Early Clearance of Plasma Epidermal Growth Factor Receptor Mutations as a Predictor of Outcome on Osimertinib in Advanced Non–Small Cell Lung Cancer; Exploratory Analysis from AURA3 and FLAURA

Osimertinib is a third-generation, irreversible, oral EGFR–tyrosine kinase inhibitor (EGFR-TKI) that potently and selectively inhibits both EGFR-TKI sensitizing [exon 19 deletion (ex19del) and L858R; EGFRm] and EGFR T790M mutations in non–small cell lung cancer (NSCLC), and has demonstrated efficacy in NSCLC including central nervous system (CNS) metastases (1–6). The drug was first approved for the treatment of EGFR T790M NSCLC following disease progression on EGFR-TKI treatment. The randomized, phase III AURA3 clinical trial (NCT02151981) compared osimertinib with platinum-based doublet chemotherapy in patients with EGFR T790M positive advanced NSCLC, whose disease had progressed on first-line EGFR-TKI therapy (5). Patients treated with osimertinib had a significantly improved median progression-free survival (mPFS) compared with platinum-pemetrexed [10.1 vs. 4.4 months; HR, 0.30; 95% confidence interval (CI), 0.23–0.41; P < 0.001; ref. 5). Subsequently, osimertinib was approved for first-line therapy based on the phase III FLAURA trial results (NCT02296125) in which patients with untreated EGFRm advanced NSCLC receiving osimertinib had significantly improved mPFS versus comparator EGFR-TKI (erlotinib/gefitinib; 18.9 vs. 10.2 months; HR, 0.46; 95% CI, 0.37–0.57; P < 0.001; refs. 2, 7, 8). This also translated to a statistically significant median overall survival (mOS) benefit (38.6 vs. 31.8 months; HR, 0.80; 95.05% CI, 0.64–1.00; P = 0.046; ref. 9).

Delivery of effective targeted therapy requires timely detection of a targetable genotype. Challenges with routine tumor genotyping have led to the emergence of robust tools for circulating tumor DNA (ctDNA) EGFRm detection; plasma ctDNA genotyping is recommended where tissue is limited (10) and has the benefit of being noninvasive, requiring only a peripheral blood draw, potentially allowing longitudinal testing on therapy to monitor outcomes and evaluate resistance. However, there remains limited trial-level data in advanced NSCLC regarding the association between changes in ctDNA detection and treatment outcomes. Because EGFR-TKI sensitizing mutations are clonal driver mutations thought to occur as a truncal event in the development of EGFRm NSCLC, they are an ideal ctDNA marker and a potential surrogate for overall tumor burden.

We hypothesized that clearance of EGFRm ctDNA from plasma after initial detection at baseline might correlate with treatment effect in EGFRm NSCLC. Serial plasma samples were analyzed independently from AURA3 and FLAURA for the presence of EGFRm before and during treatment and correlated with PFS and objective response rate (ORR).

Full details of the AURA3 study have been published previously (5). Briefly, AURA3 was a randomized, open-label, phase III trial that assessed the efficacy and safety of osimertinib versus platinum-based doublet chemotherapy as second-line therapy in patients with EGFR T790M advanced NSCLC and disease progression after first-line EGFR-TKI therapy. Eligible patients were stratified on the basis of race (Asian and non-Asian) and randomly assigned (2:1) to receive either oral osimertinib (80 mg once daily) or platinum-based doublet chemotherapy (pemetrexed 500 mg/m² + carboplatin area under curve 5 or pemetrexed 500 mg/m² + cisplatin 75 mg/m²) on day 1 of every 21-day cycle. Tumor tissue T790M status was centrally confirmed using the cobas® EGFR Mutation Test (Roche Molecular Systems).

Similarly, full details of the phase III, double-blind, randomized FLAURA study have been published previously (2). In brief, FLAURA assessed the efficacy and safety of osimertinib versus comparator EGFR-TKIs (erlotinib or gefitinib) in treatment-naïve patients with EGFRm locally advanced or metastatic NSCLC. Local or central tissue confirmation of the ex19del or L858R EGFR mutation, alone or co-occurring with other EGFR mutations, was required. Patients with CNS metastases whose condition was neurologically stable were eligible. Eligible patients were stratified on the basis of EGFR mutation type (ex19del or L858R) and race (Asian and non-Asian) and randomly assigned 1:1 to receive either oral osimertinib 80 mg once daily or comparator EGFR-TKI (gefitinib 250 mg or erlotinib 150 mg once daily).

AURA3 and FLAURA were conducted in accordance with the principles outlined in the Declaration of Helsinki, Good Clinical Practice and local regulatory requirements, and was approved by the institutional review boards or independent ethics committees of the participating study centers. All patients provided written informed consent.

The representativeness of those patients included in the reported analyses is presented in Supplementary Table S1.

