Clinical factors and molecular co-alterations impact outcomes in patients receiving first-line osimertinib for EGFR-mutated non-small cell lung cancer

Abstract

Purpose:

Until recently, osimertinib was the preferred first-line therapy for nearly all patients with EGFR-mutated advanced NSCLC. However, with approval of the FLAURA2 and MARIPOSA regimens, identifying molecular or clinical factors that confer higher risk for inferior outcomes to first-line osimertinib may better inform upfront treatment selection for newly-diagnosed patients.

Methods:

This real-world, multicenter retrospective study across 4 U.S. academic centers included adult patients with metastatic NSCLC and classic sensitizing EGFR mutations (exon 19 deletion, L858R mutation) receiving first-line osimertinib monotherapy who had tissue-based next-generation sequencing performed at diagnosis. Survival analyses for time-to-treatment discontinuation (TTD) on osimertinib and overall survival (OS) were conducted by number of co-alterations, individual gene alterations, previously defined molecular pathways, and clinical factors.

Results:

A total of 219 patients were included (median age 68, 62 % female, 51 % with baseline CNS metastases). TP53 was most frequently co-altered (58 %), followed by EGFR amplification (13 %), CDKN2A (10 %), RB1 (9 %), and PIK3CA (8 %). Number of de novo co-alterations and presence of co-alterations by molecular pathway did not significantly predict TTD or OS. Co-alteration of both TP53 and PIK3CA, but not TP53 alone, demonstrated significantly shorter median OS compared to the TP53 and PIK3CA wild-type population (21.9 vs 39.5 months, p = 0.046), with similar numerical trend for median TTD. On multivariate analysis, L858R mutation and CNS and bone metastases remained significant predictors for inferior clinical outcomes.

Conclusion:

Co-alteration of TP53 and PIK3CA may represent a high-risk molecular subgroup of EGFR-mutated NSCLC, however both clinical and molecular features should be considered in risk stratification.

1. Introduction

The pooled prevalence of epidermal growth factor receptor (EGFR) mutations in non-small cell lung cancer (NSCLC) has been estimated at approximately 20–30 %, with variation by geography and racial origin [1,2]. Osimertinib is a third-generation, irreversible oral EGFR tyrosine kinase inhibitor (TKI) that selectively inhibits classic sensitizing EGFR mutations (exon 19 deletion and exon 21 L858R) as well as the EGFR T790M resistance mutation [3-5]. Since 2018 it has been the standard-of-care first-line treatment for patients with advanced or metastatic EGFR-mutated (EGFRm) NSCLC based on the phase 3 FLAURA trial demonstrating improved progression-free survival (PFS) and overall survival (OS) compared with first-generation EGFR TKIs [6-8]. Despite the favorable outcomes and tolerability of osimertinib, there remain a subset of patients who derive less benefit, and real-world outcomes are inferior to those of the clinical trial setting [9]. Furthermore, as with other targeted therapies, all patients eventually develop resistance and disease progression.

As a result, there has been increasing interest in intensifying first-line systemic therapy for metastatic EGFRm NSCLC beyond osimertinib monotherapy. The phase 3 FLAURA2 trial demonstrated superior PFS for osimertinib combined with platinum-pemetrexed chemotherapy compared to osimertinib alone, particularly in patients with L858R mutations or central nervous system (CNS) metastases [10,11]. Analogously, the phase 3 MARIPOSA trial demonstrated superior PFS for the EGFR-MET bispecific antibody amivantamab combined with the third-generation EGFR TKI lazertinib compared to osimertinib or lazertinib alone [12]. In both FLAURA2 and MARIPOSA, the incidence of grade 3 or higher adverse events was approximately two times higher in the combination arm when compared to EGFR TKI monotherapy. While MARIPOSA has demonstrated a statistically significant OS benefit for escalated combination therapy compared to osimertinib, and a press release ahead of presented data suggests FLAURA2 will do the same, neither study addresses whether upfront combination is superior to sequential treatment in this setting [10,12-15]. As a result, to balance benefits and risks when choosing between these different treatment options, understanding the heterogeneity of EGFRm NSCLC is increasingly important.

