Purpose:

Immunotherapy has transformed the treatment of many solid tumors, with some patients deriving long-term benefit, but how to identify such patients remains unclear. Somatic mutations detected in circulating tumor DNA (ctDNA) from plasma can be an indicator of disease progression, response to therapy, and clonality of primary and metastatic lesions. Hence, ctDNA analysis can provide a valuable noninvasive and tumor-specific marker for longitudinal monitoring of tumor burden. We explored the use of ctDNA to predict survival on durvalumab, an anti-PD-L1 therapy.

Experimental Design:

Variant allele frequencies (VAF) of somatic mutations in 73 genes were assessed in ctDNA using targeted sequencing in a discovery cohort consisting of 28 patients with non–small cell lung cancer (NSCLC) and two validation NSCLC and urothelial cancer (UC) cohorts of 72 and 29 patients, respectively, to correlate ctDNA changes with clinical outcomes.

Results:

Somatic variants were detected in 96% of patients. Changes in VAF preceded radiographic responses, and patients with reduction in VAF at 6 weeks had significantly greater reduction in tumor volume, with longer progression-free and overall survival.

Conclusions:

ctDNA VAF changes are strongly correlated with duration of treatment, antitumor activity, and clinical outcomes in NSCLC and UC. Early on-treatment reduction in ctDNA VAF may be a useful predictor of long-term benefit from immunotherapy. Prospective studies should validate these findings and the value of utilizing early changes in ctDNA for therapeutic decision making by identifying nonresponders to checkpoint inhibitor monotherapies and guiding combination therapies.

This article is featured in Highlights of This Issue, p. 6105

Levels of circulating tumor DNA (ctDNA) in plasma are known to correlate with tumor burden and changes from baseline correlate with response to some immunotherapies. Here, we demonstrate that somatic mutations in ctDNA are detectable in plasma from most patients with advanced/metastatic non–small cell lung cancer and urothelial cancer. Targeted genomic sequencing shows that during treatment with durvalumab, an anti-PD-L1 immunotherapy, a reduction in ctDNA variant allele frequency at 6 weeks is associated with, and could often precede, tumor shrinkage. The reduction is also associated with improved progression-free and overall survival. This early change may be a useful and noninvasive way to predict long-term benefit from immunotherapy, opening up unique opportunities to support decision making in indications where image-based response analysis may not be reliable, and to enable early treatment decisions before availability of radiographic response.

Circulating cell-free DNA (cfDNA) is present in plasma and serum of healthy individuals as well as patients with cancer. In patients with cancer, cfDNA is seen in markedly higher quantities compared with those found in healthy individuals (1). Circulating tumor DNA (ctDNA) is the fraction of cfDNA that is derived from tumor tissues (2). ctDNA is thought to be shed into circulation by apoptotic and necrotic tumor cells in patients with cancer (3, 4), highly prevalent in most advanced solid tumors with the exception of brain tumors (5), and has a half-life ranging from 16 minutes to a few hours (6–8). Because advanced tumors, either pretreated or at progression, have a higher mitotic index and undergo more rapid cell cycling compared with normal tissue or earlier stage tumors, ctDNA constitutes a larger proportion of cfDNA in metastatic disease (9, 10). Patients with high tumor burden and aggressive disease have higher proportions of ctDNA, which may rise above 90% of cfDNA, where it becomes a negative prognostic indicator.

Analysis of ctDNA provides a significant opportunity to study tumor growth dynamics and the evolving genomic landscape of tumors. The advantages of using liquid biopsies (e.g., plasma) for genomic analysis in patients with cancer include: (i) noninvasive nature of sample collection (i.e., blood draws); (ii) the ability to obtain repeat samples throughout the duration of treatment; (iii) representation of intra- and intertumor heterogeneity in patients; and (iv) continuous monitoring of genetic alterations during therapy as a surrogate of tumor burden and tumor clonal structure in the emergence of resistance to treatment (11). Numerous studies have shown that levels of ctDNA in plasma are correlated with tumor burden and that response to certain therapies correlate with decreased levels of ctDNA in on-treatment samples. It has been shown that patients with non–small cell lung cancer (NSCLC) with somatic alterations of >5% variant allele frequency (VAF) have shortened survival (12). A study that evaluated 162 plasma samples from 18 patients with colorectal cancer demonstrated that higher ctDNA levels were associated with higher tumor burden and that ctDNA dynamics may be more sensitive than carcinoembryonic antigen (CEA) levels in monitoring tumor burden (10). In other reports, tumor progression after chemotherapy was shown to be associated with increasing plasma (13, 14) and serum (15) DNA concentrations in NSCLC. Similar associations between ctDNA and tumor burden have been reported in metastatic melanoma (16), breast cancer (9), gynecologic malignancies (17), and metastatic colorectal cancer (18), and associations with response to various targeted and systemic therapies have been shown (19–23). A recent ctDNA study demonstrated the ability to observe emergence of resistance to adjuvant chemotherapy and the ability to perform phylogenetic ctDNA profiling to track subclones during relapse and metastasis (24).

Immune checkpoint inhibitors, such as anti-PD-1 [pembrolizumab (25), nivolumab (26)] and anti–PD-L1 [atezolizumab (27), durvalumab (28), and avelumab (29)], have shown clinical benefit in multiple tumor types. Monitoring responses to immune checkpoint inhibitors can be challenging due to their mechanism of action being significantly different from other types of therapies (30). Because these therapies seek to activate the immune system, response to therapy and subsequent tumor regression can be delayed compared with chemotherapy, radiotherapy, or targeted therapies (31). ctDNA analysis of early time-course samples could be used to gauge antitumor response to immunotherapy, opening up unique opportunities in immuno-oncology. Such assessment of patient responses could inform decision making in indications where image-based response analysis may not be reliable, to enable early treatment or treatment combination decisions before availability of radiographic response, and to quickly inform whether responses to certain treatment combinations (e.g., anti-PD-L1 and chemotherapy) are transient or durable.

