Lung cancer diagnosed in the young is associated with enrichment for targetable genomic alterations and poor prognosis

Abstract

Importance

NSCLC in the young is a rare entity and the genomics and clinical characteristics of this disease are poorly understood. In contrast, young age at diagnosis has been demonstrated to define unique disease biology in other cancers. Here we report on the association of young age with targetable genomic alterations and prognosis in a large cohort of NSCLC patients.

Objective

To determine the relationship between young age at diagnosis and both the presence of a potentially targetable genomic alteration as well as prognosis and natural history.

Design

All patients with NSCLC genotyped at the Dana-Farber Cancer Institute between 2002–2014 were identified. Tumor genotype, patient characteristics and clinical outcomes were collected. Multivariate logistic regression was used to analyze the relationship between age and mutation status. Multivariate Cox proportional hazard models were fitted for survival analysis.

Setting

A National Cancer Institute (NCI) designated comprehensive cancer center.

Participants

All patients with NSCLC seen at the Dana-Farber Cancer Institute between 2002–2014 who underwent tumor genotyping.

Main Outcome Measure

The frequency of targetable genomic alterations by defined age categories as well as the association of these age groups with survival.

Results

2237 patients with NSCLC were studied. EGFR (p=0.02) and ALK (P<0.01) were associated with younger age, and a similar trend existed for HER2 (p=0.15) and ROS1 (p=0.1) but not BRAF V600E (p=0.43). Amongst patients tested for all 5 targetable genomic alterations, younger age was associated with an increased frequency of a targetable genotype (p<0.01). Those diagnosed at age 50 or younger have a 59% increased likelihood of harboring a targetable genotype. While presence of a potentially targetable genomic alteration treated with a targeted agent was associated with improved survival, the youngest and oldest age groupings had similarly poor outcomes even when a targetable genotype was present.

Conclusion & Relevance

Younger age is associated with an increased likelihood of harboring a targetable genotype and is an underappreciated clinical biomarker in NSCLC. The survival of young NSCLC patients is unexpectedly poor compared to other age groups, suggesting more aggressive disease biology. These findings underscore the importance of comprehensive genotyping including NGS in younger patients with lung cancer.

Introduction

Non-small cell lung cancer (NSCLC) is increasingly understood as a heterogeneous disease, both in its clinical presentation and its genomic make-up. While old age is a commonly considered factor in selecting treatment, young age is a relatively minor factor when planning patient care in NSCLC.^1,2 This is in contrast to a number of cancers where young age at diagnosis is understood to represent a distinct disease biology. For instance, breast cancer occurring in young individuals has been associated with both an increased frequency of BRCA1/2 mutations as well as an increased likelihood of HER2 overexpression and triple-negative disease.^3–5 Breast cancer in younger patients has also been demonstrated to exhibit a more aggressive disease biology and has been associated with higher mortality even after controlling for the effect of genetics and therapy.^6–8 Colon cancer in the young has been associated with higher rates of microsatellite instability and similarly demonstrates more aggressive disease biology.^9–12 Lastly, acute lymphoblastic leukemia in older adolescents and young adults exhibits a poorer prognosis than pediatric patients but better outcomes than older age groups and may benefit from more aggressive treatment regimens.^13–17

In contrast, NSCLC occurring in the young is a poorly studied clinical entity. The median age at diagnosis of NSCLC is 70 years of age and less than 5% of patients are younger than 50 at diagnosis.¹⁸ Recent data have suggested that ALK and ROS1-rearranged lung cancers are associated with a younger age at diagnosis.^19–22 However, these lung cancers represent only a small proportion of all NSCLC and are only two of a larger number of potentially targetable genotypes in lung cancer.^23–25 A higher incidence of EGFR mutations among young patients with advanced NSCLC has also been suggested by small retrospective studies.^21,22,26 More recent studies of the incidence of EGFR mutations and ALK rearrangements have suggested that age may not be as significant of a predictor of mutation status as previously surmised.^27–29 However, studying the relationship between age and genotype in young patients is challenging given the presence of multiple confounding factors including smoking status, sex and race.^{19,20,28–33} The relative rarity of young NSCLC patients in the aforementioned studies and the low incidence of many of these targetable genomic alterations further complicates the study of this association.¹⁸

On the basis of these studies, we hypothesized that young age at diagnosis would define a population of NSCLC patients enriched with targetable genomic alterations. The aim of this study was thus to both examine the relationship between young age, targetable genotype and prognosis as well as to establish a definition for young age which delineates this genetically and potentially biologically unique population.