The role of TP53 co-alterations in EGFRm NSCLC has been studied extensively, albeit with significant heterogeneity and predominantly in cohorts of patients treated with earlier-generation EGFR TKI [16-22]. Many clinical trials include subgroup analyses evaluating outcomes by the presence or absence of TP53 co-alterations, and most demonstrate TP53 to be a prognostic rather than predictive biomarker. A potential exception is the MARIPOSA study, where a secondary analysis reported a median PFS of 18.2 months for amivantamab plus lazertinib compared to 12.9 months for osimertinib in patients with TP53 co-alterations (HR 0.65, p = 0.003). There was no statistically significant difference for the TP53 wild-type population (HR 0.75, p = 0.114) [23]. In contrast, there was no significant difference between intensified treatment compared to osimertinib by baseline TP53 co-alteration in the FLAURA2 study [24]. Overall, with TP53 co-alterations occurring in approximately 50 % of EGFRm NSCLC [9], there remains a need to better identify the subgroup of patients likely to have inferior outcomes on first-line osimertinib monotherapy, enabling more tailored selection for upfront treatment intensification.

In this study, we retrospectively analyzed the real-world frequency and distribution of de novo co-alterations occurring in patients with metastatic EGFRm NSCLC treated with first-line osimertinib monotherapy. We then investigated the predictors of OS and time-to-treatment discontinuation (TTD) based on clinical factors, number of co-alterations, individual gene alterations, and molecular pathways.

2. Material and methods

2.1. Patients and data collection

Four academic medical centers (Stanford University, Massachusetts General Hospital, University of California, Davis [UC Davis], University of California, San Diego [UCSD]) participated in this retrospective multicenter cohort study. According to distinct institutional review board (IRB)-approved protocols at each center, patient data were collected via chart review by investigators at each site. Inclusion criteria included any adult patient (age ≥ 18) with metastatic (Stage IV) NSCLC harboring a classic sensitizing EGFR mutation (exon 19 deletion or exon 21 L858R) receiving osimertinib 80 mg daily monotherapy as first-line systemic therapy. All patients were required to have tissue-based next-generation sequencing (NGS) containing the 58 genes of interest (Supplemental Table 1) performed at the diagnosis of metastatic disease and before the initiation of osimertinib. NGS platforms included in this study were Stanford STAMP, MGH SNaPshot, FoundationOne^®, Tempus xT, StrataNGS^™, and UCSF500 [25-30]. Exclusion criteria included receipt of other systemic therapies in combination with osimertinib, previous receipt of osimertinib in the neoadjuvant or adjuvant setting for early-stage disease, or insufficient follow-up (e.g., lack of imaging or clinical follow-up after starting osimertinib, discontinuation of osimertinib for toxicity prior to first response assessment).

The following baseline variables were retrospectively collected: age, sex, race, ethnicity, smoking history, tumor histology, programmed death-ligand 1 (PD-L1) tumor proportion score (TPS), stage at initial diagnosis, and prior systemic therapies for early-stage disease. Additional variables at the time of osimertinib initiation were also collected, including performance status (Eastern Cooperative Oncology Group [ECOG]) and sites of metastatic disease (extrathoracic, CNS, liver, and bone). Reason for treatment discontinuation and types of subsequent therapy were collected at the time of osimertinib discontinuation. Dates of metastatic disease diagnosis, osimertinib initiation, osimertinib discontinuation, most recent follow-up, and death, if applicable, were collected for TTD and OS analyses. The presence or absence of pathogenic or likely pathogenic alterations (as defined by each individual sequencing platform, including mutations, rearrangements, or amplifications, if applicable) in the 58 genes shared between all platforms included was collected for all patients at baseline. With the exception of EGFR, individual genes were classified into one of nine molecular pathways of interest based on biologic mechanism and previously published work by Blakely, et al [31]. A subset of patients had additional tissue NGS results available from the time of disease progression on osimertinib.

2.2. Outcome variables

The primary outcome variables assessed were TTD and OS. Time measurements were calculated in months beginning from the initiation date of osimertinib. Treatment discontinuation was defined as discontinuation of osimertinib monotherapy for any reason, including progression of disease, treatment toxicity, or death. Patients receiving local therapies such as radiation for oligoprogression could continue osimertinib post-progression and were not defined to have treatment discontinuation until they discontinued osimertinib monotherapy as the systemic therapy. Dose modification of osimertinib during treatment was allowed so long as a patient remained on osimertinib as monotherapy. Magnetic resonance imaging (MRI) of the brain, computed tomography (CT) of the chest, abdomen, and pelvis, and positron emission tomography-CT (PET-CT) of the body were monitored on a schedule defined by individual institutions and providers.

