Late Recurrence of Completely Resected Stage I-IIIA Lung Adenocarcinoma

Abstract

Objective:

Research into the risk factors associated with late recurrence (>2 years after surgery) of lung adenocarcinoma (LUAD) is limited. We investigated the incidence of and clinicopathologic and genomic features associated with late recurrence of resected stage I-IIIA LUAD.

Methods:

We performed a retrospective analysis of patients with completely resected pathologic stage I-IIIA LUAD (2010–2019). Patients with a history of lung cancer, neoadjuvant therapy, or mucinous or noninvasive LUAD, or with follow-up of <2 years were excluded. Cox and logistic regression modeling were used to compare clinicopathologic variables among patients with no, early (≤2 years), and late recurrence. Comparisons of genomic mutations were corrected for multiple testing.

Results:

Of the 2349 patients included, 537 developed a recurrence during follow-up. Most recurrences (55% [297/537]) occurred early; 45% (240/537) occurred late. A larger proportion of late recurrences than early recurrences were locoregional (37% vs. 29%; p=0.047). Patients with late recurrence had more aggressive pathologic features (IASLC grade 2 and 3, lymphovascular invasion, visceral pleural invasion) and higher stage than patients without recurrence. Pathologic features were similar between patients with early and late recurrence, except stage IIIA disease was more common in the early cohort. No genomic mutations were associated with late recurrence.

Conclusions:

Late recurrence of LUAD following resection is more common than previously reported. Patients without disease >2 years after surgery who had aggressive pathologic features at the time of resection have an elevated risk of recurrence and may benefit from more-aggressive follow-up.

Introduction

Surgery is the mainstay of treatment for patients with early-stage non-small cell lung cancer (NSCLC) as it offers the best chance of cure. However, 5-year overall survival (OS) ranges from 73% to 90% for patients with stage I disease and is less than 50% for patients with locoregional disease.¹ An important driver of poor long-term survival after curative resection is disease recurrence. Most recurrences occur at extrathoracic sites,² and, as a result, metastasis accounts for 90% of cancer-related deaths.³

Recurrence of NSCLC, specifically lung adenocarcinoma (LUAD), is thought to be an early phenomenon; previous studies have reported that up to 80% of recurrences occur within 2 years of resection.^4–6 Thus, the risk factors associated with recurrence are largely driven by these early recurrences, and comparatively little is known about recurrences more than 2 years after surgery, as long-term follow-up is often limited. Research into risk factors for delayed recurrences may have important implications for treatment and follow-up recommendations. Therefore, the aims of this study were to investigate the incidence of late recurrence in patients with completely resected stage I-IIIA LUAD and to identify clinicopathologic and genomic risk factors associated with late recurrence (>2 years after surgery). Additionally, we aimed to compare features associated with late recurrence and features associated with early recurrence (≤2 years).

Methods

After approval from our institutional review board (#18–391; 9/7/2018), we performed a retrospective review of a prospectively maintained database. Patients with pathologic stage I-IIIA LUAD (American Joint Committee on Cancer [AJCC] eighth edition)⁷ who underwent complete resection from January 2010 to December 2019 were included. Patients were excluded if they had a history of lung cancer, neoadjuvant therapy, or mucinous or noninvasive adenocarcinoma histologic subtype (Supplemental Figure 1). Patients without recurrence were required to have at least 2 years of follow-up, with no evidence of disease at their most recent follow-up, to be included. Baseline demographic, clinicopathologic, and operative details were obtained for each patient. Tumors were graded from 1 to 3 in accordance with the 2020 International Association for the Study of Lung Cancer (IASLC) classification.⁸ A subset of the overall cohort had tumor genomic information obtained using hybridization capture-based next generation sequencing.⁹

Patient follow-up was consistent with National Comprehensive Cancer Network (NCCN) standards, with a chest CT every 6 months for 2–3 years, followed by low-dose chest CT annually thereafter.⁷ Recurrence was determined by imaging studies and/or pathology via tissue biopsy. Recurrence was differentiated from a second primary lung tumor using molecular testing and the Martini and Melamed criteria.¹⁰ The first site of recurrence was classified as locoregional (ipsilateral lung/regional lymph nodes, ipsilateral mediastinum), distant, or locoregional plus distant for patients with simultaneous locoregional and distant recurrence.² Distant recurrence included pleural effusions, contralateral lung, and extrathoracic metastases. Distant recurrence identified at >1 site concurrently was classified as multiple.

