Enter Your Email Address to Receive a
Copy of MedicExhange Member Demograhpics

Gene expression signatures predict survival in patients with early stage lung cancer, particularly when combined with clinical data.

scientists report in the July 20 online issue of Nature Medicine. The researchers' comparison of several models showed that gene cluster expression analysis performed best.

"This is the largest study to date on lung adenocarcinomas," senior author Dr. David G. Beer of the University of Michigan in Ann Arbor told Reuters Health. "It is a multi-center blinded analysis, and we tested a large number of different predictors with and without the use of clinical variables such as patient stage, age and gender."

"We were able to confirm that some predictors were able to identify high-risk, early stage lung cancer patients in the blinded cohorts tested, and that the addition of clinical variables improved their performance," Dr. Beer added. "Some prior predictors did not do as well as expected but they had not been tested in such a rigorous manner."

Based on the "urgent need to establish new diagnostic paradigms and validate in clinical trials methods for improving the selection of stage I-II patients who are most likely to benefit from adjuvant chemotherapy," Dr. Beer and colleagues formed the "Director's Challenge Consortium for the Molecular Classification of Lung Adenocarcinoma."

The group compared the performance of eight recently published models involving DNA microarray analysis to characterize 442 lung adenocarcinomas, using strict criteria for sample processing and data analysis.

"The most successful classifiers at stratifying stage 1 samples were trained on samples from all stages," the authors report, suggesting that "heterogeneity of aggressiveness exists in stage 1 tumors and that the pattern of gene expression in higher stage tumors is informative for predicting the risk of stage 1 tumors."
The most accurate model used the most genes and relied on the correlated expression of gene clusters to predict outcome.

Relatively high expression of genes in one cluster of 545 genes was associated with reduced survival. This cluster included cyclins, topoisomerases, checkpoint genes, and chromosome and spindle protein genes, "consistent with elevated cell proliferation and loss of cell cycle control being associated with poor outcomes."

Greater expression in clusters that included differentiation-related genes and genes related to immunological function was associated with better survival.

A combination of gene expression and clinical data performed better than either type of data in isolation, the authors observed, "supporting the combined use of clinical and molecular information when building prognostic models for early-stage lung cancer."

"We should be able to build a useful clinical predictor for lung adenocarcinoma and hopefully improve it with greater understanding of the heterogeneity of these tumors," Dr. Beer said. He and his colleagues plan to test these predictors in prospective clinical trials.

In the meantime, he noted, "we are molecularly characterizing the different subtypes of lung adenocarcinomas and determining if common or unique genes are better predictors for these tumors."

Nat Med 2008