Sex- and Age-Associated Differences in Genomic Alterations among Patients with Advanced Non-Small Cell Lung Cancer (NSCLC)

Simple Summary Lung cancer is the leading cause of cancer-related death. While sex and age impact outcomes in non-small cell lung cancer (NSCLC), it is not clearly understood how these factors affect tumor biology. We believe that sex and age influence the distribution of genomic alterations in NSCLC and evaluated for differences in predictive and/or prognostic alterations in individuals with advanced NSCLC. Our study is the largest, to our knowledge, to evaluate and confirm that the genomic landscape in advanced NSCLC differs by both sex and age. Abstract Genomic mutations impact non-small cell lung cancer (NSCLC) biology. The influence of sex and age on the distribution of these alterations is unclear. We analyzed circulating-tumor DNA from individuals with advanced NSCLC from March 2018 to October 2020. EGFR, KRAS, ALK, ROS1, BRAF, NTRK, ERBB2, RET, MET, PIK3CA, STK11, and TP53 alterations were assessed. We evaluated the differences by sex and age (<70 and ≥70) using Fisher’s exact test. Of the 34,277 samples, 30,790 (89.83%) had a detectable mutation and 19,923 (58.12%) had an alteration of interest. The median age of the ctDNA positive population was 69 (18–102), 16,756 (54.42%) were female, and 28,835 (93.65%) had adenocarcinoma. Females had more alterations in all the assessed EGFR mutations, KRAS G12C, and ERBB2 ex20 ins. Males had higher numbers of MET amp and alterations in STK11 and TP53. Patients <70 years were more likely to have alterations in EGFR exon 19 del/exon 20 ins/T790M, KRAS G12C/D, ALK, ROS1, BRAF V600E, ERBB2 Ex20ins, MET amp, STK11, and TP53. Individuals ≥70 years were more likely to have alterations in EGFR L861Q, MET exon 14 skipping, and PIK3CA. We provided evidence of sex- and age-associated differences in the distribution of genomic alterations in individuals with advanced NSCLC.


Introduction
Lung cancer is the third most common cancer in the U.S., after breast and prostate, with an incidence of 50.3 per 100,000 women and 64.1 per 100,000 men [1][2][3].The median age at diagnosis is 71 years, and non-small cell lung cancer (NSCLC) is uncommon among younger individuals (<50 years) [2][3][4].Lung cancer is also the leading cause of death among cancer patients.The estimated five-year overall survival is 25.4%, with an estimated death rate of 29.3 per 100,000 women and 42.2 per 100,000 men [1][2][3].
Recent advances, including the discovery of predictive and prognostic biomarkers and the development of targeted therapies, have led to improved outcomes for some patients with lung cancer.Genomic testing, particularly on tumor samples, has become an integral part of the management of NSCLC and is considered a standard of care [5].A tumor tissue biopsy offers precise histological and molecular information about the tumor, allowing for an accurate diagnosis and a comprehensive review of histology, mutational status, and tumor microenvironment.This approach is limited, however, by the need for invasive procedures and the potential sampling errors due to insufficient tissue or tumor heterogeneity, where a biopsy may fail to capture the complete genetic diversity of the tumor [5].The concept of a "liquid biopsy" dates back to the 1940s when the molecules released by the primary tumor or metastatic lesions into the peripheral blood were first detected.Modern technologies have allowed us to detect small amounts of tumor in the blood [6].Next-generation sequencing (NGS) assays are currently utilized to assess circulating-tumor DNA (ctDNA) in clinical practice.ctDNA is particularly helpful when the tumor is difficult to access safely or when repeat biopsies provide inadequate tissue for evaluation and diagnostic testing [6].ctDNA is more representative of the mutational landscape of the cancer as a whole compared to tissue alone.The testing is less invasive and can be repeated to assess changes in the tumor mutational landscape over time, i.e., real-time monitoring [6].Limitations of liquid biopsy include its inability to detect some rearrangements/fusions and to distinguish tumor alterations from contaminant leukocyte DNA.Some of the alterations detected may represent clonal hematopoiesis [6].Detection of rearrangements and fusions using circulating tumor RNA (ctRNA) is thought to be more sensitive than ctDNA.The widespread use of ctRNA has been challenging due to its instability and rapid degradation by ribonucleases and immune cells.ctRNA can be broken down even before it enters circulation [7].ctRNA may be utilized in conjunction with ctDNA in the future to enhance the detection of targetable alterations [8].
ctDNA is often utilized in NSCLC to detect predictive and/or prognostic mutations that help guide therapeutic decisions [6,9].The current understanding of these predictive and/or prognostic mutations has improved with more frequent genomic testing and advanced sequencing technologies, but little is known about the factors that influence the distribution of these mutations.Sex and age are thought to impact the genomic landscape and clinical outcomes in patients with NSCLC [4,10,11].Women with lung cancer have improved median overall survival (OS) compared to men with similar disease and treatment responses, while individuals ≤40 or >70 years tend to have a worse prognosis [4,12,13].Most of the previously published studies evaluating sex and age differences were conducted using a limited mutation panel and often included a small number of patients [4,10,11].
We conducted a comprehensive analysis to examine how sex and age influence the genomic profiles of advanced NSCLC using one of the largest real-world datasets.We believe that with further characterization of these differences, we can better stratify those with mutations that portend worse outcomes and/or predict response to therapy and improve current treatment strategies.