Provision of plasma samples in AURA3 and FLAURA was mandatory, per study protocol, for all patients who gave informed consent. Samples from patients who withdrew consent were excluded from analysis. Blood was collected into EDTA tubes prior to treatment on cycle 1, day 1 (baseline), on cycle 2, day 1 (3 weeks), and on cycle 3, day 1 (6 weeks). Within 4 hours of collection, whole blood was centrifuged for 10 minutes at 1,200 g after which the plasma supernatant was further cleared by centrifugation for 10 minutes at 3,000 g and stored at −80°C until use. Plasma ctDNA EGFR mutation analysis was conducted by droplet digital PCR (ddPCR; Biodesix, Boulder, CO). For this analysis, patients evaluable for ctDNA clearance analysis were required to have a baseline plasma sample and a plasma sample at Weeks 3 and/or 6 of treatment that returned a valid ddPCR result for EGFR mutations (ex19del, L858R). Baseline EGFRm that became non-detectable at later time points was defined as ‘clearance’. Baseline EGFRm that remained detectable at later time points was defined as ‘non-clearance’. Samples that did not pass Biodesix quality control metrics (<1% of tested samples) were excluded from analyses. The vast majority of excluded patients from this analysis were due to lack of consent at the time of the analysis or missing/uncollected samples. ctDNA next-generation sequencing (NGS; Guardant Health) was used as an orthogonal method to confirm technical performance. Samples from China were also excluded due to export restrictions.

Clinical outcomes assessed in this exploratory analysis were PFS and ORR as assessed by the investigator in accordance with Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST v1.1). OS was also assessed, defined as time from randomization until death from any cause.

Sample size was determined by availability of plasma samples collected for ctDNA analysis. Clinical outcomes were investigated on the basis of clearance of plasma EGFRm at Weeks 3 and/or 6 with HRs calculated using an unadjusted Cox proportional hazards model using profile likelihood CIs and a Breslow approach for handling ties (11). The correlation between EGFRm variant allele frequency (VAF) in baseline plasma and PFS was calculated using a simple linear regression. Among all patients randomized in AURA3 and in FLAURA, and with a valid baseline plasma ctDNA result, univariable analyses to assess impact of baseline covariates on PFS and OS were performed for each using Cox proportional hazards models; each model included treatment and one of the baseline covariates of interest [smoking history, brain or visceral metastases, presence of detectable plasma EGFRm at baseline, World Health Organization performance status (WHO PS), disease classification, race, and EGFR mutation type]. Multivariable analyses to assess impact of baseline covariates on PFS and OS were performed for each study using a Cox proportional hazards model for treatment and all the baseline covariates of interest. ORs for RECIST-based responses versus non-responses and plasma ctDNA clearance versus non-clearance were calculated, with P values determined using Fisher's exact test. Clinical outcome data cutoffs were April 15, 2016 (AURA3) and June 12, 2017 (FLAURA) for PFS and ORR, and March 15, 2019 (AURA3) and June 25, 2019 (FLAURA) for OS.

Data underlying the findings described in this manuscript may be obtained in accordance with AstraZeneca's data sharing policy described at https://astrazenecagrouptrials.pharmacm.com/ST/Submission/Disclosure. Full study protocols available at https://astrazenecagrouptrials.pharmacm.com/ST/Submission/View?id=12356.

Requests to access the data from the AURA3 and FLAURA studies described in the current manuscript can be submitted through: https://vivli.org/members/enquiries-about-studies-not-listed-on-the-vivli-platform/. Some patients/countries may need to be excluded on the basis of the informed consent form or country‐level legislation. Use of data must comply with the requirements of Human Genetics Resources Administration of China and patients who have withdrawn consent for data use will be removed from the shared dataset. Patient-level image or genetic data is not available for access.

Overall, 419 patients were enrolled in AURA3, of whom 291 (69%) had a valid plasma ddPCR result at baseline and were included in the evaluable ctDNA patient population (osimertinib arm, n = 209; platinum-pemetrexed arm, n = 82; Supplementary Fig. S1). In the evaluable ctDNA at baseline patient population, 206 (71%) and 85 (29%) patients had baseline detectable and non-detectable plasma EGFRm, respectively. Baseline demographics are presented in Table 1. The proportion of patients with baseline detectable plasma EGFRm was similar between treatment arms [osimertinib arm, 148/209 (71%), platinum-pemetrexed arm, 58/82 (71%)]. A high concordance was demonstrated for EGFRm detection and VAF at baseline between ddPCR and NGS techniques (Supplementary Fig. S2A).

