Cell-Free Circulating Tumor DNA for NonInvasive Molecular Profiling Among Patients Undergoing Definitive Chemoradiation for Locally Advanced Lung Cancer

Abstract

Purpose

We prospectively explored the utility of liquid biopsy for cell-free circulating tumor DNA (ctDNA) as a prognostic and predictive biomarker in patients with non-small cell lung cancer (NSCLC) treated with definitive chemoradiation therapy.

Methods and Materials

This prospective clinical cohort consisted of patients with unresectable, locally advanced NSCLC who had liquid biopsy testing before initiation of cancer therapy. Liquid biopsy testing was performed using an institutional assay that included 129 genes and paired white blood cell sequencing. Variant allele frequency was defined as the proportion of mutant alleles at a particular genetic locus. A US Food and Drug Administration-recognized database (OncoKB) was used to classify alterations. We evaluated progression-free survival from the start of radiation therapy using the log-rank test.

Results

Among 25 patients with prospective testing of ctDNA levels before therapy initiation, 18 patients had adenocarcinoma (72%), 7 patients had squamous cell carcinoma (28%), and 23 (92%) were former or current smokers. Twelve patients (48%) received adjuvant durvalumab. The median radiation dose was 60 Gy in 30 fractions (range, 55-66 Gy in 20-33 fractions). Seventy-six percent of patients (n = 18) had one or more alterations detected (median, 3 alterations, range, 1-8), including genomic markers of radiation response in 3 patients. The most common driver alteration detected was KRAS mutation in 24% of the cohort (n = 6). The detection of ctDNA levels was significantly associated with pretreatment 18F-fluorodeoxyglucose positron emission tomography standardized uptake value metrics, and the association was strengthened by integrating the number of mutations (compared with variant allele frequency) as the outcome variable. Among patients with baseline detectable ctDNA levels, the median progression-free survival was 21.3 months and was not reached among patients without baseline ctDNA level detection (hazard ratio, 4.54, P = .04).

Conclusions

Prospective liquid biopsy testing among patients treated with definitive chemoradiation therapy identifies driver alterations and markers of radiation response with direct implications for therapy personalization.

Introduction

Genomic profiling is the standard of care for patients with advanced lung cancer and is routinely used to direct therapy selection. Recent data suggest that patients with early-stage lung cancer may also benefit from the personalization of therapy based on genetic profiling.1, 2, 3 Patients with localized lung cancer treated with definitive radiation therapy (RT) often have a paucity of tissue, limiting the feasibility of comprehensive genomic profiling. This population may also be challenging to molecularly profile noninvasively because of decreased tumor volume and the absence of extrapulmonary disease.^4,5 These factors may in turn limit cell-free circulating tumor DNA (ctDNA) shedding and detection.⁶ We assessed the feasibility and outcomes of noninvasive genomic profiling using liquid biopsy for ctDNA among patients with localized lung cancer treated with definitive RT. We also evaluated radiographic factors associated with the detection of baseline ctDNA levels in this context.

Methods and Materials

Patients were prospectively enrolled in an institutional review board protocol if they had stage II, stage III, or locally recurrent lung cancer and were planned for definitive RT at our cancer center between September 2020 and May 2021. On protocol, patients had plasma collection for comprehensive genomic profiling before treatment initiation. A subset of patients also had tissue-based genomic profiling at the discretion of the treating clinician. Genomic alterations were categorized according to OncoKB, a US Food and Drug Administration-recognized human genetic variant database.⁷ Overall survival (OS) and progression-free survival (PFS) were defined from the start of RT. Disease recurrence within the planning target volume was considered a local recurrence, whereas recurrence outside the planning target volume was categorized as a distant recurrence. Pretreatment emission tomography (PET)-computed tomography (CT) scans were obtained for imaging analysis and registered to planning CT contours to identify gross disease.

Variant allele frequency (VAF) represents the proportion of mutant alleles at a particular genetic locus.⁸ Undetectable ctDNA level was classified as VAF of 0. Time to recurrence, OS, and PFS were defined from the start of RT. The cumulative incidence of local and distant progression was evaluated using Gray's test, considering death as a competing risk factor, with curve comparisons conducted using the log-rank test. Linear regression analysis was applied to explore the association between ctDNA levels and 9 first-order pretreatment standardized uptake value (SUV) metrics (kurtosis maximum, mean, median, minimum, ±SD, p10, p90, and skewness) of the gross tumor volumes from pretreatment 18F-fluorodeoxyglucose positron emission tomography 18F-FDG PET-CT scans. To perform this imaging analysis, the planning CT scans were deformably registered to the PET-CT scans using B-splines, and all gross tumor volumes were systematically corrected for errors caused by motion, registration, and tumor growth using a previously validated technique.⁹

Results

A total of 25 patients treated with definitive RT had noninvasive prospective genomic profiling with Memorial Sloan Kettering - Analysis of Circulating cfDNA to Examine Somatic Status (Table 1). A total of 19 patients (76%) had one or more ctDNA alterations detected at baseline. The median number of ctDNA alterations detected per patient was 1 (range, 0-8), and the median VAF was 0.45. Baseline ctDNA level burden, as defined by the number of mutations and VAF, was not significantly associated with the T stage, N stage, or overall disease stage.