Methods

The cohort for analysis was identified from an institutional database of NSCLC patients undergoing routine clinical tumor genotyping between January 1^st, 2002 and January 1^st, 2014. Patients were eligible if they consented to allow their clinical information to be used in retrospective research studies on an institutional review board (IRB)-approved protocol (Dana-Farber/Harvard Cancer Center protocol #02–180), or if they were deceased and data was made available on an IRB-approved waiver of consent. The database was queried for information on age, sex, race, smoking status, date of diagnosis, histology, stage, and date of death. Race was included given known associations between tumor genetics and patient race.

Tumor genotyping was performed per the standard clinical practice at our institutions as previously described.³⁴ Briefly, testing was performed on formalin-fixed paraffin-embedded (FFPE) biopsy samples which were pre-screened by board-certified pathologists. Specific genotyping methods utilized included validated PCR-based assays for EGFR and KRAS mutations, bidirectional Sanger sequencing of EGFR, KRAS, HER2, and BRAF, break-apart FISH and IHC assays for ALK and ROS1, and more recently a targeted next-generation sequencing (NGS) assay including each of these genes.^34,35 Standard genetic testing was performed on all patients samples, however, not all assays were able to be performed on each sample due to tissue availability. As a general practice, genotyping was performed on all patients with non-squamous NSCLC and with sufficient tissue for testing, either as part of routine clinical care or an institutional genomic analysis protocol. Patients with squamous histology that underwent clinical genotyping based on unique clinical or pathological characteristics suggesting a possible mixed histology were also included in the study.

This analysis focused on targetable genotypes for which approved targeted therapies exist or where compelling clinical trial data suggests the potential for targeted therapy. For this study, targetable genomic alterations were defined as EGFR mutations in the kinase domain,³⁰ ALK rearrangements,³⁶ ROS1 rearrangements,³⁷ HER2 mutations in the kinase domain,³⁸ and BRAF V600E mutations.³⁹ As a comparison, oncogenic mutations in KRAS were also studied as a genotype which was not targetable during the study period.⁴⁰ Other rare genotypes were studied more irregularly during the study period (e.g. RET rearrangements) or had inadequate data regarding targetability (e.g. PIK3CA mutations, non-V600E BRAF mutations) and were therefore not included in this analysis.^33,41

Statistical analysis

EGFR, KRAS, ALK, ROS1, HER2 and BRAF V600E are exclusive genotypes that have been previously demonstrated to not overlap.^29,42 As such, patients with missing results for some genotypes but possessing a known driver genomic alteration in one of these 6 genes had their missing values coded to be wildtype. Chi-square tests were used to test for association between categorical variables. Multivariate logistic regression models were fitted to determine if age was an independent predictor of mutation using backwards stepwise regression and including age as a continuous variable, never smoking status, female sex, adenocarcinoma histology, Asian race and stage in the initial models.

A composite variable using the results of EGFR, ALK, ROS1, HER2 and BRAF genotyping was used, such that anyone with an abnormality of interest in any of these 5 genes was considered to be positive for a targetable genotype; anyone with known wildtype status across all 5 genes was considered negative for this composite variable; and patients negative for some of the genes but with incomplete genotyping for others were excluded from analysis of the composite variable. Overall survival was defined as the time from date of diagnosis of metastatic disease to date of death from any cause, and patients thought to remain alive at the time of analysis have been censored at their last follow-up date. Event time distributions were estimated using the method of Kaplan and Meier, and multivariate Cox models were used to estimate adjusted hazard ratios. The logrank test was used to compare event-time distributions. All p-values are two-sided and significance was defined at the 0.05-level; no adjustments have been made for multiple comparisons. Age categories utilized in this analysis were <40, 40–49, 50–59, 60–69 and ≥70 years of age. Age cut-point analysis for patients possessing a targetable genomic alteration based on the previously defined composite endpoint was performed by recursive partitioning.

