Commission on Cancer Standards for Lymph Node Sampling and Oncologic Outcomes after Lung Resection

Abstract

BACKGROUND:

The newest Commission on Cancer (CoC) standards recommend sampling 3 mediastinal and 1 hilar lymph node station, 3(N2)1(N1), for lung cancer resections. However, the relationship between the CoC standards and outcomes has not been thoroughly investigated.

METHODS:

A prospective institutional database was queried for clinical stage I-III lung resections prior to the implementation of the new standards. The relationship between the 3(N2)1(N1) standard (“guideline concordant”) and outcomes (upstaging, complications, receipt of adjuvant therapy, locoregional/distant recurrence, and survival) were assessed using multivariable models and stratified by stage.

RESULTS:

Of 9,289 pulmonary resections 3048 (33%) were guideline concordant and 6241 (67%) were not. Compared to non-concordant, those who were guideline-concordant had higher rates of nodal upstaging (21% vs 13%; OR 1.32 [95% CI 1.14–1.51] p<0.001) and in-hospital complications (34% vs 27%), (OR 1.17 [95% CI 1.05–1.30], p=0.004), but similar adjuvant systemic therapy administration (19% vs 13%; OR 1.09 [95% CI 0.95–1.24], p=0.2), (98% chemotherapy). Locoregional and distant recurrence were not significantly improved with guideline concordance across clinical stage I, II and III subsets. Overall survival was similar in clinical stages I and II but improved survival was observed among guideline concordant clinical stage III patients (HR 0.85 [95% CI 0.74–0.97], p=0.02).

CONCLUSIONS:

Sampling 3(N2)1(N1) was associated with increased upstaging and complications but not with decreased recurrence or mortality in clinical stage I or II patients. Survival was improved among concordant, clinical stage III patients. Further study is indicated to determine the ideal lymph node sampling strategy across heterogeneous lung cancer patients.

The most recent iteration of the American College of Surgeons Commission on Cancer (CoC) standards for curative-intent pulmonary resection for primary malignancy has modified the recommendation for lymph node sampling from 10 total lymph nodes to 3 mediastinal and 1 hilar lymph node 3(N2)1(N1)¹. These recommendations are based on a number of observational studies showing variability in outcomes based on the previous standard of 10 lymph nodes^2,3 and that more thorough mediastinal station nodal sampling was associated with improved separation of survival between pN1 and pN2 cohorts⁴. These recommendations extend to all histology and stages of non-small cell lung cancer undergoing curative intent resection and do not exclude patients with prior resection or more indolent subtypes^1,5.

The relationship between the new CoC standards and oncologic outcomes has not been thoroughly validated. Recent studies have had contradictory findings in terms of the association of guideline concordance and oncologic outcomes^2,4,6,7. Prior studies were limited by small sample size, restriction to clinical stage I patients or variability of care across the different participating centers beyond just lymph node sampling practices^2,4,6,7.

Using our large institutional database, we sought to determine if historical concordance with the new CoC standards over the past 20 years (i.e., prior to their actual implementation) was associated with improved outcomes among all patients who would be scrutinized by the CoC rules (ie all curative intent resections regardless of subtype or patient characteristics).

MATERIAL AND METHODS

Study population and data collection

All consecutive patients who underwent pulmonary resection for clinical stage I-III lung cancer from January 1, 2000 to December 31, 2020 at Memorial Sloan Kettering Cancer Center were identified from a prospectively maintained and audited database. The study period was designed to end prior to the implementation of the new CoC standards on January 1, 2021 (after this time lymph node sampling was influenced by the standards and was highly compliant). Patients upstaged to pathologic M1 disease or with R2 resections were excluded as the intraoperative recognition of M1 disease or inability to obtain a macroscopic margin could exclude these from being considered curative intent resections (n=199), (consort diagram, Supplemental Figure 1).

Patient characteristics, preoperative workup, pathology, and postoperative outcomes were prospectively collected during the patient’s hospital stay and at outpatient visits. Nodal upstaging was defined as an increase in nodal staging from cN0 to pN1 or pN2, and from cN1 to pN2. Nodal upstaging was expressed as a proportion of upstaged out of the number of patients able to be upstaged (i.e. cN2 patients cannot be upstaged by nodes during ipsilateral surgery). Complications were identified and confirmed by treating physicians. Data from additional treatment or follow-up at other institutions was gathered through telephone calls and correspondence received by external care providers.

