Lung Cancer 115 (2018) 75–83

Contents lists available at ScienceDirect

Lung Cancer
journal homepage: www.elsevier.com/locate/lungcan

Long-term outcomes of surgical resection for stage IV non-small-cell lung T

cancer: A national analysis
Chi-Fu Jeffrey Yanga, Lin Gub, Shivani A. Shaha, Babatunde A. Yerokuna, Thomas A. D’Amicoa,
⁎
Matthew G. Hartwiga, Mark F. Berryc,
a
Department of Surgery, Division of Thoracic Surgery, Duke University Medical Center, 3496 DUMC, Durham, NC, United States
b
Department of Biostatistics, Duke University, United States
c
Department of Cardiothoracic Surgery, Stanford University Medical Center, 300 Pasteur Drive, Falk Building 2nd Floor, Stanford, CA 94305-5407, United States

A R T I C L E I N F O A B S T R A C T

Keywords: Objective: Treatment guidelines recommend surgical resection in select cases of stage IV non-small-cell lung
Lung cancer surgery cancer (NSCLC) but are based on limited evidence. This study evaluated outcomes associated with surgery in
Stage IV lung cancer stage IV disease.
Methods: Factors associated with survival of stage IV NSCLC patients treated with surgery in the National Cancer
Date Base (2004–2013) were evaluated using multivariable Cox proportional hazards analyses. Outcomes of the
subset of patients with cT1-2, N0-1, M1 and cT3, N0, M1 disease treated with surgery or chemoradiation were
evaluated using Kaplan-Meier analyses.
Results: The five-year survival of all stage IV NSCLC patients who underwent surgical resection (n = 3098) was
21.1%. Outcomes were related to the locoregional extent of the primary tumor, as both increasing T status (T2
HR 1.30 [p < 0.001], T3 HR 1.28 [p < 0.001], and T4 HR 1.28 [p < 0.001], respectively, compared to T1)
and nodal involvement (N1 HR 1.34 [p < 0.001], N2 HR 1.50 [p < 0.001], and N3 HR 1.49 [p < 0.001],
respectively, compared to N0) were associated with worse survival. Outcomes were also related to the extent of
surgical resection, as pneumonectomy (HR 1.58, p < 0.001), segmentectomy (HR 1.36, p = 0.009), and wedge
resection (HR 1.70, p < 0.001) were all associated with decreased survival when compared to lobectomy. The
five-year survival of cT1-2, N0-1, M1 and cT3, N0, M1 patients was 25.1% (95% CI: 22.8–27.5) after surgical
resection (n = 1761) and 5.8% (95% CI: 5.2–6.5) after chemoradiation (n = 8180).
Conclusions: Surgery for cT1-2, N0-1, M1 or cT3, N0, M1 disease is associated with a 5-year survival of 25% and
does not appear to compromise outcomes when compared to non-operative therapy, supporting guidelines that
recommend surgery for very select patients with stage IV disease. However, surgery provides less benefit and
should be considered much less often for stage IV patients with mediastinal nodal disease or more locally ad-
vanced tumors.

1. Introduction disease and otherwise early-stage local disease [3–5].

The majority (∼55%) of patients with non-small-cell lung cancer
(NSCLC) have stage IV disease with distant metastases at the time of
diagnosis [1]. The generally recommended treatment for these patients
is systemic chemotherapy, but the prognosis is very poor with a 5-year
survival of only 4.9% [1,2]. However, there are unusual NSCLC pre-
sentations where tumor biology is such that even patients with distant
metastatic disease are potentially curable with multimodality treatment
[3]. Although there is no role for local therapy of the primary tumor for
most patients with stage IV disease, aggressive local therapy may be
appropriate for selected patients with limited-site oligometastatic
Treatment guidelines reflect the potential benefit of surgery in select
stage IV patients. For patients who have single brain or adrenal me-
tastases but a primary tumor that is otherwise T1-2, N0-1 or T3, N0, a
management strategy included in the National Comprehensive Cancer
Network (NCCN) guidelines is local treatment of the metastasis fol-
lowed by resection of the lung lesion in combination with che-
motherapy either before or after lung resection [6]. The NCCN guide-
lines also recommend that patients with lung cancer and a solitary
contralateral lung nodule be treated as if they have two primary lung
tumors if both are curable [6]. This aggressive strategy has been re-
ported to have 5-year survival rates of 7–21% for brain metastases

⁎
Corresponding author at: Falk Cardiovascular Research Center, 300 Pasteur Dr, Stanford, CA 94305, United States.
E-mail address: berry037@stanford.edu (M.F. Berry).

