ASO Author Reflections: Predicting pancreatic cancer liver metastasis by integrating primary tumor clinicopathologic features and liver radiomics

Past

Pancreatic ductal adenocarcinoma (PDAC) is one of the most aggressive solid tumors, with advanced presentation the norm – due either to metastasis or locally advanced disease precluding potentially curative resection. Until recently, less than 20% of patients were eligible for resection; however advances in regimens and sequencing of systemic therapy, including neoadjuvant therapy (NAT), have enabled resection in a somewhat larger fraction¹.

Many patients harbor tumors that recur early, even after a technically sound and complete resection. Furthermore, these tumors appear to have a predilection for metastasis to specific organs, which seems to be associated with variable outcomes. The weight of evidence suggests a mechanism that involves early, organ-specific spread of microscopic disease, even before the primary tumor is evident clinically. Indeed, animal studies have demonstrated the presence of disseminated tumor cells (DTCs) even at seemingly non-invasive stages of PanIN², yet conventional cross-sectional imaging studies are unable detect DTCs or micrometastases. Clinicians treating PDAC recognize these shortcomings in defining tumor biology and the pre-metastatic niche, given our inability to objectively quantify it. In theory, NAT should target the microscopic disease and potentially improve the resulst of resection – or at least, provide a “test of time” to allow selection of patients with more indolent tumors³; however, NAT has not been shown to have a clear benefit in this regard.

Present

In the present study, we analyzed recurrence patterns after resection of localized PDAC and found that liver metastasis (LM) tends to occur early after resection and portends the worst outcome⁴. Poor differentiation, lymphovascular invasion, and larger tumor size were independent predictors of early LM, as was low Fibrosis-4 score (an indirect assessment of the pre-metastatic niche). It has been shown in several preclinical studies that primary pancreatic cancers secrete circulating factors that can remotely prepare distant organs to receive DTCs⁵. Many of these changes involved immunomodulation and remodeling of the extracellular matrix in a manner that is favorable to the survival and later growth of DTCs. We speculated that these changes in the target organ (ie, the liver) would be potentially measurable using quantitative imaging techniques. We, therefore, used radiomic analysis of preoperative contrast-enhanced CT scans to assess for changes the might indicate the presence of a ‘pre-metastatic niche’ or established micrometastases. By combining a radiomics-based model with primary tumor size and differentiation (all of which may be determined preoperatively), we developed a model that predicted LM within 6 months with reasonably high AUC and negative predictive value.

Future

Our study paves the way for a holistic approach for the use of imaging techniques to quantify the metastatic risk of PDAC patients before surgery. We show that consideration of the phenotype of the primary tumor and the status of potential metastatic organs, particularly the liver, is a promising approach for selection of patients most likely to benefit from upfront resection versus NAT. Future studies should include radiomics of the primary tumor and the liver, and integration of additional modalities that facilitate quantification of “tumor biology”, including needle- and liquid-biopsy based assays.