The lymph node stations that were sampled were determined by the pathology report of the resection specimen as outlined in CoC guidelines. Stations that were sampled at a prior mediastinoscopy procedure were not included, as they did not appear in the same pathology report for the resection, which is a requirement to comply with CoC guidelines. CoC guidelines do not allow nodal stations sampled by endobronchial ultrasound (EBUS) to be included^1,5.

This study was approved by the Memorial Sloan Kettering Institutional Review Board (#18–391, 04/24/2023); all procedures were performed in accordance with the Helsinki declaration. The need for informed consent was waived because of the retrospective nature of the analysis and the lack of identifiable patient information. The STROBE reporting guidelines were followed⁸.

Statistical analysis

Differences in patient, operative, and tumor characteristics were compared between CoC guideline concordance status using Chi-square test or Fisher’s exact tests. Relationships between guideline concordance and dichotomous outcomes (in-hospital complications, rates of upstaging and adjuvant therapy) were expressed as adjusted odds ratios (aOR) using multivariable logistic regression models adjusted for age, sex, smoking status, clinical stage (IA1, IA2, IA3, IB, IIA, IIB, IIIA, IIIB, IIIC), histology, approach (open, VATS, robotic), extent of resection (wedge/segmentectomy, lobectomy, bilobectomy, pneumonectomy), margins (R0, R1), year of surgery, FEV1, DLCO, prior lung cancer, invasive nodal staging (EBUS, mediastinoscopy) and prior systemic therapy. Cumulative incidence of locoregional recurrence was summarized using competing risk approach where distant recurrence and death without recurrence were considered as competing events. Patients were otherwise censored at the time of last follow-up. Similarly, the cumulative incidence of distant recurrence considered locoregional recurrence and death without recurrence as competing events. Cumulative incidence curves were compared between guideline concordance status using Gray’s test by clinical stage. The relationships between guideline concordance and cumulative incidence of various types of recurrence were expressed as adjusted hazard ratios (aHR) using multivariable competing risk regression models. The relationship between guideline concordance and overall survival was quantified using multivariable Cox models. The multivariable time-to-event models were adjusted for patient and tumor characteristics (as above plus adjuvant systemic therapy) and stratified by clinical stage (I, II, III) or pathologic lymph node stage (pN1, pN2).

All statistical tests were two-sided and p-values <0.05 was considered significant. All analyses were performed using Stata 15.1 (StataCorp, College Station, TX) and R 4.3.2 (R Core Team, Vienna, Austria).

Subset analyses

Two additional subset analyses were performed. A restricted subset of patients was created to address potential imbalances across key factors between the two concordance groups. To reflect a more homogeneous cohort, this subset included patients with adenocarcinoma or squamous lung cancers with no prior lung resection or cancer who underwent lobectomy with R0 resection, and excluded cancers that proved to be in situ, minimally invasive, or patients who had undergone preop systemic therapy. All outcomes were analyzed using the approaches above to compare guideline concordant to those non-concordant.

An additional subset analysis based on the previous CoC lymph node sampling recommendations of obtaining at least 10 total lymph nodes was also performed for comparison. This analysis was performed on patients who had information on total lymph nodes collected (data on total lymph nodes harvested were not reported in pathology reports prior to 2013). All outcomes were analyzed using the approaches above to compare those with at least 10 total lymph nodes excised to those with <10.

RESULTS

Patient and Tumor Characteristics

9,289 pulmonary resections for lung cancer were identified, which were performed by 19 surgeons. 3048 (33%) resections would have been concordant with the new CoC lymph node sampling standard and 6241 (67%) would not have been. (Of note, all patients underwent their resections before implementation of the new CoC lymph node sampling standard). Age, sex, smoking history and DLCO did not significantly differ between groups. As expected, patients with clinical stage I disease, carcinoid histology, wedge resection or a prior lung cancer resection were less likely to have guideline concordant lymph node sampling (58% vs 73%, p<0.001; 4.6% vs 7.8%, p<0.001, 15% vs 33%, p<0.001 and 2.4% vs 6.4%, p<0.001, respectively), (Table 1, Supplemental Table 1).

Table 1.

