Co-mutation landscape and prognostic impact of genomic complexity in EGFR-mutant non-small cell lung cancer

Abstract

Methods

Clinical and genomic data of patients with EGFR-mutant NSCLC were collected from publicly available cBioPortal cohorts. Cohort composition, stage availability, recurrent co-mutations, and study-origin heterogeneity were evaluated. Patients harboring EGFR Del19 or L858R mutations and available survival information were included for survival analyses. Genomic landscapes were analyzed using a standardized Del19/L858R survival cohort. Survival outcomes were evaluated using Kaplan-Meier curves and Cox regression models, including sensitivity analyses adjusted for study origin and available tumor stage. Co-mutation burden and TP53/ATM-defined DDR-related alterations were further assessed as exploratory prognostic features.

Results

A total of 1,024 EGFR-mutant NSCLC patients were identified. Among them, 844 patients harbored Del19 or L858R mutations. Among the 844 patients with EGFR Del19 or L858R mutations, 820 had complete survival information and were included in survival analyses (Del19, n = 429; L858R, n = 391). An apparent OS difference was observed between Del19 and L858R tumors in the pooled cBioPortal cohort; however, this association was attenuated after accounting for study-origin heterogeneity and was therefore interpreted as exploratory. TP53 mutation was the most robust adverse prognostic factor and remained independently associated with inferior survival after adjustment for study origin. PIK3CA mutation also remained associated with poorer OS in the study-origin-adjusted model, whereas EGFR subtype and CDKN2A mutation were not retained as independent prognostic factors after accounting for study-level heterogeneity. Increasing co-mutation burden was associated with progressively shorter OS. TP53/ATM-defined DDR-related alterations were associated with inferior OS in exploratory analysis, but this result was largely driven by TP53 because ATM-only alterations were rare and not significantly associated with OS.

Conclusion

Our study demonstrates that prognosis in EGFR-mutant NSCLC is strongly influenced by broader genomic context, particularly TP53 mutation and overall co-mutation burden. EGFR subtype-specific survival differences and TP53/ATM-defined DDR-related findings should be interpreted cautiously because of study-origin heterogeneity, incomplete clinical annotation, and the limited DDR gene definition.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12885-026-16715-7.

Introduction

Lung cancer remains the leading cause of cancer-related mortality worldwide, accounting for approximately 1.8 million deaths annually [1]. Non-small cell lung cancer (NSCLC) represents nearly 85% of all lung cancer cases, among which lung adenocarcinoma is the predominant histological subtype [2]. The identification of oncogenic driver mutations has fundamentally transformed the treatment landscape of NSCLC, enabling the development of precision medicine approaches and significantly improving patient outcomes [3, 4].

Among actionable driver alterations, mutations in the epidermal growth factor receptor (EGFR) gene are among the most common genomic events in NSCLC, particularly in East Asian populations, women, and never-smokers [5]. EGFR tyrosine kinase inhibitors (EGFR-TKIs) have become the standard first-line treatment for patients with EGFR-mutant NSCLC and have demonstrated substantial improvements in progression-free survival and quality of life compared with conventional chemotherapy [6, 7]. Nevertheless, considerable heterogeneity in treatment response and long-term survival remains evident among patients harboring EGFR mutations, indicating that additional molecular factors may influence disease progression and clinical outcomes [8].

The two most prevalent EGFR mutation subtypes, exon 19 deletion (Del19) and L858R, account for the majority of EGFR-mutant NSCLC cases [9, 10]. Although both mutations are generally considered sensitizing alterations, accumulating evidence suggests that they exhibit distinct biological characteristics and therapeutic responses [11]. Previous clinical studies have reported differences in survival outcomes and responses to EGFR-TKIs between Del19 and L858R tumors, suggesting that these mutations may represent biologically distinct disease entities [12]. However, the molecular mechanisms underlying these differences remain incompletely understood.

Recent genomic studies have demonstrated that EGFR-mutant NSCLC frequently harbors concurrent genomic alterations in addition to the primary driver mutation [13]. Among these, TP53, RB1, PIK3CA, CDKN2A, and ATM are recurrently altered and have been implicated in tumor progression, therapeutic resistance, and survival outcomes [14]. In particular, TP53 mutation has emerged as one of the most common co-occurring alterations in EGFR-mutant tumors and has been associated with inferior responses to EGFR-targeted therapies [15, 16]. However, the relative prognostic importance of different co-mutations and their interactions with specific EGFR mutation subtypes remain incompletely characterized.

