can clearly demonstrate such studies' true significance and allow accurate comparisons with similar studies.
A more subtle goal, says Dr. Sheila E. Taube, associate director of the NCI Cancer Diagnosis Program and a co-author of the guidelines document, is to nudge researchers into designing better studies to begin with.
"If journal reviewers and editors send a message to researchers that they have not included pertinent categories of information in their submissions and refer them to the guidelines, I think that it will not only change the reporting, but when they do their next study, change how it's designed and conducted," she says.
Thousands of studies have looked for definitive associations between clinical endpoints (such as time to disease recurrence or mortality) and biological markers (such as the levels or expression of specific proteins or genes, or gene/protein patterns). But they produced few clinically meaningful markers that can help oncologists make treatment decisions, such as whether adjuvant therapy is needed.
Part of the problem, notes Dr. Lisa McShane, the lead author and member of the NCI Biometric Research Branch, is the sometimes striking inconsistency in how studies of the same markers are conducted and analyzed. Often, the studies look at different patient populations: The patients may have different tumor characteristics and receive different treatments, and the marker may have been measured using different assay methods. These factors could influence the observed association of the marker with outcome. And, these important details - which could explain the inconsistent results - are frequently not reported.
Such shortcomings have serious implications. When an expert panel convened by the American Society of Clinical Oncology met 5 years ago to update clinical guidelines on tumor markers for colorectal and breast cancer, the inconsistency in how studies had been conducted and concerns about statistical analyses made the process a struggle, recalls panel member Dr. Nancy Kemeny, of Memorial Sloan-Kettering Cancer Center. This was true even for exhaustively studied markers such as CEA, elevated levels of which have been shown in a number of studies to indicate disease recurrence in patients with colorectal cancer.
"We spent a long time looking at CEA," Dr. Kemeny says. "We almost came out against recommending it, partly because we don't have a good randomized study."
The panel also looked at other markers in colorectal cancer that, according to some studies, offer important prognostic clues, such as levels of CA 19-9, a protein shed by tumor cells, and expression levels of the tumor suppressor gene p53. "But we couldn't come up with a positive feeling about them," Dr. Kemeny says. "It's possible that some may be positive, but the available studies just weren't good enough for us to say that the markers should be used to monitor patients."
A study in the July 20 Journal of the National Cancer Institute offered another cautionary tale about tumor marker prognostic studies - namely that many published studies are often tinged with bias.
The study involved a meta-analytic review of studies examining the relationship between p53's protein product (TP53) and prognosis after treatment for head and neck squamous cell cancer. While a review of only published studies turned up a strong association between TP53 status and survival, when the results of unpublished studies were added to the analysis and standardized definitions of TP53 status and patient outcomes were applied, "the statistical significance of the association was abrogated," the researchers concluded.
Drs. Taube and McShane believe the guidelines can help address such issues and, in so doing, have a substantial impact.
"Hopefully we will have better science being done and being reported in a way that is more interpretable and can more quickly get us to an understanding of a marker's clinical significance," says Dr. Taube. The ability to assess potential markers more quickly and accurately, she concludes, "will allow us to get things into the clinic more efficiently."
The guidelines can be found at http://www.cancerdiagnosis.nci.nih.gov/assessment/progress/clinical.html.
By Carmen Phillips