ctDNA levels have been shown to correlate with response to immunotherapies. Somatic hotspot mutations in BRAF, cKIT, NRAS, and TERT were analyzed from ctDNA in 12 patients with metastatic melanoma receiving anti-CTLA-4 or anti-PD-L1 therapies. The study showed that mutant allele frequencies in hotspot genes correlated with clinical and radiologic outcomes and in one case, preceded manifestation of relapse (16). In a study of 48 patients with metastatic melanoma receiving adoptive transfer of autologous tumor-infiltrating lymphocytes, a strong correlation was seen between the clearance of BRAFV600E mutation in serum and complete response over the next 1 to 2 years. A majority of the patients showing no clearance failed to achieve objective response (32). In another study of patients with metastatic melanoma receiving anti-PD-1 therapy, ctDNA levels at baseline correlated with lactate dehydrogenase levels, tumor burden and Eastern Cooperative Oncology Group (ECOG) scores, and longitudinal assessment of ctDNA levels correlated with tumor response, progression-free survival (PFS), and overall survival (OS; ref. 33). A similar association between ctDNA and response to immunotherapy has been reported in colorectal cancer (34). In a study of 49 patients with NSCLC, targeted next-generation sequencing (NGS) of 43 hotspots in 24 genes found 57% of baseline samples positive for ctDNA. A 50% reduction in the variant with the highest variant allele fraction was seen in patients on anti-PD-1 or anti-PD-L1 checkpoint inhibitors at a median of 24.5 days on ctDNA but response was not observed by imaging until a median of 72.5 days. The decline in ctDNA was strongly associated with time on treatment, PFS, and OS (35).

These studies are limited in that they have focused on one or a few specific mutations or a specific cancer type, have a limited number of patients in the analysis, and do not provide independent cohorts to confirm initial findings.

Here, using a broad NGS-based mutation panel, we demonstrate a strong relationship between clinical outcome in metastatic NSCLC and changes in ctDNA VAF from baseline to 6 weeks after initiation of treatment with durvalumab. We validated this finding in independent sets of patients with NSCLC and urothelial carcinoma (UC) treated with durvalumab.

Study design and patients

Study 1108 (NCT01693562) is a phase I/II, first-in-human, multicenter, open-label, dose-escalation and dose-expansion study being conducted at 70 centers worldwide. Eligible patients were ≥18 years of age with histologically or cytologically confirmed inoperable or metastatic transitional-cell UC or NSCLC who had progressed on, been ineligible for, or refused any number of prior therapies. Patients had an ECOG performance status score of 0 or 1, adequate organ and hematologic functions, and fresh tumor biopsy and/or archival tumor tissue available for PD-L1 testing. Key exclusion criteria were active autoimmune disease or inflammatory bowel disease, prior severe or persistent immune-related adverse events (AE), previous exposure to anti-PD-1 or anti-PD-L1 therapy, requirement for >10 mg/day of prednisone or equivalent, and untreated central nervous system (CNS) metastases. As of April 29, 2016, 304 patients with NSCLC and 191 patients with UC had received durvalumab (10 mg/kg i.v. twice weekly).

ATLANTIC (NCT02087423) is a multicenter, phase II open-label study enrolling patients with stage IIIB/IV NSCLC with disease progression following two or more systemic treatments, including one platinum-based chemotherapy and one tyrosine kinase inhibitor (TKI) for EGFR mut/ALK+ patients. As of June 3, 2016, 444 patients had received durvalumab (10 mg/kg i.v. twice weekly) for up to 12 months.

For the purpose of discovery of ctDNA-based biomarkers in plasma, we used samples from 28 patients with NSCLC from Study 1108. The findings from these samples were independently validated in samples from 72 patients with EGFR-wild-type NSCLC (cohort 2) of ATLANTIC and 29 patients with UC in Study 1108 for confirmation.

The studies were conducted in accordance with Good Clinical Practices, the Declaration of Helsinki, and approval by each Institution's Ethical Review Board. Patients provided written informed consent.

Plasma collection and cfDNA isolation

As ctDNA analysis was not preplanned, our analysis was limited to available samples from patients who consented to participate in an optional biomarker component of the study. We used pretreatment samples that were collected either at screening or before first dose depending on sample availability. Post-dose samples were collected prior to the fourth treatment (6 weeks after first dose). Briefly, venous blood was collected in K2-EDTA tubes during routine phlebotomy and 10 mL of blood was processed to isolate plasma by centrifugation at 1,300 g for 10 minutes. Plasma was immediately aliquoted and stored at −20 °C or colder. Cell-free DNA was extracted from 1-mL aliquots of plasma using the QIAamp Circulating Nucleic Acid Kit (Qiagen), concentrated using Agencourt Ampure XP beads (Beckman Coulter), and quantified by Qubit fluorometer (Life Technologies). All cell-free DNA isolation and sequencing was performed at Guardant Health.

Genomic analysis

Genomic alterations (mutations, insertions, deletions, and amplifications) were detected from cfDNA extracted from plasma samples using a broad targeted NGS-based 73-gene panel (Guardant360), including coverage of the most prevalent tumor suppressor genes in human cancers. After isolation of cfDNA by hybrid capture, the assay is performed using molecular barcoding and proprietary bioinformatics algorithms with massively parallel sequencing on an Illumina Hi-Seq 2500 platform in a CLIA/CAP accredited laboratory (Guardant Health). Variants in plasma ctDNA were assessed in samples collected at predose and 6 weeks after first dose of treatment with durvalumab.

PD-L1 staining

PD-L1 status was determined by IHC, using a cutoff of PD-L1 expression on ≥25% of tumor cells at any intensity in NSCLC and expression on ≥25% of tumor or immune cells at any intensity in UC. PD-L1 expression level was determined by IHC using the SP263 anti–PD-L1 antibody assay (Ventana Medical Systems) as described previously (36, 37). Samples were classified as having ≥25% or <25% of tumor cell membranes or immune cells (for UC only) staining positive for PD-L1 at any intensity. This cutoff was chosen based on a number of considerations, including the prevalence of PD-L1 expression in the population, ease of scoring by pathologists, optimizing for higher negative predictive value, and delineating between responders and nonresponders (38).