https://doi.org/10.1016/j.lungcan.2017.11.021
Received 18 May 2017; Received in revised form 17 November 2017; Accepted 22 November 2017
0169-5002/ © 2017 Elsevier B.V. All rights reserved.
C.-F.J. Yang et al.
[7–12], and 12–40% for adrenal metastases [13–18].
However, published outcomes related to these management strate-
gies are generally single-institution studies of likely highly selected
patients. The nature of these unusual clinical scenarios is such that
meta-analyses or large prospective studies of patients are limited or not
available. The practice of resection of special instances of stage IV
NSCLC is still influenced by the subjective opinion of clinicians; the
NCCN guidelines mention that some NCCN Panel Members feel that
local therapy for adrenal metastases is only advisable if the synchro-
nous lung disease is stage I or possibly stage II [6]. As such, clinicians
have extremely limited evidence to guide therapy or even provide a
patient an estimate of their prognosis with an aggressive therapy re-
gimen that does have some risk of treatment-related morbidity.
This study was undertaken to quantify outcomes associated with
surgical therapy of the primary pulmonary tumor in patients with stage
IV NSCLC using the National Cancer Database (NCDB) with the fol-
lowing objectives: (1) identify patient and tumor factors associated with
survival for patients that underwent surgical resection to potentially
improve the patient selection process and (2) test the hypothesis that
surgical resection is associated with improved survival over chemor-
adiation in patients whose local tumor extent was stage I–II (cT1-2, N0-
1 or cT3, N0).
2. Methods
2.1. National Cancer Data Base
Jointly managed by the American Cancer Society and the American
College of Surgeons (ACS) Commission on Cancer (CoC), the National
Cancer Data Base (NCDB) contains information on approximately 70%
of all newly diagnosed cases of cancer in the United States and Puerto
Rico [19]. The NCDB includes data from over 1500 cancer centers in
the U.S. with records of > 30 million patients [20]. Clinical staging
information is recorded in the NCDB using the American Joint Com-
mittee on Cancer (AJCC) 6th and 7th edition TNM classifications for the
years of study inclusion [21,22].
2.2. Study design
All patients in the NCDB diagnosed with stage IV NSCLC
(2004–2013) were identified using International Classification of
Diseases for Oncology, 3rd edition (ICD-O-3) histology (adenocarci-
noma, squamous, adenosquamous, and large cell carcinoma) and to-
pography codes. Patients were excluded if they exhibited non-malig-
nant pathology, had history of previous unrelated malignancy, or
received only palliative treatment. From this patient group, we created
two patient cohorts for analysis.
2.3. Cohort of all surgical patients
The first cohort was all patients with stage IV disease who under-
went surgical resection with or without chemotherapy and with or
without radiation. All patients with T1-4, N-3 disease were included in
this analysis to assess the impact of disease stage on survival of surgery
patients. A multivariable Cox proportional hazards model was per-
formed using this cohort to assess factors associated with survival. The
factors that were adjusted in the model included: age, sex, race, CDCC
score, clinical T status, clinical N status, facility type, histology, tumor
location, insurance type, type of adjuvant or neoadjuvant therapy
(chemotherapy, radiation, chemoradiation, or none), and extent of re-
section (wedge resection, segmentectomy, lobectomy, or pneumo-
nectomy).
2.4. Cohort of cT1-2, N0-1 and cT3, N0 patients
The second cohort were those patients with clinical T1-2, N0-1, M1
76
Lung Cancer 115 (2018) 75–83
and T3, N0, M1 disease who received the following: 1) surgery with or
without chemotherapy and with or without radiation, 2) chemotherapy
alone, 3) radiation alone, or 4) chemoradiation. Patients were divided
into four groups by the type of primary treatment described above. The
Kruskal-Wallis test for continuous variables and Pearson’s chi-square
test for discrete variables were used to compare baseline characteristics
and unadjusted outcomes. A multivariable logistic regression was per-
formed to evaluate predictors of use of surgery (vs no surgery) and
included the following covariates: age, sex, race, Charlson Deyo co-
morbidity (CDCC) score, tumor location, histology, clinical T status,
clinical N status, facility type, and insurance type. The Kaplan-Meier
product limit approach was used to estimate median survival and 5-
year survival for each subgroup of patients. Within the surgery group,
we calculated the survival stratified by each substage of stage IV NSCLC
and stratified by type of operation. The primary outcome of overall
survival was compared between groups using the log-rank test.
To address the possibility of selection bias where patients who un-
derwent surgery may simply be the healthiest patients with the most
indolent of stage IV disease, we performed an exploratory analysis by
comparing the overall survival of 1761 patients who underwent surgery
with the 1761 patients who had the longest survival after being treated
with chemotherapy and radiation using Kaplan Meier analysis. We also
performed a propensity-score-matched analysis to create a cohort of
patients who underwent surgery with similar baseline characteristics to
patients who received chemotherapy and radiation, and to attempt to
control for imbalance between the two groups. A logistic regression
model was fitted to calculate the propensity score for the probability of
a patient receiving surgical treatment. The following were adjusted as
independent variables in the model: age, sex, race, CDCC, clinical T&N
stages, histology type, treating facility type, and insurance status. Using
a greedy 1:1 matching algorithm [23,24], 3512 patients were matched
by propensity score.
All of the above-mentioned analyses were repeated in a sensitivity
analysis that focused only on patients who were diagnosed as having
stage IV disease from years 2010–2013 in accordance to the AJCC 7th
edition. We did this analysis because multiple tumors in the same lung
were originally classified in the 6th edition as being M1 disease [21]. In
AJCC 7th edition, this is now considered T4 disease [22]. In the 7th
edition, malignant pleural and pericardial effusions now have been
reclassified from T4 to M1a [22]. In these sensitivity analysis, the
variable “site of metastasis” was added to the modeling because me-
tastasis site data (data on contralateral lung, pericardial and pleural
effusion, brain, bone and liver) became available from 2010 onwards.
All statistical analyses were performed using SAS for Windows,
Version 9.4; SAS Institute Inc.; Cary, NC. A 2-sided p-value of 0.05 was
used to define the statistical significance.
3. Results
3.1. Outcomes associated with surgery
Between 2004 and 2013, surgery was utilized in 3098 patients with
clinical M1 NSCLC in the NCDB (Fig. 1). Characteristics of these pa-
tients are detailed in Table 1. Overall 5-year survival of this cohort was
21.1% (Fig. 2A). Factors associated with increased survival in these
patients who underwent surgery are detailed in Table 2 and were
consistent in the sensitivity analysis using only patients staged by AJCC
7th edition (Table A.6). Survival stratified by different sub-stages are
shown in Fig. 2B.
3.2. Impact of surgery for patients with cT1-2, N0-1 or cT3, N0 disease
Of the 94,672 patients from Fig. 1 who had cM1 disease and re-
ceived some combination of chemotherapy, radiation, or surgery, there
were 21,108 cT1-2, N0-1 or cT3, N0 patients who received treatment.
Baseline characteristics stratified by the different treatments for this
C.-F.J. Yang et al.
Fig. 1. Study Design (AJCC 6th and 7th edition cT1-4, N0-3, M1 patients).
cT1-2, N0-1 or cT3, N0 cohort are detailed in Table 3. Patients treated
with chemoradiation or surgery were younger than patients treated
with only chemotherapy or radiation. Compared to others, the surgery
group contained a higher percentage of patients who were female,
white, had higher CDCC scores, and lived further from the treating
hospital. The surgery group was most likely to have T1, N0, M1 disease
when compared to the other groups. The tumors of surgical patients
were smaller than the non-operative groups.
In the surgery group, 384 (21.8%) patients received chemotherapy,
604 (34.3%) received chemoradiation, 358 (20.3%) received radiation
and 415 (23.6%) did not receive additional therapy. Data on whether
the patient had chemotherapy and/or radiation preoperatively or
postoperatively are detailed in Table A.1. The type of radiation therapy
used is detailed in Table A.2. Most patients underwent lobectomy
(n = 1113, 63.2%) while 507 (28.8%) received wedge resection. A
small percentage of patients underwent segmentectomy (3.3%, n = 58)
and pneumonectomy (4.7%, n = 83) (Table A.3). The age and CDCC
scores of surgical patients are detailed in Table A.4. The median (IQR)
and mean (SD) length of hospital stay were 5 (3, 7) and 7 (9) days,
respectively. The overall 30-day mortality was 4.1% (n = 72) and the
overall 90-day mortality was 11.0% (n = 194). The 90-day mortality
was 8.9% (n = 99) in the lobectomy group, 16.2% (n = 82) in the
wedge resection group, 8.6% (n = 5) in the segmentectomy group, and
9.6% (n = 8) in the pneumonectomy group. The readmission rate
within 30 days of discharge was 8.7% (n = 153).
Predictors of use of surgery (Table A.5) included younger age and
increasing CDCC score. When compared to patients with
77
Lung Cancer 115 (2018) 75–83
adenocarcinoma histology, patients with adenosquamous and large cell
carcinoma were more likely to undergo surgery. Increasing T status, N1
(compared to N0) status, community and comprehensive (compared to
academic) facilities and lack of insurance or Medicare/aid (compared to
private insurance) were associated with decreased use of surgery.
Kaplan–Meier analysis demonstrated a 5-year survival of 25.1%
(95% CI: 22.8–27.5) for patients who underwent surgery, 5.8% (95%
CI: 5.2–6.5) for patients who received chemoradiation, 5.9% (95% CI:
5.1–6.8) for patients who received only chemotherapy, and 3.2% (95%
CI: 2.7–3.8) for patients who received radiation alone (Fig. 3). Among
the patients who underwent surgery, increasing substage was generally
associated with decreased survival (Fig. A.1A), while patients who
underwent lobectomy had the best survival when compared to patients
who underwent wedge resection, segmentectomy, or pneumonectomy
(Fig. A.1B).
In an exploratory analysis, we compared the overall survival of
1761 patients who underwent surgery with the 1761 patients who
underwent chemotherapy and radiation who had the longest survival
(Fig. A.2A). All patients in the chemotherapy and radiation group
survived to at least 1.8 years. The surgery group had better 5-year
survival (25.1% [95% CI: 22.8–27.5]) when compared to the chemor-
adiation group (24.1% [95% CI: 21.6–26.6]). Propensity-score
matching (Table A.6) also found surgery patients to have better survival
than those treated with chemoradiation (25.2% [95% CI: 22.8–27.6]
versus 6.2% [95% CI: 4.8–7.7], p < 0.001, Fig. A.2B).
Our current study does include a small percentage of pa-
tients—7.5% (n = 1583)—diagnosed in 2004–2009 who were
C.-F.J. Yang et al. Lung Cancer 115 (2018) 75–83

