Long-Term Outcomes of En Bloc Resection of Non-Small Cell Lung Cancer Invading the Thoracic Inlet and Spine

Transcription

1 Long-Term Outcomes of En Bloc Resection of Non-Small Cell Lung Cancer Invading the Thoracic Inlet and Spine Elie Fadel, MD, Gilles Missenard, MD, Charles Court, MD, Olaf Mercier, MD, Sacha Mussot, MD, Dominique Fabre, MD, and Philippe Dartevelle, MD Department of Thoracic and Vascular Surgery and Heart-Lung Transplantation, Hôpital Marie-Lannelongue, Le Plessis Robinson; and Institut Gustave Roussy, Paris-Sud University, Villejuif, France Background. The purpose of this study was to determine whether en bloc resection of non-small cell lung cancer (NSCLC) invading the thoracic inlet (TI) and spine can provide good long-term outcomes. Methods. We studied 54 consecutive patients treated with en bloc resection of NSCLC invading the TI and spine between 1992 and 2009 at our center. There were 36 men and 18 women with a mean age of 51 years (range, 37 to 71 years). Tumor resection involved at least 2 vertebral levels. We divided the patients into 3 groups based on whether vertebral invasion involved the transverse process only, the intervertebral foramina, requiring hemivertebrectomies with spinal fixation, or the vertebral body, requiring total vertebral body resection with spinal fixation. Results. Induction chemotherapy was given to 27 (50%) patients including 3 who also received induction radiotherapy. Nine (17%) patients were in the transverse process group, 42 (78%) in the intervertebral foramina group, and 3 (6%) in the vertebral body group. Resection involved the subclavian artery in 19 (35%) patients. Complete resection was achieved in 49 (91%) patients. There were no perioperative deaths or residual neurologic impairments. Recurrence occurred in 31 (57%) patients and was local (n 6), systemic (n 24), or both (n 1). Local recurrence was more common in patients with N2-3 disease (p ) and subclavian artery involvement (p 0.031). There was a nonsignificant increase in local recurrence in patients with positive resection margins (40% vs 10%, p 0.058). The 1-, 5-, and 10-year survival rates were 82%, 31%, and 31%, respectively. The 1-, 5- and 10-year disease-free survival rates were 63%, 28%, and 28%, respectively. Five patients are alive and free of disease 10 years after surgery. By multivariate analysis, factors that independently affected survival were incomplete (R1) resection (p 0.006; odds ratio 67; 95% confidence interval 1.5 to 11.3) and subclavian artery involvement (p 0.037; odds ratio 0.46; 95% confidence interval 0.2 to 0.9). Conclusions. Good long-term survival can be achieved in highly selected patients with NSCLC invading the TI and spine, provided complete en bloc resection is performed. (Ann Thorac Surg 2011;92: ) 2011 by The Society of Thoracic Surgeons In 2002, we reported a mixed multidisciplinary approach combining a posterior midline incision with the anterior transcervical approach to allow en bloc resection of non-small cell lung cancer (NSCLC) invading both the thoracic inlet (TI) and the intervertebral foramina (IF). The preliminary results in 17 patients showed no perioperative deaths and an overall 3-year survival of 39% [1]. During the same period, several authors described innovative surgical techniques for resecting lung cancer with vertebral invasion. These techniques fall into 2 main groups based on opposite strategies; namely, complete en bloc resection [1 4] and intralesional resection [5 7]. Although incomplete or intralesional NSCLC resection was associated with poor Accepted for publication April 15, Presented at the Forty-seventh Annual Meeting of The Society of Thoracic Surgeons, San Diego, CA, Jan 31 Feb 2, Address correspondence to Dr Fadel, Department of Thoracic and Vascular Surgery and Heart-Lung Transplantation, Hôpital Marie- Lannelongue (Paris-Sud University), 133 Ave de la Resistance, Le Plessis Robinson, France; fadel@ccml.com. survival [8], this strategy led to improved postoperative outcomes with apparently similar survival rates when incorporated into a multidisciplinary therapeutic program [5]. However, to date only preliminary results in small numbers of patients are available, and long-term local recurrence and survival rates are not available. Thus, the published data cannot determine whether en bloc resection, with its higher postoperative morbidity rate, is preferable over intralesional resection. The purpose of this study was to update our previous experience by evaluating the outcomes in a larger number of patients managed with en bloc resection and follow-up for several years. Patients and Methods Patients Between January 1992 and December 2009, 54 consecutive patients underwent en bloc resection with curative intent for NSCLC invading the TI and spine, at the Department of Thoracic Surgery, Marie Lannelongue 2011 by The Society of Thoracic Surgeons /$36.00 Published by Elsevier Inc doi: /j.athoracsur