Beyond individual genomic alterations, increasing evidence suggests that overall genomic complexity may play an important role in determining tumor behavior. The accumulation of multiple co-occurring mutations may reflect enhanced genomic instability, increased clonal diversity, and accelerated tumor evolution [17]. Consequently, co-mutation burden has emerged as a potential biomarker for prognostic stratification across multiple cancer types. In addition, pathway-level alterations, particularly those affecting the DNA damage response (DDR) pathway, may provide biologically meaningful insights beyond single-gene analyses [18]. Defects in DDR-related genes can promote genomic instability and contribute to tumor progression, yet their prognostic significance in EGFR-mutant NSCLC remains incompletely defined [19].

In the present study, we performed a comprehensive multi-cohort analysis of EGFR-mutant NSCLC using publicly available cBioPortal datasets. We systematically characterized the genomic landscape of EGFR-mutant tumors, compared co-mutation patterns between Del19 and L858R subtypes, evaluated the prognostic impact of recurrent co-mutations, and investigated the clinical significance of co-mutation burden and DDR pathway alterations. Our findings provide further insights into the molecular heterogeneity of EGFR-mutant NSCLC and identify genomic features that may improve prognostic stratification and facilitate more precise clinical management.

Materials and methods

Data collection and patient selection

Clinical and genomic data of patients with non-small cell lung cancer (NSCLC) were obtained from publicly available datasets in the cBioPortal for Cancer Genomics. The contributing studies and available clinical annotations were summarized by study origin. Because the pooled dataset was derived from multiple public cohorts with different sequencing strategies and clinical annotation depth, study origin was explicitly considered in sensitivity analyses. Cases lacking survival information or mutation annotation were excluded from survival analyses.

The two most common EGFR mutation subtypes, exon 19 deletion (Del19) and L858R, were selected for comparative and survival analyses. Patients carrying rare EGFR mutations, including G719X, L861Q, S768I, and T790M, were retained for the overall mutation-spectrum analysis but excluded from Del19/L858R subtype-specific survival analyses (Fig. 1A). For each contributing cBioPortal study, the available sequencing platform or gene panel information was summarized where recoverable from the dataset annotation and source study information (Supplementary Table S4). Because sample-level panel coverage was not uniformly available in the derived mutation matrix, mutation-negative status was not interpreted as definitive biological wild-type status for genes that may not have been assayed; this panel heterogeneity was treated as a limitation.

Fig. 1.

Study workflow and EGFR mutation spectrum in EGFR-mutant NSCLC. A Study workflow illustrating patient selection and analytical strategy. A total of 1,024 EGFR-mutant NSCLC patients were identified from public cBioPortal cohorts. Among them, 844 patients harbored Del19 or L858R mutations, and 820 patients with available survival information were included in survival analyses. Subsequent analyses included co-mutation landscape characterization, co-mutation burden assessment, TP53/ATM-defined DDR-related exploratory evaluation, and survival analyses. B Distribution of recurrent EGFR mutations along the EGFR protein structure. Exon 19 deletion (Del19) and L858R were the predominant EGFR alterations, whereas less frequent mutations included G719X, L861Q, S768I, and T790M. Functional domains of the EGFR protein are indicated below the mutation sites

Genomic landscape and co-mutation analysis

Somatic mutation data were extracted from cBioPortal and processed using R software (version 4.3.0). The frequencies of recurrent genomic alterations were calculated, and the top co-mutated genes were identified in a standardized Del19/L858R survival cohort (n = 820) to avoid inconsistent denominators across panels.

Oncoprint plots were generated to visualize the mutation landscape of EGFR-mutant tumors. Differences in co-mutation frequencies between Del19 and L858R tumors were compared using Pearson’s chi-square test or Fisher’s exact test when appropriate. Recurrent co-mutations evaluated in this study included TP53, PIK3CA, RB1, CDKN2A, ATM, KRAS, STK11, KEAP1, NF1, and BRAF.

Definition of co-mutation burden and pathway alterations

To evaluate the impact of genomic complexity, co-mutation burden was defined as the total number of recurrent co-occurring genomic alterations identified in each patient. Patients were categorized into three groups according to co-mutation burden: No co-mutations (0), One co-mutation (1), Two or more co-mutations (≥ 2).

