TP53 Co-Mutation Status Association with Clinical Outcomes in Patients with EGFR-Mutant Non-Small Cell Lung Cancer

Simple Summary There is conflicting evidence on the impact of TP53 co-mutations on survival in patients with EGFR-mutated advanced non-small cell lung cancer (aNSCLC). This observational study used data from a de-identified database to assess the role of TP53 co-mutation on survival in patients with EGFR-mutated aNSCLC (n = 356). We observed a significant decrease in survival outcomes (namely real-world progression-free survival [rwPFS] and overall survival [OS]) duration in the group of aNSCLC patients with presence of TP53 co-mutation compared to those with wild-type TP53 tumors. TP53 co-mutations were also associated with worse outcomes by subgroups, including type of EGFR-mutation and first-line treatment. Together, these findings indicated negative impact of TP53 co-mutation in outcomes for patients with EGFR-mutated aNSCLC. Abstract TP53 co-mutations have shown association with poor prognosis in various solid tumors. For EGFR-mutated advanced non-small cell lung cancer (aNSCLC), conflicting results exist regarding its impact on survival. Clinical outcomes and genomic data were obtained retrospectively from the real-world (rw) de-identified clinicogenomic database. Patients who initiated therapy for EGFR-mutated aNSCLC between January 2014 and December 2020 were identified. Clinical outcomes were evaluated by TP53-mutational status. In 356 eligible EGFR-mutated aNSCLC patients (median age 68 years), 210 (59.0%) had TP53 co-mutation and 146 (41.0%) had TP53 wild-type tumor. Unadjusted analysis showed significantly shorter survival in patients with TP53 co-mutation versus TP53 wild-type (rw progression-free survival [rwPFS]: HR = 1.4, 95% CI 1.1–1.9, p = 0.0196; overall survival [OS]: HR = 1.6, 95% CI 1.1–2.2, p = 0.0088). Multivariable analysis confirmed independent association between TP53 co-mutation and worse rwPFS (HR = 1.4, 95% CI 1.0–0.9, p = 0.0280) and OS (HR = 1.4, 95% CI 1.0–2.0, p = 0.0345). Directionally consistent findings were observed for response rates, and subgroups by EGFR-activating mutation and first-line (1 L) therapy, with more pronounced negative effect in 1 L EGFR-TKI subgroup. TP53 co-mutations negatively affected survival in patients with EGFR-mutated aNSCLC receiving standard 1 L therapy in real-world practice.


Introduction
Epidermal growth factor receptor (EGFR) mutations are among the most common oncogenic drivers in advanced or metastatic non-small cell lung cancer (aNSCLC) [1,2], which continues to be the most common cause of cancer deaths in the world [3]. Most (~85%) EGFR-activating mutations involve deletions in exon 19 (ex19del) or a substitution mutation in exon 21, specifically Leu858Arg (ex21L858R) [4] and they are classical EGFR mutations [5]. The current NCCN guideline recommendation and standard of care for aNSCLC with an

Study Design and Patients
This retrospective observational study used the nationwide de-identified Flatiron Health Foundation Medicine (FH-FMI) aNSCLC clinicogenomic database to select adults (≥18 years) with aNSCLC harboring an EGFR mutation, with or without the TP53 gene mutation, who initiated first-line (1 L) therapy between January 2014 and September 2020 in routine clinical practice. The EGFR mutations included were the exon 19 deletions or the exon 21L858R mutation, evaluated in comprehensive genomic profiling baitsets DX1, T7, T5a, or T4b. The patient data originated from approximately 280 cancer clinics (~800 sites of care). Patients with evidence of other primary tumors, other than non-melanoma skin cancer, and those with no confirmed EMR activity within 90 days of aNSCLC diagnosis, were excluded. Patients were followed from the date of initiation of 1 L therapy (index date) until death, loss to follow-up, or end of the database (December 2020), whichever occurred first. Eligible patients were divided into two groups based on the status of the TP53 gene mutation: TP53 wild-type and TP53 mutant groups.
The development, structure, and validation of the linked, de-identified FH-FMI clinicogenomic database have been described previously [39]. Briefly, the database contains retrospective longitudinal clinical data derived from EMR data, comprising patient-level structured and unstructured data (e.g., demographics, tumor stage and histology, performance status, tumor progression, death), curated via technology-enabled abstraction and linked to genomic data derived from FMI comprehensive genomic profiling (CGP) tests (e.g., next generation sequencing) in the FH-FMI clinicogenomic database by de-identified, deterministic matching. The clinicogenomic database included 12,726 patients with chartconfirmed aNSCLC receiving care at Flatiron Health's network of community (~75%) and academic (~25%) oncology practices [40].
The study was exempt from institutional review board (IRB) approval because it was a retrospective study and used de-identified data in compliance with the Health Insurance Portability and Accountability Act. Approval from WCG IRB for data collection for this real-world cohort was obtained prior to study conduct and included a waiver of informed consent.