Time to recurrence was calculated from the date of surgery to the date of biopsy confirming recurrence or imaging findings consistent with recurrence, if tissue was not acquired. Recurrence was dichotomized into early and late using the 2-year mark. We selected 2 years on the basis of the distribution of recurrences in previous studies.^4,11,12 Additionally, NCCN guidelines⁷ dictate follow-up using this time point, with more-intensive radiographic follow-up for the first 2 years postoperatively.

Clinicopathologic features were summarized as frequency (percentage) or median (interquartile range [IQR]) for the overall, no recurrence, early recurrence, and late recurrence cohorts (Table 1). The reverse Kaplan-Meier method estimated median follow-up. Univariable and multivariable Cox models were used to compare clinicopathologic variables between patients with late recurrence and those without recurrence. Following the approach implemented in previous studies,^13,14 patients with early recurrence were excluded from this analysis. Given the heterogenous distribution of stage across the cohorts, univariable models were adjusted for stage. Among patients with recurrence, logistic regression models quantified the association between clinicopathologic features and the odds of late (vs. early) recurrence. Features with p<0.05 in univariable models were included in multivariable models. OS after late recurrence was calculated from the date of recurrence diagnosis to the date of death and was estimated using the Kaplan-Meier approach. Patients were then stratified by site of recurrence (locoregional only vs. any distant [locoregional plus distant and distant only]), and OS was analyzed in a similar fashion. Cox models were used to estimate associations between genomic mutations and late recurrence. Comparisons of genomic mutations across cohorts were adjusted to correct for multiple testing using the false-discovery rate. All statistical tests were two-sided, and p<0.05 was considered statistically significant. Statistical analyses were conducted using R 4.3.1 (R Core Team, Vienna, Austria).

Table 1.

Clinicopathologic Characteristics by Recurrence Status

Characteristic	All Patients (n=2349)	No Recurrence (n=1812)	Early Recurrence (n=297)	Late Recurrence (n=240)
Age at surgery, years	69 (62–74)	69 (62–74)	69 (62–76)	68 (62–74)
Sex
Female	1497 (64)	1175 (65)	171 (58)	151 (63)
Male	852 (36)	637 (35)	126 (42)	89 (37)
Smoker
Ever	1766 (75)	1362 (75)	236 (79)	168 (70)
Never	583 (25)	450 (25)	61 (21)	72 (30)
SUVmax of the primary tumor (n=2066)	3.2 (1.7–6.6)	2.6 (1.5–5.2)	7.2 (4.0–11.1)	4.8 (2.5–7.5)
Surgical approach
Minimally invasive surgery	1854 (79)	1492 (82)	187 (63)	175 (73)
Open	495 (21)	320 (18)	110 (37)	65 (27)
Surgical resection
Wedge	526 (22)	409 (23)	63 (21)	54 (23)
Segmentectomy	312 (13)	258 (14)	29 (9.8)	25 (10)
Lobectomy	1477 (63)	1131 (62)	191 (64)	155 (65)
Bilobectomy	27 (1.1)	13 (<1)	10 (3.4)	4 (1.7)
Pneumonectomy	7 (<1)	1 (<1)	4 (1.3)	2 (<1)
IASLC grade (n=2330)
1	326 (14)	312 (17)	6 (2.1)	8 (3.4)
2	1190 (51)	963 (53)	100 (34)	127 (54)
3	814 (35)	528 (29)	184 (63)	102 (43)
Lymphovascular invasion (n=2307)	830 (36)	513 (29)	189 (65)	128 (54)
Visceral pleural invasion (n=2335)	388 (17)	208 (12)	114 (39)	66 (28)
Pathologic stage
I	1868 (80)	1581 (87)	134 (45)	153 (64)
II	306 (13)	171 (9.4)	80 (27)	55 (23)
IIIA	175 (7.4)	60 (3.3)	83 (28)	32 (13)
Lymph nodes sampled	11 (6–18)	11 (5–18)	11 (6–17)	11 (5–16)
Adjuvant therapy (n=2345)	383 (16)	202 (11)	112 (38)	69 (29)

Data are median (interquartile range) or no. (%).

IASLC=International Association for the Study of Lung Cancer; SUVmax=maximum standardized uptake value.

Results

In total, 2349 patients were included, with a median follow-up of 71 months (IQR, 48–103 months). Most patients were women (64% [1497/2349]); 75% (1766/2349) had a history of smoking (Table 1). The most common operation performed was lobectomy (63% [1477/2349]); 35% of patients (838/2349) underwent sublobar resection. Most patients had stage I disease (80% [1868/2349]). Sixteen percent of patients (383/2345) received adjuvant therapy; the most commonly administered regimen was chemotherapy only (Supplemental Table 1).