Statistical analysis

Somatic variants of unknown significance as well as variants known to be associated with cancer including single nucleotide variants (SNV), insertions/deletions (indels), and fusions were summarized per patient by calculating the mean allele frequency of all genes with a VAF ≥0.3% at predose. Synonymous and nonsynonymous variants were included in calculation of VAF. Variants with VAF < 0.3% were not included in the mean VAF calculation based on 95% to 100% limits of detection of 0.2% to 0.25% for SNVs, indels, and fusions for this technology (39). Only variants observed at predose were used for the 6-week mean VAF calculation. For variants detected at predose but not 6 weeks, the 6-week VAF was set to 0. The change in mean VAF (dVAF) was calculated as (mean VAFweek 6) – (mean VAFpretreatment), hence a dVAF < 0 indicates a decrease at week 6. We compared dVAF between baseline and week 6 in plasma samples using a paired Student t test. To test the stability of VAF measurements, we analyzed replicate samples from the same patient at screening and predose in a limited number of cases and found that VAF as well as dVAF values calculated from these replicates were highly correlated (Supplementary Fig. S1).

For two NSCLC samples in the discovery set, and seven NSCLC and two UC samples in the validation sets, either no mutations were detected or allele frequencies of variants were below the 0.3% cutoff in predose samples. This left 26 NSCLC, 65 NSCLC, and 27 UC samples, respectively, for analysis of dVAF. The prevalence of mutations in The Cancer Genome Atlas for lung adenocarcinoma, lung squamous cell carcinoma, and bladder cancer indications were calculated using MUTECT2 variant calling data downloaded on May 26, 2016.

A Fisher exact test was used to determine the association between dVAF ≥ 0 or dVAF < 0 and PD-L1 expression ≥25% or <25%.

Objective response rate was calculated according to RECIST v1.1, and a Cox proportional hazards model was calculated adjusting for baseline ECOG score, gender, age, smoking status, previous lines of therapy, and histology.

The median follow-up time for patients with NSCLC and UC was 15 and 12 months, respectively, in Study 1108 and 9 months for ATLANTIC.

Demographics and characteristics of patients included in our analysis are listed in Table 1. In Study 1108, the majority of patients had undergone fewer lines of therapy than those enrolled in ATLANTIC. Both validation datasets [patients with EGFR–wild-type NSCLC (cohort 2) of ATLANTIC and patients with UC in Study 1108] had notably higher percentages of PD-L1–positive patients compared with the discovery set (patients with NSCLC in Study 1108).

Table 1.

Patient demographics, baseline characteristics, and prior therapies

Study 1108 lungATLANTICStudy 1108 bladder
 28 72 29 
Age Mean (SD) 62.21 (13.27) 61.43 (9.79) 65.86 (8.19) 
 Range 31–87 23–78 49–81 
Gender (%) 8 (28.6) 30 (41.7) 9 (31.0) 
 20 (71.4) 42 (58.3) 20 (69.0) 
Race (%) Asian 6 (21.4) 36 (50.0) 3 (10.3) 
 African American – – 1 (3.4) 
 White 21 (75.0) 36 (50.0) 22 (75.9) 
 Other 1 (3.6) – 3 (10.3) 
Previous lines of therapy (%) 7 (25.0) – 1 (3.4) 
 4 (14.3) – 16 (55.2) 
 17 (60.7) – 6 (20.7) 
 – 31 (43.1) 5 (17.2) 
 – 18 (25.0) 1 (3.4) 
 >4 – 23 (31.9) – 
Smoking history (%) Nonsmoker 6 (21.4) 13 (22.0) 13 (44.8) 
 Smoker 22 (78.6) 59 (78.0) 16 (55.2) 
Stage at entry (%) III 3 (10.7) 16 (18.1) – 
 IV 25 (89.3) 56 (66.7) 29 (100.0) 
ECOG/WHO PS at baseline (%) 10 (35.7) 27 (37.5) 11 (37.9) 
 18 (64.3) 45 (62.5) 18 (62.1) 
PDL1 status (%) Negative 13 (46.4) 11 (15.3) 6 (20.7) 
 Positive 13 (46.4) 58 (80.6) 23 (79.3) 
 Unknown 2 (7.1) 3 (4.2) – 
Histology (%) Nonsquamous 10 (35.7) 57 (79.2) – 
 Squamous 18 (64.3) 15 (20.8) – 
 Bladder – – 29 (100.0) 
Study 1108 lungATLANTICStudy 1108 bladder
 28 72 29 
Age Mean (SD) 62.21 (13.27) 61.43 (9.79) 65.86 (8.19) 
 Range 31–87 23–78 49–81 
Gender (%) 8 (28.6) 30 (41.7) 9 (31.0) 
 20 (71.4) 42 (58.3) 20 (69.0) 
Race (%) Asian 6 (21.4) 36 (50.0) 3 (10.3) 
 African American – – 1 (3.4) 
 White 21 (75.0) 36 (50.0) 22 (75.9) 
 Other 1 (3.6) – 3 (10.3) 
Previous lines of therapy (%) 7 (25.0) – 1 (3.4) 
 4 (14.3) – 16 (55.2) 
 17 (60.7) – 6 (20.7) 
 – 31 (43.1) 5 (17.2) 
 – 18 (25.0) 1 (3.4) 
 >4 – 23 (31.9) – 
Smoking history (%) Nonsmoker 6 (21.4) 13 (22.0) 13 (44.8) 
 Smoker 22 (78.6) 59 (78.0) 16 (55.2) 
Stage at entry (%) III 3 (10.7) 16 (18.1) – 
 IV 25 (89.3) 56 (66.7) 29 (100.0) 
ECOG/WHO PS at baseline (%) 10 (35.7) 27 (37.5) 11 (37.9) 
 18 (64.3) 45 (62.5) 18 (62.1) 
PDL1 status (%) Negative 13 (46.4) 11 (15.3) 6 (20.7) 
 Positive 13 (46.4) 58 (80.6) 23 (79.3) 
 Unknown 2 (7.1) 3 (4.2) – 
Histology (%) Nonsquamous 10 (35.7) 57 (79.2) – 
 Squamous 18 (64.3) 15 (20.8) – 
 Bladder – – 29 (100.0) 

Consistency in ctDNA detection rates and recurrently mutated genes across three patient cohorts of NSCLC or UC

We first evaluated the ability to detect ctDNA mutants in pretreatment plasma from patients with NSCLC in Study 1108. At least one somatic variant was observed in 27 of 28 (96%) NSCLC discovery, 29/29 UC (100%), and 68 of 72 NSCLC (94%) validation samples, suggesting that the method used has the sensitivity required for detecting ctDNA in most patients with these advanced cancers. Figure 1 summarizes the variants detected in ctDNA discovery and validation datasets noted above. The 5 most recurrent genes containing nonsynonymous variants or copy number amplifications in the NSCLC discovery cohort were TP53 (85%), PIK3CA (26%), FGFR1 (26%), ERBB2 (26%), and EGFR (22%; Fig. 1A). Within the NSCLC validation cohort, the 5 most recurrent genes containing nonsynonymous variants or copy number amplifications were TP53 (72%), KRAS (27%), EGFR (22%), BRAF (16%), and PIK3CA (16%; Fig. 1B). For the UC validation cohort, the most recurrent variants were TP53 (69%), ARID1A (41%), TERT (41%), PIK3CA (31%), and ERBB2 (31%; Fig. 1C).