Pooled analysis across both arms showed mPFS (95% CI) of 12.4 months [8.1–not calculable (NC)] and 6.7 months (5.6–7.3) in patients with baseline non-detectable versus detectable plasma EGFRm, respectively (HR, 0.48; 95% CI, 0.33–0.68; P < 0.0001; Fig. 1A). In the osimertinib arm, mPFS (95% CI) was 14.0 months (12.3–NC) and 8.3 months (6.8–10.6) in patients with baseline non-detectable and detectable plasma EGFRm, respectively (Fig. 1B); values in the platinum-pemetrexed arm were 5.8 months (4.2–9.9) and 4.2 months (4.1–5.6), respectively (Fig. 1B).

A pooled analysis of OS across both arms showed mOS (95% CI) of 44.9 months (37.2–NC) and 21.6 months (18.9–24.1) in patients with baseline non-detectable versus detectable plasma EGFRm, respectively (HR, 0.40; 95% CI, 0.28–0.56; P < 0.0001; Supplementary Fig. S3A).

A univariable analysis of baseline characteristics found only plasma EGFRm (detectable vs. non-detectable) and WHO PS (0 vs. 1) to be prognostic factors for PFS and OS, with EGFR mutation (ex19del vs. L858R) also prognostic for OS (Table 2; Supplementary Table S2). Detection of plasma EGFRm at baseline was the only characteristic with prognostic value for PFS in the corresponding multivariable analysis while for OS, all baseline characteristics that were prognostic in the univariable analysis were also prognostic in the multivariable analysis with the addition of race (Asian vs. non-Asian) and smoking (no vs. yes; Table 2; Supplementary Table S2). There was no correlation in either treatment arm between PFS and EGFRm VAF (Supplementary Fig. S4A) showing that the association between baseline detectable plasma EGFRm and PFS did not extend to EGFRm allele frequency.

Among 189 patients in the osimertinib arm with a valid baseline (detectable or non-detectable) and Week 3 ddPCR result, 80 (42%) patients had baseline detectable plasma EGFRm that became non-detectable (clearance). Of 189 patients in the osimertinib arm with a valid baseline (detectable or non-detectable) and Week 6 ddPCR result, 83 (44%) patients had baseline detectable plasma EGFRm that cleared (Fig. 2A). In the platinum-pemetrexed arm, 80 patients and 81 patients had a valid baseline (detectable or non-detectable) and Week 3 and/or Week 6 ddPCR result, respectively. Clearance occurred at a lower frequency in the platinum-pemetrexed arm, with 14/80 (18%) patients and 21/81 (26%) patients having baseline detectable plasma EGFRm that cleared at Weeks 3 and 6, respectively (Fig. 2A). In patients with baseline detectable plasma EGFRm, clearance (non-detectable plasma EGFRm) was more frequently observed at Week 3 in patients receiving osimertinib (80/128, 63%) versus platinum-pemetrexed (14/56, 25%); OR, 5.0; 95% CI, 2.5–10.1; P < 0.00001 (Fisher's test). A similar trend was observed for Week 6 clearance: osimertinib (83/128, 65%) versus platinum-pemetrexed (21/57, 37%); OR, 3.2; 95% CI, 1.7–6.1; P = 0.0007.

In patients with baseline detectable plasma EGFRm treated with osimertinib, mPFS (95% CI) at Week 3 was 10.9 months (8.3–12.6) in patients with non-detectable plasma EGFRm (clearance; n = 80) and 5.7 months (4.1–9.7) in patients with detectable plasma EGFRm (non-clearance; n = 48; Fig. 2B). Similar outcomes were observed in patients with clearance (n = 83) or non-clearance (n = 45) at Week 6: 10.9 months (8.5–12.5) and 4.2 months (3.6–6.5), respectively (Fig. 2C). In patients with baseline detectable plasma EGFRm treated with platinum-pemetrexed, the mPFS (95% CI) in patients with clearance (n = 14) and non-clearance (n = 42) at Week 3 was 6.2 months (4.0–9.7) and 4.2 months (4.0–5.1), respectively (Fig. 2B); similar values were observed in patients with clearance (n = 21) and non-clearance (n = 36) at Week 6: 6.6 months (4.0–8.4) and 4.2 months (4.0–4.4), respectively (Fig. 2C).

PFS was significantly improved with osimertinib versus platinum-pemetrexed irrespective of clearance of plasma EGFRm at Week 3 and Week 6. mPFS for osimertinib versus platinum-pemetrexed in patients with detectable plasma EGFRm at Weeks 3 and 6 was 5.7 versus 4.2 months (HR, 0.54; 95% CI, 0.34–0.86; P = 0.009; Fig. 2B) and 4.2 versus 4.2 months (HR, 0.61; 95% CI, 0.37–1.00; P = 0.05; Fig. 2C), respectively. Among those patients with clearance of plasma EGFRm at Weeks 3 and 6, mPFS was 10.9 versus 6.2 months (HR, 0.42; 95% CI, 0.22–0.80; P = 0.008; Fig. 2B) and 10.9 versus 6.6 months (HR, 0.43; 95% CI, 0.25–0.74; P = 0.002; Fig. 2C), respectively.