To investigate pathway-level genomic alterations, recurrent mutations were further grouped according to their biological functions. Because broad DDR gene coverage was not uniformly available across contributing cohorts and because TP53 and ATM were the only recurrent DDR-related genes with sufficient representation in the downloaded matrix, DDR-related status was conservatively defined as alteration of TP53 and/or ATM. This analysis was prespecified as exploratory and was not interpreted as a comprehensive DDR pathway analysis.

Survival analysis

Overall survival (OS) was selected as the primary clinical endpoint. Kaplan-Meier survival curves were generated using the survival and survminer R packages, and differences between groups were evaluated using the log-rank test. Survival analyses were performed according to EGFR mutation subtype (Del19 vs. L858R), TP53 mutation status, combined EGFR subtype and TP53 status, co-mutation burden, non-TP53 co-mutation burden, and TP53/ATM-defined DDR-related status. Median survival times, hazard ratios (HRs), and corresponding 95% confidence intervals (CIs) were calculated when applicable.

Cox regression analysis

Univariate Cox proportional hazards regression analyses were performed to evaluate the prognostic significance of recurrent co-mutations. Variables with potential prognostic relevance were subsequently included in multivariable Cox proportional hazards regression models. The base multivariable model included EGFR subtype, TP53, PIK3CA, CDKN2A, RB1, and KRAS. Because the pooled cBioPortal dataset comprised heterogeneous studies, a study-origin-adjusted Cox model was considered the primary multivariable sensitivity model. A stage-available sensitivity model was also performed in patients with tumor stage information; KRAS was not estimable in that subset because only two KRAS-mutant cases were present. To evaluate whether the prognostic effect of TP53 differed according to EGFR subtype, an interaction term between TP53 mutation status and EGFR subtype was incorporated into the Cox model. Hazard ratios (HRs), 95% confidence intervals (CIs), and P values were reported.

Statistical analysis

All statistical analyses were performed using R software (version 4.3.0). Continuous variables were summarized as median values with interquartile ranges and compared using the Wilcoxon rank-sum test. Categorical variables were compared using Pearson’s chi-square test or Fisher’s exact test, as appropriate. All statistical tests were two-sided. A P value < 0.05 was considered statistically significant. The proportional hazards assumption for Cox regression models was evaluated using Schoenfeld residuals generated from the variables included in the current cBioPortal multivariable Cox model (Supplementary Figure S1).

Results

Patient characteristics and EGFR mutation spectrum

A total of 1,024 EGFR-mutant NSCLC patients were identified from the pooled cBioPortal datasets. Among them, 844 patients harbored the two major EGFR mutation subtypes, including Del19 and L858R. After excluding Del19/L858R patients without complete survival information, 820 patients, including Del19 (n = 429) and L858R (n = 391) cases, were included in Del19/L858R survival analyses. Baseline clinical characteristics of the Del19 and L858R survival cohorts are summarized in Table 1. Detailed comparisons of recurrent co-mutations are summarized in Supplementary Table S1. The distribution of EGFR mutation subtypes varied across the included cBioPortal cohorts, although Del19 and L858R consistently represented the predominant EGFR alterations (Fig. 1B and Supplementary Figure S2A-B). Tumor stage availability across contributing cBioPortal studies is summarized in Supplementary Figure S2C, and cohort-level sequencing information is provided in Supplementary Table S4. Exon 19 deletion and L858R represented the two most prevalent EGFR alteration subtypes, whereas less frequent mutations included G719X, L861Q, S768I, and T790M.

Table 1.