End Points
The main end point was real-world progression-free survival (rwPFS, the time until the first documented tumor progression, death, or loss to follow-up), which was chosen because it measures treatment effects without the cofounding influence of subsequent therapies. Secondary end points included overall survival (OS, time until death or loss to follow-up), real-world overall response rate (rwORR, documented complete response or partial response), real-world disease control rate (rwDCR, documented complete response or partial response or stable disease), and real-world duration of response (rwDoR: time from initial tumor response to disease progression or death, whichever occurred first). Realworld best overall response (rwBOR) was defined as best tumor response (from progressive disease, stable disease, partial response, to complete response) during treatment. All end points were indexed to the start of 1 L therapy unless otherwise noted.

Statistical Analysis
Baseline characteristics and time-to-event endpoints were analyzed and presented for all eligible patients, whereas analysis of tumor response end points was restricted to patients with post-baseline tumor response assessment data. Descriptive statistics, including frequencies and percentages for categorical variables and median and range for continuous variables, were used to summarize the demographic and clinical characteristics as well as treatments received. Differences between groups were assessed using tests appropriate for the distribution of each measure, including t-tests for continuous variables and chi-squared tests for categorical variables. The Kaplan-Meier method was used to estimate rwPFS and other time-to-event endpoints, along with median values and 95% confidence intervals (CIs). The difference between groups for time-to-event endpoints was assessed using the log-rank test.
To evaluate and identify the baseline characteristics potentially associated with outcome in the dataset, hazard ratios (HRs) for time-to-event endpoints and odds ratios (ORs) for response rates, along with the corresponding 95% CIs, were computed using univariate Cox-proportional hazards and logistic regression models, respectively. Independent variables for the univariate models (including age, sex, smoking history, Eastern Cooperative Oncology Group performance status [ECOG PS], presence of brain or bone metastases, EGFR mutation type, and TP53 mutation status) were selected a priori as having a potential prognostic impact in EGFR-mutated NSCLC based on the literature [41][42][43][44][45][46][47]. Any factor found to be statistically significant for any endpoint, p < 0.05 in a univariate model, was then included as a covariate in all final multivariable Cox and logistic regression models, with one final multivariable model for each endpoint. The final multivariable models were the primary basis for investigating the independent prognostic effect of the TP53 co-mutations. No data imputation was performed for missing values.
For rwPFS and rwDoR, patients who were alive and progression-free at the end of the dataset (December 2020) were censored at the date of their last clinic visit. For patients still alive at the end of the observation period, OS was censored at their last activity date in the EMR. Subgroup analyses were performed to assess the effect of TP53 co-occurring mutations in subsets with EGFR exon 19 deletions, EGFR exon 21L858R mutations, and those treated with EGFR-TKI monotherapy, first-, second-, and third-generation EGFR-TKIs, and other non-EGFR-TKI therapy (including chemotherapy and immunotherapy). All statistical tests were two-sided, with a p value < 0.05 considered statistically significant. Statistical analyses were performed using SAS version 9.4 software (SAS Institute Inc., Cary, NC, USA).