Table 1 Table 1 (continued)

Preoperative and Demographic Characteristics of All cT1-4, N0-3, M1 Surgical Patients.
Total (N = 3098)
Total (N = 3098)
Large cell 181 (5.8%)
Age (Years) at Diagnosis Adenosquamous 134 (4.3%)
Median, IQR 64 (56, 71)
Size of Tumor (mm)
Age, n (%) Median, IQR 30 (20, 49)
< 50 305 (9.8%)
Separate Tumor Nodules −Ipsilateral Lung, n (%)
50–69 1867 (60.3%)
No separate tumor nodules in ipsilateral lung 1217 (39.3%)
70–79 762 (24.6%)
Separate tumor nodules in ipsilateral lung, same lobe 186 (6.0%)
> 79 164 (5.3%)
Separate tumor nodules in ipsilateral lung, different lobe 193 (6.2%)
Sex, n (%) Separate tumor nodules, ipsilateral lung, same and 110 (3.6%)
Male 1593 (51.4%) different lobe
Female 1505 (48.6%) Separate tumor nodules, ipsilateral lung, unknown if same 40 (1.3%)
or different lobe
Race, n (%)
Unknown if separate tumor nodules, separate tumor 1352 (43.6%)
White 2707 (87.4%)
nodules cannot be assessed, or not documented in
Black 299 (9.7%)
patient record
Other 92 (3.0%)

Charlson-Deyo Comorbidity Score, n (%)

0 1827 (59.0%) classified under the 6th edition as having M1 disease because they had
1 927 (29.9%)
different tumors in the same lung; these patients would be considered in
2 344 (11.1%)
the 7th edition as not having M1 disease. This included 150 (8.5%)
Insurance Status, n (%)
patients in the surgery cohort. The study schema, preoperative and
Not Insured 126 (4.1%)
Private 1269 (41.0%) demographic characteristics for patients staged only by AJCC 7th edi-
Medicare/aid 1626 (52.5%) tion are shown in Fig. A.5 and Table A.9. The primary analyses per-
Other Gov. 38 (1.2%) formed above were repeated in sensitivity analyses where we only in-
Unknown 39 (1.3%) cluded patients staged by AJCC 7th edition (Tables A.10–A.13, and
Distance from Hospital (miles)
Figs. A.6–A.8). The findings of the sensitivity analyses were similar and
Median, IQR 10 (4, 25) consistent with the findings from the primary analysis.

Distance from Hospital (miles), n (%)