Baseline clinical characteristics of the Del19 and L858R survival cohorts

Characteristic	Del19 (n = 429)	L858R (n = 391)	P
Age, years, median (IQR)	64.0 (55.0-71.5)	66.0 (61.0–73.0)	< 0.001
Age missing, n (%)	130 (30.3)	116 (29.7)
Sex, n (%)			0.920
Female	307 (71.6)	282 (72.1)
Male	122 (28.4)	109 (27.9)
Smoking status, n (%)			0.114
Never/non-smoker	188 (43.8)	168 (43.0)
Ever/current/former smoker	70 (16.3)	85 (21.7)
Missing	171 (39.9)	138 (35.3)
Race/ethnicity, n (%)			0.014
Asian	150 (35.0)	155 (39.6)
White	146 (34.0)	118 (30.2)
Black	15 (3.5)	8 (2.0)
Other/unknown	34 (7.9)	14 (3.6)
Missing	84 (19.6)	96 (24.6)
Tumor stage, n (%)			0.129
I	26 (6.1)	35 (9.0)
II	13 (3.0)	10 (2.6)
III	16 (3.7)	16 (4.1)
IV	121 (28.2)	82 (21.0)
Other/unknown	0 (0.0)	1 (0.3)
Missing	253 (59.0)	247 (63.2)
Study origin, n (%)			< 0.001
luad_mskcc_2023_met_organotropism	91 (21.2)	97 (24.8)
nsclc_ctdx_msk_2022	101 (23.5)	77 (19.7)
luad_mskcc_2020	64 (14.9)	82 (21.0)
luad_oncosg_2020	56 (13.1)	63 (16.1)
luad_mskimpact_2021	58 (13.5)	27 (6.9)
luad_cas_2020	20 (4.7)	27 (6.9)
bm_nsclc_mskcc_2023	20 (4.7)	13 (3.3)
lung_msk_pdx	19 (4.4)	4 (1.0)
lung_pdx_msk_2021	0 (0.0)	1 (0.3)

Genomic landscape of EGFR-mutant NSCLC

To characterize the genomic landscape of EGFR-mutant NSCLC, we first evaluated the frequency of recurrent co-occurring genomic alterations using a standardized Del19/L858R survival cohort (n = 820). TP53 was the most frequently co-mutated gene, followed in descending order by RB1, PIK3CA, NF1, CDKN2A, ATM, BRAF, KEAP1, KRAS, and STK11 (Fig. 2A). The overall mutation landscape is illustrated in the oncoprint plot (Fig. 2B), demonstrating substantial genomic heterogeneity among EGFR-mutant tumors.

Fig. 2.

Recurrent co-mutation landscape in the Del19/L858R survival cohort. A Frequencies of recurrent co-mutations in the Del19/L858R survival cohort (n = 820). Genes are shown in descending order of mutation frequency. B Oncoprint showing recurrent co-mutations in the same survival cohort. Samples were ordered by EGFR subtype and co-mutation burden. EGFR subtype and OS event status are annotated above the mutation matrix. C Comparison of recurrent co-mutation frequencies between Del19 (n = 429) and L858R (n = 391) tumors. Numbers above the bars indicate the number of mutated cases in each subgroup

Comparative analysis between Del19 and L858R tumors revealed distinct co-mutation patterns (Fig. 2C). Detailed comparisons of recurrent co-mutations between Del19 and L858R tumors are summarized in Supplementary Table S1. Although TP53 was the most frequent co-mutation in both subtypes, no significant difference in TP53 prevalence was observed between Del19 and L858R tumors (55.0% vs. 53.5%, P = 0.674). BRAF co-mutations were enriched in Del19 tumors compared with L858R tumors (2.6% vs. 0%, P = 0.001), whereas other recurrent alterations showed no statistically significant subtype-specific differences after frequency comparison.

