Impact of patient selection and treatment strategies on outcomes after lobectomy for biopsy-proven stage IIIA pN2 non-small cell lung cancer

Abstract

OBJECTIVES

We evaluated the impact of patient selection and treatment strategies on long-term outcomes of patients who had lobectomy after induction therapy for stage IIIA pN2 non-small cell lung cancer (NSCLC).

METHODS

The impact of various patient selection, induction therapy and operative strategies on survival of patients with biopsy-proven stage IIIA pN2 NSCLC who received induction chemotherapy ± radiation followed by lobectomy from 1995 to 2012 was assessed using Cox proportional hazards analysis.

RESULTS

From 1995 to 2012, 111 patients had lobectomy for stage IIIA pN2 NSCLC after chemotherapy ± radiation with an overall 5-year survival of 39%. The use of induction chemoradiation decreased over time; from 1996 to 2007, 46/65 (71%) patients underwent induction chemoradiation, whereas from 2007 to 2012, 36/46 (78%) patients underwent induction chemotherapy. The use of video-assisted thoracoscopic surgery (VATS) increased over the time period of the study, from 0/26 (0%) in 1996–2001, to 4/39 (10%) in 2002–07 to 33/46 (72%) in 2008–12. Compared with patients given induction chemotherapy alone, patients given additional induction radiation were more likely to have complete pathologic response (30 vs 11%, P = 0.01) but had worse 5-year survival in univariable analysis (31 vs 48%, log-rank P = 0.021). Patients who underwent pathologic mediastinal restaging following induction therapy but prior to resection had an improved overall survival compared with patients who did not undergo pathologic mediastinal restaging {5-year survival: 45.2 [95% confidence interval (CI): 33.9–55.9] vs 13.9% (95% CI: 2.5–34.7); log-rank, P = 0.004}. In multivariable analysis, the particular induction therapy strategy and the surgical approach used, as well as the extent of mediastinal disease were not important predictors of survival. However, pathologic mediastinal restaging was associated with improved survival (HR 0.39; 95% CI: 0.21–0.72; P = 0.003).

CONCLUSIONS

For patients with stage IIIA pN2 NSCLC, the VATS approach or the addition of radiation to induction therapy can be selectively employed without compromising survival. The strategy of assessing response to induction therapy with pathologic mediastinal restaging allows one to select appropriate patients for complete resection and is associated with a 5-year overall survival of 39% in this population.

INTRODUCTION

Multimodality therapy is the preferred approach for patients with stage IIIA non-small cell lung cancer (NSCLC) due to N2 mediastinal lymph node metastases, but the optimal management strategy has not been definitively established by randomized, controlled data [1–3]. The use of surgery is variable, but when combined with other modalities, likely provides survival benefit over non-operative management (chemotherapy and radiation) in selected patients [2,4]. However, several questions remain regarding both the selection of patients most appropriate for surgery in this setting and the most appropriate perioperative treatment strategy. Firstly, how the extent of mediastinal nodal disease should impact the decision to use surgery is not well characterized. Although it is generally recognized that surgery will most likely benefit patients with single-station, non-bulky (<3 cm) N2 disease [2], there are studies showing both a worse survival with multistation N2 disease [5–8] and no differences in survival between single versus multistation N2 NSCLC [9–11]. Secondly, guidelines specifically recommend induction therapy be given prior to surgical intervention [2], but it is still unclear if the induction strategy should be induction chemoradiation or induction chemotherapy alone [2]. Thirdly, patients that demonstrate a response to induction therapy likely have improved survival [1], but how to best perform mediastinal restaging and then use those results to decide on whether proceeding with resection is appropriate is also unclear [12, 13]. Finally, preliminary studies have demonstrated the feasibility of video-assisted thoracoscopic surgery (VATS) lobectomy for patients who have undergone induction therapy [14–16] but there have been no long-term studies comparing lobectomy by VATS and thoracotomy approaches in stage IIIA pN2 patients.

In this study, we aimed to evaluate the impact of these various clinical factors on long-term outcomes of patients who had lobectomy after induction therapy for stage IIIA pN2 NSCLC. Our objective was to further improve the selection of patients with stage IIIA (pN2) NSCLC who are most likely to benefit from surgery by providing clinicians with quantifiable evidence that may be used in the treatment decision process for this patient population.