Baseline Demographics and Clinical Characteristics
A total of 356 patients with EGFR-mutated aNSCLC were eligible for the study (Figure 1), 210 (59.0%) of whom had a TP53 co-mutation and 146 (41.0%) had TP53 wild-type tumors. Baseline patient characteristics overall and by TP53 co-mutation status are summarized in Table 1. Overall, the median age was 68 years (interquartile range [IQR] 60.5-76.1) and the majority were white (60.4%) and female (68.5%) with stage IV disease (86.2%) at initial diagnosis, no history of smoking (56.7%), and ECOG PS ≥ 1 (70.2%). The proportion of patients with a TP53 co-mutation with exon 19 deletion or exon 21L858R mutation was 57.0% and 43.0%, respectively. Distribution of baseline characteristics was generally similar by TP53 co-mutation status, although patients with a TP53 co-mutation tended to be younger (<65 years: 43.3% vs. 31.5%) and white (62.3% vs. 57.5%).
those treated with EGFR-TKI monotherapy, first-, second-, and third-generation E TKIs, and other non-EGFR-TKI therapy (including chemotherapy and immunother All statistical tests were two-sided, with a p value < 0.05 considered statistically signifi Statistical analyses were performed using SAS version 9.4 software (SAS Institute Cary, NC, USA).

Baseline Demographics and Clinical Characteristics
A total of 356 patients with EGFR-mutated aNSCLC were eligible for the study (Figure 1), 210 (59.0%) of whom had a TP53 co-mutation and 146 (41.0%) had TP53 wild tumors. Baseline patient characteristics overall and by TP53 co-mutation status are sum rized in Table 1. Overall, the median age was 68 years (interquartile range [IQR] 60.5and the majority were white (60.4%) and female (68.5%) with stage IV disease (86.2%) at i diagnosis, no history of smoking (56.7%), and ECOG PS ≥1 (70.2%). The proportion of pat with a TP53 co-mutation with exon 19 deletion or exon 21L858R mutation was 57.0% 43.0%, respectively. Distribution of baseline characteristics was generally similar by TP5 mutation status, although patients with a TP53 co-mutation tended to be younger (< 65 y 43.3% vs. 31.5%) and white (62.3% vs. 57.5%).

Treatments Received
By inclusion criteria, all patients received 1 L systemic therapy (Table 2). However, 169 (47.5%) also received second-line (2 L) treatment during the observation period of the study. Treatment patterns were generally similar between the TP53 co-mutation and TP53 wild-type groups (Table 2). Overall, the majority received targeted therapy with EGFR-TKI monotherapy in 1 L (75.0%), with chemotherapy alone (10.0%) and a chemotherapy combination with immunotherapy (5.6%) as the next two common treatments administered in 1 L. In 2 L, EGFR-TKI monotherapy (67.5%) was most received treatment, with chemotherapy in combination with immunotherapy (7.3%), and chemotherapy alone (4.7%) as the second and third most common treatments administered in 2 L. Osimertinib was the most used EGFR-TKI in both 1 L and 2 L (43.3% and 47.9%, respectively), followed by erlotinib (18.3% and 8.9%), and afatinib (12.4% and 8.3%) (Supplementary Table S1).

EGFR Mutation Subtype
When analysis was restricted to the type of EGFR-activating mutation, the negative effect of TP53 co-mutations on the rwPFS in the overall population (HR = 1.4, 95% CI 1.0-1.9, Figure 2) was directionally consistent across the subgroups by activating the EGFR mutation subtype, but was not statistically significant in the subgroups of patients with exon 19 deletion ( Figure 4A). Similarly, the negative effect of TP53 co-mutations on OS in the overall population (HR = 1.4, 95% CI 1.0-2.0, p = 0.0345, Figure 2B) was directionally consistent with the EGFR-activating mutation subtype. However, the difference in OS was statistically significant only in the subgroup of patients with exon 19 deletions (median 25.6 [95% CI 21.7-29.7] vs. 47.1 [28.7-not achieved, NA], HR = 1.9, 95% CI 1.2-2.9, p = 0.0072, Figure 3B) and not significant in those with exon 21L858R mutations (median 25 Figure 4B).
Supplementary Figure S1 presents the four-way comparison between the survival values for the included groups. Survival analysis for patients with EGFR exon 21L858R mutations and TP53 co-mutations had the shortest median rwPFS (9.9 months) compared with 15.6 months, 14.6 months, and 13.2 months for patients with TP53 wild-type plus EGFR exon 19 deletions, TP53 wild-type plus EGFR exon 21L858R mutations, and TP53 co-mutations plus EGFR exon 19 deletions, respectively (p = 0.0243, Supplementary Figure S1A). Similarly, patients with EGFR exon 21L858R and TP53 co-mutations had the shortest median OS (25.3 months) compared with 47.1 months, 28.7 months, and 25.6 months for patients with TP53 wild-type plus EGFR exon 19 deletions, TP53 wild-type plus EGFR exon 21L858R mutations, and TP53 co-mutations plus EGFR exon 19 deletions, respectively (p = 0.0388; Supplementary Figure S1B).