Figure 1.

Mutations detected in ctDNA from 27 patients with NSCLC in the discovery cohort (A), 68 patients with NSCLC (B), and 29 patients with UC (C) in the validation cohorts, color-coded by the type of mutation.

Figure 1.

Mutations detected in ctDNA from 27 patients with NSCLC in the discovery cohort (A), 68 patients with NSCLC (B), and 29 patients with UC (C) in the validation cohorts, color-coded by the type of mutation.

Close modal

Reduction in mean VAF at 6 weeks is associated with change in tumor volume and time on study

We examined the correlation between dVAF and changes in tumor volume in response to durvalumab by comparing the percent change in tumor volume in patients showing dVAF < 0 versus those showing dVAF ≥ 0. We found that in the NSCLC discovery samples, mean tumor volume reduced by 39% in patients with dVAF < 0, whereas it increased by 36% in patients with dVAF ≥ 0 (Fig. 2A; P = 0.0001). A similar trend was observed in the validation sets: a decrease of 31% for dVAF < 0 and an increase of 11% for dVAF ≥ 0 in NSCLC (Fig. 2B, P = 0.0009), and a decrease of 38% for dVAF < 0 and a decrease of 15% for dVAF ≥ 0 in UC (Fig. 2C; P = 0.18).

Figure 2.

A–C, Changes in tumor volume plotted by dVAF < 0 versus dVAF ≥ 0 for NSCLC discovery cohort (A), NSCLC (B), and UC (C) validation cohorts. D–F, Median duration of treatment plotted by dVAF < 0 versus dVAF ≥ 0 for NSCLC discovery cohort (A), NSCLC (B), and UC (C) validation cohorts. On all plots, the horizontal bar represents the median, the box represents the interquartile range (IQR), and the whiskers represent 1.5 times the IQR above the upper quartile and below the lower quartile.

Figure 2.

A–C, Changes in tumor volume plotted by dVAF < 0 versus dVAF ≥ 0 for NSCLC discovery cohort (A), NSCLC (B), and UC (C) validation cohorts. D–F, Median duration of treatment plotted by dVAF < 0 versus dVAF ≥ 0 for NSCLC discovery cohort (A), NSCLC (B), and UC (C) validation cohorts. On all plots, the horizontal bar represents the median, the box represents the interquartile range (IQR), and the whiskers represent 1.5 times the IQR above the upper quartile and below the lower quartile.

Close modal

We also compared dVAF with the length of time patients stayed on durvalumab. In the NSCLC discovery samples, patients with dVAF < 0 had a median duration of treatment of 22 months. In contrast, duration of treatment for patients with dVAF ≥ 0 was only 6.5 months (Fig. 2D; P = 0.00001). This trend was confirmed in the validation sets: in the NSCLC cohort, the median duration of treatment for patients with dVAF < 0 was 12 months versus 8 months for dVAF ≥ 0 (Fig. 2E; P = 0.04). In the UC cohort, median duration of treatment was 13 months for dVAF < 0 versus 7 months for dVAF ≥ 0 (Fig. 2F; P = 0.03).

These observations suggest that a decrease in mean VAF within 6 weeks of initiation of durvalumab treatment may be associated with better outcomes.

Decreases in mean VAF after 6 weeks of treatment correlated with complete/partial response

ctDNA data were then evaluated for associations between dVAF and objective response by radiography. Each plot in Fig. 3 represents mean VAF changes in the three groups defined by radiographic response: complete/partial response (CR/PR), stable disease (SD), or progressive disease (PD). In the discovery NSCLC cohort (Fig. 3A), mean VAF decreased by 2.7% (P = 0.005) for patients with PR and 1.5% (P = 0.14) for patients with SD. In contrast, patients with PD showed an increase of 1.7% (P = 0.16). In the validation NSCLC cohort (Fig. 3B), the mean VAF decreased by 4% (P = 0.0009) in patients with CR/PR and 1.1% (P = 0.2) in patients with SD, whereas the mean VAF increased 1.4% (P = 0.3) in patients with PD. Similarly, in the validation UC cohort (Fig. 3C), the mean VAF decreased by 2.2% (P = 0.009) in patients with CR/PR and 1.1% (P = 0.36) in patients with SD, whereas the mean VAF increased 2.5% (P = 0.42) in patients with PD. Overall, of the 37 responders in the three studies, 36 had a reduction in mean VAF at week 6; 23 of 46 patients with PD showed an increase in mean VAF, whereas 12 patients with SD showed an increase of mean VAF and 22 patients showed a decrease.

Figure 3.

Changes in mean ctDNA VAF at 6 weeks compared with baseline, plotted by objective response status for NSCLC discovery cohort (A), NSCLC (B), and UC (C) validation cohorts (CR, complete response; PR, partial response; SD, stable disease; PD, progressive disease).

Figure 3.

Changes in mean ctDNA VAF at 6 weeks compared with baseline, plotted by objective response status for NSCLC discovery cohort (A), NSCLC (B), and UC (C) validation cohorts (CR, complete response; PR, partial response; SD, stable disease; PD, progressive disease).