METHODS

Patient selection and variables of interest

A retrospective analysis was performed of 111 consecutive patients with biopsy-proven stage IIIA pN2 NSCLC who received induction chemotherapy with or without radiation followed by lobectomy from 1996 to 2012 at Duke University Medical Center, with Institutional Review Board approval and waiving of individual patient consent. The American Joint Committee on Cancer (AJCC) 7th Edition of Lung Cancer Staging guidelines [17] were used to stage all lung cancers; patients staged with earlier versions of the AJCC guidelines were re-staged according to seventh edition standards.

Data were collected on basic demographics, comorbidities, prior thoracic surgery, Zubrod score, radiation use, induction chemotherapy regimens, clinical/pathologic T and N status before induction therapy, pathologic T and N stage after lobectomy, histology and pre- and postinduction therapy tumour size. The proportion of patients who had nodal down-staging after therapy (N2 to pN1/N0 after lobectomy) was calculated. During the study period, clinical staging techniques included positron emission tomographic (PET) scan, computed tomographic (CT) scan and magnetic resonance imaging. Pathologic evaluation of N2 lymph nodes prior to therapy was performed primarily using mediastinoscopy and less often endobronchial ultrasound (EBUS). Restaging after induction therapy but prior to resection was done via mediastinoscopy, EBUS, VATS or thoracotomy. The primary outcomes were overall and recurrence-free survival. Other perioperative outcomes analysed included perioperative complications, chest tube duration, length of hospitalization and 30-day all-cause mortality. Follow-up was complete for all patients. Post-discharge follow-up data were collected through clinic notes, direct contact with patients and physicians, a national death registry and postal questionnaire as collected by the Duke Cancer Institute.

Statistical analysis

Overall and recurrence-free survival of the entire cohort and subgroups were assessed using Kaplan–Meier analyses and log-rank test. The Kaplan–Meier graphs in the figures were trimmed to extend out to 5 years. Overall survival was determined as the time from lobectomy to death with patients censored upon last-available follow-up. Recurrence-free survival was determined from the time of lobectomy to death or recurrence as seen on imaging with patients censored at the time of last follow-up. Several a priori subgroup analyses were performed including analysis of overall and recurrence-free survival stratified by induction therapy regimen (induction chemotherapy alone versus induction chemoradiotherapy), invasive mediastinal restaging approach, surgical approach (VATS versus thoracotomy) and single-station versus multistation N2 disease prior to induction therapy. Multivariable Cox proportional hazards analysis was used to model overall survival adjusting for variables chosen in advance, based on clinical relevance, which include age, induction radiation use, pretreatment clinical T status, surgical approach (VATS versus thoracotomy), use of invasive mediastinal restaging (involving pathologic re-evaluation by frozen section analysis), single- versus multistation N2 disease (prior to induction therapy) and operative year. The proportional hazards assumption was tested for the Cox models using smooth scaled Schoenfeld residual plots and there were no violations of assumptions found. All statistical analyses were performed using the Stata Statistical Software version 13.1 (StataCorp LP, College Station, TX, USA).

RESULTS

Patient characteristics

A total of 111 consecutive patients with stage IIIA pN2 NSCLC who underwent induction chemotherapy with or without radiation followed by lobectomy were included in this study. Patient characteristics are summarized in Table 1. There were 70 total deaths and 62 total recurrences observed. The median age was 63 years and the majority of patients were male. The most common comorbidity was hypertension, but diabetes and coronary artery disease were also somewhat common in the patient population. Patients on average had mildly impaired pulmonary function as measured by forced expiratory volume in 1 s (FEV₁) and diffusing capacity for carbon monoxide (DLCO).

