Serial Monitoring of Circulating Tumor DNA by Next-Generation Gene Sequencing as a Biomarker of Response and Survival in Patients With Advanced NSCLC Receiving Pembrolizumab-Based Therapy

Abstract

Although the majority of patients with metastatic non-small-cell lung cancer (mNSCLC) lacking a detectable targetable mutation will receive pembrolizumab-based therapy in the frontline setting, predicting which patients will experience a durable clinical benefit (DCB) remains challenging.

Materials and methods: Patients with mNSCLC receiving pembrolizumab monotherapy or in combination with chemotherapy underwent a 74-gene next-generation sequencing panel on blood samples obtained at baseline and at 9 weeks. The change in circulating tumor DNA levels on-therapy (molecular response) was quantified using a ratio calculation with response defined by a > 50% decrease in mean variant allele fraction. Patient response was assessed using RECIST 1.1; DCB was defined as complete or partial response or stable disease that lasted > 6 months. Progression-free survival and overall survival were recorded.

Results: Among 67 patients, 51 (76.1%) had > 1 variant detected at a variant allele fraction > 0.3% and thus were eligible for calculation of molecular response from paired baseline and 9-week samples. Molecular response values were significantly lower in patients with an objective radiologic response (log mean 1.25% v 27.7%, P < .001). Patients achieving a DCB had significantly lower molecular response values compared to patients with no durable benefit (log mean 3.5% v 49.4%, P < .001). Molecular responders had significantly longer progression-free survival (hazard ratio, 0.25; 95% CI, 0.13 to 0.50) and overall survival (hazard ratio, 0.27; 95% CI, 0.12 to 0.64) compared with molecular nonresponders.

Conclusion: Molecular response assessment using circulating tumor DNA may serve as a noninvasive, on-therapy predictor of response to pembrolizumab-based therapy in addition to standard of care imaging in mNSCLC. This strategy requires validation in independent prospective studies.

FIG 1.

Baseline mutational profile. Pretreatment circulating tumor DNA analysis showing nonsynonymous variants (green), indels (golden brown), or both (red) detected in each gene for the 62 patients with at least one baseline mutation identified. Each row indicates a gene for which one or more patient had a mutation detected, with rows ordered from top to bottom based on decreasing prevalence of mutations. Number of variants detected in each patient is represented by the height of the bars and patients are arranged in increasing order. The four rows at the bottom indicate assessment of response at 6 months with DCB (green), NDB (beige), and one patient without a 6-month assessment (dark gray); treatment with pembrolizumab monotherapy (gold) or pembrolizumab plus chemotherapy (light purple); PD-L1 status with PD-L1 > 50% (blue); and smoking status with current (dark purple), former (teal), and never smokers (light gray). DCB, durable clinical benefit; NDR, no durable benefit; PD-L1, programmed death ligand 1; SNV, single nucleotide variant.

FIG 2.

Molecular response and response to pembrolizumab-based therapy. (A) The association between molecular response and RECIST response at week-9 on therapy for the 17 patients with a CR or PR, 22 patients with SD, nine patients with PD, and three patients without a 9-week RECIST response assessment. (B) Association between molecular response and RECIST response at 6 months for the 33 patients achieving a DCB and 17 patients with an NDB. (C) Swimmer plots for each patient showing the duration of pembrolizumab-based therapy arranged by patients with a molecular response (blue) and without a molecular response (red). Patients treated with pembrolizumab monotherapy are denoted with a gold box to the left of the horizontal bar and those treated with pembrolizumab plus chemotherapy with a purple box. CR, complete response; DCB, durable clinical benefit; NA, RECIST not available; NDR, no durable benefit; PD, progressive disease; PR, partial response; SD, stable disease.

FIG 3.

Association of molecular response with survival outcomes. Kaplan-Meier survival curves using a molecular response cutoff of 50% for (A) PFS and (B) OS. Kaplan-Meier survival curves showing patients with complete clearance of ctDNA at 9 weeks (blue), molecular response > 0% (teal), and molecular response > 50% (red) for (C) PFS and (D) OS. Dotted lines show median survival times for each group; median PFS was not reached for patients with ctDNA clearance (C). Median OS was not reached for patients with ctDNA clearance or molecular response but no clearance (D). ctDNA, circulating tumor DNA; OS, overall survival; PFS, progression-free survival.