Abbreviated^* Patient and Tumor Characteristics by Concordance with the new CoC Lymph Node Assessment Standard

	CoC concordant
	Yes, N = 3,048	No, N = 6,241	p-value
Age, n (%)			0.2
<55	323 (11)	705 (11)
55–64	787 (26)	1,530 (25)
65–74	1,237 (41)	2,476 (40)
75+	701 (23)	1,530 (25)
Sex, n (%)			0.72
Female	1,809 (59)	3,728 (60)
Male	1,239 (41)	2,513 (40)
History of smoking, n (%)			0.94
No	583 (19)	1,198 (19)
Yes	2,465 (81)	5,043 (81)
% Predicted FEV1, n (%)			<0.001
<40%	11 (0.4)	39 (0.6)
40–60%	157 (5.3)	450 (7.5)
60%+	2,807 (94)	5,534 (92)
Unknown	73	218
Year of Surgery, n (%)			<0.001
2000–2005	760 (25)	1,599 (26)
2006–2010	458 (15)	1,751 (28)
2011–2015	494 (16)	1,713 (27)
2016–2021	1,336 (44)	1,178 (19)
Clinical Stage (TNM 8th Ed., n (%)			<0.001
I	1,781 (58)	4,534 (73)
II	595 (20)	822 (13)
III	672 (22)	885 (14)
Procedure, n (%)			<0.001
Wedge	147 (4.8)	1,472 (23.5)
Segmentectomy	311 (10.2)	557 (8.9)
Lobectomy	2,326 (76)	3,899 (62)
Bilobectomy	102 (3.3)	156 (2.5)
Pneumonectomy	162 (5.3)	157 (2.5)
Approach, n (%)			<0.001
Open	1,657 (54)	3,243 (52)
Video assisted thoracic surgery	688 (23)	2,348 (38)
Robotic	703 (23)	650 (10)
Preop Systemic Therapy, n (%)			<0.001
No	2,371 (78)	5,383 (86)
Yes	677 (22)	858 (14)
Margin, n (%)			0.08
R0	2,962 (97)	6,102 (98)
R1	86 (2.8)	139 (2.2)
Histology, n (%)			<0.001
Adenocarcinoma	2,135 (70)	4,459 (71)
Carcinoid	140 (4.6)	488 (7.8)
Other	244 (8.0)	404 (6.5)
Squamous	529 (17)	890 (14)
Adjuvant Systemic Therapy, n (%)			<0.001
No	2,467 (81)	5,436 (87)
Yes	581 (19)	805 (13)
Prior Invasive Nodal Staging, n (%)			<0.001
None	2,342 (77)	5,435 (87)
Endobronchial ultrasound	483 (16)	339 (5.4)
Mediastinoscopy	223 (7.3)	467 (7.5)

^*Full Table (Supplemental Table 1)

Nodal sampling

The nodal stations with the greatest differential in sampling between the groups were the level 9 and level 7 lymph node stations (Supplemental Table 2).

Upstaging

Greater upstaging was observed in the guideline concordant group, (21% vs 13%, p<0.001). This was true for both pN1 nodes (11% vs 7%, cN0→pN1, p<0.001) and pN2 nodes (10.1% vs 5.9%, p<0.001), (Supplemental Table 3). These findings persisted in a multivariable model adjusted for patient and tumor characteristics (aOR 1.32, [95% CI 1.14–1.51] p<0.001) (Table 2).

Table 2.

Upstaging, Adjuvant Systemic Therapy and Complications by Concordance with the new CoC Lymph Node Assessment Standard

	CoC guideline concordant	Not CoC guideline concordant	P Value	Multivariable Model**
Upstaging*	534/2572 (21%)	705/5619 (13%)	<0.001	aOR 1.32 [1.14–1.51] p<0.001
Adjuvant systemic therapy	581 (19%)	805 (13%)	<0.001	aOR 1.09 [0.95–1.24] p=0.2
In-hospital complications	1024 (34%)	1664 (27%)	<0.001	aOR 1.17 [1.05–1.30] p=0.004).

^*Upstaging ratio includes upstaged patients divided by patients who could possibly be upstaged (ie cN2 patients cannot be upstaged by ipsilateral surgery)

^**Adjusted for patient and tumor characteristics as listed in methods

Adjuvant therapy

A greater proportion of patients with guideline concordant lymph node sampling received adjuvant systemic therapy (19% vs 13%, p<0.001). However, this was not significant in a multivariable model (aOR 1.09, [95% CI 0.95–1.24], p=0.2) (Table 2). Considering the time period of this study, adjuvant therapy consisted of almost exclusively chemotherapy (98%) as opposed to immunotherapy or targeted therapy.

