Background: Owing to recent advances in colonoscopic techniques, majority of submucosal colorectal cancers (T1 CRCs) can now be removed by endoscopic resection. Among these, 70% of T1 CRCs are deemed as “high risk” because they meet certain pathological risk factors including the presence of lymphovascular invasion, poor differentiation, and increased depth of tumor (>1000um). However, post-surgical pathology data indicates that in reality only 10-15% of all T1 CRCs are genuinely high risk, while all other patients undergo unnecessary surgeries. Since current pathological criteria are inadequate, availability of more robust molecular biomarkers that may help identify ‘genuine high risk patients with lymph node (LN) metastasis’ more accurately will reduce the burden of surgical overtreatments. Due to the growing interest in developing miRNA biomarkers, we undertook this study to identify a miRNA-based diagnostic signature for detecting LN metastasis in CRC.
Methods: In a biomarker discovery step using RNA-Seq data from 15 LN-positive and 104 LN-negative T1/2 CRC patients, we identified candidate miRNAs with >0.5 log fold change and a p<0.05. Thereafter, using a receiver operating curve (ROC) based backwards elimination approach, we identified a signature of miRNAs that were differentially expresses in LN positive vs. negative CRCs. We validated the performance of this miRNA signature to detect LN metastasis in 191 surgically resected CRC specimens from two independent patient cohorts by qPCR assays.
Results: We identified a panel of 10 differentially expressed miRNAs in the discovery step, which was initially validated in a training cohort of 61 T1 CRC samples, which included 8 LN-positive cases. Using a logistic regression analysis model, we deduced robust AUC values when using miRNA expression results alone (AUC=0.85, 95%CI: 0.74-0.93, p<0.001) for identifying LN-positive T1 CRCs. Thereafter using the same model parameters in an independent validation cohort of 130 T1 CRCs, which included 16 LN-positive patients, we were able to successfully confirm our results from the training cohort (AUC=0.74, 95%CI:0.66-0.781, p<0.001) for the identification of high risk T1 CRC patients with lymph node metastasis.
Conclusions: Based upon a systematic approach, we firstly report the feasibility and promise of a miRNA signature that can be used clinically for the detection of T1 CRCs with lymph node metastasis, which will reduce patient discomfort and healthcare costs.