2.3. Statistical considerations

Baseline patient characteristics were summarized via descriptive statistics. Chi-square proportion tests were used to compare frequency of co-alterations between patients with EGFR exon 19 deletion versus L858R mutation. Kaplan-Meier analysis was used to estimate time-to-event outcomes, namely TTD and OS. Cox proportional hazards regression was used to assess the statistical significance of associations between clinical or molecular factors and time-to-event outcomes. For subgroup analyses by pathway or individual gene alteration, analysis was limited to groups with at least 10 events in each arm of interest. A multivariate model (incorporating primary EGFR mutation, smoking history, sites of metastatic disease, number of co-alterations, and molecular alterations by pathway) was derived using backwards selection, starting with a model including all variables with p < 0.2 in univariate analysis, and removing non-significant variables until only significant variables remained. A paired sample t-test was used to compare number of co-alterations between pre-treatment and post-progression time points. A prespecified p-value threshold of < 0.05 was used throughout the analysis. Statistical analyses were conducted with R Version 4.3.2 [32]. The data cutoff date was July 31, 2024.

3. Results

3.1. Patient characteristics

A total of 219 patients with metastatic NSCLC and classic sensitizing EGFR mutations treated with first-line osimertinib monotherapy between September 2017 and July 2024 were identified. These patients were distributed across the four treatment centers as follows: Stanford University (n = 108), Massachusetts General Hospital (n = 42), UC Davis (n = 39), and UCSD (n = 30). Median follow-up time was 27.6 months (range: 1.1–77.0 months, interquartile range: 15.5–41.3 months) from the start of osimertinib monotherapy.

Baseline demographic and clinicopathologic characteristics of the overall cohort are summarized in Table 1. Median age at diagnosis of metastatic disease was 68 years (range: 33–89 years). Consistent with known demographic patterns for EGFRm NSCLC, most patients were female (62 %), identified as Asian race (49 %), and had never smoked (70 %). Most patients had an ECOG performance status of 0 or 1 at the time of osimertinib initiation (81 %) and had been diagnosed with de novo Stage IV disease (85 %). The primary sensitizing EGFR mutation included exon 19 deletion in 55 %, L858R in 45 %, and concurrent EGFR exon 19 deletion and L858R alterations in the tumor of a single patient. Most patients had extrathoracic metastatic disease at baseline (73 %), particularly bone metastases (52 %), CNS metastases (51 %), and liver metastases (13 %).

Table 1.

Baseline demographic and clinicopathologic characteristics of overall cohort.

Characteristic (n = 219 unless otherwise specified)	Count (Percentage)
Age at diagnosis of metastatic disease, median (range)	68 (33–89) years
Sex
Female	135 (62 %)
Male	84 (38 %)
Race (n = 213)
Asian	105 (49 %)
White	81 (38 %)
Othera	27 (13 %)
Ethnicity (n = 215)
Hispanic or Latino	14 (7 %)
Not Hispanic or Latino	201 (93 %)
Smoking status
Never	154 (70 %)
Current or former (≥100 cigarettes/lifetime)	65 (30 %)
ECOG performance status (n = 218)
0–1	176 (81 %)
2–3	42 (19 %)
Histology
Adenocarcinoma	213 (97 %)
Squamous cell carcinoma	4 (2 %)
Adenosquamous carcinoma	1 (0.5 %)
Poorly differentiated carcinoma	1 (0.5 %)
Primary EGFR mutation
Exon 19 deletion	120 (55 %)
Exon 21 L858R	98 (45 %)
Both exon 19 deletion and exon 21 L858R	1 (0.5 %)
PD-L1 TPS (n = 191)
Negative (<1%)	82 (43 %)
Low (1–49 %)	75 (39 %)
High (≥50 %)	34 (18 %)
TMB (n = 110)
<10 mut/Mb	104 (95 %)
≥10 mut/Mb	6 (5 %)
Stage at initial diagnosis (AJCC 8th edition)
I-III	32 (15 %)
IV	187 (85 %)
Location(s) of metastatic disease
Extrathoracic	160 (73 %)
Bone	114 (52 %)
CNS	111 (51 %)
Liver	29 (13 %)
Tissue NGS platform at metastatic diagnosis
Stanford STAMP	108 (49 %)
MGH SNaPshot	41 (19 %)
Foundation One	40 (18 %)
Tempus	27 (12 %)
STRATA	2 (1 %)
UCSF500	1 (0.5 %)

AJCC, American Joint Committee on Cancer; ECOG, Eastern Cooperative Oncology Group; EGFR, epidermal growth factor receptor; NGS, next-generation sequencing; TMB, tumor mutational burden; TPS, tumor proportion score.

