Genomically informed longitudinal monitoring of circulating tumor DNA (ctDNA) to predict outcomes of cancer therapy.

Abstract

3533 Background: Short fragments of ctDNA can be detected and quantified from blood samples of patients with cancer. We hypothesize that dynamic changes in quantity of ctDNA in patients with advanced solid cancers during the first few weeks of therapy can predict treatment outcomes reported by standard imaging. Methods: We enrolled patients with advanced cancers treated with experimental therapies, who had blood collection for ctDNA isolation and testing at baseline, mid-cycle and at the time of restaging imaging. Patients who were treated with multiple treatment lines were included with separate record for each therapy. Genomically informed molecular testing of ctDNA was performed using unamplified droplet digital PCR (QX200, Bio-Rad) designed based on known molecular profile of tumor tissue and ctDNA was quantified as aggregate variant allele frequency (VAF%) for detected molecular aberrations. Patients were classified based on results of their first restaging imaging as responders (complete [CR] or partial response [PR]) vs. non-responders (stable disease [SD], progressive disease [PD]) and progressors (PD) vs. non-progressors (CR, PR, SD). Results: Total of 85 patients who received 132 courses of therapies between May 2012 and June 2019 were analyzed. Breast (N = 21), melanoma (N = 14) and cholangiocarcinoma (N = 14) were most frequent tumor types. Aggregate VAF at mid-cycle was higher in non-responders (3.98%) compared to responders (0.40%, P = 0.016) and in progressors (4.40%) compared to non-progressors (2.10%, P = 0.019) as measured by 5% trimmed mean. Similarly, aggregate VAFs at first imaging restaging was higher in non-responders (5.10%) compared to responders (0.10%, P = 0.001) and in progressors (10.80%) compared to non-progressors (0.90%, P < 0.001). Progressors demonstrated increase in ctDNA VAF at the time of the first imaging restaging compared to decrease in non-progressors (0.7% vs. -4%, P = 0.015). In addition, increase in ctDNA VAF at the first imaging restaging was associated with more PD (44% vs. 8%, P = 0.019) and less PR/CR (0% vs. 31%, P < 0.001). Median time-to-treatment failure was shorter in patients with increase in ctDNA VAF at the time of the first imaging restaging (52 days vs. 89 days, P = 0.002). Conclusions: Dynamic changes in quantity of blood-derived ctDNA within the first few weeks of therapy correspond with treatment outcomes reported by the first restaging imaging and time-to-treatment failure.