Consistent with the overall patient population in AURA3, among those patients with baseline detectable plasma EGFRm and valid ctDNA results at Weeks 3 and/or 6, ORR was significantly higher in those receiving osimertinib versus platinum-pemetrexed (70% vs. 42%; OR, 3.2; 95% CI, 1.7–6.0; P = 0.0004). There was a trend towards a higher ORR in patients treated with osimertinib versus platinum-pemetrexed even in patients with baseline detectable plasma EGFRm and non-clearance at Week 3 (50% vs. 38%; OR, 1.62; 95% CI, 0.70–3.81; P = 0.2559) and at Week 6 (49% vs. 33%; OR, 1.91; 95% CI, 0.78–4.83; P = 0.1569). ORR was higher across both treatment arms in those patients with clearance versus those with non-clearance at Weeks 3 and 6. In patients receiving osimertinib versus platinum-pemetrexed with clearance at Week 3, ORR was 81% versus 57% (OR, 3.25; 95% CI, 0.95–10.81; P = 0.0603) and at Week 6 ORR was 81% versus 57% (OR, 3.14; 95% CI, 1.11–8.77; P = 0.0309).

Overall, 556 patients were enrolled in FLAURA, of whom 499 (90%) had a valid plasma ddPCR result at baseline (evaluable ctDNA patient population: osimertinib arm, n = 247; comparator EGFR-TKI arm, n = 252; Supplementary Fig. S5). In the evaluable ctDNA at baseline patient population, 352 (71%) and 147 (29%) patients had baseline detectable or non-detectable plasma EGFRm, respectively. Baseline demographics are presented in Table 1. The proportion of patients with baseline detectable plasma EGFRm was similar between treatment arms [osimertinib arm, 171/247 (69%), comparator EGFR-TKI arm, 181/252 (72%)]. Similar to that observed with the AURA3 data, good concordance was demonstrated for EGFRm detection and VAF at baseline between ddPCR and NGS techniques (Supplementary Fig. S2B).

Pooled analysis across both arms showed mPFS (95% CI) of 19.1 months (16.6–23.4) and 11.1 months (10.8–13.1) in patients with baseline non-detectable versus detectable plasma EGFRm (HR, 0.54; 95% CI, 0.41–0.70; P < 0.0001; Fig. 3A). In the osimertinib arm, mPFS (95% CI) was 23.5 months (19.1–NC) and 15.2 months (13.6–20.7) in patients with baseline non-detectable or detectable plasma EGFRm, respectively; a similar trend was observed in the comparator EGFR-TKI arm with mPFS of 15.3 months (13.7–18.3) and 9.6 months (8.3–10.2), respectively (Fig. 3B).

A pooled analysis of OS across both arms showed mOS (95% CI) of NC (47.7–NC) and 29.7 months (25.8–32.4) in patients with baseline non-detectable versus detectable plasma EGFRm, respectively (HR, 0.34; 95% CI, 0.25–0.45; P < 0.0001; Supplementary Fig. S3B).

In a univariable analysis of baseline characteristics, plasma EGFRm (detectable vs. non-detectable), in addition to presence of brain or visceral metastases, WHO PS (0 vs. 1), and EGFR mutation (ex19del vs. L858R) were found to be prognostic factors for PFS and OS, with disease classification (locally advanced vs. metastatic) also being prognostic for OS (Table 2; Supplementary Table S2). In the corresponding multivariable analysis, plasma EGFRm detection and EGFR mutation remained independent prognostic factors for PFS, with the presence of baseline detectable plasma EGFRm being a negative prognostic factor for PFS (Table 2). Plasma EGFRm, WHO PS, and EGFR mutation remained prognostic factors for OS in the multivariable analysis (Supplementary Table S2). Similar to that observed with the AURA3 data, there was no correlation in either treatment arm between PFS and EGFRm VAF (Supplementary Fig. S4B).

Among 238 patients in the osimertinib arm with a valid baseline (detectable or non-detectable) and Week 3 ddPCR result, 106 (45%) patients had baseline detectable plasma EGFRm that became non-detectable (clearance). Of 240 patients in the osimertinib arm with a valid baseline (detectable or non-detectable) and Week 6 ddPCR result, 134 (56%) patients had baseline detectable plasma EGFRm that cleared (Fig. 4A). In the comparator EGFR-TKI arm, 243 patients and 235 patients had a valid baseline (detectable or non-detectable) and Week 3 and/or Week 6 ddPCR result, respectively. Clearance occurred at a similar frequency to the osimertinib arm, with 102/243 (42%) patients and 124/235 (53%) patients having baseline detectable plasma EGFRm that cleared at Weeks 3 and 6, respectively (Fig. 4A). In patients with baseline detectable plasma EGFRm, frequency of clearance was slightly higher at Week 3 in patients receiving osimertinib (106/162, 65%) versus comparator EGFR-TKI (102/172, 59%); OR, 1.3; 95% CI, 0.8–2.1; P = 0.3 (Fisher's test). A similar trend was observed for Week 6 clearance: osimertinib (134/164, 82%) versus comparator EGFR-TKI (124/164, 76%); OR, 1.4; 95% CI, 0.8–2.6; P = 0.2.