^aIncludes Native Hawaiian or Other Pacific Islander (n = 7), Black or African American (n = 4), American Indian or Alaska Native (n = 1), and Other (n = 15) per U.S. Census Bureau categories.

^bAdjuvant afatinib for 3 months in both patients, as part of a clinical trial protocol.

3.2. Frequency and distribution of de novo co-alterations

In the overall cohort, 20 % of tumors did not have any de novo co-alterations by tissue-based NGS at the time of metastatic disease diagnosis. Most tumors had one (40 %) or two (25 %) co-alterations, while 15 % had three or more co-alterations (range 3–7) (Fig. 1A). Alterations in the P53 pathway were most frequent (61 %), followed by cell cycle (22 %), PI3K/AKT/mTOR (13 %), and WNT/beta-catenin (9 %) (Fig. 1B).

Fig. 1.

(A) Frequency and distribution of de novo co-alterations by (B) molecular pathway and (C) individual gene.

Analogously, TP53 was the most common individual gene altered (58 %), followed by EGFR amplification (13 %), CDKN2A (10 %), RB1 (9 %), PIK3CA (8 %), CTNNB1 (7 %), and EGFR mutations other than exon 19 deletion or L858R (6 %) (Fig. 1C). STK11 (2 patients, 0.9 %) and KEAP1 (0 %) alterations were uncommon in this EGFRm patient population. A full breakdown of the individual gene co-alteration frequency by molecular pathway is summarized in Supplemental Table 2.

A second, non-classic, EGFR mutation was more common among tumors with EGFR L858R compared to exon 19 deletion (12.2 versus 1.7 %, p = 0.002). For tumors with EGFR exon 19 deletion, other EGFR mutations included 1 case with T790M and 1 case with V834L. For those with EGFR L858R, other EGFR mutations included 3 cases with T790M, 2 cases with V834L, 1 case with V834L and L718Q, and 1 case each with G729A, A289T, A289V, L718Q, E709A, T725M, and L62R. None of the other pathway co-alteration frequencies were significantly different between EGFR L858R and exon 19 deletion (Table 2).

Table 2.

Frequency of pathway co-alterations by primary sensitizing EGFR mutation.

Pathwaya	Exon 19 deletion (n = 120) Count (%)	Exon 21 L858R (n = 98) Count (%)	p-value (χ2)
P53 pathway	73 (60.8 %)	60 (61.2 %)	0.95
Cell cycle	29 (24.2 %)	18 (18.4 %)	0.30
PI3K / AKT / mTOR	16 (13.3 %)	13 (13.3 %)	0.99
EGFR amplification	16 (13.3 %)	12 (12.2 %)	0.81
WNT / β-catenin	7 (5.8 %)	13 (13.3 %)	0.06
Other EGFR mutationb	2 (1.7 %)	12 (12.2 %)	0.002 *
Epigenetic modification / transcription factor	3 (2.5 %)	7 (7.1 %)	0.10
DNA repair	4 (3.3 %)	3 (3.1 %)	0.91
RAS / RAF / MAPK	3 (2.5 %)	3 (3.1 %)	0.80
RTK	3 (2.5 %)	2 (2.0 %)	0.82
Ubiquitin-proteasome system	2 (1.7 %)	0 (0 %)	NE

^*Significant at p < 0.05 threshold.

EGFR, epidermal growth factor receptor; NE, not evaluable; NGS, next-generation sequencing.

^aPathogenic or likely pathogenic alterations were defined by each individual sequencing platform (including mutations, rearrangements, or amplifications, if applicable), among the 58 genes shared between all included NGS platforms. With the exception of EGFR, individual genes were classified into one of nine molecular pathways of interest based on biologic mechanism and previously published work by Blakely, et al.²⁸.

^bFor exon 19 deletion, other EGFR mutations included 1 case with T790M and 1 case with V834L. For exon 21 L858R, other EGFR mutations included 3 cases with T790M, 2 cases with V834L, 1 case with V834L and L718Q, and 1 case each with G729A, A289T, A289V, L718Q, E709A, T725M, and L62R.