In-hospital complications

In-hospital complications were greater in the guideline concordant group (34% vs 27%; aOR 1.17, [95% CI 1.05–1.30] p=0.004) (Table 2). Clavien-Dindo Grade 3 or greater complications were also more frequent amongst the guideline concordant group (7.5 vs 4.7% , aOR 1.32, [95% CI 1.08–1.61] p=0.006). Chylothorax (1.2% vs 0.4%, p<0.001), recurrent nerve injury (1.5% vs 0.5%, p<0.001) and bleeding complications (1.8% vs 1.2%, p=0.03) were all more frequent in the guideline concordant group, (Supplemental Table 4).

Cumulative Incidence of Recurrence

For clinical stage I, and III patients, the adjusted cumulative risk of locoregional (LR) or distant recurrence (DR) analyzed by guideline concordance was not significantly different. Among clinical stage II patients, CoC concordance was not significantly associated with hazard of LR (aHR0.93 [95% CI 0.67–1.29] p=0.7 but was significantly associated with greater hazard of DR (aHR1.28 [95% CI 1.02–1.62] p=0.03) (Table 3). Within the pN1 and pN2 subsets of patients, guideline concordance was not significantly associated with LR or DR. (Supplemental Table 5 and 6).

Table 3.

Adjusted Cumulative Risk of Locoregional Recurrence, Distant Recurrence and Overall Survival by Concordance with the new CoC Lymph Node Assessment Standard, Stratified by Clinical Stage

	aHR* (aHR<1 favors CoC guideline concordant group)	95% CI	P Value
Clinical Stage I
Locoregional Recurrence	1.02	0.84–1.24	0.8
Distant Recurrence	1.15	0.97–1.35	0.10
Overall Survival	0.98	0.89–1.08	0.7
Clinical Stage II
Locoregional Recurrence	0.93	0.67–1.29	0.7
Distant Recurrence	1.28	1.02–1.62	0.03
Overall Survival	0.99	0.85–1.15	0.9
Clinical Stage III
Locoregional Recurrence	0.83	0.62–1.11	0.2
Distant Recurrence	0.88	0.74–1.06	0.2
Overall Survival	0.85	0.74–0.97	0.02

^*Adjusted for patient and tumor characteristics as listed in methods

Overall Survival

For clinical stage I and II patients, guideline concordance was not significantly associated with overall survival (Table 3). However, for clinical stage III patients, guideline concordance was significantly associated with lower hazard of death (aHR 0.85 [95% CI 0.74–0.97] p=0.02). A similar relationship was observed in the patients with pN2 disease (aHR: 0.83 [95% CI 0.71–0.96] p=0.02), (Supplemental Table 7).

Restricted Subset Analysis

Of the 4,170 patients who underwent R0 lobectomy for invasive adenocarcinoma or squamous lung cancers with no prior treatment or lung cancer (characteristics Supplemental Table 8), 3048 (33%) were CoC guideline concordant. The results of analyses using this restricted cohort of patients were similar to the results observed using the entire cohort. There was no significant recurrence or survival benefit associated with guideline concordance among the subsets of patients with clinical stages I and II; however, there were a few differences in conclusions between the entire cohort and the restricted cohort (Supplemental Tables 9–18). Notably, the multivariable models indicate that guideline concordance was no longer significantly associated with in-hospital complications (p=0.06) and distant recurrence for clinical stage II (p=0.16); however, the effect direction and size were similar to the entire cohort for both. Additionally, overall survival for clinical stage III patients no longer favored the CoC concordant group (aHR: 0.96 [95% CI 0.70–1.33] p=0.8) in a much smaller cohort (n=284 vs n=1557) after all patients with preop systemic therapy were excluded.

Subset analyzed by the old CoC recommendation of collecting at least 10 lymph nodes

Of 5,263 patients identified with data on total lymph nodes sampled (years 2013–2020, 67% had at least 10 lymph nodes sampled. Patients with at least 10 lymph nodes sampled had greater odds of upstaging (32% vs 16%; aOR 1.58 [95% CI: 1.27–1.97] p<0.001) and greater odds of receiving adjuvant systemic therapy (25% vs 13%; aOR 1.35 [95% CI: 1.11–1.65] p=0.002). The hazard of locoregional recurrence, distant recurrence, and death were similar across clinical stage I and II regardless of whether at least 10 lymph nodes were collected. Among the subset of patients with clinical stage III disease, those with at least 10 lymph nodes collected had lower hazard of distant recurrence (aHR 0.77 [95% CI 0.60–0.99], p=0.04) and death (aHR: 0.69 [95% CI 0.57–0.85], p<0.001). (Supplemental Tables 19–28).

