Background – Molecular biomarkers that quantify changes in circulating tumor DNA (ctDNA) can predict clinical outcomes. However, there is no consensus on classification of molecular responders (MRs) and molecular non-responders (nMRs), who may benefit from early treatment intervention. We explored the classification of MRs vs. nMRs in a real-world pan-cancer cohort of patients treated with immune checkpoint inhibitor (ICI) therapy in two ways. (1) As previously described, a ≥ 50% decrease in ctDNA tumor fraction (TF) between baseline and on-treatment time points (MR50), and, (2) hypothesizing that consistently low TF may represent a favorable subgroup with prolonged outcomes despite small increases in TF, MRlow, we evaluated a “low TF” threshold, above the TF detection rate, classifying patients as MRlow if TF remains below the TF threshold at both timepoints, regardless of magnitude change.

Methods – The cohort consisted of advanced cancer patients from the Tempus de-identified clinicogenomic database who received a liquid biopsy at pre-treatment baseline and within 21-180 days after starting ICI therapy. TF was quantified via an ensemble algorithm, xM for treatment response monitoring, that incorporates pathogenic variant allele frequencies, copy number information, and germline information. Evaluable patients had TF ≥ limit of blank in at least one blood sample. Real-world overall survival (rwOS) was defined as the time from on-treatment testing to death or last known clinical record in event-free patients. Cox proportional hazards models were used to estimate the hazard ratio (HR) for MR status (MR vs. nMR) and tests for significance were conducted using 1-sided Wald tests at a 5% significance level.

Results – The evaluable cohort consisted of 71 advanced pan-cancer patients with >10 cancer types; most commonly NSCLC (38.0%, n=27) and small cell lung cancer (19.7%, n=14). Twenty-five patients (35.2%) were treated with ICI monotherapy and 46 (64.8%) were treated with ICI-chemotherapy combination. Using a definition of MR of a 50% decrease in TF alone, 44 patients were MRs and 27 patients were nMRs, with MRs having significantly longer rwOS than nMRs (HR=0.44, P=0.012). We next evaluated various low TF thresholds and found that a 1% low TF threshold further improved the association between MR and rwOS (HR = 0.32, P = 0.0008). 53 patients were classified as MRs [18 MRlow + 35 MR50] and 18 patients were classified as nMRs. Of the MRlow patients, 6 had increasing TF. Median rwOS for nMRs was seven months and not reached at 18 months for both MRlow and MR50.

Conclusions – This study suggests that patients with consistently low TF below 1% while on ICI therapy (MRlow), regardless of TF changes, represent a unique, biologically differentiated, prognostic group. Our study suggests that MRlow patients have prolonged rw-survival, similar to molecular responders based on a 50% decrease in TF (MR50). Prospective validation for this threshold is planned.

VIEW THE PUBLICATION