3.3. Time to treatment discontinuation and overall survival outcomes

In univariate, unadjusted survival analyses for the overall cohort, median OS was 39.5 months (95 % CI: 32.9–48.6 months) and median TTD was 19.6 months (95 % CI: 15.8–25.4 months). EGFR exon 19 deletion was associated with improved median OS and TTD compared to L858R mutation (median OS [mOS]: 41.2 versus 30.9 months, p = 0.005; median TTD [mTTD]: 25.8 versus 14.2 months, p = 0.002) (Fig. 2A). The presence of baseline CNS, liver, and bone metastases were all significantly associated with inferior OS and TTD compared to the absence of these factors, respectively (Fig. 2B-D).

Fig. 2.

Overall survival (OS) and time-to-treatment discontinuation (TTD) by (A) primary EGFR mutation, (B) CNS metastases at diagnosis, (C) liver metastases at diagnosis, and (D) bone metastases at diagnosis.

There were no significant differences in outcomes between patients with one or more baseline co-alterations compared to those with no baseline co-alterations (mOS: 37.6 months versus not reached, p = 0.17; mTTD: 18.8 versus 23.1 months, p = 0.54) (Fig. 3A). An alternative cutoff threshold of 2 or more baseline co-alterations (compared to 0 or 1) also yielded no notable outcome differences. The presence of a P53 pathway co-alteration demonstrated a numerical trend towards inferior OS and TTD, though these findings were not statistically significant (mOS: 34.0 versus 41.6 months, p = 0.14; mTTD: 16.5 versus 25.8 months, p = 0.074) (Fig. 3B). Overall, none of the molecular pathways of interest was a significant predictor of the primary OS or TTD outcomes (Supplemental Table 3).

Fig. 3.

Overall survival (OS) and time-to-treatment discontinuation (TTD) by (A) number of co-alterations, (B) P53 pathway co-alteration, and (C) TP53 and PIK3CA co-alterations.

By multivariate analysis incorporating primary EGFR mutation, smoking history, sites of metastatic disease, number of co-alterations, and molecular alterations by pathway, primary EGFR mutation (exon 19 deletion versus L858R) (HR 0.56, 95 % CI: 0.38–08.1, p = 0.0024) and presence of bone metastases (HR 1.80, 95 % CI: 1.23–2.63, p = 0.0026) were significant predictors of OS. Primary EGFR mutation (exon 19 deletion versus L858R) (HR 0.61, 95 % CI: 0.44–0.84, p = 0.0028) and presence of CNS metastases were significant predictors of TTD. Neither number of co-alterations (≥1 versus none) nor co-alterations by molecular pathway were significant predictors of OS or TTD on multivariate analysis (Supplemental Table 4). Smoking history was not associated with adverse OS or TTD on either univariate or multivariate analysis.

As TP53 co-alterations were observed at a high frequency in this population, the effect of secondary co-alterations in combination with TP53 was analyzed. On univariate analysis, co-alteration of TP53 and PIK3CA (TP53mut / PIK3CAmut) was associated with significantly inferior OS and numerically inferior TTD compared to the reference wild-type population (TP53wt / PIK3CAwt) (mOS: 21.9 versus 39.5 months, p = 0.046; mTTD: 13.7 versus 25.8 months, p = 0.071). In contrast, there were no statistically significant differences when TP53 but not PIK3CA was co-altered (TP53mut / PIK3CAwt), relative to the TP53 and PIK3CA wild-type population (TP53wt / PIK3CAwt) (mOS: 39.8 versus 39.5 months, p = 0.26; mTTD: 16.9 versus 25.8 months, p = 0.11) (Fig. 3C). Prevalence of bone (5/10), liver (2/10), and CNS (5/10) metastases in the TP53mut / PIK3CAmut group was comparable to that of the overall population (Table 1). PIK3CA alteration alone was not associated with either OS or TTD. As with PI3K/AKT/mTOR pathway alterations more generally, this difference was not significant on multivariate analysis, albeit likely underpowered due to limited sample size. There were no significant differences in OS or TTD for the other TP53 co-alteration pairs (TP53 co-altered with EGFR amplification, CDKN2A, or RB1 = mut / mut), when compared to their respective wild-type reference groups (wt / wt) (Supplemental Table 5).