Table 1:

Patient characteristics

Characteristic
Sex (n, %)
Male	70 (63)
Female	41 (37)
Age (years)
Mean ± SD	62 ± 9.6
Median (25th–75th percentile)	63 (55–68)
Ethnicity (n, %)
White	95 (86)
Black	13 (12)
Native American	3 (3)
FEV1 (%)
Mean ± SD	76 ± 20
Median (25th–75th percentile)	78 (60–89)
DLCO (%)
Mean ± SD	74 ± 18
Median (25th–75th percentile)	72 (62–86)
BMI at operation
Mean ± SD	27.6 ± 5.1
Median (25th–75th percentile)	27.7 (23.7–30.5)
History of diabetes (n, %)	19 (17)
Renal insufficiency (n, %)	1 (1)
Hypertension (n, %)	54 (49)
COPD (n, %)	32 (29)
PVD (n, %)	6 (5)
Prior thoracic surgery (n, %)	32 (29)
Zubrod score (n, %)a
0	41 (37)
1	68 (61)
2	2 (2)
History of cancer (n, %)	19 (17)
Radiation (n, %)	56 (50)
TIA/CVA (n, %)	9 (8)

FEV₁: forced expiratory volume in 1 s; DLCO: diffusing capacity of carbon monoxide; BMI: body mass index; COPD: chronic obstructive pulmonary disease; PVD: peripheral vascular disease; TIA: transient ischaemic attack; CVA: cerebral vascular accident; SD: standard deviation.

^aZubrod score of 0: fully active with no restrictions; 1: restricted in strenuous activity but ambulatory and can carry out light work; 2: ambulatory and mobile for >50% of waking hours but cannot perform work activities.

Practice patterns

The induction therapy strategy and surgical approach to resection both evolved over the course of the study period. The use of induction radiation decreased over time (Fig. 1A). From 1996 to 2007, the majority of patients [46/65 (71%)] underwent induction chemoradiation. From 2007 to 2012, the majority of patients [36/46 (78%)] underwent induction chemotherapy. The use of VATS increased over the time period of the study (Fig. 1B). No VATS lobectomies were performed from 1996 to 2001, 4 VATS lobectomies were performed from 2002 to 2007 and 33 VATS lobectomies were performed from 2008 to 2012.

Figure 1:

(A) Trends in use of induction radiation. (B) Trends in use of VATS. VATS: video-assisted thoracoscopic surgery.

Throughout the study period, most patients [n = 93 (84%)] underwent pathologic mediastinal restaging after induction therapy but prior to resection. Regarding invasive mediastinal restaging strategy in this group, 21 (23%) patients underwent a mediastinoscopy (of which 5 were redo mediastinoscopies), 13 (14%) underwent VATS restaging and 59 (63%) underwent restaging by thoracotomy approach.

Post-treatment characteristics and outcomes

Tumour characteristics are specified in Table 2. Adenocarcinoma, seen in 54 (49%) patients, was the most commonly observed preinduction therapy histology. The majority of patients were down-staged from pN2 to pN1/N0. Complete pathologic response was observed in 21% of patients. More patients in the induction chemoradiation subgroup had complete pathologic response when compared with patients in the induction chemotherapy group (P = 0.01). As depicted in Table 3, perioperative complications were observed in 53 (48%) patients. The most common complication was atrial fibrillation. The 30-day mortality was 4%. The median length of hospitalization was 6.3 days.

Table 2:

Treatment and tumour details

Characteristic
Video-assisted thoracoscopic surgery (n, %)	37 (33)
Induction chemotherapy (n, %)
Epidermal growth factor receptor inhibitor	1 (1)
Taxane only	2 (2)
Dasatinib	1 (1)
Unknown	9 (8)
Platinum-doublet therapy	98 (88)
Multistation N2 disease prior to induction therapy (n, %)	27 (24)
Pre-induction therapy histology (n, %)
Adenocarcinoma	54 (49)
Squamous	27 (24)
Large cell	8 (7)
Non-small cell, not otherwise specified	22 (20)
Pre-induction therapy tumour sizea, mean (SD)	3.8 ± 2.3
Post-induction therapy tumour size (n, %), mean (SD)	2.4 ± 2.3
Down-staged from pN2 to pN1/N0 (n, %)	71 (64)
Down-staged from pN2 to pN0 (n, %)	57 (51)
Complete pathologic response (ypT0N0) (n, %)	23 (21)
Received induction chemoradiation	17 (30b)
Received induction chemotherapy	6 (11c)

SD: standard deviation.

^aData only available for 76/111 patients.