Of the overall cohort, 537 patients developed recurrence during follow-up (Table 2). Recurrence was confirmed pathologically in 87% of patients (469/537). Most recurrences (55% [297/537]) occurred early (≤2 years); 45% (240/537) occurred late. Moreover, 25% of late recurrences (59/240) occurred >5 years postoperatively. The rate of recurrence peaked in year 2, but the risk of recurrence remained high in year 3, before gradually declining thereafter (Figure 1). Recurrence rates for stage II and IIIA LUAD peaked early, whereas recurrence rates for stage I disease persisted at a relatively consistent rate throughout the study period (Figure 2). A larger proportion of late recurrences than early recurrences were locoregional or locoregional plus distant (60% [144/240] vs. 49% [147/297]; p=0.047) (Table 2). In the late recurrence cohort, 10% of recurrences (23/240) occurred at the pleura or contralateral lung. Isolated brain metastasis was the first site of recurrence in 13% of late recurrences (31/240).

Table 2.

Location of First Recurrence by Cohort

Site	All Recurrences (n=537)	Early Recurrence (n=297)	Late Recurrence (n=240)	P valuea
Locoregional	174 (32)	86 (29)	88 (37)
Locoregional plus Distant	117 (22)	61 (21)	56 (23)	0.047
Distant	246 (46)	150 (51)	96 (40)
Brain	68 (13)	37 (12)	31 (13)
Bone	43 (8.0)	24 (8.1)	19 (7.9)
Pleura	30 (5.6)	18 (6.1)	12 (5.0)
Contralateral lung	28 (5.2)	17 (5.7)	11 (4.6)
Liver	10 (1.9)	9 (3.0)	1 (<1)
Adrenal	10 (1.9)	7 (2.4)	3 (1.3)
Other	18 (3.4)	14 (4.7)	5 (2.1)
Multiple	38 (7.1)	23 (7.7)	14 (5.8)

Data are no. (%).

^aPearson’s chi-squared test.

Figure 1.

Rates of recurrence after resection among patients with stage I-IIIA lung adenocarcinoma. Recurrence rate expressed as events per 100 person-years.

Figure 2.

Rates of recurrence after resection of stage I-IIIA lung adenocarcinoma stratified by stage. Recurrence rate expressed as events per 100 person-years.

Patients with late recurrence were more likely to have aggressive pathologic features, including IASLC grade 2 and 3 (vs. grade 1), lymphovascular invasion (LVI), and visceral pleural invasion (VPI) compared with patients without recurrence (Table 3). Additionally, higher stage was associated with late recurrence (stage II: hazard ratio [HR], 1.99 [95% confidence interval {CI}, 1.40–2.83]; p<0.001; stage IIIA: HR, 2.81 [95% CI, 1.84–4.29]). Receipt of adjuvant therapy was not associated with late recurrence when stage was adjusted for in the analysis (HR, 1.12 [95% CI, 0.72–1.76]; p=0.6).

Table 3.

Univariable and Multivariable Analyses for Late Recurrence Versus No Recurrence

Variable	Univariable Analysis		Multivariable Analysis
	HR (95% CI)	P value	HR (95% CI)	P value
Age at surgery	1.01 (0.99–1.02)	0.5	—	—
Male sex	1.02 (0.79–1.33)	0.9	—	—
Ever smoker	0.79 (0.60–1.04)	0.087	—	—
SUVmax of the primary tumor	1.04 (1.01–1.06)	0.003	1.01 (0.98–1.04)	0.4
Sublobar resectiona (vs. lobectomy)	1.26 (0.95–1.69)	0.11	—	—
IASLC grade (vs. grade 1)
2	4.80 (2.35–9.82)	<0.001	4.67 (1.89–11.5)	<0.001
3	5.37 (2.59–11.1)	<0.001	4.41 (1.75–11.1)	0.002
Lymphovascular invasion	1.97 (1.49–2.59)	<0.001	1.56 (1.14–2.13)	0.005
Visceral pleural invasion	2.01 (1.50–2.70)	<0.001	1.61 (1.17–2.20)	0.003
Pathologic stage (vs. I)
II	3.03 (2.23–4.13)	<0.001	1.99 (1.40–2.83)	<0.001
IIIA	4.35 (2.97–6.37)	<0.001	2.81 (1.84–4.29)	<0.001
Adjuvant therapy	1.12 (0.72–1.76)	0.6	—	—

Cox regression model. Univariable models adjusted for stage.

CI=confidence interval; HR=hazard ratio; IASLC=International Association for the Study of Lung Cancer; SUVmax=maximum standardized uptake value.

