Incidence of occult pN2 disease following resection and mediastinal lymph node dissection in clinical stage I lung cancer patients

Abstract

OBJECTIVES: Early clinical stage (T1 and T2) non-small cell lung cancer (NSCLC) is commonly treated with anatomic lung resection and lymph node sampling or dissection. The aims of this study were to evaluate the incidence and the distribution of occult N2 disease according to tumour location and the short- and long-term outcomes.

METHODS: We performed a retrospective review of patients with clinical stage I NSCLC who underwent anatomic lung resection and lymphadenectomy. Mediastinal lymphadenectomy (ML) was defined as resection of at least 2 mediastinal stations, always including station 7 lymph nodes. Patients who had a lobe-specific lymphadenectomy were excluded.

RESULTS: One thousand six hundred and sixty-seven consecutive patients met inclusion criteria and were included. Overall, 9% (146/1667) of the patients had occult pN2 disease. At multivariable analysis, adenocarcinoma histology and vascular invasion were independently associated with greater risk of occult pN2 disease. In left and right upper lobe tumours, station 7 nodes were involved in 5 and 13% of pN2 positive cases, respectively. Station 5 and station 2/4 nodes were involved in 29 and 18% of left and right lower lobe pN2 tumours, respectively. There was no postoperative mortality, and postoperative morbidity was 28%. The median overall survival was 77.4 months. N0 patients had a median overall survival of 83.7 months vs 48.0 months and 37.9 months in N1 and N2 populations, respectively (P < 0.001).

CONCLUSIONS: Sixteen percent of pN2 patients had mediastinal lymph node metastasis beyond the lobe-specific lymphatic drainage. We recommend a complete lymphadenectomy be performed, even in clinical stage I NSCLC.

INTRODUCTION

The standard of care for the treatment of early stage non-small-cell lung cancer (NSCLC) remains an anatomic lung resection with a sampling or systematic mediastinal lymph node dissection [1]. Some studies have suggested that for early clinical stage disease, selective mediastinal node removal according to known patterns of nodal spread may be appropriate [2].

Supporting this approach are studies that show occult mediastinal nodal disease is rare in early stage NSCLC [3], and when occult mediastinal nodal disease does occur, it preferentially occurs in a predictable pattern [4, 5]. Upper lobe tumours follow predictable lobe specific pathways: right upper lobe tumours involve primarily station 4R (>60% of cases), left upper lobe tumours involve primarily level 5–6 (>70% of cases), and both lower lobes involve primarily station 7 (65% of cases) [5–7]. Riquet et al. [8] report that more than 70% of right upper lobe tumours and more than 95% of left upper lobe tumours only involve station 2–4 or 5–6 respectively, while more than 50% of lower lobe tumours involve only station 7. Based on these observations, a lobe-specific lymphadenectomy has been proposed as an alternative approach with a low risk for under-sampling N2 disease in patients with early clinical stage NSCLC [9–11]. However, the majority of studies supporting a lobe-specific lymphadenectomy are post-hoc analyses based on final pathologic information. Therefore, the groups of patients who may benefit from a lobe-specific lymphadenectomy and the relevance of this lymphadenectomy in early clinical stage lung cancer remain unclear.

To answer whether a systematic lymphadenectomy is appropriate in early clinical stage NSCLC, we retrospectively reviewed patients with cT1/T2N0M0 lung cancer who underwent a systematic mediastinal lymphadenectomy (ML), and evaluated the incidence and the distribution of occult pN1 and pN2 disease, the incidence of pN2 disease beyond the lobe-specific lymphatic drainage and the short- and long-term outcomes.

MATERIALS AND METHODS

We performed a retrospective review from 2000–12 of patients with cT1/T2aN0M0 NSCLC who underwent a surgical resection with curative intent with a mediastinal node dissection. The seventh edition of the International Association for the Study of Lung Cancer was used for TNM staging [12]. Inclusion criteria were clinical stage I disease as determined by computed tomography (CT) and positron emission tomography (PET) scans, the performance of a hilar lymphadenectomy, and a mediastinal nodal dissection.