In patients with baseline detectable plasma EGFRm treated with osimertinib, mPFS (95% CI) in patients with clearance (n = 106) and non-clearance (n = 56) at Week 3 was 19.8 months (15.1–NC) and 11.3 months (9.5–16.5), respectively (Fig. 4B). Similar mPFS (95% CI) with osimertinib was observed in patients with clearance (n = 134) and non-clearance (n = 30) at Week 6: 19.8 months (15.1–NC) and 11.1 months (6.8–13.8), respectively (Fig. 4C). In the comparator EGFR-TKI arm, the mPFS (95% CI) in patients with clearance (n = 102) and non-clearance (n = 70) at Week 3 was 10.8 months (9.7–11.1) and 7.0 months (5.6–8.3), respectively (Fig. 4B). Similar mPFS (95% CI) with comparator EGFR-TKI was observed in patients with clearance (n = 124) and non-clearance (n = 40) at Week 6: 10.2 months (9.5–11.1) and 8.2 months (5.0–9.6), respectively (Fig. 4C).

In the subgroup of patients with non-clearance of plasma EGFRm at Week 3, mPFS was significantly longer with osimertinib compared with comparator EGFR-TKI (11.3 vs. 7.0 months; HR, 0.50; 95% CI, 0.33–0.76; P = 0.001; Fig. 4B); however, the difference was not significant when comparing the subgroup of patients with non-clearance at Week 6 (11.1 vs. 8.2 months; HR, 0.69; 95% CI, 0.40–1.17; P = 0.164; Fig. 4C). In the subgroup of patients with clearance of plasma EGFRm, mPFS was significantly longer with osimertinib compared with comparator EGFR-TKI at both Week 3 (19.8 vs. 10.8 months; HR, 0.41; 95% CI, 0.29–0.58; P < 0.0001; Fig. 4B) and Week 6 (19.8 vs. 10.2; HR, 0.40; 95% CI, 0.29–0.55; P < 0.0001; Fig. 4C). Baseline plasma EGFRm status and clearance at Weeks 3 and 6 were strong predictors of PFS regardless of treatment group (Supplementary Fig. S6).

The ORR was similar between the osimertinib and comparator EGFR-TKI treatment groups in patients with non-clearance of plasma EGFRm at Week 3 (80% vs. 76%; OR, 1.31; 95% CI, 0.56–3.16; P = 0.5319) and at Week 6 (73% vs. 73%; OR, 1.04; 95% CI, 0.36–3.11; P = 0.9381), respectively. ORR was also similar between the treatment groups in patients with clearance of plasma EGFRm at Week 3 (86% vs. 88%; OR, 0.81; 95% CI, 0.35–1.82; P = 0.6083) and at Week 6 (86% vs. 90%; OR, 0.65; 95% CI, 0.29–1.38; P = 0.2643).

Tumors that shed detectable ctDNA at baseline represent a poor prognosis subset, thus baseline assessment of detectable plasma ctDNA may be a valuable prognostic marker. Further, monitoring plasma ctDNA changes over time may offer insight into treatment effects in patients with EGFRm NSCLC. In both the AURA3 and FLAURA trials, pooled analyses across treatment arms showed patients with baseline non-detectable plasma EGFRm had longer PFS and OS compared with patients with baseline detectable plasma. Multivariable analyses showed baseline detectable plasma EGFRm to be the only baseline characteristic that was an independent prognostic factor for PFS in both trials with EGFR mutation type also being prognostic in FLAURA. Notably, in both trials WHO PS was not found to be an independent prognostic factor for PFS. Baseline detectable plasma EGFRm was also an independent prognostic factor for OS in both trials, as were WHO PS and EGFR mutation type.

