Using a pretreatment serum microrna signature to identify non-small cell lung cancer patients who benefit from radiation dose escalation

Purpose/Objective(s): While biologically equivalent doses (BED) >100 Gy have shown high local tumor control in early stage non-small cell lung cancer (NSCLC), the RTOG 0617 trial showed no overall benefit to escalating radiation dose > 60 Gy in locally advanced (LA) NSCLC treated with conventional fractionation. Using institutional data from prospective studies, we explored whether a pretreatment serum MicroRNA signature could be used to identify a subgroup of patients that benefit from dose escalation in NSCLC. Materials/Methods: From 2004-2013, 85 NSCLC patients from prospective imaging and dose escalation lung cancer trials with available serum were included in analysis. SBRT patients were excluded. The miRNA profiling and quantification was performed using Serum & Plasma miScript. The miRNA Array (SABiosciences/Qiagen), which is a realtime PCR-based array containing a panel of 84 cancer-related miRNA assays and spiked-in synthesized cel-miR-39 as a normalization signal. Tumor dose was captured by gEUD (aZ≤10, a/bZ10). Elastic net Cox regression was performed using clinical factors, dose and dose∗miRNA interaction terms to predict overall survival. From the fitted model, we estimated each patient's individual hazard ratio for dose as a function of their miRNA values and placed patients into 1 of 2 groups based on the median value of these hazard ratios. Rigorous and full cross validation was used, so that prediction of each patient's group status was based on completely refitting the elastic net model while excluding that patient's data. Results: Median age was 66y with mostly male (78%) and stage III (83%) patients. 84.3% of patients had KPS>80 and median gEUD was 68.2 Gy. In stepwise Cox regression, older age, lower KPS and greater tumor volume were associated with decreased OS (P<0.05). The Cox Elastic Net model selected 18 miRNA∗dose interaction terms. In the miRNA group predicted to benefit from high dose (NZ42), higher doses were associated with longer survival. The hazard ratio, adjusted for clinical factors, was .72 (PZ.04) while unadjusted median OS for patients receiving < vs >Z 68 Gy was 13 vs 35 months (PZ.03). In the group predicted not to benefit from high dose (NZ43), there was no observed OS benefit from high dose (adjusted hazard ratio = 1.01, P=.92). Conclusion: Our data identified a specific pretreatment miRNA signature that predicts a subset of LA-NSCLC that may benefit from RT dose escalation. These results are compelling and should be investigated in a prospective dose escalation trial.