Lymph nodes with a short axis diameter of less than 10 mm and non-FDG-avid on PET (SUVmax <2.5) were considered cN0. Patients with any clinical suspicion of N2 disease based on PET or CT findings underwent surgical staging of their mediastinum (i.e. mediastinoscopy, endobronchial ultrasound biopsy) and were excluded from this analysis regardless of the results of those staging studies. Patients with carcinoid tumours, synchronous cancers without a complete lymphadenectomy (n = 15), patients who had a lobe-specific lymphadenectomy (n = 1228), a bilobectomy (n = 66) or a pneumonectomy (n = 70) were excluded. In addition to operative and clinical staging details, patient demographics and clinicopathologic details were obtained.

ML was defined as resection of at least 2 mediastinal stations, always including station 7. For right sided tumours, additional stations included stations 2 or 4 R and/or 8 or 9; and for left sided tumours this included stations 5 or 6 and/or stations 8 or 9 [2]. Of note, this category also includes patients who underwent a more extensive ML, defined as removal of 3 or more mediastinal lymph node stations according to the ESTS and IASLC definition [1, 13]. Skip metastases were defined as N2 lymph node involvement without N1 involvement. Given the retrospective analysis and surgeon preference, our definition of ML includes patients who had en bloc lymphadenectomies which includes the surrounding fat along the anatomic boundaries.

Statistics

Descriptive statistics such as frequencies, medians and ranges were utilized for patient and tumour characteristics. Associations between patient and tumour characteristics with spread to lymph nodes (N0/N1 vs N2) were analysed using univariable logistic regression and Wald tests. A multivariable model was built using factors with P < 0.10 in univariable analyses and guided by clinical judgment. Associations between tumour characteristics and skip metastases in the N2 subset cohort were analysed using the Wilcoxon rank sum test for continuous variables and Fisher’s exact test for categorical variables.

Recurrence was analysed using competing risks methods. The risk of recurrence, the event of interest, was quantified using a cumulative incidence function that accounted for 2 competing events: new primary and death without recurrence. Time was calculated from date of surgery until recurrence (event of interest), new primary (competing event) or death (competing event), whichever occurred first. Patients who were alive without recurrence or new primary at the end of the study period were censored at the date of last available follow-up. Gray’s test was used to compare the cumulative incidence (of recurrence) functions of subgroups [11, 12].

Overall survival (OS) was estimated using the Kaplan–Meier method. Time was calculated from date of surgery until death. Patients were followed in the survivorship program [14] and data prospectively collected. Patients who were still alive at the end of the study period were censored at the date of last available follow-up. The log-rank test was used to compare the OS between subgroups. The association of positive N2 station and skip metastases with OS was analysed in the N2 subset cohort.

All P-values were two-sided, and P < 0.05 was considered significant. All statistical analyses were done in R (version 3.2.0, The R Foundation for Statistical Computing) including the ‘survival’ and ‘cmprsk’ packages.

RESULTS

One thousand six hundred and forty-eight consecutive patients who met inclusion criteria underwent 1667 resections with ML between January 2000 and December 2012. Nineteen patients had 2 operations for a new synchronous (n = 14) or metachronous (n = 5) primary lung cancer. Median age was 69 and 41% of the patients were male (Table 1). Lobectomy was performed in 1329 (80%), segmentectomy in 199 (12%) and a wedge resection in 139 (8%) patients. The median number of lymph node stations resected was 4 (range 2–9). Patient characteristics are summarized in Table 1. The incidence of N1 disease and N2 disease in our study population was 8 and 9%, respectively.