Results

Patient characteristics

A total of 2237 eligible patients with NSCLC and tumor genotyping results were identified (Table 1). The majority of patients had adenocarcinoma histology (87%) or NSCLC NOS (12%) and only a small minority had squamous histology (1%). Most patients were stage IIIB/IV (63%). There was a slight preponderance of females (62%). A significant minority of patients in this institutional experience were never-smokers (27%). Patients were primarily Caucasian (90%); the remainder was Asian (5%), Black (4%), or Hispanic (1%).

Table 1.

patient demographic information and genotype.

		Total		<40		40–49		50–59		60–69		70+		p-value
Patients		2237	100%	81	4%	252	11%	597	27%	597	31%	610	27%
Sex	Male	858	38%	28	35%	81	32%	214	36%	262	38%	273	45%	0.002
	Female	1379	62%	53	65%	171	68%	383	64%	435	62%	337	55%
Race	White	1975	90%	59	76%	211	85%	521	89%	626	92%	558	93%	<0.001
	Asian	98	4%	8	10%	18	7%	28	5%	29	4%	15	2%
	Black	87	4%	5	6%	15	6%	31	5%	17	3%	19	3%
	Hispanic	31	1%	6	8%	4	2%	6	1%	6	1%	9	1%
Smoker	Never	594	27%	54	67%	106	43%	165	28%	138	20%	131	22%	<0.001
	Light	192	9%	6	7%	29	12%	57	10%	47	8%	53	9%
	Heavy	1399	63%	21	26%	110	44%	365	61%	495	83%	408	67%
Stage at diagnosis	I	376	17%	8	10%	33	13%	94	16%	125	18%	116	19%	<0.001
	II	175	8%	3	4%	20	8%	35	6%	64	9%	53	9%
	IIIA	288	13%	7	9%	25	10%	82	14%	99	14%	75	12%
	IIIB	133	6%	10	12%	22	9%	29	5%	41	6%	31	5%
	IV	1263	57%	53	65%	152	60%	357	60%	367	53%	334	55%
Histology	Adeno	1939	87%	68	84%	212	84%	518	87%	630	90%	511	84%	<0.001
	Squamous	29	1%	3	4%	3	1%	7	1%	6	1%	10	2%
	NSCLC NOS	269	12%	10	12%	37	15%	72	12%	61	9%	89	14%
Mutation Status†
EGFR Mutation	Mutant	540	26%	25	32%	80	34%	149	26%	157	24%	129	23%	0.02
	Non- mutant	1538	74%	53	68%	155	66%	418	74%	487	76%	425	77%
ALK	Mutant	84	5%	13	19%	26	13%	25	5%	16	3%	4	1%	<0.001
	Non- mutant	1699	95%	55	81%	177	87%	466	95%	566	97%	435	99%
HER2 Exon 20	Mutant	44	3%	1	2%	10	5%	13	3%	10	2%	10	2%	0.15
	Non- mutant	1630	97%	62	98%	182	95%	451	97%	536	98%	399	98%
ROS1	Mutant	20	1%	3	6%	3	2%	5	1%	4	1%	5	1%	0.1
	Non- mutant	1348	99%	51	94%	151	98%	371	99%	436	99%	339	99%
BRAF V600E	Mutant	24	1%	0	0%	1	1%	6	1%	8	1%	9	2%	0.43
	Non- mutant	1671	99%	62	100%	194	99%	465	99%	545	99%	405	98%
KRAS Mutation	Mutant	493	27%	6	9%	28	13%	131	26%	197	34%	131	27%	<0.001
	Non- mutant	1353	73%	60	91%	182	87%	373	74%	390	66%	348	73%

^†Patients not tested for a given gene or with indeterminate results excluded from each sub-category.

The median age of patients included in this study was 62 years (20–95). Younger age was associated with increased likelihood of being a never smoker (p<0.01), female sex (p<0.01) and initial presentation with stage IV disease (p<0.01). 81 patients (4%) were diagnosed with NSCLC at age 40 or younger (Table 1). These patients were more likely to be never smokers (66%, p<0.01), female (65%, p<0.01) and present with stage IV disease (77%, p<0.01) compared to older patients.