Close modal

Reduction in mean VAF after 6 weeks of treatment with durvalumab may precede radiographic tumor shrinkage

The ability to detect early clinical response or nonresponse through the course of therapy can help inform clinical decisions. Therefore, we investigated whether a decrease in mean VAF after 6 weeks of treatment preceded radiographic determination of a decrease in tumor volume by ≥30%. Figure 4 shows each patient who received durvalumab colored by outcome as determined by radiography. The plot also shows each patient's duration of treatment with durvalumab, their mean VAF status at 6 weeks (increase or decrease), and the time at which the first radiographic response was recorded. In the discovery samples, out of the 10 patients who showed a ≥30% decrease in tumor volume, 7 showed a corresponding decrease in mean VAF 1 to 12 months prior to radiographic confirmation (Fig. 4A). Similar trends were noted in the validation samples. In the NSCLC cohort, 24 patients showed a ≥30% decrease in tumor volume and 22 of them had a corresponding decrease in mean VAF. In 9 of those 22 patients, the decreases in mean VAF preceded radiographic change in tumor volume by 3.6 to 9 months (Fig. 4B). Similarly, in the UC cohort, 14 patients had a ≥30% decrease in tumor volume; 11 of them had a corresponding decrease in mean VAF. In 4 of those 11 patients, the decreases in mean VAF preceded radiographic reduction in tumor volume by 1.3 to 6 months (Fig. 4C).

Figure 4.

Individual swimmer plots for each patient showing the duration of treatment for NSCLC discovery cohort (A), NSCLC (B), and UC (C) validation cohorts. Lanes are colored by objective responses PR/CR (), SD (), and PD (). Increase () or decrease () in dVAF is marked for each patient. Timepoints at which a radiographic response was confirmed () following the RECIST criteria are marked on the plot.

Figure 4.

Individual swimmer plots for each patient showing the duration of treatment for NSCLC discovery cohort (A), NSCLC (B), and UC (C) validation cohorts. Lanes are colored by objective responses PR/CR (), SD (), and PD (). Increase () or decrease () in dVAF is marked for each patient. Timepoints at which a radiographic response was confirmed () following the RECIST criteria are marked on the plot.

Close modal

Interestingly, 2 patients with UC had an increase in mean VAF and a ≥30% increase in tumor volume at 6 weeks. Both stayed on durvalumab and showed a ≥30% reduction in tumor volume 2 and 6 months later, respectively. Overall, reductions in mean VAF 6 weeks after initiation of treatment preceded radiographic responses in 20 of 48 (42%) patients.

Patients with a reduction in VAF after 6 weeks of treatment with durvalumab had improved survival

Next, we evaluated whether dVAF correlated with PFS and/or OS. Figure 5A shows that the patients with NSCLC in the discovery cohort with dVAF ≥ 0 had markedly shorter PFS compared with patients with dVAF < 0 [median PFS (mPFS) 1.45 vs. 13.7 months, HR 0.008 (95% CI, 0.0007–0.09)]. The median OS (mOS) was also shorter in patients with dVAF ≥ 0 versus dVAF < 0: 9.07 months versus 28.13 months, respectively, HR 0.001 (95% CI, 0–0.09). Consistent with the discovery cohort, an increase in mean VAF correlated with shorter mPFS for patients with NSCLC in the validation cohort (Fig. 5B): mPFS 1.9 months for dVAF ≥ 0 versus 5.6 months for dVAF < 0, HR 0.26 (95% CI, 0.12–0.54). Likewise, mOS for dVAF ≥ 0 versus dVAF < 0 was 8.7 months versus NR, HR 0.23 (95% CI, 0.09–0.61). Similarly, for the UC cohort (Fig. 5C), patients with an increase in mean VAF had markedly shorter PFS [mPFS of 1.6 months for dVAF ≥ 0 vs. 13.8 months for dVAF < 0, HR 0.21 (95% CI, 0.05–0.96)] and OS [mOS of 8.2 months for dVAF ≥ 0 vs. NR for dVAF < 0, HR 0.001 (95% CI, 0–0.28)]. Overall, patients with a decrease in mean VAF had significantly longer PFS and OS compared with patients with an increase in mean VAF.

Figure 5.

Kaplan–Meier curves of median PFS and OS in relation to increase or decrease in mean VAF: NSCLC discovery cohort (A), NSCLC (B), and urothelial carcinoma (UC) (C) validation cohorts.

Figure 5.

Kaplan–Meier curves of median PFS and OS in relation to increase or decrease in mean VAF: NSCLC discovery cohort (A), NSCLC (B), and urothelial carcinoma (UC) (C) validation cohorts.

Close modal

The relationships observed between PFS, OS, and dVAF were not significantly impacted by baseline ECOG score, gender, age, smoking status, previous lines of therapy, or histology.

Tumor PD-L1 expression can predict clinical outcomes for durvalumab and other anti-PD-1/PD-L1 agents (40). For this reason, we examined the correlation between pretreatment PD-L1 status (≥25% or <25% as measured by IHC) and dVAF. Patients in all three discovery and validation cohorts showed no statistical association between dVAF and PD-L1 status (P > 0.1 for all contrasts, data not shown).

Interestingly, the analysis of patients with PD after 6 weeks of therapy showed the emergence of new EGFR mutations in 7 patients and increase in mean VAF in 5 patients. These EGFR mutations include V765M, T638M, and R973Q, which can sensitize the tumors to approved EGFR-targeting agents like osimertinib and erlotinib (41–43).

Circulating biomarkers such as ctDNA offer significant promise as valuable tools for monitoring tumor burden and antitumor response, and providing a real-time snapshot of tumor evolution in a metastatic setting as patients relapse and are treated with multiple lines of therapy over time. Previous studies have shown that ctDNA can be used to monitor responses to conventional as well as targeted therapies (9, 16–23). In this study, we investigated the potential utility of changes in ctDNA as an early predictor of efficacy during anti–PD-L1 therapy. Using an NGS-based 73-gene panel, we detected mutations in 124/129 (96%) patients. Mutation frequencies were generally consistent with frequencies reported in COSMIC (44), except for PIK3CA and FGFR1 for lung cancer, and ARID1A and TERT in UC, which we detected more frequently. These differences may be related to differences in disease stage, as well as better coverage of these genes on our panel.

Patients who had a decrease in mean VAF (dVAF < 0) after 6 weeks of durvalumab stayed on therapy significantly longer than patients who had an increase (dVAF ≥ 0). This suggests that a change in mean VAF is directly related to antitumor activity and may have clinical significance. Decreases in mean VAF also correlated well with tumor response (CR/PR). In addition, we found that patients with a decrease in mean VAF had a markedly improved PFS and OS compared with those with dVAF ≥ 0, and that dVAF does not significantly correlate with PD-L1 status. These data strongly support the potential use of ctDNA dVAF as an endpoint to inform drug development and/or treatment decisions.