Table 1.

Baseline patient and treatment characteristics

Patient Characteristic	No. (%)
Total	25 (100%)
Sex
Male	14 (56%)
Female	11 (44%)
KPS score
90-100	15 (60%)
70 -80	10 (40%)
Histology
Adenocarcinoma	18 (72%)
Squamous cell carcinoma	7 (28%)
AJCC 8th edition staging
IIB	2 (8%)
IIIA	6 (24%)
IIIB	11 (44%)
IIIC	3 (12%)
Locally recurrent	3 (12%)
Smoking history
Current/former	23 (92%)
Never	2 (8%)
Radiation dose, median (range)	60 Gy (55- 66 Gy)
Concurrent systemic therapy	21 (84%)
Adjuvant durvalumab	22 (88%)
Race
White	12 (48%)
Asian	7 (28%)
Black	3 (12%)
Other	3 (12%)

Abbreviations: AJCC = The American Joint Committee on Cancer; KPS = Karnofsky Performance Status

Following plasma collection for ctDNA level testing, all patients received definitive RT to a median dose of 60 Gy (range, 55-63 Gy). Additionally, 22 patients (88%) received concurrent systemic therapy, and 12 patients (48%) received adjuvant durvalumab.

Figure 1 contains an Oncoprint of the most commonly detected ctDNA alterations. Oncogenic driver alterations were most commonly detected in KRAS. One patient with a KRAS G12V driver mutation had a co-occurring pathogenic KEAP1 alteration (KEAP1 Q282*).

Figure 1.

Oncoprint of most common alterations detected in circulating tumor DNA.

Abbreviation: LR = local recurrence.

Association between ctDNA levels and outcomes and pretreatment SUV metrics

Among the 7 patients with disease progression, 4 patients had distant failure, 2 patients had isolated locoregional recurrence, and 1 patient had combined locoregional and distant recurrence.

Among the 19 patients with detectable ctDNA levels at baseline, the 24-month cumulative incidence of distant failure was 27% (95% CI, 9.2%-48%), whereas none of the 6 patients with undetectable baseline ctDNA levels had a distant failure (P = .2; Fig. 2A). All 3 patients who had a locoregional failure also had baseline detectable ctDNA levels, whereas none of the patients without baseline detectable ctDNA levels had a locoregional failure. The 24-month incidence of local failure was 11% among patients with detectable disease at baseline, whereas none of the 6 patients with undetectable ctDNA levels had local failure (P = .4; Fig. 2B). The VAF was not associated with the cumulative incidence of locoregional or distant failure or with the incidence of local failure, distant failure, PFS, or OS.

Figure 2.

(A) Cumulative incidence of distant failure with complete risk of death using cell-free circulating tumor DNA (ctDNA) level detection. (B) Cumulative incidence of local failure with complete risk of death using ctDNA detection.

Correlation between imaging features and ctDNA levels

The detection of ctDNA levels was significantly and linearly associated with 4 of 9 pretreatment 18F-fluorodeoxyglucose PET SUV metrics on linear regression analysis which used 2 approaches to quantify ctDNA level (number of mutations and maximum VAF). A stronger association using the number of mutations versus VAF as the outcome variable (R² = 0.17-0.29 vs 0.14; P = .004-.03 vs .04) (Fig. 3; Table 2)

Figure 3.

The linear regression curves for the association between detected cell-free circulating tumor DNA (ctDNA), (Max VAF) and SUV_Kurtosis (A) and between ctDNA (number of mutations) and SUV_Max (B).

The linear regression equation is inserted above the regression line and the coefficient of determination (P-value) is inserted below the corresponding regression line.

Abbreviations: SUV = standardized uptake value; VAF = variant allele frequency.

Table 2.

The linear regression analysis results between circulating tumor DNA (ctDNA) detection (max variant allele frequency [VAF] or number of mutations) and each of the 9 first-order histogram standardized uptake value metrics of the gross tumor volume from pretreatment 18F-fluorodeoxyglucose positron emission tomography-computed tomography scans

	R2	R2	P-value	P-value	B0	B0	B1	B1
SUV metric	Max VAF	No. Mutations	Max VAF	No. mutations	Max VAF	No. Mutations	Max VAF	No. Mutations
Mean	−0.03	0.15	.56	.04	0.78	0.18	0.17	0.49
±SD	0.00	0.24	.34	.009	0.57	0.36	0.32	0.63
Median	−0.02	0.01	.46	0.28	0.70	1.02	0.29	0.34
Mini	0.04	−0.01	.18	.39	2.91	2.42	−2.28	−0.57
Maxi	0.11	0.29	.06	.004	−0.69	−0.22	0.15	0.16
P10	−0.03	−0.02	.64	.46	1.73	1.65	−0.22	0.24
P90	−0.04	0.17	.66	.03	1.15	0.53	0.04	0.21
Skewness	0.07	−0.02	.12	.44	−0.04	1.78	1.19	0.25
Kurtosis	0.14	−0.01	.04	.39	0.61	2.02	0.45	0.06

