In the largest study of its kind, researchers have identified a panel of genes that can predict which lung cancer patients will have the worst survival and therefore need the most aggressive treatment. However, developing a clinically useful genetic test for this indication will not be as simple as some would like to believe, according to the research team.
The study, published in Nature Medicine, looked at 442 lung cancer tissue samples collected from six hospitals in North America, and discovered that the expression of certain genes, when considered with clinical factors such as patients’ tumor stage, age, and gender, can help determine their likelihood of survival from the disease.
"Our findings suggest that there is a potential for successfully predicting lung cancer prognosis based on gene expression, but it is likely to be more difficult to develop a clinically useful test than has been suggested by previous studies,” said Kerby Shedden, study author and associate professor of statistics at the University of Michigan, in a statement.
Before a test can be developed for the indication, researchers will need to conduct “more assay standardization and a large prospective study to identify a signature that is ready for clinical use," Shedden added.
The researchers did not provide a timeline for when a test may be developed.
The National Cancer Institute funded the study, conducted by four academic institutions: the University of Michigan Comprehensive Cancer Center, H. Lee Moffitt Cancer Center and Research Institute, Memorial Sloan-Kettering Cancer Center, and Dana-Farber Cancer Institute.
“Several models [out of eight models] examined produced risk scores that substantially correlated with actual subject outcome. Most methods performed better with clinical data, supporting the combined use of clinical and molecular information when building prognostic models for early-stage lung cancer,” the study authors reported.
“This study also provides the largest available set of microarray data with extensive pathological and clinical annotation for lung adenocarcinomas,” they said.
All the study sites used Affymetrix 133A arrays to measure gene expression in patients, utilized the same reagents, and practiced uniform study protocols. “The only major differences [were] the tumors themselves,” David Beer, study author and co-director of the cancer genetics program at the University of Michigan Comprehensive Cancer Center, told Pharmacogenomics Reporter this week.
Any test developed would need to be “robust, relatively inexpensive, and accurate.”
Having collected lung cancer tissue samples from six centers, the researchers grouped these samples into four sets based on the laboratories where the samples were processed. Two sets were designated as the "training sets," in which researchers tested various gene expression methods to gauge whether they were predictive of patient outcome.
Meanwhile, the outcome data for the other two additional tumor sets were blinded to the researchers. Researchers used these "validation sets," to discern whether the outcome of the training sets could be replicated.
“Eight classifiers producing either categorical or continuous risk scores were developed by investigators using the training data and were tested for effectiveness on the two remaining data sets,” the researchers reported. “Most of these classifiers incorporated techniques that have repeatedly been applied in gene expression-based prognosis and found to work well in at least some instances.”
The classifier, which showed the best predictive ability in the study, called Method A, looked at all tumor samples or stage I samples alone, both with and without clinical covariates.
Researchers found that Method A was able to predict patient outcome based on the expression of 100 gene “clusters.”
“Relatively higher expression of genes in cluster 6 of method A (545 genes) was associated with poor subject outcome,” the study authors write in the paper. The genes in cluster 6 included cell proliferation-related genes, such cyclin A (CCNA2) and other cyclins, BUB1B, topoisomerases, checkpoint genes (CHEK1), and chromosomal and spindle protein genes.
According to the researchers, the genes in this classifier and a few others “may provide insight into the biology of aggressive tumors.”
Although the study findings hold the potential for a diagnostic test that can determine the aggressiveness of treatment based on the likelihood of survival, no plans have yet been made to develop such a test.
“We are not currently discussing a specific test with an industry partner, nor have we decided an array will be used,” Beer said. However, he noted that any test developed would need to be “robust, relatively inexpensive, and accurate.”
Still the introduction of such a diagnostic could potentially be an improvement from the standard treatment in lung cancer.
Currently, lung cancer patients receive chemotherapy after surgery to lessen the risk of the cancer returning.
However, it is generally understood that some Stage I lung cancer patients have an aggressive form of the illness with poor prognosis and can benefit from chemotherapy. In contrast, more advanced lung cancer patients have better prognosis and therefore do not usually receive adjuvant chemotherapy. However, studies suggest that this latter population could possibly benefit from adjuvant chemotherapy.
“[T]here is an 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,” the study authors write in the paper.
However, developing a genetic test for lung cancer is complicated since there are multiple types and subtypes of the disease. Furthermore, since smoking is a common factor among lung cancer patients, exposure to tobacco smoke can produce numerous genetic alterations.
“To be able to offer one simple gene test for the disease, scientists would need to accurately model the known cellular diversity and the potential differences underlying the aggressiveness between lung cancers,” the researchers said in a statement.
Factoring in the clinical data along with gene expression, as researchers did in this study, can help more accurately model these differences between patients.
“Gene expression is not just a black box approach, which a lot of researchers think it is,” Beer said. “Sometimes knowing the context actually helps you use that information more efficiently. We found that looking at clinical data along with gene expression can be a more reliable indicator.”
According to the American Cancer Society, 215,020 people will be diagnosed with lung cancer this year and 161,840 will die from the disease.
Lung cancer, the leading cause of cancer-related death in the US, has a 15 percent five-year, overall survival rate. According to the study authors, the survival rate for lung cancer patients has not improved over many decades.
The potential economic savings to the healthcare system from being able to determine which patients should not receive adjuvant chemotherapy, and thus avoid serious side effects, could be significant. However, researchers in this study have not conducted a pharmacoeconomic analysis of introducing such a test into the lung cancer treatment paradigm.
The next step in this research will be to test the gene predictors in a larger prospective trial. According to the University of Michigan, enrollment for this trial has not begun.