3.4. Osimertinib discontinuation and subsequent management

Among the 149 patients who discontinued osimertinib monotherapy, most did so for disease progression (67 %), followed by reasons of death (15 %) and toxicity (12 %). One hundred and ten patients (74 %) received subsequent systemic therapy after osimertinib monotherapy. Osimertinib was continued with second-line systemic therapy in 56 % of cases. Platinum-doublet chemotherapy was most commonly included as part of subsequent therapy (61 %), followed by bevacizumab (17 %) and other EGFR TKIs (11 %, most frequently in patients who had stopped osimertinib due to toxicity) (Table 3).

Table 3.

Osimertinib discontinuation and subsequent clinical management.

Characteristic	Count (Percentage)
Reason for discontinuation of first-line osimertiniba (n = 149)
Disease progression	100 (67 %)
Death	22 (15 %)
Toxicity	18 (12 %)
Patient preference	3 (2 %)
Other	6 (4 %)
Repeat tissue biopsy performed at progression (n = 151)	66 (44 %)
Repeat tissue NGS performed at progression (n = 151)	46 (30 %)
Tissue NGS platform at progression (n = 46)
Stanford STAMP	23 (50 %)
MGH SNaPshot	14 (30 %)
Tempus	5 (11 %)
Foundation One	3 (7 %)
Caris	1 (2 %)
Received second-line systemic therapy after osimertinib (n = 149)	110 (74 %)
Components of subsequent second-line systemic therapy (n =110)
Osimertinib (continuation)b	61 (56 %)
Other EGFR TKIc	12 (11 %)
Platinum-doublet chemotherapy	67 (61 %)
Single-agent chemotherapy	5 (5 %)
Bevacizumab	19 (17 %)
Checkpoint inhibitor immunotherapy	7 (6 %)
Amivantamab	2 (2 %)
Otherd	16 (15 %)

ADC, antibody-drug conjugate; CNS, central nervous system; EGFR, epidermal growth factor receptor; HER3, human epidermal growth factor receptor 3; NGS, next-generation sequencing; TIL, tumor-infiltrating lymphocyte; TKI, tyrosine kinase inhibitor.

^aPatients receiving local therapies (e.g., radiation) for oligoprogression could continue osimertinib post-progression and were not defined to have treatment discontinuation until they discontinued osimertinib monotherapy as the systemic therapy.

^bOsimertinib was continued with other systemic therapies.

^cIncludes erlotinib (n = 5) and gefitinib (n = 4), as well as 1 patient each receiving lazertinib, furmonertinib, and BLU-945 (all as part of clinical trials).

^dIncludes capmatinib (n = 1), crizotinib (n = 1), alectinib (n = 1), and trastuzumab deruxtecan (n = 1) as part of routine care; also includes necitumumab (n = 5), savolitinib (n = 3), HER3 ADC (n = 2), selumetinib (n = 1), and TIL therapy (n = 1) as part of clinical trials.

Repeat tissue sampling was performed in 66 patients (44 %) at the time of radiographic progression, of which 46 patients (30 %) had tissue NGS sent on the biopsy (Table 3). Results of this repeat tissue NGS were available for 45 patients. Among this subgroup, there was no significant difference in the number of co-alterations between the pre-treatment and post-progression time points (p = 0.76). At progression, alterations in the P53 pathway were most frequent (78 %), followed by cell cycle (29 %), and RAS/RAF/MAPK (13 %). TP53 was the most common individual gene altered (69 %), followed by RB1 (13 %), CCNE1 (9 %), PIK3CA (7 %), MDM2 (7 %), and CDKN2A (7 %). Frequency of alterations in the RAS/RAF/MAPK (2 % at diagnosis, 11 % at progression) and RTK pathways (0 % at diagnosis, 9 % at progression) increased between diagnosis and progression, whereas frequency of alterations in the PI3K/AKT/mTOR pathway decreased between diagnosis and progression (20 % at diagnosis, 9 % at progression). A full breakdown of the individual gene co-alteration frequency by molecular pathway at the time of disease progression is summarized in Supplemental Table 6. There was only a single case of small cell transformation in this cohort, representing 1 % of patients who discontinued osimertinib for disease progression.

