Researchers from the Medical College of Wisconsin, the Mayo Clinic, and Washington University in St. Louis have identified two microRNA expression signatures predictive of prognosis in patients with stage I non-small cell lung cancer.
In their study, the researchers were able to validate one signature that was predictive independent of cancer subtype and another specific to adenocarcinoma. Both profiles could accurately predict which stage I lung cancer patients might benefit from more aggressive therapy, the authors wrote.
The study appeared in an advanced access offering in the journal Carcinogenesis last week.
According to the study authors, gene expression analysis has failed to identify expression-based prognostic signatures in stage I NSCLC robust enough for clinical applications. Meanwhile, the disease recurs in about 30 percent of patients treated with surgical resection at stage I, highlighting the need for biomarkers that can reliably identify those at higher risk of relapse for additional or more aggressive first line therapy.
With several recent studies correlating miRNA expression with outcome in lung cancer, the investigators set out to try to identify miRNA signatures that could accurately predict prognosis in what they claim is the largest study of its kind to date.
In their study, they studied FFPE tissue from the tumors of 357 patients who were diagnosed with stage I NSCLC and received surgery between 1995 and 2005 at the Washington University School of Medicine.
They then validated the predictive signatures found in the first group in a second set of 170 independent samples, split between 85 FFPE and 85 fresh-frozen tumor tissue samples from the Mayo clinic.
Subtypes in the discovery cohort were spread among adenocarcinoma, squamous cell, large cell, bronchioloalveolar, adeonsquamous, and large cell neuroendocrine carcinomas, the team reported.
Using an Illumina human miRNA expression profiling V2 panel, the group measured miRNA expression levels from all 357 samples, comparing the group of patients with recurrence in two years against those who didn't recur for more than seven years.
They established which miRNAs were differentially expressed regardless of cancer subtype, as well as signatures for both adenocarcinoma and squamous cell carcinoma specifically, after finding that the miRNAs strongly differed between those two groups.
The researchers ended up with a signature of 34 miRNAs complied from all 357 FFPE samples that could predict five-year relapse-free survival. The researchers also identified a 27-miRNA predictive signature from only the adenocarcinoma samples, and a 17-miRNA signature from only those with squamous cell carcinoma.
Four miRNAs overlapped between the full stage I signature and the ADC-specific group. Another five overlapped between the full signature and the SCC signature, the authors reported. Additionally, the group identified 10 miRNAs that were associated specifically with brain metastasis.
To determine the predictive power of the signatures, the team derived risk scores using a partial Cox regression method, and performed receiver operating curve analysis.
The group found that Cox model risk scores estimated by miRNA expression gave better prediction that tumor stage information, with areas under the curve of 86 percent to 91 percent for the miRNA signature and only 64 percent to 69 percent for staging.
"This clearly demonstrates that miRNA expression signatures combined with stage information has higher classification power than the staging method to predict survival," the authors wrote.
The researchers then tested the robustness of the signatures in a second sample set from the Mayo Clinic, made up of 85 fresh-frozen, and 85 preserved tumor samples. Because only six of these samples were SCC, the team was only able to validate the all-type and ADC-specific signatures.
Using their all-type signature to predict risk in these samples, the researchers measured an AUC of 81 percent, while stage information alone gave only a 61 percent AUC. The ADC-specific signature also showed better performance than staging in the 110 ADC samples, the group reported.
According to the researchers, several miRNAs in their predictive signatures are associated with important cancer-related pathways, including miR-31, and miR-34.
If validated in prospective studies, the authors wrote that the signatures could allow "immediate and widespread use in the clinic."
Have topics you'd like to see covered in Gene Silencing News? Contact the editor at mashford [at] genomeweb [.] com.