Plasma ctDNA levels are variable and reflective of tumor burden and progression, with other studies also showing detectable ctDNA to be predictive of worse outcomes (12–14). Furthermore, circulating tumor fraction, measured prior to treatment initiation, was found to be an independent prognostic factor for OS in advanced NSCLC in a large retrospective study of patients with different tumor types (15). However, the prognostic power of baseline ctDNA is not fully understood, as demonstrated by the lack of correlation observed between EGFRm VAF and PFS in our analysis, which was an unexpected outcome (Supplementary Fig. S4). These results suggest that allele frequency is not key for predicting PFS when EGFRm is detected at baseline, and it is likely that the variable impact of EGFR amplifications on EGFRm allele frequency confounds any association with PFS. These data suggest that there are likely additional underlying biological differences more complex than just overall disease burden that influence whether ctDNA is detectable or not. Furthermore, the sum of target lesions does not fully account for all known disease and the detectable ctDNA signal is also likely influenced by micrometastatic disease that scans do not detect. Ongoing research aims to understand the correlation between ctDNA detection and clinical features (16) indicative of more aggressive disease. The potential to identify patients with a poor prognosis and increased risk of progression may help support baseline ctDNA analysis as a clinical tool for risk stratification.

In patients with baseline detectable plasma EGFRm (approximately two thirds of patients in both trials), early plasma EGFRm clearance (Week 3 and/or 6) after treatment with a first- or third-generation EGFR-TKI, was associated with favorable PFS compared with patients without clearance of plasma EGFRm. Similar outcomes were observed across samples from both trials suggesting that plasma EGFRm clearance is not specific for line of therapy and may be applicable for all patients with EGFRm NSCLC treated with EGFR-TKIs. Frequency of clearance was significantly lower in patients treated with platinum-pemetrexed compared with those treated with osimertinib in AURA3, but clearance predicted improved PFS compared with patients without clearance of plasma EGFRm across both treatment arms. These data suggest that plasma EGFRm clearance may be predictive of outcomes in patients with EGFRm NSCLC regardless of treatment received. Our data are also consistent with other studies showing that on-treatment plasma EGFRm clearance is predictive of survival outcomes with first- or second-line osimertinib or other EGFR-TKIs (17–22).

Notably, in the subgroup of patients without clearance of plasma EGFRm, osimertinib treatment was associated with a PFS benefit compared with comparator EGFR-TKI or platinum-pemetrexed. Therefore, non-clearance of plasma EGFRm does not support stopping of osimertinib therapy but may represent a patient population suitable for treatment intensification to further improve treatment response beyond single-agent EGFR-TKI, such as the addition of chemotherapy, currently under investigation in ongoing clinical trials (23, 24), or the addition of radiotherapy. The lack of ctDNA clearance might also be a useful biomarker for local consolidative therapies (25, 26). It is possible that adapting treatment approaches for patients without clearance at these early time points could be used to improve outcomes. These patients who may be expected to have poor prognosis could be enrolled to trials of intensified therapy, such as the addition of chemotherapy or mechanistically-based combination approaches to determine if clearance could be achieved and potentially improve their treatment outcome. The subgroup of patients without clearance of plasma EGFRm ctDNA at Week 3, but who then have clearance at Week 6 after treatment with a first- or third-generation EGFR-TKI potentially represent an intermediate response group and this also warrants further study.

It is interesting to note that the proportion of patients with plasma EGFRm clearance by Weeks 3 or 6 was similar between osimertinib and comparator EGFR-TKI treatment arms in the FLAURA trial (Week 3, 45% and 42%; Week 6, 56% and 53%). This reflects the similar ORR between treatment arms and is likely indicative of the established therapeutic potential of these compounds (2). Of further interest is that while ORR was similar, clearance of plasma EGFRm in the osimertinib arm was associated with longer PFS versus comparator EGFR-TKI. This may be indicative of the duration and depth of response to osimertinib and the ability to delay emergence of resistance mechanisms. Therefore, early ctDNA dynamics may be a potential pharmacodynamic biomarker that reflects a drug's initial activity but potentially not longer term benefits, for example due to the blocking of resistance mechanisms.

In contrast, a greater proportion of patients treated with osimertinib in AURA3 had plasma EGFRm clearance by Weeks 3 or 6 compared with patients treated with platinum-pemetrexed (Week 3, 42% vs. 18%; Week 6, 44% vs. 26%). These data suggest early on-treatment ctDNA dynamics may predict clinical benefit, as indicated by significantly better ORR with osimertinib versus platinum-pemetrexed in patients with baseline detectable plasma EGFRm (70% vs. 42%; OR 3.2, 95% CI, 1.7–6.0; P = 0.0004). Further, ORR was higher across both treatment arms in patients with plasma EGFRm clearance than in those without plasma EGFRm clearance. This is an important study demonstrating predictive value for ctDNA in targeted therapy and chemotherapy treated patients from two phase III trials and suggests that on-treatment ctDNA dynamics may be a useful tool across treatment modalities.