<1 101 (3.3%) 4. Discussion
1–20 2044 (66.0%)
21–40 467 (15.1%)
41–60 178 (5.7%) In this national analysis of patients with stage IV NSCLC in the
> 60 244 (7.9%) NCDB who underwent surgical resection, nodal involvement and higher
Unknown 64 (2.1%) T status were found to be associated with worse survival. The use of any
Facility Type, n (%) surgery other than lobectomy (wedge resection, segmentectomy and
Community Cancer Program 278 (9.0%) pneumonectomy) was also associated with worse survival. These results
Comprehensive Community Cancer Program 1370 (44.2%) indicate that patients with less locally advanced tumors have better
Academic/Research Program 1160 (37.4%)
long-term outcomes, which is consistent with what would be expected
Integrated Network Cancer Program 290 (9.4%)
based on clinical intuition and experience. These results also suggest
Tumor Location, n (%)
that utilizing surgery only when patients have small tumors or when
RUL 983 (31.7%)
RML 129 (4.2%) nodal disease is not present may be the optimal balance of benefits and
RLL 469 (15.1%) risks of therapy. Indeed, in the subset of patients who underwent
LUL 844 (27.2%) treatment for cT1-2, N0-1, M1 or cT3, N0, M1 NSCLC, we found that
LLL 386 (12.5%) surgical resection of the primary tumor was associated with a five-year
Main Bronchus 21 (0.7%)
Overlapping Lesion 59 (1.9%)
survival of 25.1%, which was better than the “best surviving” patients
Other 207 (6.7%) who underwent chemotherapy and radiation as well as a group of
chemoradiation patients created using propensity-score matching. In
TNM Stage, n (%)
T1, N0 615 (19.9%) both unadjusted and multivariable analysis, long-term outcomes were
T2, N0 625 (20.2%) best for patients in the subset of the surgery group who were younger
T3, N0 201 (6.5%) and whose primary lung tumors were small, N0, and were treated with
T4, N0 315 (10.2%) lobectomy. These are all characteristics that clinicians can consider
T1, N1 110 (3.6%)
T2, N1 210 (6.8%)
when choosing a treatment regimen for these types of patients. The
T3, N1 63 (2.0%) study findings suggest that having an aggressive treatment strategy that
T4, N1 75 (2.4%) includes surgery is not unreasonable in highly select patients.
T1, N2 164 (5.3%) While the NCDB does not detail the nature and distribution of pa-
T2, N2 240 (7.7%)
tients’ metastatic disease, our analysis of prognostic factors such as age,
T3, N2 120 (3.9%)
T4, N2 216 (7.0%) sex, and TNM stage is consistent with previous findings. Previous stu-
T1, N3 38 (1.2%) dies that have shown that prognoses of patients with metastases to the
T2, N3 29 (0.9%) brain were better for patients who were younger, female, have lower T
T3, N3 24 (0.8%) status, good performance status, and metachronous presentation
T4, N3 53 (1.7%)
[8,9,12,25,26]. There are also some findings in our study that are dif-
Histology, n (%) ferent from previously reported results. We found that patients under-
Adenocarcinoma 2205 (71.2%)
going surgery with large cell carcinoma had worse survival. This is in
Squamous 578 (18.7%)
contrast to previous studies that did not find associations between
histologic subtype and survival [3,14].

78
C.-F.J. Yang et al. Lung Cancer 115 (2018) 75–83

Fig. 2. Overall Survival of cT1-4, N0-3, M1 Surgical Patients (Fig. 2A) and Stratified by Substage (Fig. 2B).

Most NSCLC patients have stage IV disease with distant metastases approximately 4.9% [1,2]. However, aggressive local therapy may be
at the time of diagnosis [1]. Treatment is generally palliative in these appropriate for selected patients with limited-site oligometastatic dis-
circumstances and prognosis is very poor with a 5-year survival of ease and otherwise early-stage local disease [3–5]. Retrospective

79
C.-F.J. Yang et al. Lung Cancer 115 (2018) 75–83

Table 2 guidelines, but possibly are more influenced by individual or institu-

Cox Proportional Hazard Multivariable Modeling for All Patients Who Underwent tional beliefs or preferences.
Surgery.
The clinical scenario where lung resection may be appropriate in the
Hazard Ratio 95% CI p-value setting of stage IV disease is very uncommon. In our NCDB analysis, of
the 123,985 patients with M1 disease who met selection criteria, only
Age (unit = 10yrs) 1.01 1.01, 1.02 < 0.001 27,868 (25%) had a clinical stage of cT1-2, N0-1, M1 or cT3, N0, M1.
Female vs. male 0.75 0.69, 0.82 < 0.001
The extremely limited number of patients with potentially operable
Race (ref = white) stage IV disease is a main reason why current evidence for this topic is
Black 0.92 0.80, 1.07 0.27 generally limited to small single-institution studies. A strength of this
Other 0.62 0.47, 0.83 0.001
study is that by using the NCDB, we could assemble a cohort of over
Neo or adjuvant therapy (ref = none) 3000 stage IV patients who were treated with lung resection. In addi-
Chemotherapy (only) 0.64 0.57, 0.72 < 0.001
tion to demonstrating long-term survival comparable to what has been
Chemoradiation 0.81 0.72, 0.92 < 0.001
Radiation (only) 1.31 1.15, 1.50 < 0.001 reported by previous small studies, [7–17,27–30] we were able to
perform multivariable analyses to identify patient and tumor char-
Surgery (ref = lobectomy)
Pneumonectomy 1.58 1.31, 1.90 < 0.001 acteristics that are associated with survival and that can be used in the
Segmentectomy 1.36 1.08, 1.71 0.009 patient selection process.
Wedge 1.70 1.55, 1.88 < 0.001 This study has several limitations. First, detailed information on the
CDCC Score nature and distribution of patients’ metastatic disease is not available.
1 1.05 0.96, 1.16 0.29 Most studies examining outcomes after lung resection in the setting of
2 1.11 0.97, 1.27 0.13 stage IV disease have typically only included patients with solitary
Clinical T Status (ref = T1) metastatic disease in the brain or adrenal gland [3,4]. In this study, we
T2 1.30 1.17, 1.45 < 0.001 are not able to account for detailed information regarding the extent of
T3 1.28 1.11, 1.47 < 0.001
stage IV disease although we were able to broadly identify which pa-
T4 1.28 1.13, 1.45 < 0.001
tients had pleural or pericardial effusions and which patients had me-
Clinical N Status (ref = N0)
tastases to the brain, bone, liver and contralateral lung in sensitivity
N1 1.34 1.19, 1.51 < 0.001
N2 1.50 1.35, 1.66 < 0.001
analyses. It is possible that some patients who underwent surgery had
N3 1.49 1.23, 1.81 < 0.001 stage IV disease due to malignant pericardial or pleural effusions or
Tumor Location (ref = RUL)
more wide-spread disease in multiple sites, although general practice
RML 1.14 0.92, 1.42 0.23 guidelines would discourage surgery in those settings [6]. The NCDB
RLL 1.33 1.17, 1.51 < 0.001 also does not record if and how the metastatic disease was treated with
LUL 1.02 0.92, 1.14 0.70 other local therapy such as radiation or surgical resection. Further,
LLL 1.04 0.91, 1.20 0.55
because detailed information on the metastases are unavailable (e.g.
Other 1.13 0.96, 1.32 0.14
size of brain metastases), it is possible that among patients with the
Histology (ref = adenocarcinoma)
same T and N status, patients who underwent surgery had less disease
Adenosquamous 1.21 0.99, 1.48 0.06
Large cell 1.40 1.17, 1.67 < 0.001
burden when compared to patients undergoing non-operative treat-
Squamous 1.05 0.94, 1.18 0.37 ment.
Facility Type (ref = academic)
A second limitation of our study is that the NCDB does not explicitly
Community 1.39 1.20, 1.62 < 0.001 record how stage IV disease is established. Some patients with other-
Comprehensive 1.25 1.14, 1.37 < 0.001 wise early-stage disease may have been over-staged as stage IV based on
Integrated Network 1.18 1.01, 1.37 0.04 radiographic studies without pathologic confirmation. Patients may
Insurance (ref = private) also have had synchronous earlier-stage disease but were over-staged as
Medicare/aid 1.08 0.98, 1.20 0.13 having stage IV NSCLC without pathologic confirmation. Although the
Uninsured 1.08 0.87, 1.34 0.50 situation is uncommon, pathologic confirmation of stage IV disease
Other Gov. 1.07 0.74, 1.57 0.71
Unknown 1.02 0.71, 1.48 0.91
should be considered in all patients with otherwise early-stage lung
primaries to ensure that potentially curative therapy is not in-
appropriately withheld due to clinical over-staging of stage IV disease.
studies of small sample sizes, ranging from 12 to 144 patients, have Third, a small percentage of patients in the study did not have stage
shown reasonable long-term outcomes when pulmonary resection is IV disease as classified by the AJCC 7th edition. As noted above, the
utilized in patients who have single brain or adrenal metastases but an study duration was from 2004 to 2013. During this time period, there
otherwise early-stage primary tumor, with 5-year survival of approxi- were changes made from the AJCC 6th to 7th edition [31]. In the AJCC
mately 10–20% for patients with oligometastases disease to the brain 6th edition, the presence of separate tumors in a different lobe was
[7,8,27–29] and 5-year survival of approximately 25% for patients with considered M1 disease [21] whereas in the 7th edition [22], this was
isolated metastasis to the adrenal gland [14,17]. As such, the NCCN considered T4 disease. In this study, only 7.5% of patients in the entire
guidelines consider surgical resection of the lung primary along with cohort, and 8.5% of patients in the surgery group, had tumors that
local treatment of the metastasis in combination with systemic che- would now be classified as T4 according to AJCC 7th edition guidelines.
motherapy to be an acceptable treatment in these patients. The results We did perform a sensitivity analysis of patients diagnosed using the
of our current study using the NCDB and patients across a wide spec- 7th edition and the findings were consistent with those from the pri-
trum of institutions appear to support current practice and NCCN mary analyses.
guidelines. In this study, the five-year survival of 1761 patients treated Fourth, a significant limitation is that selection bias may have im-
with lung resection for stage IV disease was 25%, which is much higher pacted the results. Of note, the NCDB does contain Charlson-Deyo co-
than the 5-year survival of 4.9% typically observed for stage IV patients morbidity scores, and we included this variable in our multivariable
[1,2]. However, our study also found that 1337 stage IV patients un- and propensity-score matched analysis. However, detailed data re-
derwent surgery despite having N2 or N3 mediastinal nodal disease, T4 garding pulmonary function or performance status is not available in
tumors, or T3 tumors with nodal disease. This finding suggests that the NCDB. Thus, the better outcomes we observed with surgery could
clinicians are not choosing surgery for stage IV patients based on NCCN reflect that surgery was reserved for patients with better performance
status, although, in this study, patients who underwent surgery had