Discussion
In this large retrospective clinicogenomic dataset, real-world progression-free and overall survival were significantly shorter among treated EGFR-mutated aNSCLC patients with a TP53 co-mutation versus those with a TP53 wild-type tumor, even after adjustment for known clinicopathologic prognostic factors. A non-significant trend toward lower tumor response rates was observed among patients with a TP53 co-mutation. This negative contribution was generally independent of the type of 1 L treatment or EGFR mutation subtype. These findings confirm and strengthen available evidence from other datasets with limited follow-up [28,[30][31][32][33]48,49], supporting the independent negative prognostic value of TP53 co-mutation status. This may also help to identify a subset of patients with anticipated worse outcomes on 1 L aNSCLC treatment, particularly with EGFR-TKI monotherapy. While this study represents a large dataset, larger cohorts with EGFR and TP53 co-mutated aNSCLC are needed to validate the current findings and to further clarify the contribution and underlying mechanism of a TP53 co-mutation to primary or acquired resistance to EGFR-TKIs.
The study dataset was drawn predominantly from community oncology centers and restricted to patients with genomic testing. However, there was limited representation from academic practices where broad-based genomic sequencing may be part of research protocols, and from vulnerable patients with disproportionate access to physicians with distinct testing and practice patterns. The dataset is nationally illustrative of demographic (e.g., age, sex, and geographic location) and clinical (e.g., stage of disease) characteristics and distribution of patients across community and academic practices [50][51][52]. For example, the incidence of a TP53 co-mutation in our study (59%) was within the range reported in other observational and clinical datasets (30-72%) of patients with EGFR-mutated aNSCLC, using various methods of EGFR detection with different sensitivity [28]. As also previously reported [31,48], TP53 co-mutations were generally not associated with other measured prognostic factors (e.g., age, sex, smoking status, ECOG PS, baseline brain metastases), EGFR mutation subtype, or treatment assignment. Lastly, consistent with guideline recommendations [7,8] and prior studies [53], most patients with EGFR-mutated aNSCLC received EGFR-TKI monotherapy (predominantly osimertinib) in the 1 L and 2 L settings in current study, with few patients receiving 1 L EGFR-TKI combinations with either anti-angiogenics or chemotherapy.
In line with previous studies linking a TP53 co-mutation with reduced sensitivity to EGFR-TKIs [29,49], a TP53 co-mutation was associated with worse outcomes in the current dataset, even after adjustment for known aNSCLC prognostic factors including age, sex, smoking history, performance status, and baseline brain or bone metastases. Our findings also suggest that TP53 co-mutations in the context of the EGFR mutation subtype may impact response to treatment and the survival outcomes, as the negative survival effect associated with TP53 co-mutations has been observed to occur less among patients with EGFR exon 19 deletions than exon 21L858R mutations [31,32] and is likely because EGFR exon 19 deletions are known to be more responsive to EGFR-TKIs [53,54], as also demonstrated by our results. Notably, this finding appears to be driven by the robust OS in patients with TP53 wild-type tumors plus the exon 19 deletion subgroup (47.1 months), as the OS was similar between the exon 19 deletion subgroup plus a TP53 co-mutation (25.6 months) and an exon 21L858R mutation plus the TP53 co-mutation subgroup (28.7 months). Sufficiently powered subgroup analyses on the effect of TP53 co-mutations by the EGFR mutation subtype are warranted to validate these associations.
One of the strengths of this study is that it was conducted between 2014-2020, where the treatment landscape for EGFR-mutated NSCLC had dramatically changed, spanning the approval of multiple first-, second-, and third-generation EGFR-TKIs, and allowing for selection of a relatively large cohort with a long follow-up to assess the prognostic impact TP53 co-mutations in this setting. Despite the large sample size and directional consistency in survival endpoints and response, as well as subgroup analysis, the relatively small numbers and limited follow-up among some subgroups may have (1) underestimated the number of patients (<50% in the current dataset) who ultimately did receive 2 L treatment, and (2) limited statistical power to detect significant differences. For instance, as FDA approval of osimertinib in the frontline setting was in April 2018 [55], findings from subgroup analysis by third-generation EGFR-TKIs had limited follow-up (less than 2 years) and should be validated in future studies as cohorts of patients treated with 1 L osimertinib mature in routine practice.