TP53 co-mutation is associated with inferior survival in EGFR-mutant NSCLC

Kaplan–Meier analysis showed a difference in overall survival between patients with EGFR Del19 and L858R mutations in the pooled cBioPortal survival cohort (Fig. 3A). Although patients with L858R mutations appeared to have longer overall survival in this dataset, this observation should be interpreted cautiously because the pooled public cohorts were heterogeneous and lacked uniform treatment information. Therefore, EGFR subtype-specific survival differences were considered exploratory rather than definitive in the present study (Fig. 3A). Given that TP53 was the most prevalent co-mutated gene, we further evaluated its prognostic significance. TP53-mutant patients experienced significantly shorter overall survival than TP53 wild-type patients (P < 0.001, Fig. 3B). Subgroup analyses demonstrated that TP53 mutation remained associated with unfavorable prognosis within the Del19 subgroup; similar trends were observed across EGFR mutation subtypes (Fig. 3C). In the base multivariable Cox model, EGFR L858R was associated with a lower hazard of death compared with Del19 (HR = 0.733, 95% CI: 0.561–0.957, P = 0.023). TP53 mutation was the strongest adverse prognostic factor (HR = 2.248, 95% CI: 1.695–2.981, P < 0.001), while PIK3CA and CDKN2A mutations were also associated with poorer OS in the base model (Table 2). However, after adjustment for study origin, the association between EGFR subtype and OS was completely attenuated (HR = 1.01, 95% CI: 0.77–1.33, P = 0.947), and CDKN2A was no longer statistically significant (HR = 1.75, 95% CI: 0.85–3.62, P = 0.130). In contrast, TP53 mutation remained independently associated with inferior OS (HR = 1.84, 95% CI: 1.39–2.45, P < 0.001), and PIK3CA mutation also remained associated with poorer OS (HR = 1.63, 95% CI: 1.08–2.46, P = 0.019). Therefore, the apparent EGFR subtype-specific survival difference and the CDKN2A association were interpreted as exploratory and potentially influenced by study-level heterogeneity (Fig. 3D and Supplementary Figure S3A-B). Univariate Cox regression identified TP53 mutation as the strongest adverse prognostic factor among recurrent co-mutations (HR = 2.291, 95% CI: 1.740–3.016, P < 0.001), followed by RB1 mutation (HR = 1.804, 95% CI: 1.215–2.680, P = 0.003) (Supplementary Table S2). To further assess the combined prognostic impact of EGFR subtype and TP53 status, patients were stratified into four groups according to EGFR mutation subtype and TP53 mutation status. The worst survival outcome was observed in patients harboring concurrent L858R and TP53 mutations, whereas Del19 tumors without TP53 mutation exhibited the most favorable prognosis (Fig. 3E). No significant interaction was observed between EGFR subtype and TP53 mutation status (interaction HR = 1.12, 95% CI: 0.63–1.99, P = 0.693), suggesting that the adverse prognostic effect of TP53 mutation was generally consistent across EGFR mutation subtypes (Supplementary Figure S4A-B and Supplementary Table S3)

Fig. 3.

Prognostic impact of TP53 co-mutation in EGFR-mutant NSCLC. A Kaplan–Meier analysis comparing overall survival (OS) between patients harboring EGFR Del19 and L858R mutations. B Kaplan-Meier analysis of OS according to TP53 mutation status in the Del19/L858R survival cohort. C Kaplan–Meier analysis evaluating the prognostic impact of TP53 mutation within the EGFR Del19 subgroup. D Forest plot comparing multivariable Cox models before and after adjustment for study origin. The study-origin-adjusted model was considered the primary sensitivity model because of heterogeneity among contributing cBioPortal studies. E Kaplan-Meier analysis stratifying patients according to combined EGFR mutation subtype and TP53 mutation status. Patients were categorized into four groups: Del19/TP53-wild type, Del19/TP53-mutant, L858R/TP53-wild type, and L858R/TP53-mutant

Table 2.

Multivariable Cox regression analyses of overall survival before and after adjustment for study origin

Variable	Base model HR (95% CI)	P	Study-origin-adjusted HR (95% CI)	P
EGFR L858R vs. Del19	0.733 (0.561–0.957)	0.023	1.01 (0.77–1.33)	0.947
TP53 mutation	2.248 (1.695–2.981)	< 0.001	1.84 (1.39–2.45)	< 0.001
PIK3CA mutation	1.716 (1.143–2.576)	0.009	1.63 (1.08–2.46)	0.019
CDKN2A mutation	2.111 (1.036–4.303)	0.040	1.75 (0.85–3.62)	0.130
RB1 mutation	1.321 (0.878–1.989)	0.182	1.09 (0.72–1.65)	0.673
KRAS mutation	2.649 (0.971–7.229)	0.057	1.19 (0.42–3.34)	0.747

Increased co-mutation burden predicts poor clinical outcome

To determine whether the overall genomic complexity of EGFR-mutant tumors influences prognosis, patients were categorized according to co-mutation burden (0, 1, or ≥ 2 co-mutations).

Kaplan–Meier analysis demonstrated a stepwise reduction in overall survival with increasing co-mutation burden. Patients with ≥ 2 co-mutations exhibited the worst prognosis (Fig. 4A). Patients harboring two or more co-mutations exhibited the poorest clinical outcomes, whereas patients without detectable co-mutations experienced the longest survival.

