[gdpawel]

According to government researchers, genetic tests designed to predict how well lung cancer patients will fare after treatment do very little to guide doctors. Their review of studies aiming to show a gene signature for lung tumors found serious problems in the design and analysis of the studies.

Jyothi Subramanian and Richard Simon of the National Cancer Institute in Bethesda, Maryland say that researchers need some clear guidelines to follow in what is becoming a booming industry of using genes to predict who needs more treatment.

"None of the studies reviewed were successful in showing clear usefulness for the gene expression signatures over and above the known risk factors," they wrote in the Journal of the National Cancer Institute.

Researchers hoped to be able to use DNA sequences to predict risk of disease, to help choose the best drug for a patient, and to follow how well patients are doing on various therapies.

However, so many different stretches of DNA are involved in disease that experts say it will take years to sort out. And there is no clear agreement on how to design studies to show whether such tests predict what they are supposed to.

Subramanian and Simon reviewed 16 studies published from 2002 to 2009 on tests looking for active genes in non-small cell lung tumors that would predict whether a patient's tumors would return or spread.

Diagnosing lung cancer is tricky. At Stage I, before the tumor has spread, surgery can get the tumor, if the surgeon can get to it without destroying the lung. Most patients are not diagnosed until Stage IV, when the cancer has spread throughout the body, because symptoms are vague.

In between, it is hard to assess how much the tumor has spread and whether a patient would benefit from chemotherapy.

Even when a surgeon gets all the tumor and patients are designated as stage IA, meaning no spread, 30 percent of the patients relapse. So the pressure is on for a way to test the tumor to predict who needs extra treatment.

But none of the tests so far do this adequately, Subramanian and Simon said. And most of the studies did not even take into account known risk factors, such as whether the edges of the tumor looked fuzzy, they said.

"Hence, we again emphasize the fact that care must be taken to collect and use as much clinical information about patients as possible when developing prognostic signatures," they concluded.

Source: MedlinePlus U.S. National Library of Medicine