Association between genotype and age

Across the entire patient cohort (n=2237, Table 1), 32% of patients possessed a targetable genomic alteration (EGFR kinase mutation, ALK or ROS1 rearrangement, HER2 kinase mutation, BRAF V600E). EGFR kinase mutations (p=0.02) and ALK rearrangements (p<0.01) in particular were both associated with an increased likelihood in patients diagnosed with NSCLC at a younger age on univariable analysis (Figure 1, Table 1). In contrast, the likelihood of harboring a KRAS mutation decreased with younger age at diagnosis (p<0.01) (Figure 1, eFigure 1). No significant association with age at NSCLC diagnosis was found among the three rarest targetable genotypes (ROS1, HER2, BRAF V600E), though a non-significant trend towards younger age at diagnosis was seen for ROS1 (p=0.10) and HER2 (p=0.15) (Figure 1, Table 1).

Figure 1.

The frequency of potentially targetable genomic alterations is highest among young patients and steadily decreases with age (A) whereas KRAS exhibits the opposite trend. Complete genotyping refers to testing for all 5 targetable alteration. Association with younger age is apparent when each individual alteration is considered including EGFR mutations (B), ALK rearrangements (C), ROS1 rearrangements (D) and HER2 exon 20 insertions (E) whereas BRAF V600E appears more common among older patients (F). Mt: mutation wt: wild-type.

Studying the composite variable of presence of a targetable alteration, the likelihood of harboring such an alteration increased with younger age at diagnosis in patients who had undergone testing for all 5 alterations (n=1325, p=<0.01) (Figure 1, eTable 1). A multivariate analysis of the relationship between age and presence of a targetable genotype (composite variable) was then performed. As expected, smoking status (p<0.01), female sex (p=0.02) and Asian race (p=0.03) were significantly associated with presence of a targetable genotype. Controlling for these factors, this model maintained a significant association between age and the presence of a targetable genotype (p<0.01) (eTable 1). Examining the individual targetable mutations using this multivariate model, the only genotype associated with a younger age at diagnosis was ALK rearrangement (p<0.01) whereas the association between EGFR mutations and age was no longer significant (p=0.28). In this multivariate analysis, KRAS mutations (p=0.04) and, unexpectedly, BRAF V600E mutations (p<0.01) were both associated with older age at diagnosis.

Defining a criteria for young age at NSCLC diagnosis

The frequency of targetable genomic alterations across small age groupings (5 years each) was studied to look for an age cutpoint where the likelihood of harboring a targetable genotype changes dramatically (eFigure 2), and demonstrated an apparent drop in the incidence of targetable genotypes above age 50. We therefore studied whether an age at diagnosis of 50 was a cutpoint that could differentiate young patients with an increased chance of possessing a targetable alteration. Age quartiles were explored in the 1325 patients tested for all 5 targetable genotypes, and this similarly identified that those in the 25^th percentile (≤54 years) had a 46% higher frequency of targetable genotypes compared to those in the upper quartiles. A recursive partitioning model was subsequently fitted to help independently validate the age cutpoint observed in the data. An age cut-point of 50 years was identified using this methodology with 78% of patients younger than 50 years of age harboring a genomic alteration (160/206) compared to 49% of patients 50 years and older (547/1119). Thus, there existed a 59% increased chance of detecting a targetable alteration in a patient younger than 50 compared to an older patient.

