76P - Improving access to whole genome sequencing for patients with cancer of unknown primary using formalin-fixed paraffin embedded tissues and cell-free DNA

Background

Cancer of unknown primary (CUP) is a metastatic cancer without a primary site diagnosis after standardized investigations. Genomic testing can inform both treatment targets and identify the primary tissue of origin (TOO). Whole-genome sequencing (WGS) can increase diagnostic yield over DNA panel testing but a systematic comparison of these methods has not been reported in CUP. Furthermore, WGS has been largely restricted to using fresh tissues, limiting test access in many centres.

Methods

Whole-genome and transcriptome sequencing (WGTS) was applied to 75 CUP tumours (71 with matched panel), with 61 (81%) involving formalin-fixed paraffin embedded (FFPE) tissues. Diagnostic features (e.g. driver mutations, mutation signatures) were used to inform TOO diagnosis with clinicopathological review. Treatment targets were identified based on standard of care and clinical trials. The CUP Prediction Algorithm (CUPPA) utilising genome-wide DNA and RNA features was used to determine likely TOO and results compared to genome-informed pathology review. WGS and CUPPA was tested on 22 cell-free DNA (cfDNA) samples with tumour-fraction >8%.

Results

WGTS was successful for 73/75 tissue samples finding additional diagnostic features over panel in 20/71 (28%) cases and assisting a clinicopathological diagnosis in 48/73 (66%) overall. Treatment targets were found in 50/73 (68%) cases including mutation features in 21/71 (30%) not found by panel. High-likelihood CUPPA predictions were concordant with pathology review in 41/53 (77%) cases and none were discordant with pathology opinion. Additional high-likelihood CUPPA predictions were made in 9/20 (45%) cases otherwise deemed diagnostically unresolved. Applying WGS to cfDNA, CUPPA predictions were concordant with a pathological diagnosis in 9/11 (81%) cases, while 4/11 (36%) had high-likelihood predictions but TOO was pathologically undiagnosed.

Conclusions

WGTS is feasible using FFPE samples and diagnostically superior to panel sequencing. The curation of diagnostic features combined with CUPPA prediction can inform TOO in 74% CUPs by WGTS. Use of high tumour-fraction cfDNA may increase access to WGS and TOO prediction in some patients.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

Peter MacCallum Cancer Centre.

Funding

National Health and Medical Research Council, Victorian Cancer Agency, Medical Research Future Fund/Department Industry (Australian Federal Government).

Disclosure

D.D.L. Bowtell: Financial Interests, Personal, Other, Personal consulting fees (that are outside the submitted work): Exo Therapeutics; Financial Interests, Institutional, Other, Patent application to be submitted (unrelated to the submitted work).: Henry Jackson Foundation INOVA HealthCare University of Melbourne; Financial Interests, Institutional, Funding, Research support paid to my institution.: Roche/Genentech, AstraZeneca, BeiGene. S.M. Grimmond: Financial Interests, Personal, Invited Speaker, Precision Oncology Symposium Speaker 2024: AstraZeneca. L. Mileshkin: Financial Interests, Personal, Advisory Board, Participation in Dostarlimab advisory board: GSK; Financial Interests, Institutional, Coordinating PI, Institutional funding from BeiGene for an investigator-initiated trial: BeiGene; Financial Interests, Personal, Coordinating PI, Support for flights to attend ESMO meetings in Madrid and Singapore to present results of the CUPISCO trial: Roche; Financial Interests, Personal, Other, Medical writing support for publications related to the CUPISCO trial: Roche; Non-Financial Interests, Other, Co-chair of the Steering Committee for the CUPISCO trial in CUP (non-remunerated): Roche; Non-Financial Interests, Member, Member of multiple other cancer organisations as above: ASCO, MOGA, COSA, IGCS, GCIG. All other authors have declared no conflicts of interest.