^bOf 56 patients who underwent induction chemoradiation.

^cOf 55 patients who underwent induction chemotherapy.

Table 3:

Perioperative outcomes

Characteristics
30-day mortality (n, %)	4 (4)
Perioperative complications (n, %)	53 (48)
Atrial fibrillation (n, %)	23 (21)
Prolonged air leaks (n, %)	15 (14)
Postoperative bleeding requiring blood transfusion (n, %)	10 (9)
Pneumonia (n, %)	6 (5)
Postoperative bleeding requiring reoperation (n, %)	3 (3)
Respiratory failure (n, %)	2 (2)
Other major complications (n, %)	10 (9)
Chest tube duration (days)
Median (25th–75th percentile)	3 (3–5)
Length of hospitalization (days)
Median (25th–75th percentile)	5 (4–6)

Overall survival

The median overall survival and 5-year overall survival of the overall cohort were 30.1 months [95% confidence interval (CI), 22.0–56.4] and 39% (95% CI, 29–49), respectively, with a median follow-up of 24 months (first quartile to third quartile, 12–51) (Fig. 2). Ninety-seven percent of all patients had complete follow-up (deceased or alive with known clinical status) beyond 12 months.

Figure 2:

Overall survival of patients with stage IIIA-pN2 who underwent induction therapy followed by lobectomy.

In univariable analysis, patients treated with induction chemoradiation exhibited a worse overall survival (log-rank P = 0.021) compared with patients treated with induction chemotherapy alone [5-year survival: 31.1% (95% CI, 19.0–43.9) versus 47.6% (95% CI, 31.8–61.9)] (Fig. 3). Patients who underwent VATS had significantly better overall survival compared with those who had surgery by thoracotomy approach [5-year survival: 56.6% (95% CI, 33.9–74.1) vs 31.4% (95% CI, 20.8–42.5); log-rank, P = 0.007]. There were no significant differences in overall survival (log-rank, P = 0.065) between patients with single-station N2 and multistation N2 disease [5-year survival: 37.3% (95% CI, 25.8–48.7) vs 42.8% (95% CI, 22.5–61.7)], nor was there any significant difference in overall survival between patients who were down-staged compared with those who were not down-staged [5-year survival: 44.7% (95% CI, 31.9–56.7) vs 31.3% (95% CI, 16.4–47.4); log-rank P = 0.18].

Figure 3:

Overall survival stratified by induction therapy strategy.

Patients who underwent pathologic mediastinal restaging following induction therapy but prior to resection had an improved overall survival compared with patients who did not undergo pathologic mediastinal restaging [5-year survival: 45.2% (95% CI, 33.9–55.9) vs 13.9% (95% CI, 2.5–34.7); log-rank P = 0.004] (Fig. 4A). Of the patients who did not get invasive mediastinal restaging, 13 of 18 (72%) patients had residual N2 disease, including 6 patients with multistation disease. In contrast, of the 93 patients who did undergo invasive mediastinal restaging, 84 had no residual N2 disease found at the time of restaging whereas 9 had residual microscopic N2 disease (of these 9 patients, 8 patients had microscopic N2 disease on final pathology, whereas 1 did not).

Figure 4:

(A) Overall survival stratified by invasive mediastinal restaging approach. (B) Recurrence-free survival stratified by invasive mediastinal restaging approach.

Recurrence-free survival

The median recurrence-free survival and 5-year recurrence-free survival were 17.5 months (95% CI, 13.7–27.5) and 28% (95% CI, 18–39), respectively.

The induction chemoradiotherapy group did not have a significantly different recurrence-free survival when compared with the induction chemotherapy group [5-year recurrence-free survival: 24.7% (95% CI, 13.1–38.2) vs 22.5% (95% CI, 10.0–38.0), log-rank P = 0.36]. No significant difference was observed in recurrence-free survival between VATS and thoracotomy approaches [5-year recurrence-free survival: 27.3% (95% CI, 12.0–45.2) vs 22.3% (95% CI, 12.7–33.7); log-rank P = 0.17], nor between the single-station group and multistation group [5-year recurrence-free survival: 26.1% (95% CI, 15.4–38.2) vs 17.0% (95% CI, 4.9–35.6); log-rank P = 0.30]. Patients who were down-staged did show an improved recurrence-free survival compared with those who were not down-staged [5-year recurrence-free survival: 37.0% (95% CI, 22.7–51.2) vs 9.8% (95% CI, 1.9–25.7); log-rank P < 0.001]. Patients who underwent pathologic mediastinal restaging also had an improved recurrence-free survival when compared with patients who did not undergo pathologic mediastinal restaging [5-year recurrence-free survival: 31.9% (95% CI, 20.1–44.4%) vs 7.8% (0.5–29.5%); log-rank P = 0.008] (Fig. 4B).