Survival analysis

Given the findings that younger patients with advanced NSCLC exhibited a unique biology enriched for targetable genomic alterations, we next sought to determine whether prognosis was similarly unique. Overall survival by pre-defined age categories revealed the lowest median overall survival occurring for patients younger than 40 years of age (18.2 months, 95%CI: 13.6–25.7) and those older than 70 (13.6 months, 95%CI: 11.4–15.7) compared to those 40–49 years old (22.9 months, 95%CI: 19.1–28.3), 50–59 years old (21.3 months, 95%CI: 18.5–25.1) and 60–69 years old (18.3 months, 95% CI: 16.5–20.6) (p=<0.01). Under the hypothesis that targetable genotype may be associated with survival,²⁹ the relationship between targetable genotype and overall survival was tested in univariate analysis and found to be significant (p=<0.01, eFigure 3). Overall survival analysis by age category was then explored separately in those patients with and without targetable genotypes (EGFR, ALK, ROS1, HER2 or BRAF V600E). In patients harboring a targetable genotype, a significant difference was found between age categories (p<0.01) (Figure 2A). In this subset analysis, lowest median overall survival was found in patients 40 years of age or younger (21.4 months, 95%CI: 13.6–47.3) and those older than 70 (22.3 months, 95%CI: 16.9–28.6). The longest median overall survival occurred among those 50–59 years of age (35.4 months, 95%CI: 29.6–41.4). In contrast, no difference in overall survival was found between age categories in patients that did not harbor a targetable genotype (p=0.41) (Figure 2B).

Figure 2.

The survival of patients possessing a targetable alteration is lowest among patients >70 and those <40 as demonstrated by these KM plots (A). All other age groups that possess a targetable alteration have significantly improved survival compared to the oldest age group (A) (p<0.01). There is no significant difference in survival between age groups among patients that do not possess a potentially targetable alteration (B, p=0.41).

A survival analysis was then performed by Cox-regression examining the effect of age on survival and controlling for the presence of a targetable genomic alteration, use of targeted therapy, metastatic disease at diagnosis, sex, adenocarcinoma histology, presence of brain metastases and the year of metastatic disease. This revealed that the oldest age category had a significantly poorer chance of survival compared to all age categories except the youngest; but there was no significant difference in overall survival between the oldest and youngest age group (HR 0.73 95% CI 0.32–1.67, p=0.46) (eTable 2). Presence of a targetable genomic alteration that received targeted therapy was associated with improved survival in this model (HR 0.66, p<0.01) whereas a targetable alteration that did not receive targeted therapy was not (HR1.06, p=0.71). Metastatic disease at diagnosis (HR 1.48, p<0.01) and the presence of brain metastases were associated with poorer survival (HR 1.25, p<0.01).

Discussion

This study tested the hypothesis that young age at diagnosis is an under-appreciated marker of disease biology in NSCLC. Studying 5 targetable genotypes across 1325 patients, we identified on univariate and multivariate analysis that younger age at diagnosis was significantly associated with the presence of a targetable alteration. Specifically, patients diagnosed under the age of 50 have a 59% increased likelihood of harboring a targetable genotype. This supports the proposition that NSCLC in the young may represent a biologically distinct subgroup of lung cancer that is enriched for targetable genomic alterations. Further, it suggests that age 50 constitutes a useful age cut-off by which NSCLC in the young may be defined. This finding is consistent with previous studies demonstrating an association between young age and specific targetable genomic alterations including EGFR, ALK and ROS1.^19–22,26 The findings of this study are particularly notable given recent studies that have suggested that this association may be more tenuous than previously demonstrated. It is important to note that these studies have not directly addressed the frequency of such alterations by age group given the rarity of young patients with NSCLC.^27–29

The suggestion that younger patients with NSCLC represent a genetically unique subgroup has potential implications for the treatment of lung cancer. As the list of rare but potentially targetable genomic alterations increases, the potential cost and complexity of comprehensive genetic testing will continue to grow in parallel.⁴³ A recent study by Drilon et al examined the utility of comprehensive genotyping with targeted NGS among patients with advanced lung adenocarcinoma and a light smoking history – another population enriched for targetable genomic alterations. This study demonstrated that comprehensive genomic profiling utilizing NGS could detect potentially targetable alterations missed by standard genotyping in 65% of these patients including 26% where a targeted agent based on NCCN guidelines was readily available..⁴⁴ This highlights that clinical characteristics can be used to better apply a tool like tumor NGS for maximal utility. Given our finding that patients diagnosed at a younger age are similarly enriched for targetable genotypes, one could advocate for age at diagnosis as an appropriate clinical characteristic to consider when determining the potential utility of comprehensive genotyping for a NSCLC patient. We are not suggesting substandard genotyping in older patients, but rather are advocating for more aggressive efforts to routinely search for rare and even potentially novel targetable alterations in the young. Because not all patients have access to comprehensive tumor genotyping, a prospective study to test the utility of targeted NGS for young lung cancer (age of diagnosis under 40) has been initiated through a collaboration between the Addario Lung Cancer Medical Institute and Foundation Medicine (NCT02273336).