Abbreviations: B₀ = regression coefficient for intercept; B₁ = regression coefficient for analyzed standardized uptake value metric; Max = maximum; Min = minimum; p10 = the 10th percentile; p90 = the 90th percentile; R² = coefficient of determination; SD = standard deviation.

Bold = reached statistical significance of p<0.05.

Discussion

Surgical resection for localized lung cancer provides ample tissue for biomarker testing. However, patients who are not candidates for surgical resection based on the extent of disease or medical comorbidities receive definitive RT. These patients often have insufficient tissue, leading to challenges with comprehensive genomic profiling. In this study, we demonstrate the feasibility of prospective, noninvasive, comprehensive genomic profiling using liquid biopsy for cell-free ctDNA among patients with localized lung cancer treated with definitive RT.

Genomic profiling is critical for patients with locally advanced lung cancer treated with definitive chemoradiation. The long advanced unresectable regional areas (LAURA) trial demonstrated a significant PFS benefit among patients receiving adjuvant osimertinib following chemoradiation therapy in stage III unresectable lung cancer harboring an EGFR mutation as detected on tissue or blood-based ctDNA level testing (39.1 months with osimertinib vs 5.6 months with placebo).¹⁰ This result highlights the critical clinical need for molecular testing to guide personalized adjuvant therapy based on each patient's unique genomic profile. Given the limited tissue availability among patients with unresectable disease who have only biopsy specimens, blood-based biomarker testing is an important option in this patient population.

In addition to targeting oncogenic driver alterations, there is increasing recognition of the prognostic and predictive impact of co-occurring alterations,¹¹ which also have the potential for therapeutic targeting.¹² Profiling of all patients will be increasingly important to identify rare genetic subgroups of patients with localized lung cancer to increase enrollment in protocols and access to personalized therapies.

Serial monitoring ctDNA levels using an assay designed to detect minimal residual disease may best identify a subset of patients with localized lung cancer who are at the highest risk of disease progression.^13,14 While our institutional assay was not developed for the detection of minimal residual disease, we nonetheless observed an association between the baseline ctDNA levels and the development of local and distant disease progression. Compared with patients identified using tissue-based genomic profiling, patients identified through blood-based ctDNA levels profiling may disproportionately have higher stages or more aggressive underlying disease biologies, resulting in more shedding of ctDNA levels into peripheral circulation and worse prognoses.⁴

We noted that several pretreatment PET characteristics, including maximum standardized uptake value, are associated with the level of ctDNA. PET characteristics are known to be prognostic in locally advanced lung cancer.¹⁵ In metastatic lung adenocarcinoma, ctDNA levels provide prognostic information independent of radiologic features.⁴ The extent to which baseline ctDNA levels may provide independent prognostic information in locally advanced disease needs to be evaluated in larger patient cohorts.

There are limitations to this analysis. While prospective and conducted at a large institution with facilities throughout the surrounding community, our data enrolled patients at a single institution. Our cohort size of 25 patients limits the ability to evaluate specific genetic subgroups and includes patients with a variety of disease stages. In addition, patients only had ctDNA analysis at a single timepoint (diagnosis), and therefore dynamic changes in ctDNA levels over time could not be evaluated as a biomarker in this study. Nonetheless, we demonstrated the feasibility and utility of prospective genomic profiling in a population of tissue-limited patients treated with definitive RT for localized lung cancer and presented data that can be applied to justify further analyses on the impact of blood-based prognostication and predictive value in the context of unresectable, locally advanced NSCLC.

Disclosures

Daphna Y. Gelblum reports equity in Doximity. Charles B. Simone, II reports honoraria from Varian and Novocure and serves in a leadership position for the Proton Collaborative Group, American Society for Radiation Oncology, NRG Oncology, American Radium Society, and Annals of Palliative Medicine. Annemarie F. Shepherd reports equity in Doximity and Arcell X. Puneeth Iyengar reports funding from AstraZeneca and Incyte. Bob T. Li reports funding from AstraZeneca, Amgen, and Daiichi Sankyo and consulting fees from Bolt Biotherapeutics, Inc. James Isbell reports funding from Intuitive Surgical Inc, AstraZeneca, and Merck Sharp & Dohme and equity in LumaCyte. Daniel R. Gomez reports funding from AstraZeneca, Grail, Johnson & Johnson, Med Learning Group, Medtronic, Regeneron, and Varian Medical Systems.