2 Ann Thorac Surg FADEL ET AL 2011;92: LUNG CANCER INVADING THE THORACIC INLET AND SPINE 1025 Hospital, France. Our Institutional Review Board approved this retrospective study and waived the need for informed consent. Inclusion criteria were histologically documented NSCLC and invasion of the first rib and spine and (or) IF suggested by the preoperative investigations and confirmed surgically. Patients were not included when spinal or foraminal invasion was not confirmed during surgery. The preoperative workup consisted of a standard pulmonary and cardiac evaluation; chest radiography, bronchoscopy, computed tomography and magnetic resonance imaging (MRI) of the head, neck, and chest, and computed tomography of the abdomen. Positron emission tomography was performed routinely starting in June Surgery was contraindicated in patients with substantial involvement of the spinal canal, clinical symptoms or radiologic findings indicating involvement of the brachial plexus higher than C8, distant metastasis, or clinical N2 or N3 disease on the preoperative work-up. Mediastinoscopy was performed in patients with mediastinal node enlargement by computed tomography or mediastinal uptake by positron emission tomography. A supraclavicular biopsy was obtained when nodes were palpable in the neck. Surgical Technique All patients underwent one-stage en bloc resection of the NSCLC invading the TI and spine using the previously described mixed multidisciplinary approach combining a posterior midline incision with the anterior transcervical approach [1]. Schematically, the first step of the procedure was always the anterior approach that allows en bloc resection of the TI cancer and the involved transverse processes (TP). For more extended spinal invasion, the anterior approach was followed by the posterior approach during the same operative procedure to remove the tumor and the involved spine and to perform spinal stabilization. Briefly, with the patient in the supine position, an L-shaped cervical thoracotomy was performed [1, 9]. If the tumor was deemed resectable, the medial half of the clavicle was removed and preserved as a bone autograft. All involved vessels or nerves were resected with tumorfree margins. En bloc lung resection with mediastinal dissection was performed through the transcervical approach. Tumor spread to brachial plexus branches above C8 was considered a contraindication to surgery. The extent of vertebral resection required by the tumor spread was evaluated on preoperative MRI scans and confirmed by gentle digital palpation of the spine. At the end of the transcervical approach, the specimen was left in place, with all resected structures still adhering en bloc to the invaded spine. The patient was placed in the prone position, and a vertical midline incision extending from the spinal process of C7 to that of T5 was performed. Based on vertebral invasion as assessed by preoperative MRI and on the extent of spinal resection required to obtain en bloc tumor resection, we divided the patients into 3 groups (Fig 1), as follows: involvement of the TP only, involvement of the IF, and involvement of the vertebral bodies (VB). In the TP group, the tumor was dissected free from the VB and IF through the anterior incision, and an osteotome was then used to cut the TP in the anterior-toposterior direction (Fig 1A). A wedge of VB was removed with the costovertebral joint, exposing the ipsilateral nerve roots involved with the tumor. These roots were then ligated proximal to the dorsal root ganglion. The last step of the procedure consisted in cutting the TP. As patients in this group did not require laminectomy or Fig 1. (A) Right non-small cell lung cancer (NSCLC) invading thoracic inlet (TI) and the transverse processes. (B) Right NSCLC invading the TI and the intervertebral foramina. (C) Right NSCLC invading the TI and the cancellous bone. (D) Massive invasion of the vertebral body by a right NSCLC preventing en bloc resection.