Fig. 4.

Prognostic significance of co-mutation burden in EGFR-mutant NSCLC. A Kaplan–Meier analysis of overall survival according to co-mutation burden. Patients were categorized as having no co-mutations, one co-mutation, or two or more co-mutations (≥ 2). B Kaplan–Meier analysis stratified by both EGFR mutation subtype and co-mutation burden, demonstrating the combined impact of genomic complexity and EGFR subtype on survival outcomes

When stratified by both EGFR subtype and co-mutation burden, a consistent prognostic trend was observed (Fig. 4B). Regardless of EGFR subtype, increasing co-mutation burden was associated with progressively worse survival, highlighting the importance of global genomic complexity as a prognostic determinant in EGFR-mutant NSCLC. To determine whether this association was driven solely by TP53, we repeated the analysis after excluding TP53 from the co-mutation burden definition; non-TP53 co-mutation burden remained associated with poorer OS in sensitivity analysis (Supplementary Figure S5A-B). However, the subgroup with ≥ 2 non-TP53 co-mutations was small (n = 19), and the number at risk decreased rapidly during follow-up; therefore, this sensitivity analysis should be interpreted cautiously.

Exploratory analysis of TP53/ATM-defined DDR-related alterations

Because broad DDR gene coverage was not uniformly available across contributing cohorts, we conservatively evaluated a TP53/ATM-defined DDR-related alteration group. This analysis was intended as an exploratory pathway-oriented assessment rather than a comprehensive DDR pathway analysis.

Patients harboring TP53 and/or ATM alterations exhibited significantly shorter overall survival compared with patients without either alteration (Fig. 5). Because TP53 represented the dominant DDR-related alteration in this cohort, we further performed an ATM-only sensitivity analysis. ATM mutation alone was infrequent and was not significantly associated with OS, indicating that the TP53/ATM-defined DDR-related finding should be interpreted as exploratory and largely driven by TP53 alterations (Supplementary Figure S6).

Fig. 5.

Exploratory survival analysis of TP53/ATM-defined DDR-related alterations in EGFR-mutant NSCLC. Kaplan-Meier analysis comparing overall survival between patients with TP53 and/or ATM alterations and those without either alteration. Because ATM-only alterations were rare and were not significantly associated with OS, this TP53/ATM-defined DDR-related analysis should be interpreted as exploratory and largely TP53-driven

Collectively, our results indicate that TP53 mutation and broader genomic complexity, particularly co-mutation burden, are important determinants of survival among patients with EGFR-mutant NSCLC. In contrast, EGFR subtype-specific survival differences, CDKN2A findings, and TP53/ATM-defined DDR-related results require cautious interpretation in view of study-origin heterogeneity, panel heterogeneity, and incomplete clinical annotation.

Discussion

In this multi-cohort cBioPortal analysis of EGFR-mutant NSCLC, we identified several clinically relevant findings. First, Del19 and L858R tumors exhibited distinct genomic landscapes, whereas their apparent survival difference in the pooled cohort was attenuated after accounting for study origin. Second, TP53 mutation was the most frequent co-occurring genomic alteration and represented the most robust adverse prognostic factor. Third, increasing co-mutation burden was associated with progressively worse overall survival, highlighting the importance of genomic complexity in EGFR-mutant NSCLC. Finally, TP53/ATM-defined DDR-related alterations were associated with unfavorable prognosis in exploratory analysis, but this result was largely driven by TP53 and should not be interpreted as a comprehensive DDR pathway effect.

The prognostic significance of EGFR mutation subtype remains an area of ongoing investigation. Previous clinical studies have consistently demonstrated superior responses to EGFR tyrosine kinase inhibitors (EGFR-TKIs) among patients harboring Del19 mutations compared with those carrying L858R mutations [20, 21]. In the present study, patients harboring L858R mutations appeared to have longer overall survival than those carrying Del19 mutations in the pooled cBioPortal cohort. However, this finding contrasts with multiple previous studies reporting superior outcomes among Del19 tumors treated with EGFR-TKIs. Therefore, the observed survival difference should be interpreted as an exploratory observation rather than a definitive biological conclusion [22]. This discrepancy may reflect the distinction between treatment-specific efficacy and overall survival in heterogeneous public datasets. Overall survival may be influenced by cohort composition, treatment exposure, subsequent therapies, follow-up duration, and missing clinical variables. Therefore, the subtype-specific survival finding should not be interpreted as evidence that L858R is intrinsically associated with better prognosis than Del19, but rather as an exploratory observation requiring validation in uniformly treated cohorts.

