A stratified two-stage tumor molecular profiling algorithm to identify clinically actionable molecular alterations in pancreatic cancer☆

Abstract

Background

Tumor molecular profiling (TMP) for pancreatic cancer (PC) is recommended by current international guidelines, yet no testing standards exist. Moreover, the magnitude of benefit and the cost-effectiveness of comprehensive next-generation sequencing panels for PC are under debate.

Materials and methods

We implemented a stratified two-stage TMP algorithm for advanced PC. Stage 1 comprised immunohistochemistry for mismatch repair deficiency and targeted sequencing employing a 33-gene next-generation sequencing panel covering common PC drivers and DNA damage response genes. Based on pre-specified events (KRAS wild type, mismatch repair deficiency, molecular tumor board recommendation), subsequent comprehensive molecular testing was carried out (stage 2). We report molecular findings and patient outcomes.

Results

A total of 94 PC patients were included in the study. Some 63/94 (67.0%) patients underwent TMP according to the algorithm, of which 5/63 (7.9%) fulfilled criteria for subsequent stage 2 comprehensive testing. A total of 31/94 (33%) patients underwent upfront comprehensive molecular testing outside the algorithm based on referring physician’s request. Compared with algorithm testing, upfront comprehensive testing detected a higher number of pathogenic molecular alterations/patient (median: five versus three, P = 0.0005), however no additional actionable alterations. Actionable alterations were identified in 25/94 (26.6%) cases, including DNA damage response gene alterations, KRAS G12C and targetable drivers in KRAS wild type tumors. Patients receiving targeted therapy based on molecular profile showed superior survival (progression-free survival, overall survival) compared with patients without targeted treatment.

Conclusions

Stratified two-stage TMP reliably identifies actionable alterations in PC patients, with potential therapeutic benefit. The proposed TMP algorithm might be as effective, yet more feasible and economic compared with comprehensive upfront testing.