80
C.-F.J. Yang et al. Lung Cancer 115 (2018) 75–83

Table 3
Preoperative and Demographic Characteristics (cT1-2, N0-1, M1 and cT3, N0, M1).

Chemoradiation Chemotherapy Radiation Surgery p value*

(N = 8180) (N = 5044) (N = 6123) (N = 1761)

Age at Diagnosis, n (%) < 0.001

Median, IQR 64 (56, 71) 68 (61, 75) 70 (61, 77) 64 (55, 71)

Age, n (%) < 0.001

< 50 721 (8.8%) 261 (5.2%) 271 (4.4%) 193 (11.0%)
50–69 4977 (60.8%) 2473 (49.0%) 2764 (45.1%) 1059 (60.1%)
70–79 1990 (24.3%) 1685 (33.4%) 1915 (31.3%) 418 (23.7%)
> 79 492 (6.0%) 625 (12.4%) 1173 (19.2%) 91 (5.2%)

Sex, n (%) 0.003

Male 4390 (53.7%) 2767 (54.9%) 3384 (55.3%) 890 (50.5%)
Female 3790 (46.3%) 2277 (45.1%) 2739 (44.7%) 871 (49.5%)

Race, n (%) 0.003

White 6998 (85.6%) 4277 (84.8%) 5222 (85.3%) 1566 (88.9%)
Black 911 (11.1%) 583 (11.6%) 704 (11.5%) 153 (8.7%)
Other 271 (3.3%) 184 (3.6%) 197 (3.2%) 42 (2.4%)

Charlson-Deyo Comorbidity Score, n (%) < 0.001

0 5633 (68.9%) 3250 (64.4%) 3766 (61.5%) 1043 (59.2%)
1 1860 (22.7%) 1264 (25.1%) 1549 (25.3%) 528 (30.0%)
2 687 (8.4%) 530 (10.5%) 808 (13.2%) 190 (10.8%)

Insurance Status, n (%) < 0.001

Not Insured 451 (5.5%) 195 (3.9%) 276 (4.5%) 78 (4.4%)
Private 3074 (37.6%) 1433 (28.4%) 1380 (22.5%) 727 (41.3%)
Medicare/aid 4445 (54.3%) 3313 (65.7%) 4251 (69.4%) 914 (51.9%)
Other Gov. 106 (1.3%) 41 (0.8%) 125 (2.0%) 18 (1.0%)
Unknown 104 (1.3%) 62 (1.2%) 91 (1.5%) 24 (1.4%)

Distance from Hospital (miles) < 0.001

Median, IQR 9 (4, 22) 9 (4, 21) 9 (4, 21) 10 (4, 26)

Distance from Hospital (miles), n (%) 0.002

<1 297 (3.6%) 204 (4.0%) 264 (4.3%) 59 (3.4%)
1–20 5637 (68.9%) 3526 (69.9%) 4205 (68.7%) 1157 (65.7%)
21–40 1178 (14.4%) 645 (12.8%) 815 (13.3%) 252 (14.3%)
41–60 388 (4.7%) 250 (5.0%) 323 (5.3%) 108 (6.1%)
> 60 501 (6.1%) 321 (6.4%) 388 (6.3%) 149 (8.5%)
Unknown 179 (2.2%) 98 (1.9%) 128 (2.1%) 36 (2.0%)