Recent single-cell, spatial transcriptomic, and machine-learning studies have further shown that lung adenocarcinoma prognosis is shaped by malignant epithelial programs, immune and stromal microenvironmental states, spatial macrophage niches, and cellular heterogeneity, complementing mutation-based prognostic stratification [23, 24]. Importantly, our genomic analyses further revealed differences in co-mutation patterns between these two subtypes, supporting the concept that Del19 and L858R represent biologically distinct molecular entities rather than interchangeable forms of EGFR-mutant disease [25, 26]. These findings suggest that the clinical behavior of EGFR-mutant NSCLC may be influenced not only by the primary driver mutation but also by the broader genomic context in which it occurs.

Among all recurrent genomic alterations, TP53 emerged as the most prevalent co-mutation and the strongest adverse prognostic factor [27, 28]. TP53 mutations have been reported in approximately 40–60% of EGFR-mutant NSCLC cases and have been associated with shorter progression-free survival, inferior overall survival, and accelerated development of resistance to EGFR-targeted therapies [29, 30]. Consistent with these observations, TP53-mutant tumors in our cohort demonstrated significantly worse survival outcomes compared with TP53 wild-type tumors [31, 32]. Moreover, the adverse prognostic effect of TP53 mutation remained evident within the Del19 subgroup and was consistently observed across EGFR mutation subtypes. Patients harboring concurrent L858R and TP53 mutations exhibited the poorest survival, suggesting that TP53 may function as a major prognostic modifier in EGFR-mutant NSCLC. The prevalence of TP53 co-mutation observed in our study is consistent with previous genomic analyses of EGFR-mutant NSCLC, in which TP53 alterations occur in approximately 50–60% of cases. The high frequency and strong prognostic impact of TP53 suggest that this alteration may represent a key molecular event driving tumor evolution beyond the primary EGFR mutation.

An important finding of this study is the prognostic value of co-mutation burden. While most previous studies have focused on the impact of individual genomic alterations, our analysis demonstrated a stepwise deterioration in survival with increasing numbers of co-occurring mutations [33]. Patients with two or more co-mutations experienced the shortest overall survival, whereas those without detectable co-mutations had the most favorable outcomes [34]. These findings indicate that cumulative genomic complexity may better capture the biological aggressiveness of EGFR-mutant tumors than any single genomic event. The observed relationship between co-mutation burden and prognosis further supports the concept that tumor evolution is driven by the cooperative effects of multiple genomic alterations rather than isolated mutations.

In addition to TP53, several recurrent alterations identified in our cohort have been implicated in resistance to targeted therapies and tumor progression. PIK3CA mutations remained associated with poor survival after adjustment for study origin and may contribute to therapeutic failure through activation of the PI3K/AKT/mTOR signaling pathway [35, 36]. Activation of this pathway can restore downstream proliferative signaling despite EGFR inhibition, thereby promoting resistance to EGFR-TKIs. Similarly, alterations affecting cell-cycle regulators such as RB1 and CDKN2A may facilitate uncontrolled proliferation and genomic instability [37]. Previous studies have linked RB1 loss to aggressive tumor behavior and histologic transformation following EGFR-targeted therapy, highlighting the biological importance of cell-cycle dysregulation in EGFR-mutant NSCLC [38, 39]. Notably, BRAF co-mutations were significantly enriched in the Del19 subgroup in our cohort. Although the absolute frequency was low, BRAF alterations have been reported as bypass pathway events that may contribute to acquired resistance to EGFR-TKIs. This subtype-specific enrichment should therefore be interpreted cautiously but may warrant further validation in larger, uniformly sequenced EGFR-mutant cohorts.

Our pathway-level analysis identified TP53/ATM-defined DDR-related alterations as an adverse prognostic feature in exploratory analysis. The DDR pathway plays a central role in maintaining genomic integrity through the detection and repair of DNA damage. However, for the purpose of this study, DDR-related status was restricted to TP53 and ATM because broader DDR gene coverage was not uniformly available across all contributing cohorts. ATM-only alterations were infrequent and were not significantly associated with OS, indicating that the observed survival difference was mainly driven by TP53 alterations and should be interpreted cautiously. These findings suggest that pathway-oriented genomic classification may provide complementary prognostic information, but this result requires validation in clinically annotated cohorts with broader and harmonized DDR gene coverage.