Materials and Methods
We conducted a retrospective analysis utilizing de-identified data from the Guardant Health database (Palo Alto, CA, USA) from 1 March 2018 to 26 October 2020.Genomic profiles from patients with advanced NSCLC (stages IIIB and higher), who underwent molecular profiling using the plasma-based ctDNA NGS assay Guardant360, were included for the initial review.Single nucleotide variants (SNV), fusions, indels, and copy number variations (CNV) of up to 83 genes were analyzed.We focused on clinically relevant genomic alterations.Synonymous mutations, variants of undetermined significance (VUS), and sub-clonal mutations were excluded.While we did not have data regarding prior or current treatment, we included only the first serial sample in our analysis to limit the effects that any prior treatment may have on the genomic profiles and to capture newly diagnosed NSCLC patients.To minimize confounding factors that impact the accuracy of liquid biopsy (e.g., low tumor burden), samples without detectable alterations were excluded from the analysis [14,15].We focused on alterations with known prognostic and/or predictive significance ("alterations of interest") including EGFR exon 19 deletion (ex19 del), EGFR exon 20 insertion (ex20 ins), EGFR G719X, EGFR L858R, EGFR T790M, EGFR S7681, EGFR L861Q, KRAS G12C, KRAS G12D, KRAS G12V, ALK fusion, ROS1 fusion, BRAF V600E, NTRK 1 fusion, ERBB2 exon 20 insertion (ex20 ins), RET fusion, MET exon 14 (ex14) skipping, MET amplification (amp) medium, MET amp high, any PIK3CA mutation, any STK11 mutation, and any TP53 mutation.The frequency of these alterations was analyzed according to sex (female and male) and age (<70 and ≥70) using Fisher's exact test.We selected a cut-off of 70 years of age based on the median age of NSCLC and worse outcomes reported in individuals >70 [4,12].A p-value < 0.05 was considered statistically significant.We also assessed the co-occurrence and mutual exclusivity of these alterations using cBioPortal https://www.cbioportal.org,accessed on 9 January 2022 [16,17].

Co-Occurrence of Mutations of Interest
Most individuals (15,093 or 75.76%) had one alteration of interest; however, 4830 (24.24%) had two or more co-existing mutations of interest (Supplemental Figure S2).The individual mutation profiles are depicted in the Oncoprint (Supplemental Figure S3).The heatmap demonstrates co-occurring alterations among those with an alteration of interest (Figure 2).We found KRAS and STK11 alterations often co-occurred (p < 0.001).The other alterations tended to occur in isolation (Supplemental Table S1).