4. Discussion

Our study represents a novel analysis of the impact of de novo (baseline) genetic co-alterations in relation to real-world clinical outcomes of patients with metastatic EGFRm NSCLC treated with osimertinib monotherapy in the first-line setting. In this study, we showed that (1) neither number of co-alterations nor presence of co-alterations in any individual molecular pathway significantly predicted TTD or OS; (2) co-alteration of TP53 and PIK3CA, but not TP53 alone, was a poor molecular prognostic factor warranting further validation; and (3) primary EGFR mutation and clinical factors, namely sites of metastatic involvement, remain important prognostic considerations that should not be overlooked in the era of molecularly-targeted therapy. In the current state, first-line treatment selection in metastatic EGFRm NSCLC remains a complex discussion between patients, families, and providers. Aside from clinical contraindications, as of yet there are no prospectively defined features to guide treatment selection, particularly with respect to which patients may be best suited for osimertinib monotherapy over the combinations (MARIPOSA and FLAURA2), both of which have reported aggregate survival benefit [14,15]. Together, both molecular and clinical biomarkers have the potential to enable more complete prognostic assessments and treatment decisions in the front-line setting.

The distribution of de novo co-alterations in this study was generally consistent with previously reported data, with alterations in TP53, RB1, PIK3CA, and CTNNB1 identified most commonly [16,17,19,33]. Alterations in CDKN2A were observed in 10 % of this patient cohort, higher than the 2–6 % frequency reported in previous studies evaluating the impact of tumor suppressor gene alterations on TKI response in EGFRm NSCLC receiving mainly earlier-generation EGFR TKI [34,35]. Skakodub, et al have demonstrated a significant difference in the frequency of CDKN2A/B alterations in biopsies of brain metastasis compared to those of the primary tumor, albeit in a mixed population of patients with and without driver mutations and prior systemic treatment [36]. Information on site of biopsy was not available in our cohort, and in general is not consistently reported across similar studies. We observed compound EGFR mutations more frequently in patients with L858R mutation compared to exon 19 deletion, which has been reported elsewhere as well [37,38]. Given this was a treatment-naïve patient cohort, the incidence of de novo T790M mutation was very low. Furthermore, all patients received osimertinib, which has known selectivity against T790M resistance mutations [3]. The presence of other EGFR mutations did not independently predict TTD or OS in this study.

Despite the early prevailing view of EGFRm NSCLC as a singleoncogene disease, existing work has highlighted the prevalence and importance of co-occurring genetic alterations both de novo and after exposure to targeted therapy [31,39]. There is precedent for considering molecular pathways such as cell cycle gene alterations as predictors of differential osimertinib response. Blakely, et al demonstrated that cell cycle alterations detected by pre-treatment cell-free DNA (cfDNA) analysis predicted primary osimertinib resistance in a cohort of 41 patients, 80 % of whom had concurrent EGFR T790M mutations, indicating high frequency of pre-treatment with earlier-generation EGFR TKI [31,40]. Stockhammer, et al reported that patients with co-occurring TP53 and tumor suppressor gene alterations had inferior outcomes on EGFR TKI, with RB1 (a key regulator of the G1 to S phase transition) included among the five tumor suppressor genes assessed. Other groups have identified TP53 and RB1 co-alteration as a risk factor for histologic transformation in patients with NSCLC receiving EGFR TKI, though many of these cohorts were comprised primarily of patients receiving first-line erlotinib [41-43]. Notably in our study, we did not identify a significant association between cell cycle pathway alterations and inferior clinical outcomes, despite observing these events in 47 patients (approximately one-fifth of the cohort). Importantly and in contrast to most of the previously referenced works, our study was restricted to patients receiving osimertinib in the first-line setting. Whether this difference is related to the lack of exposure to earlier-generation EGFR TKI is not clear.

While we found no association between co-mutations in particular molecular pathways and outcomes, we did identify co-alteration of TP53 and PIK3CA as a particularly high-risk molecular subset associated with decreased OS and non-significant trend towards decreased TTD on osimertinib on univariate analysis. No significant differences in OS or TTD were identified for the TP53 alteration alone, though again, TTD on osimertinib numerically favored the subset without TP53 alterations. Several prior cohorts have identified TP53 alterations as a negative prognostic factor in EGFRm NSCLC, though these data are largely limited to mixed cohorts treated with earlier-generation EGFR TKIs [16-22]. Importantly, there are also several studies which have not identified TP53 as a negative prognostic marker in EGFRm NSCLC, thus its application as a biomarker remains an area of active investigation [44,45]. Mechanistically, the p53 protein is a tumor suppressor gene and transcription factor regulating several critical cellular responses to stress, among these cell-cycle arrest, apoptosis, and genomic stability [46,47]. Recent work by Vokes, et al demonstrated that tumors with TP53 alterations had higher TMB at baseline, which became more pronounced following EGFR TKI therapy. These authors conclude that TP53 mutations may promote resistance evolution among patients receiving targeted therapy for EGFRm NSCLC, even among patients with initially favorable response to therapy [22].