The realization that NSCLC in the young is a genetically unique disease naturally lends itself to the question of whether the natural history and underlying disease biology of NSCLC is also distinct in this subgroup. Our study has found evidence that the prognosis of metastatic patients 40 years and younger is no better than those patients older than 70 years of age whereas other age categories exhibit improved prognosis compared to the oldest age group. This association was found when the presence of a potentially targetable genomic alteration was controlled, which itself was associated with an improved prognosis as demonstrated in previous studies.³² This finding would not normally be expected if disease biology was similar across age groups even when the small size of this patient group is considered – particularly given the markedly lower rate of co-morbidities and functional impairment in younger patients.^1,2 The poor survival associated with NSCLC in this age group potentially suggests that disease biology is distinctly aggressive in the youngest patients and may support the role of young age as a marker of lung cancer biology. Other factors including late diagnosis, disease awareness and financial challenges obtaining optimal medical care amongst young patients represent potential contributors or possibly alternative explanations for this phenomenon. Although the rarity of young patients with NSCLC limits makes such analyses challenging, further examination of the prognosis of young patients with NSCLC in larger cohorts of patients is warranted in order to better understand this phenomenon.

Although the focus of this study was young age, the incidental findings of an association between older age and BRAF V600E alterations is noteworthy. Although BRAF V600E is clearly targetable in melanoma,⁴⁵ the use of BRAF inhibitors has exhibited mixed results in cancers of the colon and lung.³⁹ Preliminary results of a phase II trial of dabrafenib in advanced NSCLC possessing a BRAF V600E mutation demonstrated a modest overall response rate (ORR) of 32%, in contrast to a 53% ORR in melanoma and approximately 70% ORR with EGFR kinase inhibitors in EGFR-mutant lung cancer.^39,45,46 Of note, other non-V600E BRAF mutations may not be targetable at all.³³ These characteristics suggest that BRAF V600E may be less easily targeted in NSCLC compared to other rare genotypes like ALK and ROS1.

In interpreting these findings, several limitations inherent in this study must be considered. The data utilized in this study is limited given its retrospective nature and limited comprehensive data on individual patient treatment. The patients included in this study are also drawn from a single institution and thus are subject to referral bias. Notably, the overall rate of targetable genomic alterations was higher in the study population than expected from the general population given the large number of Caucasian patients included in the study. Although genetic testing was performed based on institutional standard of care for all patients, this standard has evolved over time and was limited by the availability of tissue for comprehensive genotyping. Thus, all patients in this study did not undergo identical genotyping and genotyping data was not completely comprehensive for all patients. A composite endpoint examining patients that had undergone comprehensive genetic testing was utilized to account for these differences.

Despite the aforementioned limitations, the findings of this study expand the current understanding of the genetics and biology of lung cancer in the young. These patients possess a uniquely high incidence of targetable genomic alterations which is paired with an unexpectedly poor prognosis. This combination of opportunity and risk defines the treatment of NSCLC in the young and requires unique therapeutic and research strategies. These data suggest that exhaustive genotyping methods such as NGS should be utilized when available for young lung cancer patients in order to detect targetable alterations which may guide therapy.⁴⁴ These methods may also facilitate detection of rare uncharacterized genomic drivers which may become targets in the future. Such an approach will both maximize the chance of these young patients having access to the most appropriate targeted therapies, and provide more comprehensive knowledge of the genomics of lung cancer in the young.

Supplementary Material

Supplemental

eFigure 1. The frequency of KRAS mutations by age category.

eFigure 2. Frequency of targetable genomic alterations by age group among patients where complete testing was performed (EGFR, ALK, ROS1, HER2, BRAF). The frequency of targetable genomic alterations appears to increase below the age of 50. This age was subsequently confirmed by recursive partitioning to represent a distinct age boundary between two populations with unique rates of targetable alterations.

eFigure 3. The survival of patients possessing a targetable alteration compared with those not possessing a targetable alteration. OS is significantly improved in patients possessing a targetable alteration (p<0.01).