Multivariable analysis

As illustrated in Table 4, the impact of important patient characteristics and treatment strategies were assessed in a multivariable analysis. Pathologic mediastinal re-evaluation prior to resection was associated with improved survival (HR, 0.39; 95% CI, 0.21–0.72; P = 0.003). The operative year, extent of mediastinal disease prior to induction therapy and particular induction therapy strategy or surgical approach used were not important predictors of survival.

Table 4:

Cox proportional hazards model

	Hazard ratio	95% CI	P-value
Age (years)	1.02	1.00, 1.05	0.11
Male	0.95	0.57, 1.57	0.84
Induction radiation	1.48	0.87, 2.53	0.15
Pretreatment clinical T status	1.01	0.85, 1.20	0.89
VATS	0.52	0.22, 1.19	0.12
Pathologic mediastinal re-evaluation prior to resection	0.39	0.21, 0.72	0.003
Multistation N2	0.75	0.42, 1.34	0.33
Operative year	1.01	0.93, 1.08	0.87

VATS: video-assisted thoracoscopic surgery; CI: confidence interval.

DISCUSSION

The present study evaluated the impact of patient selection and multiple treatment strategies on perioperative outcomes and survival of 111 patients with biopsy-proven stage IIIA (N2) NSCLC who underwent induction therapy followed by lobectomy. During the study period, there was an evolution in practice patterns. Initially, induction chemoradiation and lobectomy by thoracotomy approach were more commonly used. Over time, practice evolved to predominant use of induction chemotherapy and VATS without any apparent compromise in outcomes. Pathologic mediastinal restaging prior to resection was routinely used at a high frequency throughout the study period, and was the only factor associated with improved survival in multivariable analysis.

In the present study, the extent of mediastinal disease was not an important predictor of survival and this finding is consistent with the results reported by several studies [9–11] although there are also a significant number of studies that have shown worse survival with multistation N2 disease [5–8]. At our institution, the common practice was to operate on patients with single-station N2 disease and on patients without gross disease seen on pathologic mediastinal re-evaluation; most of the patients either had microscopic N2 or no N2 disease after induction therapy as determined by a pathologist from frozen sections. Patients with multistation N2 disease after induction therapy were typically not resected, so patients with multistation disease who did have surgical resection may have been considered by their surgeons to have characteristics that conferred a better prognosis and warranted the risk of performing resection. This selection bias could have accounted for our finding that there were no significant differences in overall survival between patients with single- and multistation N2 disease.

Although at our institution we operate primarily on patients with single-station N2 disease, a recent survey of 373 medical oncologists found that 48% would include surgery as part of the treatment plan for bulky multistation N2 disease [18] whereas another survey of 2539 surgeons found that 62% would operate on single-station macroscopic N2 disease only if the patient's disease was down-staged after induction therapy [19]. The differences in patient selection related to N2 status (microscopic versus macroscopic, single- versus multistation) between institutions may explain the discrepancy between our findings and the results of other studies that reported worse survival with multistation N2 disease.

