COMPASS trial suggests that selecting treatment based on subtypes classification may impact patients’ outcomes, but the approach is hard to implement in clinical practice
Pancreatic ductal adenocarcinoma (PDAC) is a difficult-to-treat malignancy with poor prognosis. One of the main challenges and unmet needs is the lack of biomarkers to guide first-line treatment selection, that is instead based on comorbidities, performance status, and local practice (Ann Oncol. 2023 Nov;34(11):987–1002 ).
In this landscape, different classifications and phenotypes/subtypes of PDAC have been published. One of the best-known is the so-called Moffit classification, which comprises two specific subtypes - “basal” and “classic” - based on the different gene profiling and GATA-6 expression (Nat Genet. 2015 Oct;47(10):1168–78 ). According to this classification, classic phenotype is more frequent in resected primary tumors (71% versus 29% for basal-like), compared to metastatic tumors, and is associated with a better prognosis with a 1-year survival rate almost 30% higher compared to the basal phenotype.
Recently, the COMPASS trial has reinforced the prognostic and predictive value of the Moffitt classification, and the relationship between molecular and systemic factors and patient outcomes (Nat Commun. 2025 Jul 1;16(1):5919 ).
The aim of this prospective observational study was to establish the feasibility of a comprehensive real-time genomic analysis of PDAC using whole genome sequencing (WGS) and RNA sequencing (RNASeq), to offer a molecularly-driven approach for the selection of the most suitable chemotherapy regimen. The primary objective was the feasibility of achieving a whole genome within 8 weeks of biopsy in over 80% of patients, which was reported in the first analysis of the study (Clin Cancer Res. 2018 Mar 15;24(6):1344–54 ).
Most of the patients recruited in COMPASS received modified FOLFIRINOX (m-FFX), with 36% of the patients receiving gemcitabine/nab-paclitaxel (GP). This contrasts with most real-world studies in PDAC, favouring GP (J Clin Oncol. 2020 Feb 1;38(4_suppl):666–666 ).
A classic PDAC phenotype was reported in 81% of the patients – a percentage that is slightly superior to the previously reported data. KRAS mutation was present in 93% of the patients, being KRASG12D the most frequent (43%), followed by G12V and G12R (29% and 13%, respectively). These data are in line with those previously described (Cancer Discov. 2022 Apr 1;12(4):924–937 ). Finally, CDKN2A loss of function was described in 84% of the patients, higher than that described in other reports (Cancer Cell. 2017 Aug;32(2):185-203.e13 ). Notably 26% of patients with CDKN2A homozygous deletions had intact MTAP (with 38% of deleted MTAP in the global cohort).
Findings reported a median overall survival (mOS) in the intention-to-treat (ITT) cohort of 9.2 months (95% CI 8.0–10.2), with 10.6 months (95% CI 9.6–12.46) for m-FFX and 8.4 months (95% CI 7.4–10.39) for GP. Differences in mOS were observed across Moffitt phenotypes, and subsequent chemotherapy regimens. Basal-like tumors had a worse mOS compared to the classic subtype [HR 2.21 (95% CI 1.522–3.217); p < 0.001], being 6.5 months (95% CI 4.6–10) when treated with m-FFX versus 7.6 months (95% CI 6.3–10.2) for the ones treated with GP. On the other hand, mOS for the patients with classic phenotype treated with m-FFX was 11.7 months (95% CI 10.52–14.24) compared to 9.94 months (95% CI 7.3–11.51) for those who received GP. Of the 249 patients who started chemotherapy, 10% (n = 25) received a matched targeted therapeutic treatment, without differences in OS compared to those who did not undergo a tailored approach (mOS 14.7 versus 12.8 months, HR 0.79; 95% CI 0.49–1.27, p = 0.32).
There were no statistically significant differences in mOS according to KRAS specific mutants although KRASG12R was associated to the longest mOS (12.1 months), and KRAS G12D to the shortest (8.3 months). Tumors with an HRDetect positive score were associated with improved OS, and those with a high Gustave Roussy Immuno-score (GRIm-s) had worse OS. Liver metastasis from basal-like tumors exhibited an inflamed genotype, with enriched CD8+ and NK cells as well as a high amount of co-inhibitory molecules, indicative of the immunosuppressive microenvironment.
Although the study confirms the utility of phenotypic profiling in terms of prognosis value, and suggests that selecting the treatment based on this information may impact outcomes – even though this is not fully confirmed – the proposed approach is hard to be implemented in daily practice. A major challenge is the time to get WGS results before starting a systemic treatment, that can extend beyond 1.5-2 months – this is an unacceptable time gap to wait. Another key aspect is that the clinical impact of genomic profiling is uncertain, as improved survival was not demonstrated in those in whom an actionable alteration was detected compared to those who received standard treatment.
Despite this, there may be room for this approach in the evolving therapeutic landscape, but only if the diagnostic-therapeutic algorithm priorities rapid genomic sequencing to identify patients for inclusion in clinical trials with targeted therapy, rather than relying on phenotype-based chemotherapy regimens, showing no clear impact on survival.