Previous work has shown that the cobas® EGFR Mutation Test v2 (cobas plasma), ddPCR (Biodesix), and NGS (Guardant360, Guardant Health) are suitable for detecting EGFRm in plasma (27). Our data demonstrate that ddPCR has a similar level of EGFRm detection to NGS and would be a feasible and practical option for serially assessing treatment response particularly in the early course of treatment. Furthermore, a recent consensus statement from the International Association for the Study of Lung Cancer proposed a ‘plasma first’ approach for biomarker evaluation at the time of diagnosis and for monitoring the efficacy of targeted therapies (28). However, newer technologies, and improved versions of the previously mentioned technologies that claim increased sensitivity may be superior at detecting very low frequency mutant alleles. Application of these newer technologies in this setting may identify additional patients with detectable ctDNA and further support the clinical potential for ctDNA analysis to stratify high- and low-risk patient subsets in the future. Additional work is needed to expand the EGFRm strategy used here to an NGS approach more broadly applicable to tumors not known to contain clonal driver mutations.

A strength of our study is the prospectively defined collection of plasma samples at baseline and thereafter at clearly defined time points while on therapy in all treatment arms. However, a limitation of our study is the exploratory nature of the analysis, meaning that results should be interpreted with caution and definitive conclusions cannot be made. In addition, the optimal time point for testing is not known. The data show that if patients cleared at Week 3, it was likely that this clearance persisted through Week 6. Some patients did not clear at Week 3 but did clear at Week 6. However, ongoing work is exploring if earlier clearance, before 3 weeks, could be indicative of a population with an improved prognosis. Furthermore, the results do not take into account the impact of other genetic components, and their temporal on-treatment dynamics, that co-occur with EGFRm, which could be independent factors influencing clinical outcomes (29). For example, concomitant mutations at baseline in TP53, RB1, and PTEN have been shown to result in a shorter PFS in patients with EGFRm tumors (30). The impact of any differences between genetic profiles in patients receiving first-line versus later-line osimertinib (31) and acquired resistance mechanisms is also not considered in these analyses. Further analysis is ongoing to investigate the mechanism underlying the high risk of early progression in patients with detectable plasma EGFRm following EGFR-TKI treatment. In addition, serial ctDNA profiling during the entire course of therapy may further inform early progression and emergence of acquired resistance mechanisms (32).

In conclusion, analysis of plasma EGFRm as early as 3 weeks into EGFR-TKI treatment has potential to be used as a marker to predict clinical outcome in EGFRm positive NSCLC, and thus may allow modification of therapy to optimize treatment outcome; confirmation will be needed from future clinical studies.