FIG 4.

ctDNA molecular response in pseudoprogression versus true progressive disease. (A) Patient with metastatic non–small-cell lung cancer and a dominant 7.0-cm RUL mass (red line) initiated pembrolizumab monotherapy; a 9-week (T1) scan showed an increase in the size of the RUL mass to 7.8 cm along with the development of mediastinal adenopathy. 9-week ctDNA analysis showed a reduction in identified variants with an AF > 0.3% and associated reduction the mean VAF with a molecular response = 17%. Pembrolizumab was continued and a follow-up scan at 16 weeks (T2) showed a reduction in the RUL mass to 6.2 cm and resolution of mediastinal adenopathy. The overall clinical picture was consistent with radiologic pseudoprogression to anti–PD-1 therapy. (B) Patient with newly diagnosed advanced NSCLC with a dominant 1.7-cm nodule in the RUL (red asterisk) initiated pembrolizumab monotherapy; a 9-week (T1) scan showed an increase in this lesion to 3.1 cm suggestive of progressive disease. 9-week ctDNA analysis shows an increase in ctDNA (molecular response = 189%). Pembrolizumab was continued and a repeat chest computed tomography at 13 weeks (T2) showed continued radiographic progression, and pembrolizumab was subsequently discontinued. ctDNA, circulating tumor DNA; RUL, right upper lobe; VAF, variant allele fraction.

FIG A1.

Somatic variant landscape at baseline and 9 weeks. Pretreatment (left panel) and on-treatment (right panel) ctDNA analysis showing nonsynonymous variants (green), indels (golden brown), or both (red) detected in each gene for the 62 patients with at least one baseline mutation identified. Number of variants detected in each patient is represented by the height of the bars and patients are arranged in increasing order. Rows indicate assessment of response at 6 months with DCB (green), NDB (beige), and one patient without a 6-month assessment (dark gray); treatment with pembrolizumab monotherapy (gold) or pembrolizumab plus chemotherapy (light purple); smoking status with current (dark purple), former (teal), and never smokers (light gray); and PD-L1 status with PD-L1 > 50% (dark blue). DCB, durable clinical benefit; NDR, no durable benefit; PD-L1, programmed death ligand 1.

FIG A2.

Comparison between molecular response evaluable and unevaluable cohorts. Response and outcomes for the 50 patients with a molecular response evaluable cohort (baseline Max-VAF > 0.3%) compared to the 16 patients in the unevaluable cohort (baseline Max-VAF < 0.3%). One patient in the evaluable cohort was lost to follow-up. (A) Baseline Max-VAF detected and RECIST response at 6 months. Kaplan-Meier survival curves using a cutoff baseline Max-VAF of > 0.3% for (B) progression-free survival and (C) overall survival. HR, hazard ratio; VAF, variant allele fraction.

FIG A3.

Assessment of differing methods to calculate molecular response to predict response and outcome. Comparison of different molecular response calculations to predict 9-week RECIST response and outcome. ROC curves using the different molecular response calculations to predict 6-month RECIST response. Three previously published molecular response calculations were assessed for correlation with response and outcome. Delta mean VAF (blue) = mean VAF on treatment − mean VAF at baseline. Relative delta (red) mean VAF = (mean VAF on treatment − mean VAF at baseline)/mean VAF at baseline. Ratio mean VAF Score (teal) = mVAF on-treatment/mVAF baseline. AUC, area under the curve; CR, complete response; HR, hazard ratio; OS, overall survival; PD, progressive disease; PFS, progression-free survival; PR, partial response; SD, stable disease; VAF, variant allele fraction.

FIG A4.

Association of molecular response with survival outcomes in treatment cohorts. Kaplan-Meier survival curves using a molecular response cutoff of 50% for (A) progression-free survival and (B) overall survival for patients treated with pembrolizumab monotherapy. Kaplan-Meier survival curves using a molecular response cutoff of 50% for (C) progression-free survival and (C) overall survival for patients treated with pembrolizumab plus chemotherapy. HR, hazard ratio.