TP53 co-mutation status did not appear to influence treatment selection of therapy for EGFR-mutated aNSCLC, likely because there are no approved therapies targeting TP53 mutations. Although treatments targeting TP53 co-mutations remain under investigation, our findings may have therapeutic implications given the expansion of precision oncology and increased adoption of next generation sequencing and multi-gene panels [7,8,56], resulting in increased availability of TP53 status in this EGFR-mutated aNSCLC patient population. Specifically, these findings strengthen the rationale for clinicians to consider available or emerging EGFR-TKI combination strategies with other therapies such as antiangiogenics to delay resistance and prolong duration of targeted therapy for EGFR-mutated aNSCLC. Notably, in contrast to the body of evidence [28,[30][31][32][33]48], including our findings, showing attenuated benefit with EGFR-TKI monotherapy in patients with EGFR and TP53 comutated aNSCLC, a biomarker and subgroup analysis of the RELAY trial demonstrated that the addition of ramucirumab (anti-VEGF receptor inhibitor) to erlotinib (first-generation EGFR-TKI) was associated with prolonged PFS for EGFR and TP53 co-mutated aNSCLC, both in patients with EGFR exon 19 deletions or exon 21L858R mutations [57,58]. TP53 mutations have also been linked to VEGF expression across different solid tumor cancers therefore increasing the sensitivity of these tumors to anti-VEGF therapies [59]. While promising [57,60], however, the efficacy of EGFR-TKI combinations in EGFR and TP53 co-mutated NSCLC is not established and needs further study in larger datasets.
The limitations of this study are inherent in the nature of the analyzed data, collected during routine care mainly for clinical documentation, management, and billing, and not for research. The potential for bias exists from: (1) misclassification due to data entry and reporting errors in the EMR and use of database algorithms (e.g., line of therapy, adherence to oral EGFR-TKIs), (2) missing or incomplete data (e.g., ECOG performance status, comorbidities, tumor response to therapy), and (3) unmeasured or unknown confounding bias despite the use of multivariable regression to control for measured prognostic factors. How-ever, there is no reason to believe that the validity of the dataset would be systematically biased according to TP53 co-mutation status, although the association in this study should not imply causation. In addition, this study did not capture information on different types of TP53 co-mutations and only TP53 co-mutation status at baseline was considered. This is because the information on TP53 mutation subtypes was not collected or available in the Flatiron database. As earlier reports suggest that mutations in TP53 may induce resistance through different mechanisms in the cancer cell [61], further classification of distinct TP53 mutation subtypes (e.g., disruptive vs. non-disruptive, missense vs. non-missense, at primary vs. metastatic lesions), both alone and in combination with other EGFR mutation subtypes (e.g., EGFR exon 20 T790M, the most common mechanism of resistance for 1st and 2nd generation EGFR-TKIs), may predict more precisely the impact of TP53 co-mutation on prognosis and survival on EGFR-TKIs.

Conclusions
This large observational study confirms and strengthens the available evidence demonstrating that a TP53 co-mutation identifies a subset of EGFR-mutated aNSCLC patients with worse outcomes on EGFR-TKI monotherapy as compared to those with TP53 wild-type EGFR-mutated aNSCLC who might benefit from more intensified treatment schedules. Tailoring the treatment to individual patient characteristics (e.g., EGFR mutation subtype and co-occurring mutations) at baseline might be helpful to further improve treatment outcomes for patients with EGFR-mutated aNSCLC.

Institutional Review Board Statement:
The study was exempt from institutional review board (IRB) approval because it was a retrospective study and used de-identified data in compliance with the Health Insurance Portability and Accountability Act.

Informed Consent Statement:
Approval from WCG IRB for data collection for this real-world cohort was obtained prior to study conduct and included a waiver of informed consent.

Data Availability Statement:
The data that support the findings of this study have been originated by Flatiron Health, Inc. and Foundation Medicine, Inc. These de-identified data may be made available upon request and are subject to a license agreement with Flatiron Health and Foundation Medicine; interested researchers should contact cgdb-fmi@flatiron.com and dataaccess@flatiron.com to determine licensing terms.