3 1026 FADEL ET AL Ann Thorac Surg LUNG CANCER INVADING THE THORACIC INLET AND SPINE 2011;92: spinal fixation, no posterior approach was performed. The specimen was removed en bloc with the lung, ribs, and vessels through the anterior incision. In the IF group, hemivertebrectomies were required to obtain en bloc resection with tumor-free margins, and spinal fixation was then performed. In these patients, during the anterior approach, a midline cut was made through the prevertebral planes to facilitate section of the involved VBs (Fig 1B). During the posterior midline approach, after unilateral multilevel laminectomy extending from the vertebra above to the vertebra below the invaded IF, the nerve roots were exposed and divided inside the spinal canal at their emergence from the external spinal cord sheath, proximal to the dorsal root ganglion. An osteotome was used to cut the vertebral bodies along the midline, working in the posterior-toanterior direction. The osteotome was directed toward the operator s fingers placed anteriorly in the midline cut in the vertebral body. This avoided osteotomy through unsafe areas. The specimen was removed en bloc with the involved vertebrae, lung, ribs, and vessels through the posterior incision. Spinal fixation was performed as previously described with titanium instrumentation (Neuro France Implants, Boursay France) [1]. In all our patients, autologous bone harvested from the resected clavicle was used for spinal stabilization. In the VB group, tumor invasion of the cancellous bone through the cortex was suspected on preoperative imaging studies. Because frozen-section examination to assess the margins cannot be performed on bone, we removed at least 1 entire VB then performed spinal fixation. During the posterior midline approach, bilateral multilevel laminectomy of all involved vertebrae was performed to expose the nerve roots on both sides of the spinal cord sheath, which were then divided as described above. The osteotome entered the medial cortex of the contralateral pedicle, toward the noninvaded IF (Fig 1C). Spinal fixation was performed as described above, with interposition of the resected clavicle between the remaining vertebral bodies to strengthen the fixation. Patients with NSCLC responsible for massive VB involvement visible on the preoperative MRI (Fig 1D) were contraindicated for surgery. Statistical Analysis Continuous variables are described as median and range. Categoric variables are reported as proportion (%). Categoric variables were compared using the 2 test or Fisher exact test and continuous variables using the unpaired two-tailed t test. Survival after surgery and disease-free survival were analyzed using the Kaplan-Meier method. A log-rank test was performed to compare groups regarding survival estimates. All statistical analyses were run on Statview V (Abacus Concept, Berkeley, CA). The p values less than 0.05 were considered significant. Cox multivariate proportional hazards regression methods were used to identify risk factors for mortality after surgery. All factors with p values less than 0.05 by univariate analysis were included in the multivariate model. The following variables were entered into the multivariate model: sex, age, side, preoperative serum carcinoembryonic antigen level, completeness of resection (R0/R1), lymph node status, tumor histology, tumor size, extent of lung resection, extent of vertebral resection (TP, IF, or VB group), subclavian artery or vein resection, phrenic nerve resection, brachial plexus root resection, postoperative complications, preoperative and postoperative chemotherapy, and preoperative and postoperative radiation therapy. Results There were 36 men and 18 women with a mean age of 52 years (range, 37 to 71 years). Serum carcinoembryonic antigen was elevated in 12 patients, normal in 32, and unknown in 10. Induction treatment was given to 27 (50%) patients and consisted of platinum-based chemotherapy alone (n 24) or with radiation in a mean dose of 35 Gy (range, 30 to 45 Gy) (n 3). The tumor was on the right side in 31 (57%) patients and on the left side in 23 (43%). Spinal resection involved 2 levels in 5 (9%) patients (T1 to T2, n 3; and T2 to T3, n 2), 3 levels in 35 (65%) patients (C7 to T2, n 1; T1 to T3, n 27; T2 to T4, n 5; and T3 to T5, n 2), and 4 levels in 14 patients (26%) (T1 to T4). The TP group comprised 9 (17%) patients, the IF group 40 (74%) patients, and the VB group 5 (9%) patients. The lower brachial plexus root (T1) was involved and resected in all patients; 2 patients required removal of the C8 brachial plexus root and 16 (30%) of the phrenic nerve. In 19 (35%) patients, the subclavian artery was involved. The mean length of the resected subclavian artery was 3.1 cm (range, 2.5 to 4.8 cm). Revascularization was performed by end-to-end anastomosis (n 11), often possible because the first 2 ribs were always removed with the tumor or polytetrafluoroethylene ringed graft implantation (n 8) when the segment of artery removed was so long that the 2 ends were under tension. The subclavian vein and vertebral artery were resected in 22 patients and 5 patients, respectively, and were not reconstructed. Lung resection consisted of a wedge resection in 4 (7%) patients, lobectomy in 47 (87%), sleeve lobectomy in 2 (4%), and pneumonectomy in 1 (2%), and was followed by mediastinal