COMMENT

In this study of over nine thousand pulmonary resections for lung cancer at a dedicated cancer center, collecting at least 3 mediastinal and 1 hilar lymph node was associated with greater upstaging but was not significantly associated with lower recurrence or improved survival in clinical stage I or II patients. The current findings of no significant recurrence or survival benefit associated with sampling 3(N2)1(N1) in early-stage lung cancer patients are consistent with one recent study but in contrast to others^4,6,7. Specifically, a recent single-institutional study investigated the current CoC standards by comparing a 3(N2)1(N1) group with a 2(N2)1N1 group and did not find a difference in upstaging, adjuvant therapy use, recurrence or survival among 581 clinical stage I patients⁷. However, a previous study of over 9,000 clinical stage I patients across multiple Veterans Administration hospitals found that adhering to 3(N2)1(N1) sampling correlated with increased upstaging and improved recurrence-free and overall survival rates compared to cases with no lymph node sampling⁶. Additionally, an observational study across 11 hospitals in northern Mississippi, eastern Arkansas, and western Tennessee, found that overall survival rates were greater in patients proving to have positive N1 nodes when at least 3(N2) and 1(N1) nodes were collected⁴.

The current study extends and clarifies these findings with the unique perspective of a large sample of patients undergoing care at a single dedicated cancer center. A possible explanation for the differences in findings relates to differences in the populations that were studied and compared. In the current and previous single institution studies, recurrence and survival were not improved among patients who underwent CoC concordant sampling when compared to a group of patients who were partially sampled but did not achieve the new CoC standards. However, in a previous study of heterogenous centers, outcomes were improved when comparing to no lymph node sampling⁶. No lymph node sampling was not used as a comparison in the current study because this was not a standard strategy at the institution and would represent non-standard care. Having zero lymph nodes sampled could be associated with other factors that led to decreased survival (e.g., poor health status of the patient or additional treatment inadequacies that were unaccounted for). Additionally, the heterogeneity of centers sampled in previous studies^4,6 my provide another confounding association by which patients undergoing non-standard surgical care were subject to other factors related to worse health outcomes. The current study, in contrast, focuses on patients treated at a single, dedicated cancer center, where care is likely to be more uniform.

One should also consider the theoretical framework behind which sampling 3 (N2) and 1 (N1) would improve recurrence and survival. In theory, removing more lymph nodes leads to more accurate staging, which leads to more appropriate use of adjuvant therapy and thus improved recurrence and survival⁴. In the current study there was an 8% increase in upstaging and a 6% increase in adjuvant systemic therapy (although this was not independently significant in a multivariable model) associated with collecting at least 3 N2 nodes and 1 N1 node. Although there may be some benefit in resecting involved nodes, the primary benefit of accurate nodal staging is thought to be related to the identification and eradication of nodal and associated occult micro-metastatic disease with the use of effective adjuvant systemic therapies. However, adjuvant platinum-based chemotherapy is only associated with around a 5% increase in 5-year absolute survival compared to no systemic therapy in an appropriate population^9,10. In the current study, given the time period that was studied, less than 2% of patients received adjuvant immunotherapy or targeted therapy (98% receiving chemotherapy). Therefore, one may expect a survival difference of only around 0.3% at 5 years due to the 6% difference in administration of platinum-based adjuvant therapy. Additionally, it is not clear that increased upstaging actually resulted in the receipt of more adjuvant therapy in this cohort, as this association was not found to be statistically significant in the multivariable model. Consequently, it is not surprising that an improvement in recurrence or survival was not detected in this study for clinical stage I and II lung cancer.