Table 1:

Clinicopathologic characteristics (n = 1667)

Characteristic	Number	%
Sex
Male	681	41
Female	986	59
Age, years, median (range)	69 (28–89)
≤70	944	57
>70	723	43
Laterality
Right	1094	66
Left	573	34
Location
Right upper lobe	474	28
Right middle lobe	59	4
Right lower lobe	452	27
Left upper lobe	400	24
Left lower lobe	282	17
Median diameter of lesion, cm (range)	2.3 (0–14)
Size 0–2 cm	743	45
Size 2.1–4 cm	694	42
Size >4 cm	228	14
Missing	2	0.1
Median SUVmax (range)a	4.65 (0–42)
Surgical approach
VATS	487	29
Open	1180	71
Type of resection
Wedge resection	139	8
Segmentectomy	199	12
Lobectomy	1329	80
Pathological N status
N0	1384	83
N1	137	8
N2	146	9
Histology
Adenocarcinoma	738	44
Lepidic-predominant	548	33
Squamous	269	16
Other	113	7
Vascular invasion
No	1014	61
Yes	566	34
Unknown	87	5
Visceral pleural involvement
No	1360	82
Yes	258	15
Postoperative complication
No	1196	72
Yes	471	28
Median length of stay, days (range)	5 (1–225)

Patients with parietal pleural involvement (n = 16) or with unknown involvement of visceral pleura (n = 33) were excluded.

VATS: video-assisted thoracic surgery.

^aSUVmax = 0, patients with no avid lesions.

Incidence of occult N2 disease

Nine percent (146/1667) of patients had occult pN2 disease. In our study population, 888 had only 2 mediastinal lymph node stations removed and 779 had 3 or more lymph node stations removed: 74 (8.3%) and 72 (9.2%) pN2 were discovered, respectively (P = 0.5). The proportion of pN2 disease in stage IA and IB was 6.7 and 9.3% (P = 0.18). On univariable analysis, comparing risk factors for patients with pN2 disease, tumours located in the left side, adenocarcinoma histology, T size, vascular invasion, pleural invasion and a SUVmax > 4.6 (median value of our population) were significantly associated with higher risk of occult pN2 disease (Table 2). Multivariable analysis confirmed that histology (adenocarcinoma) and vascular invasion were independently associated with greater risk of occult pN2 disease (Table 3).

Table 2:

Univariable analysis of associations with pN2 disease in clinical stage I NSCLC

Characteristics	OR	95% CI	P-value
Age (>70 vs ≤ 70)	0.96	(0.68–1.35)	0.82
Sex (female vs male)	0.84	(0.59–1.18)	0.32
Approach (VATS vs open)	0.32	(0.19–0.52)	<0.001
Type of procedure
Lobectomy	Ref.
Segmentectomy	0.33	(0.15–0.71)	0.004
Wedge	0.33	(0.13–0.78)	0.018
Location of the primary tumour
Left upper lobe	Ref.
Left lower lobe	0.90	(0.53–1.53)	0.70
Right lower lobe	0.73	(0.45–1.18)	0.20
Right middle lobe	0.86	(0.32–2.27)	0.76
Right upper lobe	0.95	(0.60–1.49)	0.81
SUVmax (>4.65 vs ≤ 4.65)	1.85	(1.24–2.77)	0.003
Size
≤2 cm	Ref.
>2 and ≤4 cm	1.48	(1.01–2.17)	0.044
>4 cm	1.86	(1.14–3.05)	0.014
Histology
Adenocarcinoma	Ref.
Other	0.43	(0.18–1.00)	0.051
Squamous cell	0.32	(0.17–0.61)	<0.001
Lepidic	0.62	(0.42–0.91)	0.016
Visceral pleura involvement (yes vs no)	2.54	(1.73–3.72)	<0.001
Laterality (left vs right)	0.66	(0.47–0.94)	0.020
Vascular invasion (yes vs no)	5.58	(3.78–8.23)	<0.001
Perineural invasion	3.56	(1.93–6.57)	<0.001

OR: odds ratio; CI: confidence interval.