From a clinical perspective, our findings support the integration of genomic co-alterations into risk stratification strategies for EGFR-mutant NSCLC. Traditional prognostic assessment has largely focused on EGFR mutation subtype alone; however, our results indicate that TP53 mutation status and co-mutation burden provide complementary prognostic information. The TP53/ATM-defined DDR-related result may generate hypotheses for future analyses but should not be used as a definitive pathway-level biomarker without broader DDR gene coverage and prospective validation.

Several limitations should be acknowledged. First, this study was based on retrospective analyses of publicly available cBioPortal datasets, which may introduce selection bias and cohort heterogeneity. Second, several key clinical variables, including tumor stage, ECOG performance status, smoking history, and detailed treatment information, were not uniformly available across cohorts, limiting comprehensive adjustment for potential confounders. Third, treatment-related variables, including specific EGFR-TKI regimens and subsequent lines of therapy, were not uniformly available across cohorts; therefore, the survival associations should not be interpreted as treatment-specific effects. Fourth, although study-origin and stage-available sensitivity analyses were performed, residual confounding caused by study-level heterogeneity and missing clinical annotation cannot be excluded. Fifth, sequencing panel heterogeneity across cBioPortal studies may affect mutation frequency and co-mutation burden estimates. Because sample-level gene-panel coverage was not uniformly available in the derived mutation matrix, mutation-negative calls for selected genes could not always be distinguished from lack of assay coverage. Sixth, the non-TP53 co-mutation burden sensitivity analysis was limited by the small size of the ≥ 2 non-TP53 co-mutation subgroup (n = 19), resulting in unstable long-term Kaplan-Meier estimates. Finally, the biological mechanisms underlying the observed associations were not experimentally validated. Future prospective studies integrating genomic, transcriptomic, spatial, and functional analyses are warranted to further elucidate the molecular basis of prognosis in EGFR-mutant NSCLC.

In conclusion, this multi-cohort analysis demonstrates that TP53 mutation and co-mutation burden are important determinants of survival in EGFR-mutant NSCLC. The apparent EGFR subtype-specific survival association was attenuated after study-origin adjustment, and TP53/ATM-defined DDR-related findings should be considered exploratory and largely TP53-driven. These findings highlight the prognostic value of genomic complexity and support the development of molecularly informed risk stratification strategies for patients with EGFR-mutant lung cancer.

Authors’ contributions

Xiangdi Yang: Data curation, Methodology, Funding Acquisition, Writing - review & editing. Guojian Huang: Formal analysis, Validation, Writing - review & editing. Kejun Liu: Data curation, Formal analysis, Visualization, Writing - original draft. Dongxia Wang: Funding, Software, Writing - review & editing.June Wang: Software, Data curation, Writing - review & editing.Meixia Huang: Validation, Visualization, Writing - review & editing.Linxuan Huang: Conceptualization, Methodology, Supervision, Funding Acquisition, Writing - original draft, Writing - review & editing. All authors have read and approved the final version of the manuscript.

Funding

This work was supported by funds from the Guangdong Basic and Applied Basic Research Foundation, China (No. 2021B1515140031). Dongguan Science and Technology of Social Development Program, China (No. 20231800936452). The Tenth Affiliated Hospital of Southern Medical University (Dongguan People’s Hospital) (No. Z202412). School-Level Scientific Research Affiliated Joint Project of Xiangnan University (No. 2022JX138 and 2025XJ87) and the 2025 Hunan Provincial Natural Science Foundation Regional Joint Fund Project (No. 2025JJ70537).

Data availability

The data that support the findings of our study are available on request from the corresponding author.

Declarations

Ethics approval and consent to participate

This study was reviewed and approved by the Medical Ethics Committee of Dongguan People’s Hospital (The Tenth Affiliated Hospital of Southern Medical University), Dongguan, China (approval No. LW2026-017). The present study involved only the secondary analysis of publicly available, de-identified cBioPortal data, and no identifiable participant information was accessed. Therefore, additional informed consent to participate was not required for the present study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.