Improved ctDNA detection in early stage non-small cell lung cancer

Katrin Heider ^1,2 Jonathan CM Wan ^1,2 Davina Gale ^1,2 Andrea Ruiz-Valdepenas ^1,2 Florent Mouliere ^1,2,3 James Morris ^1,2 Wendi Qian ⁴ Jerome Wulff ⁴ Nikolaos Demiris ⁴ Alexandra Williams ⁴ Tim Eisen ^5,6 Christopher G Smith ^1,2 Doris M Rassl ^2,7 Susan Harden ⁵ Robert C Rintoul ^2,7,8 Charles Massie ^1,2,8 Nitzan Rosenfeld ^1,2

¹Cancer Research UK Cambridge Institute, University of Cambridge, Cambridge, UK
²Cancer Research UK Cambridge Centre, University of Cambridge, Cambridge, UK
³Department of Pathology, Amsterdam UMC, Vrije Universiteit Amsterdam, Amsterdam, Netherlands
⁴Cambridge Clinical Trials Unit – Cancer Theme, Addenbrooke's Hospital, Cambridge, UK
⁵Department of Oncology, Addenbrooke’s Hospital, Cambridge, UK
⁶Oncology Early Clinical Development, AstraZeneca, Cambridge, UK
⁷Royal Papworth Hospital NHS Foundation Trust, Royal Papworth Hospital NHS Foundation Trust, Cambridge, UK
⁸Department of Oncology, University of Cambridge Hutchison–MRC Research Centre, Cambridge, UK

Overall survival of non-small-cell lung cancer (NSCLC) patients remains poor as patients are frequently diagnosed at late stage. The evaluation of circulating tumour DNA (ctDNA) has shown to offer a non-invasive method for cancer detection. However, detection rates of ctDNA in patients with early stage cancers have been limited due to sampling and sensitivity issues. We developed a novel algorithm for INtegration of VAriant Reads (INVAR), which uses sequencing data across hundreds to thousands of tumour-mutated loci to detect ctDNA in plasma samples at high sensitivity. We applied this to a cohort of stage I-III NSCLC patients recruited in the LUng cancer - CIrculating tumour DNA study (LUCID). In this prospective and observational study 100 stage I-IIIB NSCLC patients underwent radical treatment (surgery or radiotherapy +/- chemotherapy) with curative intent. Plasma samples were collected before and after treatment.

We analysed plasma samples from 90 patients using patient specific-sequencing panels. Four patients had less than 20,000 analysable molecules and were excluded. Of 86 patients with early-stage NSCLC, signals were observed in 70%. 65% of patients passed a detection threshold with 95% specificity at ctDNA fractions as low as 1.7x10^-5. For 17 of those patients staging information was available at time of writing and ctDNA was detected in 50% of stage I patients and 100% of stage II and III patients. Stage and subtype specific detection rates for the remaining cohort will be presented.

Our findings highlight an opportunity to improve ctDNA detection for early stage NSCLC and small tumours in general by using patient specific sequencing information. Additionally, our algorithm has the potential to aid in longitudinal cancer monitoring and is applicable to a variety of sequencing data types. We aim to apply this approach to serial samples obtained through the LUCID study to investigate its application in the treatment managment.