^aSegmentectomy or wedge resection.

Pathologic features were similar between the late and early recurrence cohorts (Supplemental Table 2). However, a history of smoking, higher maximum standardized uptake value of the primary tumor on PET scan, and stage IIIA disease were associated with early recurrence.

In the late recurrence cohort, nearly half of stage I recurrences (72/153) were locoregional only, whereas the majority of stage IIIA recurrences were locoregional plus distant or distant only (87% [28/32]) (Supplemental Table 3). Local therapy (surgery and/or radiation) was the most common treatment modality in the late recurrence cohort (41% [99/240]) (Supplemental Table 4).

Genomic information was available for 33% of patients (780/2349). Mutations present in at least 5% of patients with late recurrence are displayed in Supplemental Figure 2. The most common mutations in the late recurrence cohort were TP53, EGFR, and KRAS. Overall, no genomic mutations were significantly associated with late recurrence.

The median OS after diagnosis of late recurrence was 40 months (95% CI, 34–55 months) (Figure 3A), and 5-year OS was 38% (95% CI, 30%-48%). Five-year OS differed on the basis of recurrence location (Figure 3B): patients with late locoregional recurrence had a 5-year post-recurrence OS of 55% (95% CI, 42%-73%), whereas patients with any distant recurrence had a 5-year post-recurrence OS of 31% (95% CI, 22%-43%) (p=0.033).

Figure 3.

(A) Overall survival after development of late recurrence. (B) Overall survival after development of late recurrence by recurrence location. (CI 95%)

Discussion

Recurrence after complete resection of lung cancer continues to be a burden despite advances in treatment. Many studies have investigated clinical and pathologic features associated with recurrence-free survival (RFS) and OS; however, long-term follow-up is often lacking, leading to an overrepresentation of early recurrences. In this analysis of patients with stage I-IIIA LUAD who underwent complete resection, recurrence more than 2 years after surgery (late recurrence) was more common than previously reported. While prior studies estimated that 80% of recurrences occurred within 2 years of surgery,^4,6 we found that only 55% of recurrences occurred during this time (Figure 4). Aggressive pathologic features, such as poor IASLC grade, LVI, and VPI, and higher pathologic stage, were associated with late recurrence. Interestingly, specific genomic mutations were not associated with late recurrence.

Figure 4.

Late recurrence of completely resected stage I-IIIA lung adenocarcinoma.

NCCN guidelines recommend routine chest CT every 6 months for 2–3 years after surgical resection of lung cancer, followed thereafter by low-dose chest CT annually.⁷ The rationale for this schedule is that most recurrences occur within 2 years of surgical resection.^4,6 In a recent analysis of data from the National Lung Screening Trial, the highest rate of recurrence was in the first year after surgery.¹⁵ However, in this cohort of patients with stage I and II NSCLC (AJCC eighth edition) with a median follow-up of 4 years, 65% of recurrences were identified 1–4 years after surgery. In a similar analysis of early-stage lung cancer, the risk of recurrence was highest in the first 2 years after resection but remained elevated for up to 4 years.¹² Our results mirror these findings, as the majority of our patients (55%) had recurrence within 2 years of resection. Our median follow-up of ~6 years allowed us to capture patients with late recurrences who would have previously been classified as recurrence free with more-limited follow-up. Therefore, although the proportion of patients with late recurrence in our study is higher than previous studies, it is likely a more accurate representation of the distribution of recurrence events that may have been identified in previous studies if follow-up had been longer.

Recurrence in patients with NSCLC poses a particular challenge, as the majority of recurrences occur at extrathoracic sites.^16,17 In a study that investigated the recurrence dynamics of NSCLC after resection, 74% of recurrences in patients with stage I-IIIA disease (AJCC seventh edition) were distant, which peaked at 9 months postoperatively.¹⁷ In our cohort, early recurrences occurred at locoregional plus distant or distant sites with a similar frequency (71%). Another recent study of site-specific recurrences found that recurrences at distant sites peaked early, whereas intrathoracic recurrences increased over time.¹⁸ In our study, the late recurrence cohort had a greater proportion of locoregional recurrences than the early recurrence cohort (37% vs. 29%; p=0.047). This finding has important implications for surveillance and treatment. First, a large proportion of these late recurrences can be identified with surveillance before the onset of symptoms, which has been associated with improved outcomes.¹⁹ Second, these locoregional recurrences are amenable to local therapy – surgery or radiation, which has been associated with prolonged survival.^20,21