Facility Type, n (%) < 0.001

Community Cancer Program 972 (11.9%) 628 (12.5%) 686 (11.2%) 157 (8.9%)
Comprehensive Community Cancer Program 3871 (47.3%) 2397 (47.5%) 2998 (49.0%) 782 (44.4%)
Academic/Research Program 2523 (30.8%) 1529 (30.3%) 1793 (29.3%) 641 (36.4%)
Integrated Network Cancer Program 814 (10.0%) 490 (9.7%) 646 (10.6%) 181 (10.3%)

Tumor Location, n (%) < 0.001

RUL 2566 (31.4%) 1425 (28.3%) 1850 (30.2%) 618 (35.1%)
RML 345 (4.2%) 235 (4.7%) 262 (4.3%) 72 (4.1%)
RLL 1168 (14.3%) 847 (16.8%) 885 (14.5%) 250 (14.2%)
LUL 2200 (26.9%) 1339 (26.5%) 1594 (26.0%) 489 (27.8%)
LLL 1104 (13.5%) 660 (13.1%) 846 (13.8%) 225 (12.8%)
Main Bronchus 252 (3.1%) 112 (2.2%) 198 (3.2%) 11 (0.6%)
Overlapping Lesion 53 (0.6%) 35 (0.7%) 52 (0.8%) 27 (1.5%)
Other 492 (6.0%) 391 (7.8%) 436 (7.1%) 69 (3.9%)

cTNM Stage, n (%) < 0.001

T1, N0, M1 1711 (20.9%) 1098 (21.8%) 1537 (25.1%) 615 (34.9%)
T2, N0, M1 2851 (34.9%) 1766 (35.0%) 2251 (36.8%) 625 (35.5%)
T1N1M1 771 (9.4%) 472 (9.4%) 501 (8.2%) 110 (6.2%)
T2N1M1 1702 (20.8%) 1006 (19.9%) 1023 (16.7%) 210 (11.9%)
T3N0M1 1145 (14.0%) 702 (13.9%) 811 (13.2%) 201 (11.4%)

cTNM Stage for 2010–2013 (with M1a and M1b), n (%) < 0.001
T1, N0, M1a 46 (1.4%) 205 (9.2%) 40 (1.8%) 59 (9.3%)
T1, N0, M1b 626 (19.5%) 277 (12.5%) 537 (24.0%) 173 (27.3%)
T2, N0, M1a 108 (3.4%) 340 (15.3%) 66 (2.9%) 61 (9.6%)
T2, N0, M1b 988 (30.8%) 440 (19.8%) 721 (32.2%) 150 (23.7%)
T1, N1, M1a 9 (0.3%) 50 (2.3%) 7 (0.3%) 7 (1.1%)
T1, N1, M1b 285 (8.9%) 120 (5.4%) 159 (7.1%) 19 (3.0%)
T2, N1, M1a 40 (1.2%) 96 (4.3%) 20 (0.9%) 16 (2.5%)
T2, N1, M1b 506 (15.8%) 243 (11.0%) 276 (12.3%) 40 (6.3%)
T3, N0, M1a 91 (2.8%) 206 (9.3%) 64 (2.9%) 56 (8.8%)
T3, N0, M1b 510 (15.9%) 241 (10.9%) 351 (15.7%) 52 (8.2%)

Histology, n (%) < 0.001

Adenocarcinoma 5973 (73.0%) 3643 (72.2%) 4170 (68.1%) 1230 (69.8%)
Squamous 1709 (20.9%) 1137 (22.5%) 1547 (25.3%) 321 (18.2%)
(continued on next page)

81
C.-F.J. Yang et al. Lung Cancer 115 (2018) 75–83

Table 3 (continued)

Chemoradiation Chemotherapy Radiation Surgery p value*

(N = 8180) (N = 5044) (N = 6123) (N = 1761)

Large cell 367 (4.5%) 206 (4.1%) 295 (4.8%) 126 (7.2%)
Adenosquamous 131 (1.6%) 58 (1.1%) 111 (1.8%) 84 (4.8%)

Size of Tumor (mm) < 0.001

Median, IQR 37 (25, 53) 36 (25, 50) 36 (25, 51) 30 (20, 46)

Separate Tumor Nodules −Ipsilateral Lung, n (%) < 0.001

No separate tumor nodules in ipsilateral lung 3958 (48.4%) 2458 (48.7%) 2789 (45.5%) 856 (48.6%)
Separate tumor nodules in ipsilateral lung, same lobe 300 (3.7%) 230 (4.6%) 210 (3.4%) 83 (4.7%)
Separate tumor nodules in ipsilateral lung, different lobe 107 (1.3%) 101 (2.0%) 76 (1.2%) 42 (2.4%)
Separate tumor nodules, ipsilateral lung, same and different 63 (0.8%) 54 (1.1%) 39 (0.6%) 13 (0.7%)
lobe
Separate tumor nodules, ipsilateral lung, unknown if same or 103 (1.3%) 93 (1.8%) 57 (0.9%) 12 (0.7%)
different lobe
Unknown if separate tumor nodules, separate tumor nodules 3649 (44.6%) 2108 (41.8%) 2952 (48.2%) 755 (42.9%)
cannot be assessed, or not documented in patient record

P-values are from 2-sided Krukal-Wallis test for continuous variables, and from Chi-square test for categorical variables.

more co-morbidities than non-surgical groups, which was an un-
expected finding. In this study, we also found that patients who un-
derwent wedge resection and segmentectomy had worse survival when
compared to lobectomy. This finding could be a reflection of worse
performance status and cardiopulmonary function in the sublobar re-
section groups (e.g., healthier patients were selected to have lobectomy
while sicker patients were selected to have sublobar resection).
However, several studies, including the Lung Cancer Study Group
randomized trial [32], have demonstrated a benefit of lobectomy over
sublobar resection for early stage NSCLC and it is possible that the
benefit with lobectomy extends to patients with stage IV disease.
Selection bias may have also occurred in that patients who under-
went surgery had more favorable disease distribution or biology. We
did attempt to account for some of the selection bias by matching the
surgery patients to a cohort of “best surviving” patients who underwent
chemotherapy and radiation. Our hypothesis was that patients in the
chemotherapy and radiation group who had the longest survival may
have also had more favorable disease distribution or biology that would
be more comparable to the surgery group. In this comparison, surgery
was associated with better long-term survival than chemoradiation. We
additionally performed a propensity-score matched analysis to try to
account for selection bias and the findings were consistent with the

Fig. 3. Overall Survival Stratified by Treatment Group for cT1-2N0-1M1 and cT3N0M1 Patients.