Although PFS HRs can be obtained by conventional radiographic response (CR/PR/SD vs. PD), in this study we demonstrated that in 20 of 48 (42%) patients, a reduction in VAF at 6 weeks is an early indicator of patient response or nonresponse to durvalumab therapy. Hence, changes in VAF at 6 weeks could potentially be used to guide early treatment decisions. This finding is consistent with earlier reports utilizing limited ctDNA testing (35). Patients showing an increase in VAF at 6 weeks could be moved on to combination therapies or other treatments in indications where multiple options are available. Such information may also aid in early decision making in pan-tumor trials and trials with adaptive design. Furthermore, ctDNA dVAF data and imaging data may not necessarily provide redundant information, and the two are likely to complement each other in future clinical practice. ctDNA dVAF assessment may aid in the identification of pseudoprogressions, may be particularly relevant in indications where evaluation of radiographic response is challenging (e.g., pancreas and liver), and may help predict relapse in an adjuvant setting (24, 45).

It is noteworthy that the dVAF and objective response were not concordant in a minority of patients in our study. These patients represent a particularly interesting area for future research, to better understand whether ctDNA assessment helps improve the accuracy of clinical response assessments in such cases. Longitudinal monitoring of changes in tumor burden with liquid biopsies through the course of therapy may be particularly relevant for developing combination therapies in immuno-oncology, as it may predict and/or confirm radiographic responses by an orthogonal method and may help to distinguish durable versus transient responses.

In our study, we also saw the emergence of new EGFR mutations in patients with PD at week 6, which is consistent with activating mutations in EGFR conferring resistance to immunotherapy (46–50). Such patients, therefore, could potentially benefit from combination therapies with EGFR-targeted agents.

In conclusion, we have shown strong correlations between ctDNA dVAF and duration of treatment, clinical activity, PFS, and OS. This supports the use of ctDNA dVAF to monitor antitumor activity and clinical response to immunotherapies. Future studies should validate our findings through prospective analysis. Identification of ctDNA mutations associated with emerging resistance may provide an opportunity to test rational combinations.

R. Raja, M. Kuziora, P. Z. Brohawn, B. W. Higgs, and K. Ranade are employees of MedImmune and have ownership interest (including patents) in AstraZeneca. P.A. Dennis holds ownership interest (including patents) in AstraZeneca. A. Gupta holds ownership interest (including patents) in AstraZeneca and Bristol Myers Squibb.

Conception and design: R. Raja, M. Kuziora, P. Z. Brohawn, A. Gupta, K. Ranade

Development of methodology: R. Raja, M. Kuziora, B.W. Higgs

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Raja, P. Z. Brohawn, A. Gupta

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Raja, M. Kuziora, P. Z. Brohawn, B.W. Higgs, A. Gupta, P.A. Dennis, K. Ranade

Writing, review, and/or revision of the manuscript: R. Raja, M. Kuziora, P. Z. Brohawn, B.W. Higgs, A. Gupta, P.A. Dennis, K. Ranade