Table 3:

Multivariable analysis of associations with pN2 disease

Characteristics	OR	95% CI	P-value
Laterality (left vs right)	0.73	(0.48–1.12)	0.15
Histology
Adenocarcinoma	Ref.
Other	0.42	(0.16–1.11)	0.080
Squamous cell	0.32	(0.15–0.71)	0.005
Lepidic	1.06	(0.65–1.71)	0.82
Size
≤2 cm
>2 and ≤4 cm	0.88	(0.54–1.42)	0.59
>4 cm	0.97	(0.50–1.87)	0.92
Visceral pleura involvement (yes vs no)	1.36	(0.85–2.20)	0.20
Vascular invasion (yes vs no)	6.24	(3.76–10.35)	<0.001
SUVmax (>4.65 vs ≤ 4.65)	1.58	(0.98–2.55)	0.062

OR: odds ratio; CI: confidence interval.

Distribution of N2 disease

Occult pN2 occurred in 146 (146/1667, 9%) patients who had a systematic ML according to the ESTS definition (n = 779). These patients were used to analyse the distribution of nodal disease based on the tumour location (Table 4). Left upper lobe tumours (n = 39) metastasized in 37 (95%) patients to level 5–6 only, in 2 patients (5%) to level 7, in 1 case skipping level 5–6. Left lower lobe tumours (n = 25) metastasized to level 7 in 17 (68%) patients, but in 8 (32%) patients there was disease at level 5–6. In 5 of these patients, there was only involvement of stations 5–6; in 3 (12%) patients pN2 disease was detected at level 8–9 only. On the right, lower lobe tumours (n = 33) involved level 7 in 28 cases (85%), involved level 4R in 6 cases (18%), while skipping level 7 in 3 of them, and involved level 7 and 8–9 in 2 cases. In middle lobe tumours (n = 5), 4 cases metastasized to level 7, and level 4R was involved in 1 case. Right upper lobe tumours (n = 44) involved level 4R in 42 (95%) cases and level 7 in 6 cases (14%), skipping level 4R in 2 cases. There was no significant association between OS and specific N2 positive station location (P = 0.67).

Table 4:

Positive pN2 disease relative to tumour location

LN station	Right upper	Right middle	Right lower	Left upper	Left lower
2/4R	38	1	3	0	0
2/4R and 7	4	0	3	0	0
7	2	4	23	1	10
8/9	0	0	2	0	3
7 and 8/9	0	0	2	0	4
5/6	0	0	0	37	5
5/6 and 7	0	0	0	1	3

LN: lymph node.

Skip metastases

Forty-seven of 139 (34%) patients with pN2 disease did not have associated pN1 disease identified. For 7 patients, there was no information about skip metastases. The proportion of patients with visceral pleural invasion was significantly lower in patients with skip metastases than in patients with N1 and N2 disease, 17 vs 38% (P = 0.012). There was no association between skip metastases and age, sex, laterality, tumour location, histology, tumour size, vascular invasion or SUVmax (Table 5). All 139 patients with skip metastases and N2 disease died with a median survival of 37.5 months (range 32–51). There was no difference in survival between patients with or without skip metastases: median OS 38.2 vs 37.5 months (P = 0.34).

Table 5:

Associations of patient factors and skip N2 disease

Characteristics	No skip N2 (n = 92) n (%)	Skip N2 (n = 47) n (%)	P-value
Age			0.36
≤70	50 (54)	30 (64)
>70	42 (46)	17 (36)
Sex			0.71
Female	60 (65)	29 (62)
Male	32 (35)	18 (38)
Laterality			0.59
Right	49 (53)	28 (60)
Left	43 (47)	19 (40)
Primary tumour location			0.28
Left upper lobe	25 (27)	14 (30)
Left lower lobe	19 (21)	5 (11)
Right lower lobe	17 (18)	13 (28)
Right middle lobe	2 (2)	3 (6)
Right upper lobe	29 (32)	12 (26)
SUVmax median (range)	5.95 (0.0-27.0)	6.2 (1.4-25.0)	0.42
Size (cm)a			0.87
≤2 cm	28 (31)	16 (34)
>2 and ≤4 cm	43 (48)	23 (49)
>4 cm	19 (21)	8 (17)
Histology			0.49
Adenocarcinoma	56 (61)	27 (57)
Other	3 (3)	3 (6)
Squamous cell	5 (5)	5 (11)
Lepidic	28 (31)	12 (26)
Visceral pleura involvement			0.012
No	57 (62)	39 (83)
Yes	35 (38)	8 (17)
Vascular invasionb			0.68
No	22 (25)	13 (29)
Yes	65 (75)	32 (71)

^aExcludes 2 patients with missing data.

^bExcludes 7 patients with missing data.

Short- and long-term outcomes

In our study population, the median length of stay in hospital was 5 days (range 1–225), 28% of patients had postoperative complications, and there was no in-hospital mortality. The 30-day mortality was 3.6% (n = 61) and the 90-day mortality was 2.5% (n = 42). The most common complications were atrial fibrillation 7.5% (n = 125), prolonged air leak 6.2% (n = 103), pneumonia 4.8% (n = 81), chylothorax 1% (n = 17) and postoperative bleeding 0.4% (n = 6).

OS was evaluated in 1621 patients who had an R0 resection, excluding 24 R1 resections and 3 R2 resections. Two hundred and fifty-seven patients received adjuvant treatment: 205 chemotherapy alone, 27 radiotherapy alone and 25 chemoradiation. Data on any adjuvant therapy are complete for 1593/1621(98.3%) with data missing for 28 patients. The median follow-up was 34.7 months (range 5 days–172.2 months), and there were 722 deaths. No patient was lost to follow-up. The median OS and 5-year OS were 77.4 months [95% confidence interval (CI) 72.6–82.2 months] and 61%, respectively. Patients who had 2 mediastinal lymph node stations resected had a median OS of 75.5 months (95% CI 67.7–83.3 months) and patients who had a complete mediastinal lymph node dissection with at least 3 mediastinal lymph node stations resected had a median OS of 78.5 months (95% CI 71.5–85.5 months). pN0 patients had a median OS of 83.7 months vs 48.0 months and 37.9 months in pN1 and pN2 populations, respectively (P < 0.001). The 5-year OS was 65.7% for pN0, 43.7% for pN1, and 36.4% for pN2 patients (Fig. 1).

Figure 1:

Overall survival of patients according to N status (N = 1621).

Three hundred and sixty-five patients experienced recurrence, and 114 patients developed a new primary lung cancer. The 3-year cumulative incidence of recurrence was 21.9%, and it was higher for distant compared to loco-regional recurrence: 16.5 vs 5.4%. The 3-year cumulative incidence of recurrence was higher in the pN2 compared to the pN0/N1 population: 49.4% vs 19.0% (Fig. 2).

Figure 2:

Cumulative incidence of recurrence according to N status (N = 1667).

DISCUSSION

The standard treatment for early stage lung cancer is surgery: lobectomy and mediastinal lymph node dissection or sampling. The role and the extension of the lymphadenectomy on survival remain controversial, but accurate mediastinal staging is important for prognostic reasons as well as to determine the need for adjuvant therapies. Now that high resolution CT and PET scans are available to better detect lymph node metastasis, several groups have proposed a lobe-specific mediastinal lymph node assessment based on tumour location [5, 7].

Our study evaluated the patterns of occult N2 disease and skip N2 metastasis in early clinical stage NSCLC and the short- and long-term results after complete ML. The incidence of occult pN2 disease in our series did not vary significantly between clinical T1a and T1b NSCLC. Independent risk factors for occult pN2 disease were histology and the presence of vascular invasion. Several studies have confirmed that adenocarcinoma histology is associated with a higher risk for occult pN2 disease [15, 16]. SUVmax was only marginally significant, but tumour size and location of the tumour were not found to be independent risk factors for pN2 disease in contrast to a prior report [17]. SUVmax and tumour pT2a were found to be prognostic factors to develop pN2 disease in a recent publication [18]. Finally, we did not find an association with younger age and increased risk of pN2 disease as previously reported [19].