Importantly, there are now a rapidly increasing array of options for adjuvant systemic therapy, including targeted and immunotherapies which can be many times more effective than platinum-based chemotherapy in appropriate populations^11,12. If modern adjuvant therapies are becoming increasingly effective, then it stands to reason that accurate nodal staging becomes even more critical. Therefore, although there was no recurrence or overall survival benefit detected in this study for clinical stage I and II patients, there is reason to believe with our modern advancements in adjuvant therapy that an increased rate of upstaging with more systematic lymph node assessment could lead to improved recurrence and survival. Interestingly, in a subset analysis of patients who had the total number of lymph nodes recorded, the sampling of at least 10 lymph nodes was associated with a numerically greater improvement in upstaging and adjuvant therapy use than the 3(N2)1(N1) distinction (16% difference vs 8% difference in upstaging, 12% vs 6% difference in adjuvant therapy). These results could imply that systematic lymph node sampling or dissection of more total nodes (rather than just sampling a specific minimal number of nodes per station, as the new CoC standard requires) may result in more accurate nodal staging.

Another interesting finding of the current study was that there were greater complications observed in the group of patients with at least 3(N2)1(N1) sampled. These complications included a few that presumably could be linked directly to the lymph node sampling/dissection itself, such as chylothorax, recurrent nerve injury and bleeding. These results can be put in the context of the Z030 randomized trial of lymph node dissection versus sampling where, although there was no significant increase in complications, in the group undergoing more aggressive lymph node excision, there were more than double the amount of chylothoraxes and recurrent nerve injuries¹³. Certainly, the possible benefit of greater upstaging with more lymph node sampling should be balanced carefully with the possible harm of increased complications, especially in cases where it is unlikely to make a difference in oncologic outcomes. This deserves more careful study.

Clinical stage III patients with at least 3(N2) and 1(N1) sampled did not have a lower incidence of recurrence but did have improved overall survival. Interestingly, overall survival was also increased in those with pathologic N2 nodes, and those who had at least 10 lymph nodes sampled, which was coincident with a lower rate of distant recurrences. However, after eliminating patients with preop systemic therapy, the survival benefit was not detected in a much smaller cohort. It is not entirely clear if the loss of effect is due to decreased sample size or a confounding interaction. Greater survival with more aggressive lymph node excision has been shown previously in a post-hoc analysis of the ECOG 3590 trial of adjuvant therapy in patients with completely resected stages II and IIIA NSCLC¹⁴. Further study is certainly indicated in this group of patients who are also currently undergoing a shift in treatment paradigms with the introduction of neoadjuvant chemo-immunotherapy^11,15,16.

There was counterintuitively worse distant recurrence in clinical stage II patients who had at least 3(N2) and 1(N1) sampled in this study. This relationship was not statistically significant in the restricted subset of patients, although the size and direction of the effect was similar. Thus there is some lack of clarity about this subgroup.

Limitations

The foremost limitation of the current study is its retrospective nature. Although the data in this study were prospectively collected on consecutive patients it would be impossible to identify all the confounders in a retrospective fashion. However, the known, clinically important confounders were addressed and accounted for in the adjusted models. Unfortunately, tumor solidity and percentage lepidic were not available for the entire database so these could not be added to an adjustment. One could argue that patients with more indolent cancers (ground glass lesions that prove to be minimally invasive or in situ cancers, lepidic predominant or carcinoid tumors) would have had less sampling due to surgeon bias, or that wedges, pneumonectomy, preop therapy and prior resections were unbalanced in the cohorts (although all non-pathologic characteristics were adjusted for) and therefore a restricted subset was created eliminating these patients, which was consistent with the overall cohort. We chose not to make such exclusions in the primary analyses as the CoC rule does not allow for exclusions of these patients. Another limitation is that the present study represents a single dedicated cancer center, and therefore reflects a very specific group of patients and clinical practice patterns. Although this is in some way a strength of the study (in that it eliminates some potential confounding) it does detract from the generalizability of the study. Perhaps most importantly, as previously emphasized, modern adjuvant therapy is rapidly changing since the studied period, and this could lead to differences in survival associated with greater upstaging in the future.

Conclusions

Sampling 3 mediastinal and 1 hilar node was associated with increased upstaging and complications but was not associated with decreased recurrence or increased survival in clinical stage I or II patients. Survival was improved among concordant clinical stage III patients and patients with proven positive mediastinal nodes. Higher complication rates were also observed with greater sampling. These results suggest that the optimal lymph node assessment strategy may be dependent on patient and tumor characteristics. Further study is warranted to determine the impact on clinical stage III patients and the potential differences in recurrence and survival in the era of immune and targeted adjuvant therapies.