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B.W. Higgs

Study supervision: R. Raja, A. Gupta, K. Ranade

We thank the patients who participated in this study and their families. Editorial support, which was in accordance with Good Publication Practice (GPP3) guidelines, was provided by Susanne Gilbert, MA, of Cirrus Communications and was funded by MedImmune. This study was funded by MedImmune/AstraZeneca.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Leon
SA
,
Shapiro
B
,
Sklaroff
DM
,
Yaros
MJ
. 
Free DNA in the serum of cancer patients and the effect of therapy
.
Cancer Res
1977
;
37
:
646
50
.
2.
Shu
Y
,
Wu
X
,
Tong
X
,
Wang
X
,
Chang
Z
,
Mao
Y
, et al
Circulating tumor DNA mutation profiling by targeted next generation sequencing provides guidance for personalized treatments in multiple cancer types
.
Sci Rep
2017
;
7
:
583
.
3.
Snyder
MW
,
Kircher
M
,
Hill
AJ
,
Daza
RM
,
Shendure
J
. 
Cell-free DNA comprises an in vivo nucleosome footprint that informs its tissues-of-origin
.
Cell
2016
;
164
:
57
68
.
4.
Stroun
M
,
Lyautey
J
,
Lederrey
C
,
Olson-Sand
A
,
Anker
P
. 
About the possible origin and mechanism of circulating DNA apoptosis and active DNA release
.
Clin Chim Acta
2001
;
313
:
139
42
.
5.
Bettegowda
C
,
Sausen
M
,
Leary
RJ
,
Kinde
I
,
Wang
Y
,
Agrawal
N
, et al
Detection of circulating tumor DNA in early- and late-stage human malignancies
.
Sci Transl Med
2014
;
6
:
224ra224
.
6.
Lo
YM
,
Zhang
J
,
Leung
TN
,
Lau
TK
,
Chang
AM
,
Hjelm
NM
. 
Rapid clearance of fetal DNA from maternal plasma
.
Am J Hum Genet
1999
;
64
:
218
24
.
7.
Minchin
RF
,
Carpenter
D
,
Orr
RJ
. 
Polyinosinic acid and polycationic liposomes attenuate the hepatic clearance of circulating plasmid DNA
.
J Pharmacol Exp Ther
2001
;
296
:
1006
12
.
8.
Yu
SC
,
Lee
SW
,
Jiang
P
,
Leung
TY
,
Chan
KC
,
Chiu
RW
, et al
High-resolution profiling of fetal DNA clearance from maternal plasma by massively parallel sequencing
.
Clin Chem
2013
;
59
:
1228
37
.
9.
Dawson
SJ
,
Tsui
DW
,
Murtaza
M
,
Biggs
H
,
Rueda
OM
,
Chin
SF
, et al
Analysis of circulating tumor DNA to monitor metastatic breast cancer
.
N Engl J Med
2013
;
368
:
1199
209
.
10.
Diehl
F
,
Schmidt
K
,
Choti
MA
,
Romans
K
,
Goodman
S
,
Li
M
, et al
Circulating mutant DNA to assess tumor dynamics
.
Nat Med
2008
;
14
:
985
90
.
11.
Jamal-Hanjani
M
,
Wilson
GA
,
McGranahan
N
,
Birkbak
NJ
,
Watkins
TBK
,
Veeriah
S
, et al
Tracking the evolution of non-small-cell lung cancer
.
N Engl J Med
2017
;
376
:
2109
21
.
12.
Schwaederlé
MC
,
Patel
SP
,
Husain
H
,
Ikeda
M
,
Lanman
RB
,
Banks
KC
, et al
Utility of genomic assessment of blood-derived circulating tumor DNA (ctDNA) in patients with advanced lung adenocarcinoma
.
Clin Cancer Res
2017
;
23
:
5101
11
.
13.
Kumar
S
,
Guleria
R
,
Singh
V
,
Bharti
AC
,
Mohan
A
,
Das
BC
. 
Plasma DNA level in predicting therapeutic efficacy in advanced nonsmall cell lung cancer
.
Eur Respir J
2010
;
36
:
885
92
.
14.
Pan
S
,
Xia
W
,
Ding
Q
,
Shu
Y
,
Xu
T
,
Geng
Y
, et al
Can plasma DNA monitoring be employed in personalized chemotherapy for patients with advanced lung cancer?
Biomed Pharmacother
2012
;
66
:
131
7
.
15.
Gautschi
O
,
Bigosch
C
,
Huegli
B
,
Jermann
M
,
Marx
A
,
Chassé
E
, et al
Circulating deoxyribonucleic acid as prognostic marker in non–small-cell lung cancer patients undergoing chemotherapy
.
J Clin Oncol
2004
;
22
:
4157
64
.
16.
Lipson
EJ
,
Velculescu
VE
,
Pritchard
TS
,
Sausen
M
,
Pardoll
DM
,
Topalian
SL
, et al
Circulating tumor DNA analysis as a real-time method for monitoring tumor burden in melanoma patients undergoing treatment with immune checkpoint blockade
.
J Immunother Cancer
2014
;
2
:
42
.
17.
Pereira
E
,
Camacho-Vanegas
O
,
Anand
S
,
Sebra
R
,
Catalina Camacho
S
,
Garnar-Wortzel
L
, et al
Personalized circulating tumor DNA biomarkers dynamically predict treatment response and survival in gynecologic cancers
.
PLoS One
2015
;
10
:
e0145754
.
18.
Tie
J
,
Kinde
I
,
Wang
Y
,
Wong
HL
,
Roebert
J
,
Christie
M
, et al
Circulating tumor DNA as an early marker of therapeutic response in patients with metastatic colorectal cancer
.
Ann Oncol
2015
;
26
:
1715
22
.
19.
Frenel
JS
,
Carreira
S
,
Goodall
J
,
Roda
D
,
Perez-Lopez
R
,
Tunariu
N
, et al
Serial next-generation sequencing of circulating cell-free DNA evaluating tumor clone response to molecularly targeted drug administration
.
Clin Cancer Res
2015
;
21
:
4586
96
.
20.
Gray
ES
,
Rizos
H
,
Reid
AL
,
Boyd
SC
,
Pereira
MR
,
Lo
J
, et al
Circulating tumor DNA to monitor treatment response and detect acquired resistance in patients with metastatic melanoma
.
Oncotarget
2015
;
6
:
42008
18
.
21.
Momtaz
P
,
Gaskell
AA
,
Merghoub
T
,
Viale
A
,
Chapman
PB
. 
Correlation of tumor-derived circulating cell free DNA (cfDNA) measured by digital PCR (DigPCR) with tumor burden measured radiographically in patients (pts) with BRAFV600E mutated melanoma (mel) treated with RAF inhibitor (RAFi) and/or ipilimumab (Ipi)
.
J Clin Oncol
32
:
15s
, 
2014
(
suppl; abstr 9085
).
22.
Murtaza
M
,
Dawson
SJ
,
Tsui
DW
,
Gale
D
,
Forshew
T
,
Piskorz
AM
, et al
Non-invasive analysis of acquired resistance to cancer therapy by sequencing of plasma DNA
.
Nature
2013
;
497
:
108
12
.
23.
Panka
DJ
,
Buchbinder
E
,
Giobbie-Hurder
A
,
Schalck
AP
,
Montaser-Kouhsari
L
,
Sepehr
A
, et al
Clinical utility of a blood-based BRAF(V600E) mutation assay in melanoma
.
Mol Cancer Ther
2014
;
13
:
3210
8
.
24.
Abbosh
C
,
Birkbak
NJ
,
Wilson
GA
,
Jamal-Hanjani
M
,
Constantin
T
,
Salari
R
, et al
Phylogenetic ctDNA analysis depicts early-stage lung cancer evolution
.
Nature
2017
;
545
:
446
51
.
25.
Farina
MS
,
Lundgren
KT
,
Bellmunt