J.E. Gray reports personal fees from AbbVie, Axiom HC Strategies, Blueprint Medicines, Celgene, Daiichi Sankyo, EMD Serono - Merck KGaA, Inivata, Janssen Pharmaceuticals, Jazz Pharmaceuticals, Loxo Oncology Inc., OncoCyte Biotechnology, Takeda Pharmaceuticals, and Sanofi Pharmaceuticals; grants and personal fees from AstraZeneca, Genentech, Merck & Co., and Novartis; and grants from Boehringer Ingelheim, G1 Therapeutics, Ludwig Institute of Cancer Research, Pfizer outside the submitted work. M.-J. Ahn reports personal fees from AstraZeneca, Merck, Arcus, Amgen, Yuhan, Alpha Pharmaceuticals, Pfizer, Roche, ONO, Takeda, and Daiichi Sankyo outside the submitted work. G.R. Oxnard reports personal fees from Foundation Medicine and Roche outside the submitted work. F.A. Shepherd reports other support from AstraZeneca during the conduct of the study, as well as personal fees from AstraZeneca outside the submitted work. F. Imamura reports grants from AstraZeneca, Chugai, MSD, Amgen, Eisai, Sanofi, Johnson & Johnson, Merus, Ono, Taiho, Daiichi Sankyo, Astellas, Merck, AbbVie, and Boehringer Ingelheim during the conduct of the study, as well as grants from Takeda outside the submitted work. I. Okamoto reports grants and personal fees from AstraZeneca during the conduct of the study. I. Okamoto also reports grants and personal fees from Chugai Pharmaceutical Co. Ltd., Taiho Pharmaceutical Co. Ltd., and Ono Pharmaceutical Co. Ltd.; grants from Takeda Pharmaceutical Company Ltd., Eli Lilly Japan K.K., and Novartis Pharma K.K.; and personal fees from Nippon Boehringer Ingelheim Co. Ltd. outside the submitted work. B.C. Cho reports grants from MOGAM Institute, LG Chem, Oscotec, Interpark Bio Convergence Corp, GI Innovation, GI-Cell, Abion, AbbVie, AstraZeneca, Bayer, Blueprint Medicines, Boehringer Ingelheim, Champions Onoclogy, CJ Bioscience, CJ Blossom Park, Cyrus, Dizal Pharma, Genexine, Janssen, Lilly, MSD, Novartis, Nuvalent, Oncternal, Ono, Regeneron, Dong-A ST, Bridgebio Therapeutics, Yuhan, ImmuneOncia, Illumina, Kanaph Therapeutics, Therapex, JINTSbio, and Hanmi, as well as other support from Abion, BeiGene, Novartis, Boehringer Ingelheim, Roche, BMS, CJ, CureLogen, Ono, Onegene Biotechnology, Yuhan, Pfizer, Eli Lilly, GI-Cell, Guardant, HK Inno-N, Imnewrun Biosciences Inc., Janssen, Takeda, MSD, Medpacto, Blueprint Medicines, RandBio, Hanmi, Yonsei University Health System, Kanaph Therapeutic Inc., Bridgebio Therapeutics, Cyrus Therapeutics, Oscotec Inc., ASCO, AstraZeneca, ESMO, IASLC, Korean Cancer Association, Korean Society of Medical Oncology, Korean Society of Thyroid-Head and Neck Surgery, Korean Cancer Study Group, The Chinese Thoracic Oncology Society, TheraCanVac Inc., Gencurix Inc., DAAN Biotherapeutics, Interpark Bio Convergence Corp., J INTS BIO, Champions Oncology, Crown Bioscience, and Imagen outside the submitted work. Y.-L. Wu reports personal fees from AstraZeneca, BeiGene, Boehringer Ingelheim, Bristol Myers Squibb, Eli Lilly and Company, Merck Sharp & Dohme, Pfizer, Roche, and Sanofi, as well as grants from AstraZeneca, Boehringer Ingelheim, Bristol Myers Squibb, Hengrui, and Roche outside the submitted work. M. Majem reports grants and personal fees from Roche and AstraZeneca; grants from BMS; and personal fees from MSD, Takeda, Pfizer, and Novartis outside the submitted work. M. Boyer reports grants from AstraZeneca during the conduct of the study, as well as grants from Genentech/Roche, Boehringer Ingelheim, and Novartis outside the submitted work. K.C. Bulusu reports to be a full-time employee of AstraZeneca and owns company stock. A. Markovets reports other support from AstraZeneca outside the submitted work. J.C. Barrett reports other support from AstraZeneca during the conduct of the study. R. Hodge reports other support from AstraZeneca during the conduct of the study, as well as other support from AstraZeneca outside the submitted work. A. McKeown reports other support from AstraZeneca during the conduct of the study. R.J. Hartmaier reports personal fees from AstraZeneca during the conduct of the study; in addition, R.J. Hartmaier has a patent for US11066709B2 issued to Genentech Inc. and Foundation Medicine Inc. J. Chmielecki reports employee of and shareholder in AstraZeneca. V.A. Papadimitrakopoulou reports employee and stockholder of Pfizer, Inc. S.S. Ramalingam reports grants and personal fees from AstraZeneca during the conduct of the study. No disclosures were reported by the other authors.

J.E. Gray: Conceptualization, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. M.-J. Ahn: Writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. G.R. Oxnard: Conceptualization, data curation, formal analysis, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. F.A. Shepherd: Conceptualization, data curation, formal analysis, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. F. Imamura: Conceptualization, data curation, formal analysis, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. Y. Cheng: Data curation, formal analysis, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. I. Okamoto: Data curation, formal analysis, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. B.C. Cho: Conceptualization, supervision, investigation, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. M.-C. Lin: Data curation, formal analysis, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. Y.-L. Wu: Conceptualization, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. M. Majem: Data curation, formal analysis, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. O. Gautschi: Data curation, formal analysis, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. M. Boyer: Data curation, formal analysis, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. K.C. Bulusu: Data curation, formal analysis, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. A. Markovets: Data curation, formal analysis, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. J.C. Barrett: Conceptualization, data curation, formal analysis, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. R. Hodge: Data curation, formal analysis, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. A. McKeown: Data curation, formal analysis, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. R.J. Hartmaier: Conceptualization, data curation, formal analysis, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. J. Chmielecki: Conceptualization, data curation, formal analysis, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. V.A. Papadimitrakopoulou: Data curation, formal analysis, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work. S.S. Ramalingam: Conceptualization, writing–original draft, writing–review and editing, final approval of the manuscript, accountable for all aspects of the work.

The authors would like to thank all the patients involved in the FLAURA and AURA3 studies and their families, the study investigators, and the team at AstraZeneca. The studies (NCT02296125, NCT02151981) were funded by AstraZeneca, the manufacturer of osimertinib. The authors also acknowledge Ken Thress, for generating proof of concept data around EGFRm ctDNA clearance dynamics.

The authors would like to acknowledge Bernadette Tynan, MSc, and Alexandra Webster, MSc, of Ashfield MedComms, an Inizio company, for medical writing support that was funded by AstraZeneca in accordance with Good Publication Practice guidelines (https://www.ismpp.org/gpp-2022).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.