Background: Tumor recurrence is the event that has a major impact on survival after local resection of submucosal colorectal cancers (T1 CRC). However, currently, there are no established biomarkers to predict disease-free survival (DFS) in T1 CRC. We previously reported tissue-based transcriptomic biomarkers (5 miRNAs and 8 mRNAs; Ozawa et al. Gastroenterology, 2018; and Kandimalla et al. Gastroenterology, 2019) for identifying lymph node metastasis in T1 CRC patients treated with curative oncological surgical resection. Given our previous findings, we hypothesized that these biomarkers might help identify DFS in T1 CRC following local resection, hence allowing us to identify high-risk patients who could be candidates for curative oncological surgery and/or adjuvant therapy.
Methods: This study analyzed 138 patients with T1 CRC from a multicenter clinical cohort who underwent local resection. Among this cohort, we analyzed 101 patients: patients who had tumor recurrence within 3 years and who had no recurrence for more than 3 years. We evaluated the clinical significance of tissue-based transcriptomic biomarkers to identify DFS in a clinical training cohort (n= 55) and subsequently validated their performance in an independent validation cohort (n= 46). Finally, we established a risk-prediction model by combining the molecular markers with key clinical factors.
Results: In the clinical training cohort, we confirmed that 13 transcriptomic markers correlated significantly with vascular invasion (p< 0.001, AUC: 0.81), indicating their ability to detect potential metastasis including not only via lymphatic but also via blood. Next, in the same cohort, a clinically feasible reduced panel of five transcriptomic markers (2 miRNAs and 3 mRNAs) exhibited a significantly robust overall diagnostic accuracy for tumor recurrence after local resection (AUC: 0.84). Furthermore, we successfully validated its recurrence prediction accuracy in an independent validation cohort (AUC: 0.82). Finally, we established a risk stratification model by combining the transcriptomic markers and key clinical characteristics, which yielded an even superior predictive accuracy for tumor recurrence (AUC: 0.91). The prognostic impact of this risk stratification model revealed that high-risk patients exhibited significantly worse DFS vs. low-risk (p < 0.001). Finally, we established a risk-assessment nomogram based on these transcriptomic biomarkers for an easier clinical translation.
Conclusion: We have successfully established a transcriptomic-based signature in patients with T1 CRC, which allows for a refined risk assessment for predicting tumor recurrence after local resection. This model could provide an individualized clinical approach for identifying high-risk patients who require oncologic curative surgical resection or adjuvant therapy after localized resection of a T1 CRC.

Katsuki Miyazaki

Appearances