lymphadenectomy. Histologic types were adenocarcinoma (n 33), squamous cell carcinoma (n 11), large cell carcinoma (n 5), mixed carcinoma (n 3), or poorly differentiated carcinoma (n 2). On final histologic examination, 10 (20%) patients had N3 (supraclavicular) disease, 2 (4%) had N2 disease, 4 (7%) had N1 disease, and 38 (70%) had N0 disease. Median tumor size was 5 cm (range, 2.2 to 8 cm). On gross anatomy, the spinal canal was free of disease in all patients, confirming the preoperative MRI findings. Frozen sections of the distal margins of the tumor-bearing areas were examined intraoperatively. The definitive histologic evaluation showed complete microscopic resection (R0) in 49 (91%) patients and microscopic tumor invasion (R1) in 5 patients (0 in the TP group, 4 in the IF group, and 1 in the VB group; p 0.442). The residual tumor was in the brachial plexus in 2

4 Ann Thorac Surg FADEL ET AL 2011;92: LUNG CANCER INVADING THE THORACIC INLET AND SPINE 1027 patients, VB in 2, and spinal dura mater in 1. No residual macroscopic disease (R2) was left. The median number of packed red blood cells transfused during the procedure was 9 (range 0 to 31 packs). This number was similar in the IF and VB groups and significantly lower in the TP group than in the other two groups (p ). Operative times ranged from 3 hours and 40 minutes to 12 hours (median, 6 hours and 30 minutes). There were no perioperative deaths or residual neurologic impairments except neurologic sequelae after division of the C8 root in 2 patients (permanent ulnar nerve palsy). Major postoperative complications occurred in 30 patients (55%) and included pneumonia (n 15), empyema requiring surgical debridement (n 10), bleeding necessitating reoperation (n 3), cerebrospinal fluid leakage requiring ventriculoperitoneal shunt implantation (n 1), and massive acute pulmonary embolism requiring embolectomy under cardiopulmonary bypass (n 1). The median duration of mechanical ventilation was 3.1 days (range, 0 to 74 days) and the median hospital stay length was 36 days (range, 9 to 106 days), with no significant differences across the groups. No brace for spine immobilization was necessary in our series, and all patients were mobilized immediately after surgery. Adjuvant platinum-based chemotherapy was given alone (n 7) or with radiation therapy (n 25). Five patients received isolated adjuvant radiation therapy. The median cumulative radiation dose was 45 Gy (range, 30 to 60 Gy). Adjuvant therapy was not given to patients with wide tumor-free margins and pn0 disease or to patients who struggled to recover from the procedure. Among the 27 patients who received induction treatment, only 1 had a complete pathologic response on the final histologic specimen. The completeness of tumor resection was not affected by preoperative treatment as 3 of the 26 patients had a microscopic positive margin (R1) by definitive histologic examination. All patients were available for follow-up. During follow-up, spinal instability developed in 2 patients because of radiation-related osteonecrosis. One of these patients underwent successful surgery to realign the spine and change the fixation 1 year after tumor resection. The other patient was a 71-year-old man who died from acute quadriplegia with severe cervical spinal luxation after falling at home 5 months after surgery and 3 months after 60 Gy of radiation therapy. During the study, 30 (55%) patients died of tumor recurrence. Median survival was 24 months. Overall 1-year, 5-year, and 10-year survival rates were 82%, 31%, and 31%, respectively (Fig 2). The 1-year, 5-year, and 10-year disease-free survival rates were 63%, 28%, and 28%, respectively. Five (9%) patients are alive and free of disease 10 years after surgery. Median follow-up was 58 months (range, 3 to 181 months). By multivariate analysis, independent factors affecting survival were incomplete resection R1 (p 0.006; odds ratio [OR], 4.67; 95% confidence interval [95% CI] 1.5 to 11.3), and subclavian artery involvement (p 0.037; OR 0.46; 95% CI, 0.2 to 0.9) (Table 1). Fig 2. Survival curve for patients with non-small cell lung cancer invading thoracic inlet and the spine. Recurrences were seen in 31 (57%) patients and were local in 6 (11%) patients, systemic in 24 (44%) patients, and both in 1 (2%) patient. There was a nonsignificant trend toward a higher local recurrence rate in patients with positive resection margins (40% vs 10%, p 0.058). Local recurrences were more common in patients with N2-3 disease compared with patients with N0-1 disease (42% vs 5%, p ), and in patients with subclavian artery involvement (26% vs 6%, p 0.031). The overall recurrence rate was higher in patients with N2-3 disease (83% vs 50%, p ). Comment We previously reported a preliminary study in 17 patients that aimed to describe a new combined surgical approach for en bloc resection of NSCLC invading the TI and spine [1]. The current study is an extension of our experience and includes a larger number of patients with longer follow-ups. It clearly confirms the safety and efficacy of our combined anterior transcervical and posterior midline approach for complete en bloc resection of NSCLC invading both the TI and the spine. Although postoperative morbidity remained high, this aggressive surgical procedure provided good long-term outcomes with no operative mortality. The 91% rate of complete resection (R0) and 13% rate of local recurrence indicate good local control of the disease. Furthermore, for the first time, we obtained a 10-year survival rate of 31% with 5 patients alive and free of disease 10 years after tumor resection. These results indicate a major role for en bloc surgical resection as part of a multimodality curative treatment program for NSCLC invading both the TI and the spine. In the current study, we classified our patients into 3 groups based on the extent of spinal involvement. These groups are different from the groups described by Bolton and colleagues [5], who used the intralesional surgical strategy. Thus, patients with NSCLC extending along the intercostal nerve into the IF, without spread to the bone, did