Pathologic features previously associated with late recurrence were also identified in our cohort. Maeda and colleagues¹⁴ found that vascular invasion was a significant predictor of late recurrence after complete resection (HR, 2.74; p=0.038), which was confirmed in our results. We also identified VPI as a significant predictor of late recurrence in our population. Similar findings were reported by Neri et al., who identified VPI as an independent predictor of worse 5-year RFS.²²

Pathologic TNM staging has been the backbone of prognosis of NSCLC, and higher stage has been associated with late recurrence.^11,19 Mizuno et al. found that nodal positivity was associated with poor RFS more than 5 years after surgery.¹³ Similarly, Maeda et al. found that late recurrence (vs. no recurrence) was more common in patients with stage III NSCLC (20%) than in patients with stage I disease (9%).²³ Our results confirm these findings, as both stage II and IIIA disease (vs. stage I) were associated with late recurrence in our cohort.

Few studies have attempted to compare early and late recurrences in lung cancer. In a recent study comparing early recurrences (≤2 years after surgery) with late recurrences (>2 years after surgery) in patients with completely resected stage I-III LUAD, higher pT and pN stage were independent predictors of early recurrence.¹⁸ Among patients with stage IIIA disease in our cohort, early recurrence was more common than late recurrence. Interestingly, the other pathologic features predictive of late recurrence (vs. no recurrence) in our cohort, were not significantly different between the early and late recurrence cohorts.

The frequency of genomic mutations in our cohort parallels observations from previous comprehensive genomic analyses.^24,25 In our cohort, there were no statistically significant differences in genomic alterations between tumors that recurred >2 years after surgery and those that did not recur. This finding is supported by the results of Deng and colleagues, who examined genomic alterations and their association with timing to recurrence using a 9-gene panel and found that none of the genomic mutations were associated with late recurrence.¹⁸ Therefore, specific genomic mutations do not appear to be reliable biomarkers for late recurrence.

One of the main implications of our findings is that there is a subpopulation of long-term survivors who need more-aggressive surveillance. This is further supported by the fact that most late recurrences in our cohort were intrathoracic and could be detected by routine surveillance with chest CT. Physicians managing patients with completely resected NSCLC should consider obtaining a chest CT every 6 months for 5 years postoperatively in patients with risk factors for late recurrence (poor IASLC grade, presence of LVI or VPI, higher pathologic stage), as these patients continue to have an elevated risk of recurrence even multiple years after surgery.

The limitations of our study include its retrospective nature. While we aim to follow the surveillance schedule outlined by NCCN guidelines,⁷ we were unable to track the proportion of patients who had strict adherence to the guidelines. We included LUAD only, and our results may not apply to other NSCLC histologic subtypes. Also, not all patients had molecular confirmation of their recurrence, which can be important with LUAD. However, 87% of recurrences were diagnosed with tissue biopsy. Although our study timeline (2010–2019) offers the important benefit of long-term follow-up, adjuvant therapy has recently changed, with the approval of targeted therapy and immunotherapy, which may alter recurrence dynamics moving forward. Last, widespread genomic testing was not available at our institution until 2016, which may have introduced potential selection bias

Late recurrence of LUAD after complete resection of pathologic stage I-IIIA disease is more common than previously reported. Although most recurrences occur within 2 years of surgery, 45% of recurrences occur beyond this point. Aggressive pathologic features such as poor IASLC grade, LVI, VPI, and higher stage continue to place long-term survivors at risk of recurrence. Thus, patients with select pathologic features may require more-diligent long-term follow-up.

Supplementary Material

1

Supplemental Figure 1. CONSORT flow diagram.

Supplemental Figure 2. Genomic mutations by recurrence status. Genes mutated in at least 5% of tumors in the late recurrence cohort were included.

Central Picture.

Late recurrence after resection of LUAD is common and driven by aggressive pathology.

Central Message.

Late recurrence of LUAD is more common than previously reported and is associated with higher stage and aggressive pathologic features (poor grade, lymphovascular invasion, visceral pleural invasion).

Perspective Statement.

Recurrence after complete resection of LUAD is typically considered an early event, occurring within 2 years of surgery. With long-term follow-up, our study demonstrates that late recurrence (>2 years after surgery) is more frequent than previously reported, and aggressive pathologic features continue to place patients at risk of recurrence.

Abbreviations:

AJCC

American Joint Committee on Cancer

CI

confidence interval

HR

hazard ratio

IASLC

International Association for the Study of Lung Cancer

IQR

interquartile range

LUAD

lung adenocarcinoma

LVI

lymphovascular invasion

NCCN

National Comprehensive Cancer Network

NSCLC

non-small cell lung cancer

OS

overall survival

RFS

recurrence-free survival

VPI

visceral pleural invasion