J
. 
Immunotherapy in urothelial cancer: recent results and future perspectives
.
Drugs
2017
;
77
:
1077
89
.
26.
Sharma
P
,
Callahan
MK
,
Bono
P
,
Kim
J
,
Spiliopoulou
P
,
Calvo
E
, et al
Nivolumab monotherapy in recurrent metastatic urothelial carcinoma (CheckMate 032): a multicentre, open-label, two-stage, multi-arm, phase 1/2 trial
.
Lancet Oncol
2016
;
17
:
1590
8
.
27.
Powles
T
,
Eder
JP
,
Fine
GD
,
Braiteh
FS
,
Loriot
Y
,
Cruz
C
, et al
MPDL3280A (anti-PD-L1) treatment leads to clinical activity in metastatic bladder cancer
.
Nature
2014
;
515
:
558
62
.
28.
Mehta
K
,
Patel
K
,
Parikh
RA
. 
Immunotherapy in genitourinary malignancies
.
J Hematol Oncol
2017
;
10
:
95
.
29.
Katz
H
,
Wassie
E
,
Alsharedi
M
. 
Checkpoint inhibitors: the new treatment paradigm for urothelial bladder cancer
.
Med Oncol
2017
;
34
:
170
.
30.
Pardoll
DM
. 
The blockade of immune checkpoints in cancer immunotherapy
.
Nat Rev Cancer
2012
;
12
:
252
64
.
31.
Wolchok
JD
,
Hoos
A
,
O'Day
S
,
Weber
JS
,
Hamid
O
,
Lebbé
C
, et al
Guidelines for the evaluation of immune therapy activity in solid tumors: immune-related response criteria
.
Clin Cancer Res
2009
;
15
:
7412
20
.
32.
Xi
L
,
Pham
TH
,
Payabyab
EC
,
Sherry
RM
,
Rosenberg
SA
,
Raffeld
M
, et al
Circulating tumor DNA as an early indicator of response to t-cell transfer immunotherapy in metastatic melanoma
.
Clin Cancer Res
2016
;
22
:
5480
6
.
33.
Lee
JH
,
Long
GV
,
Boyd
S
,
Lo
S
,
Menzies
AM
,
Tembe
V
, et al
Circulating tumour DNA predicts response to anti-PD-1 antibodies in metastatic melanoma
.
Ann Oncol
2017
;
28
:
1130
6
.
34.
Kitahara
M
,
Hazama
S
,
Tsunedomi
R
,
Takenouchi
H
,
Kanekiyo
S
,
Inoue
Y
, et al
Prediction of the efficacy of immunotherapy by measuring the integrity of cell-free DNA in plasma in colorectal cancer
.
Cancer Sci
2016
;
107
:
1825
9
.
35.
Goldberg
SB
,
Narayan
A
,
Kole
AJ
,
Decker
RH
,
Teysir
J
,
Carriero
NJ
, et al
Early assessment of lung cancer immunotherapy response via circulating tumor DNA
.
Clin Cancer Res
2018
April 15 [Epub ahead of print]
.
36.
Powles
T
,
O’Donnell
PH
,
Massard
C
,
Arkenau
HT
,
Friedlander
TW
,
Hoimes
CJ
, et al
Efficacy and safety of durvalumab in locally advanced or metastatic urothelial carcinoma: updated results from a phase 1/2 open-label study
.
JAMA Oncol
2017
;
3
:
e172411
.
37.
Garassino
MC
,
Cho
BC
,
Kim
JH
,
Mazieres
J
,
Vansteenkiste
J
,
Lena
H
, et al
Durvalumab as third-line or later treatment for advanced non-small-cell lung cancer (ATLANTIC): an open-label, single-arm, phase 2 study
.
Lancet Oncol
2018
;
19
:
521
536
.
38.
Rebelatto
MC
,
Midha
A
,
Mistry
A
,
Sabalos
C
,
Schecter
N
,
Li
X
, et al
Development of a programmed cell death ligand-1 immunohistochemistry assay validated for analysis of non-small cell lung cancerand head and neck squamous cell carcinoma
.
Diag Pathol
2016
;
11
:
95
.
39.
Pal
SK
,
Brooks
C
,
Chudova
D
,
Odegaard
J
,
Gandara
DR
,
Mack
P
, et al
Clinical implications of genomic variants identified in over 30,000 advanced-stage cancer patients by next-generation sequencing of circulating tumor DNA
.
Ann Oncol
2017
;
28
:
v595
v604
.
40.
Patel
SP
,
Kurzrock
R
. 
PD-L1 expression as a predictive biomarker in cancer immunotherapy
.
Mol Cancer Ther
2015
;
14
:
847
56
.
41.
Kim
HS
,
Kim
SM
,
Kim
H
,
Pyo
KH
,
Sun
JM
,
Ahn
MJ
, et al
Phase II clinical and exploratory biomarker study of dacomitinib in recurrent and/or metastatic esophageal squamous cell carcinoma
.
Oncotarget
2015
;
6
:
44971
84
.
42.
Miron
B
,
Peled
N
,
Tarcic
G
,
Barbash
Z
,
Edelheit
O
,
Vidne
M
, et al
Functional profiling of oncogenic mutations in lung cancer patients (NCT02274025) - interim results
.
Ann Oncol
2016
;
27
:
1189P
.
43.
Barretina
J
,
Caponigro
G
,
Stransky
N
,
Venkatesan
K
,
Margolin
AA
,
Kim
S
, et al
The Cancer Cell Line Encyclopedia enables predictive modelling of anticancer drug sensitivity
.
Nature
2012
;
483
:
603
7
.
44.
Forbes
SA
,
Beare
D
,
Gunasekaran
P
,
Leung
K
,
Bindal
N
,
Boutselakis
H
, et al
COSMIC: exploring the world's knowledge of somatic mutations in human cancer
.
Nucleic Acids Res
2015
;
43
:
D805
11
.
45.
Chaudhuri
A
,
Chabon
JJ
,
Lovejoy
AF
,
Newman
AM
,
Stehr
H
,
Azad
TD
. 
Analysis of circulating tumor DNA in localized lung cancer for detection of molecular residual disease and personalization of adjuvant strategies
.
J Clin Oncol
35
:
15s
, 
2017
(
suppl; abstr 8519
).
46.
Borghaei
H
,
Paz-Ares
L
,
Horn
L
,
Spigel
DR
,
Steins
M
,
Ready
NE
, et al
Nivolumab versus docetaxel in advanced nonsquamous non-small-cell lung cancer
.
N Engl J Med
2015
;
373
:
1627
39
.
47.
Fehrenbacher
L
,
Spira
A
,
Ballinger
M
,
Kowanetz
M
,
Vansteenkiste
J
,
Mazieres
J
, et al
Atezolizumab versus docetaxel for patients with previously treated non-small-cell lung cancer (POPLAR): a multicentre, open-label, phase 2 randomised controlled trial
.
Lancet
2016
;
387
:
1837
46
.
48.
Hellmann
MD
,
Rizvi
NA
,
Goldman
JW
,
Gettinger
SN
,
Borghaei
H
,
Brahmer
JR
, et al
Nivolumab plus ipilimumab as first-line treatment for advanced non-small-cell lung cancer (CheckMate 012): results of an open-label, phase 1, multicohort study
.
Lancet Oncol
2017
;
18
:
31
41
.
49.
Herbst
RS
,
Baas
P
,
Kim
DW
,
Felip
E
,
Pérez-Gracia
JL
,
Han
JY
, et al
Pembrolizumab versus docetaxel for previously treated, PD-L1-positive, advanced non-small-cell lung cancer (KEYNOTE-010): a randomised controlled trial
.
Lancet
2016
;
387
:
1540
50
.
50.
Lee
CK
,
Man
J
,
Lord
S
,
Links
M
,
Gebski
V
,
Mok
T
, et al
Checkpoint inhibitors in metastatic EGFR-mutated non-small cell lung cancer - a meta-analysis
.
J Thorac Oncol
2017
;
12
:
403
7
.

Supplementary data