Discussion
Outcomes in patients with advanced NSCLC have improved in recent years due to the identification of targetable alterations and the development of novel therapies [2, 3,18,19].While 71% of patients with NSCLC have actionable alterations (e.g., EGFR, ALK, BRAF, ERBB2, MET, ROS1, RET, and KRAS), some have non-actionable mutations with prognostic and/or predictive implications (e.g., PIK3CA, STK11, and TP53) [20][21][22].It is important that we understand the factors that influence the mutational landscape of NSCLC because these alterations play a critical role in therapeutic decision-making and affect outcomes.Smaller studies have suggested that sex and age impact the distribution of these alterations [4,10,11].To our knowledge, our study represents the largest to date to investigate and provide evidence of sex-and age-associated differences in patients with advanced NSCLC.Clinicians can personalize treatment with better understanding of how these differences impact tumor biology.
There are conflicting data regarding the distribution of KRAS mutations in NSCLC.While some studies suggest that KRAS is more common in younger women, others have failed to demonstrate these differences [4,28,[30][31][32][33][34][35].In our study, alterations in KRAS G12C are more frequent in females and in younger individuals, while KRAS G12D mutations are more common in those <70 with no sex differences.We did not find differences in the distribution of KRAS G12V by sex or age.
As reported and confirmed in our study, ALK fusions are more common among young patients.Data regarding sex differences are mixed, and we found no difference by sex [4,10,[36][37][38][39][40][41][42].ROS1 fusions are reported more often among younger individuals and females [10,36,43].In our study, we did not find any significant difference in the distribution by sex, but we confirmed that ROS1 fusions are significantly more common in younger individuals.Studies have suggested that BRAF mutations are more common among females without a clear association with age [4,44,45].While we found no significant difference by sex in BRAF V600E mutations, we did note an association with age <70.
Given the rarity of NTRK fusions, sex and age differences are unknown [46][47][48].We did not observe statistically significant differences by sex or age either, though the sample size was small.ERBB2 ex20 insertions are reported more frequently among females and younger individuals, and our findings are consistent with these reports [4,49,50].Studies have failed to demonstrate sex differences in individuals with RET fusions, but a trend towards younger age has been reported [51,52].We found no statistically significant differences by sex or age.
MET ex14 skipping mutations are reported more frequently among females and older individuals, while MET amplifications are reported more often in males without an association with age [53][54][55].While we did not find any association with sex, we confirmed that MET ex14 skipping mutations are more frequent in individuals ≥70.In addition, we found that MET amplifications were associated with male sex and younger age.PIK3CA mutations are thought to occur more often in males and older individuals [38,41].While our results confirm these mutations are more common among men, we did not find any differences by age.
Lastly, our findings are consistent with data suggesting that STK11 mutations are more common in males and younger individuals and TP53 mutations are more frequent in males [56][57][58][59][60].We also found an association with TP53 mutations and younger age.
While many of our results are consistent with prior studies, several differences exist.Factors such as inclusion of early-stage disease, variability in the genomic testing (e.g., surgical specimens rather than ctDNA), inconsistency in the age cut-offs used to define older and younger populations, and differences in the populations evaluated (U.S. vs. non-U.S.; single institution vs. multi-center) may account for the discrepancies.
Our study highlights the influence sex and age have on the genomic landscape of NSCLC.We found that females were more likely to have alterations with a known targeted therapy (EGFR, KRAS G12C, ERBB2 ex20), while males were more likely to have mutations in STK11 or TP53 which are associated with poor response to therapy and worse outcomes [61,62].Similarly, we found that younger patients are more likely to have targetable alterations than older individuals, but their tumors are also more likely to harbor poor prognostic mutations, including STK11 and TP53.These differences may impact the tumor biology, treatment, and outcomes for individuals with NSCLC and help explain why males and individuals ≤40 or >70 tend to have worse prognosis [4,[10][11][12].As we work to improve upon existing treatments, it is important to consider the co-occurrence of these alterations to predict response and/or to develop novel combination strategies.In general, these mutations are thought to be mutually exclusive [63].While we confirmed that most of the evaluated alterations occur in isolation, we found that STK11 and KRAS were likely to co-occur.This has been described previously and is associated with worse clinical outcomes and poor response to immunotherapy [61,62].
We acknowledge several limitations of our study.Race and tobacco use are known to influence the genomic landscape of NSCLC [64,65].For example, EGFR L858R is three times more common in Asian individuals, while KRAS G12C is most common among white individuals [64].Tobacco use is more common among males and individuals <70) [65].Tumor mutational burden increases with tobacco use, while driver mutations are more common in never-smokers [66,67].Unfortunately, we did not have access to this demographic information and could not assess how race and tobacco use influenced the genomic landscape of our study population.In addition, information regarding the specific histology was limited and clinical outcome data were not available.While we attempted to identify patients with newly diagnosed advanced NSCLC and minimize the effect of prior treatment on the genomic profile by using only the first serial sample, we were unable to distinguish the mutational landscape differences between treatment naïve individuals and those who received therapy.We recognize that prior treatment may have influenced the tumor mutational landscape as a mechanism of resistance to therapy.Finally, cross comparison of our study to others in the literature was challenging given the different age cut-offs used to define older and younger adult populations, differences in the tissue testing (plasma, tumor, or other), inclusion of early-stage cancers, and differences in the populations evaluated (e.g., Asian or European vs. North American; single institution vs. multi-center).