82
C.-F.J. Yang et al.
primary findings of the study.
In conclusion, in patients undergoing surgery for stage IV NSCLC,
increasing T and N status is associated with worse survival. For the
subset of patients with cT1-2, N0-1, M1 or cT3, N0, M1 NSCLC, a
treatment regimen that included surgical resection of the primary
tumor was associated with a 5-year survival of 25% in the NCDB.
Although this study does not prove a definite benefit to surgery over
other therapies, these results are comparatively much better than re-
sults observed with non-surgical therapy. These relatively good long-
term outcomes appear to support current practice and NCCN guidelines,
and suggest that multi-disciplinary evaluation that considers surgery is
indicated for patients with cT1-2, N0-1, M1 and cT3, N0, M1 disease,
especially if lobectomy can be utilized to achieve complete resection.
Surgery for stage IV patients with more locally advanced tumors or
mediastinal nodal involvement provides less benefit, and should be
considered much less often than for patients with more limited local
primary disease.
Conflicts of interest and source of funding
T.A.D. serves as a consultant for Scanlan.
Acknowledgments
Through a Business Associate Agreement, the ACS has a data use
agreement with each of its CoC accredited hospitals. All data used in
this study is derived from a de-identified NCDB file. The ACS and the
CoC have not confirmed and are not responsible for the analytic or
statistical methodology used or the conclusions drawn from these data.
We would also like to thank John Deng for his assistance with the
manuscript.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in the
online version, at https://doi.org/10.1016/j.lungcan.2017.11.021.
References
[1] A. Noone, N. Howlader, M. Krapcho, D. Miller, K. Bishop, S.F. Altekruse, C.L.
Kosary, M. Yu, J. Ruhl, Z. Tatalovich, A. Mariotto, D.R. Lewis, H.S. Chen, E.J. Feuer,
K.A. Cronin, SEER Cancer Statistics Review, 1975-2013, National Cancer Institute,
Bethesda, MD, http://seer.cancer.gov/csr/1975_2013/, based on November 2015
SEER data submission, posted to the SEER web site, April 2016.
[2] G.A. Masters, S. Temin, C.G. Azzoli, G. Giaccone, S. Baker Jr., J.R. Brahmer,
P.M. Ellis, A. Gajra, N. Rackear, J.H. Schiller, T.J. Smith, J.R. Strawn, D. Trent,
D.H. Johnson, P. American Society of Clinical Oncology Clinical, Systemic therapy
for stage IV non-small-cell lung cancer: American Society of Clinical Oncology
Clinical Practice Guideline Update, J. Clin. Oncol. 33 (30) (2015) 3488–3515.
[3] B.D. Kozower, J.M. Larner, F.C. Detterbeck, D.R. Jones, Special treatment issues in
non-small cell lung cancer: diagnosis and management of lung cancer, 3rd ed:
American College of Chest Physicians evidence-based clinical practice guidelines,
Chest 143 (5 Suppl) (2013) e369S–e399S.
[4] A.B. Ashworth, S. Senan, D.A. Palma, M. Riquet, Y.C. Ahn, U. Ricardi,
M.T. Congedo, D.R. Gomez, G.M. Wright, G. Melloni, M.T. Milano, C.V. Sole,
T.M. De Pas, D.L. Carter, A.J. Warner, G.B. Rodrigues, An individual patient data
metaanalysis of outcomes and prognostic factors after treatment of oligometastatic
non-small-cell lung cancer, Clin. Lung Cancer 15 (5) (2014) 346–355.
[5] T. de Vin, B. Engels, T. Gevaert, G. Storme, M. De Ridder, Stereotactic radiotherapy
for oligometastatic cancer: a prognostic model for survival, Ann. Oncol. 25 (2)
(2014) 467–471.
[6] D.S. Ettinger, D.E. Wood, W. Akerley, L.A. Bazhenova, H. Borghaei, D.R. Camidge,
R.T. Cheney, L.R. Chirieac, T.A. D’Amico, T.L. Demmy, T.J. Dilling,
M.C. Dobelbower, R. Govindan, F.W. Grannis, Jr., L. Horn, T.M. Jahan, R. Komaki,
L.M. Krug, R.P. Lackner, M. Lanuti, R. Lilenbaum, J. Lin, B.W. Loo Jr., R. Martins,
G.A. Otterson, J.D. Patel, K.M. Pisters, K. Reckamp, G.J. Riely, E. Rohren,
S.E. Schild, T.A. Shapiro, S.J. Swanson, K. Tauer, S.C. Yang, K. Gregory, M. Hughes,
Non-small cell lung cancer, version 6.2015, J. Natl. Compr. Cancer Netw. 13 (5)
(2015) 515–524.
83
Lung Cancer 115 (2018) 75–83
[7] P.S. Billing, D.L. Miller, M.S. Allen, C. Deschamps, V.F. Trastek, P.C. Pairolero,
Surgical treatment of primary lung cancer with synchronous brain metastases, J.
Thorac. Cardiovasc. Surg. 122 (3) (2001) 548–553.
[8] P. Bonnette, P. Puyo, C. Gabriel, R. Giudicelli, J.F. Regnard, M. Riquet,
P.Y. Brichon, T. Groupe, Surgical management of non-small cell lung cancer with
synchronous brain metastases, Chest 119 (5) (2001) 1469–1475.
[9] N. Girard, V. Cottin, F. Tronc, B. Etienne-Mastroianni, F. Thivolet-Bejui,
J. Honnorat, J. Guyotat, P.J. Souquet, J.F. Cordier, Chemotherapy is the corner-
stone of the combined surgical treatment of lung cancer with synchronous brain
metastases, Lung Cancer 53 (1) (2006) 51–58.
[10] P. Granone, S. Margaritora, A. D’Andrilli, A. Cesario, K. Kawamukai, E. Meacci,
Non-small cell lung cancer with single brain metastasis: the role of surgical treat-