The side of tumour was correlated with occult N2 disease only at the univariable analysis, but was not confirmed to be an independent predictor of pN2 disease on multivariable analysis. The number of patients with occult pN2 disease was lower on the left compared to the right side. As previously reported, this may be related to the challenges of performing a more extensive ML of the left L4 and L2 paratracheal area for anatomical reasons, which make access more difficult, particularly when performing minimally-invasive approaches [20].

The impact of approach (VATS versus open surgery) and type of resection on occult pN2 disease was not an independent prognostic factor in our series. While some studies have suggested that upstaging from N0 to N1 and from N0 to N2 is higher in open compared to video-assisted thoracic surgery (VATS) resections [21, 22], most groups have found no difference in nodal station dissection rates [23] and nearly all studies have shown equivalent overall survival rates between these 2 approaches [24, 25]. While not specifically analysed in our study, we believe any lung cancer resection procedure should incorporate a mediastinal nodal dissection/sampling regardless of the approach.

As expected, we did observe an association between nodal status and survival. The 5-year overall survival in our series was 61%. pN2 patients had a 5-year survival of 36% compared to 66% for patients with N0 disease. Asamura et al. and Riquet et al. published similar survival rates for pN2 patients [6, 26]. Occult pN2 disease was also associated with a higher risk of recurrence, mainly distant recurrence. When occult N2 disease is encountered intraoperatively we favour complete resection of all nodal tissue in addition to performing the lung resection as suggested by Detterbeck [27]. Benefits of this approach are prognostic, allow appropriate selection for adjuvant therapies, and, perhaps most importantly, achieve an R0 resection.

According to the published literature, lymphatic drainage from each lobe usually follows a well-described anatomic route. This concept of a ‘fixed’ lymphatic drainage path is the underlying premise to perform a lobe-specific lymphadenectomy. Upper lobe tumours usually drain to the lymph nodes of the upper mediastinum (stations 2–6), and middle and lower lobe tumours drain to nodes in the lower mediastinum (stations 7–9) [5, 28]. An important observation of our study is that 16% of clinically node-negative, stage I NSCLC patients had N2 disease that did not follow a lobe-specific lymphatic drainage pattern. Restated, this means that if a lobe-specific lymphadenectomy had been performed in this series, 16% of pN2 patients could have been understaged, not offered evidence-based adjuvant therapy, and did not have a R0 resection. This result compares favourably to the study by Riquet et al. [26]. Sun and colleagues demonstrated that the survival in patients with lymph node metastasis beyond lobe-specific nodes had a significantly worse prognosis with median survival and 5-year survival of 24 months and 11.7% [28]. Collectively, our study and others do not support a lobe-specific lymphadenectomy as advocated by other groups [5, 29].

Limitations

Limitations of this study include that it is a retrospective, single centre analysis. This introduces potential selection biases which cannot be controlled for retrospectively. While our institution has a long history of performing mediastinal node dissection [30] we cannot control for the exact types of lymphadenectomy which recently have been more codified [1]. Despite these limitations, our study is a large series that examines the incidence of occult N2 disease of clinically node-negative, stage I NSCLC of which all patients had CT and PET imaging.

CONCLUSIONS

In conclusion, based on our observations we recommend a systematic mediastinal nodal dissection be performed even in early, clinical stage I NSCLC. Current intraoperative sentinel nodal imaging studies are not refined to guide the extent of N2 nodal dissection. Moreover, specific tumour biologies, genomic signatures, histologies and subtypes may be more prone to N2 disease, but evidence to selectively sample or dissect N2 nodes based on these criteria is absent.