Our finding that induction chemoradiation was associated with increased pathologic response but not improved survival when compared with induction chemotherapy is consistent with results previously reported by other retrospective studies, randomized trials and by a recent meta-analysis [20]. There are several hypotheses that need to be further tested in future studies which could explain our finding. One possible explanation is that whether a patient has improved survival depends primarily on whether the patient has micrometastatic disease that is responsive to induction chemotherapy and that there was a similar number of patients with a similar degree of micrometastatic disease responsive to chemotherapy in both the induction chemoradiation and chemotherapy groups. A second possibility is that the induction chemoradiation group may have been comprised of more patients with bulkier N2 and more aggressive disease when compared with the induction chemotherapy group. A third consideration is that we did not have data on how many patients were initially assigned induction chemotherapy but later did not proceed with resection; it is possible that patients in the induction chemotherapy group selected for surgery had tumours that were particularly chemoresponsive. Finally, during the study period, there was a transition in our practice from predominant use of induction chemoradiation to predominant use of induction chemotherapy; chemotherapy regimens and patient care may have been improved over the study time period as well, although in multivariable analysis, operative year was not a significant factor associated with survival.

The present study also assessed the outcomes of VATS lobectomy for stage IIIA pN2 NSCLC and found that VATS does not appear to compromise oncologic outcomes. This is generally consistent with a few other preliminary reports finding VATS to be a safe and feasible approach for patients who have undergone induction therapy followed by lobectomy [14–16], although the present study is the first to specifically evaluate the outcomes in a population of stage IIIA pN2 patients. Our finding is consistent with other studies demonstrating that VATS does not compromise oncologic efficacy for earlier stage disease [21–23].

Importantly, we found that the use of pathologic mediastinal restaging following induction therapy but prior to resection was associated with improved survival. Of note, the study results do not suggest that the procedure of pathologic restaging improves survival but rather suggest that pathologic restaging allowed for more appropriate selection of patients for surgery. The majority of patients who went straight to resection without restaging had residual N2 disease and poor long-term survival. Having a strategy of restaging could prevent patients who have non-chemoresponsive tumours from undergoing an operation that does not significantly increase their chance of cure. Of note, restaging also allowed selection for not only patients who were down-staged but also for patients with residual N2 disease that was thought to be minimal enough to portend a relatively good prognosis. In our practice, a small but significant number of patients underwent VATS restaging and several of these patients were part of the CALGB 39803 Trial demonstrating the feasibility of VATS restaging [12].

Currently, there are different treatment algorithms for invasive mediastinal staging and for treatment of biopsy-proven N2 disease [24, 25]. Our current practice is to have patients undergo mediastinal evaluation with PET followed by EBUS when N2 disease is suspected. Patients pathologically proven to have N2 disease then undergo induction chemotherapy with platinum-doublet chemotherapy followed by restaging with mediastinoscopy and then lung resection. Alternatively, if mediastinoscopy is used for primary mediastinal staging, consistent with previously described algorithms [24], the strategy is to use either VATS or repeat mediastinoscopy for mediastinal staging (the latter being particularly valuable if the left paratracheal nodes are involved).

This study has several limitations. First, it is a single-institution retrospective study with a relatively small number of patients with findings that may not be generalizable to other institutions. Second, the follow-up was relatively short, with a median follow-up of 2 years. Third, although we have tried to account for practice changes over the course of the study period by including operative year as a co-variate in the multivariable analysis, there likely have been changes in care and practice unaccounted for. Fourth, we did not have data on the number of patients that underwent induction therapy initially but who did not proceed with surgery. As the initial denominator is not known and analysis is limited to patients who underwent an operation, the results cannot be generalized to the whole patient population with clinical N2 disease. Fifth, although we had data on whether the patient had single- vs multistation disease prior to induction therapy, we do not have complete data on whether the N2 disease prior to induction therapy was microscopic or macroscopic.

In this retrospective analysis of 111 patients with stage IIIA pN2 NSCLC who underwent induction therapy followed by lobectomy, the extent of mediastinal disease, the particular induction therapy strategy used, and the surgical approach taken were not significant predictors of survival. However, the strategy of assessing response to induction therapy with pathologic mediastinal restaging allows one to select appropriate patients for complete resection and is associated with a 5-year overall survival of 39% in this population.

Funding

Chi-Fu Jeffrey Yang is supported by the American College of Surgeons Resident Research Scholarship. Robert Ryan Meyerhoff is supported by a MSTP T32 grant (Medical Scientist Training Program NSRA T32GM007171). Matthew G. Hartwig is supported by the NIH funded Cardiothoracic Surgery Trials Network, 5U01HL088953-05.

Conflict of interest: Thomas A. D'Amico is a consultant for Scanlan International, Inc.