ment, Eur. J. Cardiothorac. Surg. 20 (2) (2001) 361–366.
[11] H. Nakagawa, Y. Miyawaki, T. Fujita, S. Kubo, K. Tokiyoshi, K. Tsuruzono,
K. Kodama, M. Higashiyama, O. Doi, T. Hayakawa, Surgical treatment of brain
metastases of lung cancer: retrospective analysis of 89 cases, J. Neurol. Neurosurg.
Psychiatry 57 (8) (1994) 950–956.
[12] M. Wronski, E. Arbit, M. Burt, J.H. Galicich, Survival after surgical treatment of
brain metastases from lung cancer: a follow-up study of 231 patients treated be-
tween 1976 and 1991, J. Neurosurg. 83 (4) (1995) 605–616.
[13] O. Mercier, E. Fadel, M. de Perrot, S. Mussot, F. Stella, A. Chapelier, P. Dartevelle,
Surgical treatment of solitary adrenal metastasis from non-small cell lung cancer, J.
Thorac. Cardiovasc. Surg. 130 (1) (2005) 136–140.
[14] H. Porte, J. Siat, B. Guibert, F. Lepimpec-Barthes, R. Jancovici, A. Bernard,
A. Foucart, A. Wurtz, Resection of adrenal metastases from non-small cell lung
cancer: a multicenter study, Ann. Thorac. Surg. 71 (3) (2001) 981–985.
[15] D.J. Raz, M. Lanuti, H.C. Gaissert, C.D. Wright, D.J. Mathisen, J.C. Wain, Outcomes
of patients with isolated adrenal metastasis from non-small cell lung carcinoma,
Ann. Thorac. Surg. 92 (5) (2011) 1788–1792 discussion 1793.
[16] V.E. Strong, M. D’Angelica, L. Tang, F. Prete, M. Gonen, D. Coit, K.A. Touijer,
Y. Fong, M.F. Brennan, Laparoscopic adrenalectomy for isolated adrenal metastasis,
Ann. Surg. Oncol. 14 (12) (2007) 3392–3400.
[17] T. Tanvetyanon, L.A. Robinson, M.J. Schell, V.E. Strong, R. Kapoor, D.G. Coit,
G. Bepler, Outcomes of adrenalectomy for isolated synchronous versus metachro-
nous adrenal metastases in non-small-cell lung cancer: a systematic review and
pooled analysis, J. Clin. Oncol. 26 (7) (2008) 1142–1147.
[18] T.P. Wade, W.E. Longo, K.S. Virgo, F.E. Johnson, A comparison of adrenalectomy
with other resections for metastatic cancers, Am. J. Surg. 175 (3) (1998) 183–186.
[19] A. Rittmeyer, V. Gorbunova, A. Vikstrom, A. Scherpereel, J.H. Kim, M.J. Ahn,
A. Chella, C. Chouaid, A.K. Campbell, F. Barlesi, Health-related quality of life in
patients with advanced nonsquamous non-small-cell lung cancer receiving bev-
acizumab or bevacizumab-plus-pemetrexed maintenance therapy in AVAPERL
(MO22089), J. Thorac. Oncol. 8 (11) (2013) 1409–1416.
[20] A. Stewart, K.Y. Bilimoria, D.P. Winchester, C.Y. Ko, The National Cancer Data
Base: a powerful initiative to improve cancer care in the United States, Ann. Surg.
Oncol. 15 (3) (2008) 683–690.
[21] F.L. Greene, American Joint Committee on Cancer, American Cancer Society, AJCC
Cancer Staging Manual, 6th ed., Springer, New York, 2002.
[22] S.B. Edge, American Joint Committee on Cancer. AJCC Cancer Staging Manual, 7th
ed., Springer, New York, 2010.
[23] P.R. Rosenbaum, D. Rubin, The central role of the propensity score in observational
studies for causal effects, Biometrika 70 (1983) 41–55.
[24] L.S. Parson, Reducing bias in a propensity score matched-pair sample using greedy
matching techniques, Proceedings of the Twenty-sixth Annual SAS Users Group
International Conference, SAS Institute Inc., 2001, pp. 214–216.
[25] A. Brichet, N. Penel, B. Prevost, A. Duhamel, R. Assaker, F. Dubois, J.J. Lafitte,
Pronostic factors of synchronous brain metastases from lung cancer, Lung Cancer 33
(2–3) (2001) 143–154.
[26] T. Rice, N. Moazami, L.A. Rybicki, D.J. Adelstein, S.C. Murthy, M.M. DeCamp,
G.H. Barnett, M.A. Chidel, J.H. Suh, E.H. Blackstone, Stage III non-small cell lung
cancer and metachronous brain metastases, J. Thorac. Cardiovasc. Surg. 124 (1)
(2002) 113–122.
[27] A. Arame, P. Mordant, F. De Dominicis, C. Pricopi, C. Foucault, A. Dujon, F. Le
Pimpec-Barthes, M. Riquet, Which metastasis management allows long-term sur-
vival of synchronous solitary M1b non-small cell lung cancer? Eur. J. Cardiothorac.
Surg. 41 (3) (2012) 617–622.
[28] I. Troján, J. Furák, T. Szöke, L. Agócs, A. Csekeö, J. Kas, E. Svastics, J. Eller,
L. Tiszlavicz, Lung cancer and its operable brain metastasis: survival rate and sta-
ging problems, Ann. Thorac. Surg. 79 (1) (2005) 241–247.
[29] M. Pistolesi, A. Mussi, M. Lucchi, A. Janni, A. Chella, G. Parenti, G. Rossi,
C.A. Angeletti, Resection of single brain metastasis in non-small-cell lung cancer:
prognostic factors, J. Thorac. Cardiovasc. Surg. 112 (1) (1996) 146–153.
[30] M. Burt, M. Wronski, E. Arbit, J.H. Galicich, Resection of brain metastases from
non-small-cell lung carcinoma. Results of therapy. Memorial Sloan-Kettering Cancer
Center Thoracic Surgical Staff, J. Thorac. Cardiovasc. Surg. 103 (3) (1992)
399–410.
[31] AJCC, Summary of Changes: Understanding the Changes from the Sixth to the
Seventh Edition of the AJCC Cancer Staging Manual, 13.
[32] R.J. Ginsberg, L.V. Rubinstein, Randomized trial of lobectomy versus limited re-
section for T1 N0 non-small cell lung cancer. Lung Cancer Study Group, Ann.
Thorac. Surg. 60 (3) (1995) 615–622 discussion 622-3.