Conclusions
To our knowledge, this is one of the largest studies to date evaluating the genomic landscape differences in individuals with advanced NSCLC by sex and age.We demonstrated significant differences in the distribution of predictive and/or prognostic alterations according to sex and age.This could explain differences in outcomes in otherwise similar patients (i.e., stage, histology, and performance status).Further research is needed to understand how these mutations interact with one another, affect response to therapy, and how they can be used to expand on existing therapies to improve outcomes.The influence that race, tobacco use, and other environmental stressors have on the distribution of these alterations by sex and age should also be evaluated.

Figure 1 .Figure 1 .
Figure1.Study Population.The basic study design and population are depicted above.We evaluated the individual profiles of patients with advanced NSCLC who underwent molecular profiling using ctDNA from the Guardant Health database.Individuals without an alteration detected were excluded.Among those with an alteration, we focused on individuals with variants of predictive or prognostic significance including EGFR exon 19 deletion, EGFR exon 20 insertion, EGFR G719X,

Figure 2 .
Figure 2. Heatmap of Alterations of Interest.This heatmap shows the co-occurring mutation profiles for patients with an alteration of interest (n = 19,923).

Table 2 .
ctDNA+ population with alterations of interest according to sex.
Differences in the genomic alterations by sex in the ctDNA-positive population are depicted above.A p-value < 0.05 indicates statistical significance.Abbreviations: ctDNA, circulating tumor DNA.

Table 3 .
ctDNA+ population with alterations of interest according to age.
Differences in the genomic alterations by age in the ctDNA-positive population are depicted above.Only n = 30,732 instead of n = 30,790 were included in this analysis because of missing age data for 58 patients.A p-value < 0.05 indicates statistical significance.Abbreviations: ctDNA, circulating tumor DNA.

Table 4 .
ctDNA+ population with alterations of interest according to sex and age.
Differences in the genomic alterations by sex and age in the ctDNA-positive population (n = 30,790) are depicted above.Only n = 30,732 were included in this analysis because of missing age data for 58 patients.A p-value < 0.05 indicates statistical significance.Abbreviations: ctDNA, circulating tumor DNA.
Differences in the genomic alterations by sex and age in the ctDNA-positive population (n = 30,790) are depicted above.Only n = 30,732 were included in this analysis because of missing age data for 58 patients.A p-value < 0.05 indicates statistical significance.Abbreviations: ctDNA, circulating tumor DNA.
: Most Common Alterations of Interest.Supplemental Figure S2: Number of Mutations Per Individual with an Alteration of Interest.Supplemental Figure S3: Oncoprint.Supplemental Table S1: Analysis of Co-Occurrence of Alterations of Interest Using cBioPortal.