The PI3K/AKT/mTOR pathway is an important intracellular signaling pathway implicated in numerous cancer types [48]. The frequency of PIK3CA alterations in EGFRm NSCLC has been estimated to be between 3 and 10 % [17,49,50]. Available evidence on the impact of PIK3CA co-alteration on TKI efficacy in EGFRm NSCLC is variable, limited by small sample size and inconsistent results [51-53]. Much of this work has been in the setting of acquired resistance to EGFR TKI, rather than in the de novo treatment-naïve setting [49,54]. Our findings suggest that co-alteration of PIK3CA and TP53 may contribute to worse outcomes on osimertinib, and that the combination of these two co-alterations is more important than either one alone. A retrospective series of 89 patients with advanced NSCLC receiving first-line chemotherapy in China described the negative prognostic impact of PIK3CA and TP53 co-alteration, with approximately one-third of the cases having an EGFR mutation [55]. Similar findings have also been described in the breast cancer literature [56,57]. Ultimately, further work is needed to corroborate this finding in larger cohorts of patients with EGFRm NSCLC.

Many studies investigate clinical and molecular prognostic factors independently, making it a challenge in clinical practice to weigh these factors for any individual patient. A recent secondary analysis from the MARIPOSA trial of amivantamab plus lazertinib compared to osimertinib identified both clinical (baseline brain or liver metastases) and molecular (TP53 co-mutations, lack of ctDNA clearance) high-risk features, with the observation that 89 % of patients had at least one such feature [33]. In a real-world analysis of 1323 patients treated with first-line osimertinib, 95 % were identified as having a “high risk” factor (defined as age ≥ 65, brain or liver metastasis, TP53 mutation, EGFR L858R, or ECOG ≥ 2) [9]. Yet designating nearly the entire population of interest as high-risk has limited utility when trying to select between first-line osimertinib and intensified systemic therapy. Understanding the relative contributions of each risk factor, both clinical and molecular, are important. In this study, we showed that primary EGFR mutation and sites of baseline metastatic disease (CNS metastases for TTD, bone metastases for OS) were significant predictors of OS and TTD on first-line osimertinib, while additional features such as number of co-alterations or presence of co-alterations by molecular pathway were no longer predictive when controlling for clinical contributors.

There are several limitations of this study, namely the retrospective cohort design limiting our ability to interrogate and control for sources of potential bias. While the inclusion of multiple academic institutions with varied catchment areas contributes to diversity of the patient sample, there was consequently a variety of NGS panels utilized, each of which individually classify pathogenic or likely pathogenic alterations or define cutoffs for gene amplification. We accounted for variability in panel breadth by restricting to 58 genes shared amongst all included NGS panels and required all NGS assays to be tissue-based to account for detection differences between tissue and plasma [58,59]. Most NGS panels utilized at the baseline timepoint of this study were DNA based only (Stanford STAMP, MGH SNaPshot, FoundationOne^®), though two assays (Tempus xT, StrataNGS^™) included both DNA and RNA analysis [25-29]. Analyses of acquired resistance at time of progression on osimertinib were limited by number of paired pre- and post-treatment tissue NGS samples. The proportion of patients who received a subsequent second-line systemic therapy (74 %) in this study exceeded that of comparable clinical trial populations and real-world analyses (48–60 %), potentially due to the more contemporary study period or study site restriction to four academic medical centers in the United States [9,10,60,61]. Lastly, the moderate sample size, particularly in the context of small molecular subgroups, and generalizability of our study findings should be considered, and warrant further validation.

De novo co-alterations occurring simultaneously with EGFR have the potential to negatively impact prognosis as well as predict for lower likelihood of durable response to EGFR TKI therapy. This study further supports the impact of de novo genetic co-alterations on patient outcomes, and specifically identifies co-alteration of TP53 and PIK3CA as a potential high-risk molecular sub-group. Prospective identification of patients with high-risk features, both clinical and molecular, should be considered as a stratification or randomization factor in front-line clinical trials of EGFRm NSCLC. Furthermore, differentiating between prognostic and predictive factors should be carefully considered both in further clinical study and in interpretation of the current body of literature.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.lungcan.2025.108747.