5 1028 FADEL ET AL Ann Thorac Surg LUNG CANCER INVADING THE THORACIC INLET AND SPINE 2011;92: Table 1. Risk Factors for Death Risk Factor N Median (Months) 1 year 5 years 10 years p Value Spinal resection TP IF VB Neoadjuvant therapy Yes No Adjuvant therapy Yes No Node status N0-N N2-N Completeness of resection R R Subclavian artery resection Yes No IF intervertebral foramen; TP transverse process; VB vertebral body. not undergo VB resection with spinal fixation. In contrast, we performed en bloc resection in all patients. In patients with IF involvement but no bone invasion by preoperative MRI, multilevel hemivertebrectomy with subsequent spinal fixation was required to allow resection of the nerve roots inside the spinal canal and removal of the tumor-adherent VB aspect with safe margins. As frozen-section histology cannot be performed on bone sections, patients with preoperative MRI evidence of cortical invasion (VB group) underwent complete VB excision to maximize the chances of achieving R0 resection. With this surgical strategy, we achieved R0 resection in 91% of patients, which compares favorably with the 56% R0 rate reported with the intralesional resection technique [5]. This major difference in the surgical strategy may explain the difference regarding the proportion of patients who underwent total corporectomy, which was 9% of 54 patients in our study (VB group) and 39% of 39 patients in the study by Bolton and colleagues. In this last study, the R0 resection rate was lower and the survival time shorter than in the other groups. The extent of bone invasion consistent with intralesional resection is probably greater than that consistent with en bloc resection. Thus, the lower rate of vertebrectomy in our study reflects tighter patient selection, with exclusion of patients who were not eligible for en bloc tumor resection. In our opinion, in patients with NSCLC responsible for massive VB involvement visible on the preoperative MRI and with no response to induction chemotherapy, the risk to benefit ratio of surgical treatment is unacceptably high. Patient selection may contribute to explain the high R0 resection rate in our study. Surgical resection combined with chemoradiotherapy given either preoperatively or postoperatively has become the treatment of choice for locally advanced NSCLC [10]. Complete resection (R0) is the most widely reported predictive factor for long-term survival [1, 2, 5]. The main benefit of induction therapy is to increase the complete resection rate [2, 11]. Thus, in the Southwest Oncology Group Trial 9416, 5-year survival was 44% in patients who had superior sulcus tumors treated with preoperative chemotherapy and 45 Gy of radiation [11]. Furthermore, combined preoperative chemoradiation therapy is associated with a higher pathologic response rate than that reported in our study [2, 11]. However, to date, no randomized studies comparing neoadjuvant with adjuvant treatment are available. Whether induction therapy benefits patients undergoing en bloc resection is questionable, as the resection is done at a distance from the tumor edges and therefore removes all previously invaded tissues. Wide margins are crucial even in patients with a complete pathologic response to induction treatment [2], which is achieved in 6% to 29% of patients [5, 11, 12]. Moreover, if the maximum radiation dose is given preoperatively and the definitive histologic study shows incomplete resection, the patient is in a therapeutic impasse. Completeness of the resection may be difficult to predict as the response to induction therapy is inconsistent and frozen-section histology is not feasible on bone. We found no difference in long-term survival between patients given neodjuvant therapy and those given adjuvant therapy, in keeping with the results reported by Bolton and colleagues [5]. Therefore, when the preoperative findings indicate that complete resection is probably feasible, we prefer to start with surgical treatment and to give adjuvant treatment subsequently. Patients with positive margins can then receive adequate postoperative radiation therapy. In addition, as pointed out by Bolton and colleagues, high-dose preoperative radiation therapy can complicate the surgical resection by inducing fibrosis and obliteration of normal tissue planes [5]. Spinal stabilization failed in 2 of our patients, of whom 1 was treated successfully and the other died from spinal cord compression. Both patients were older than 60 years and received a cumulative radiation dose of 60 Gy, consistent with radiation-induced osteonecrosis as a cause of the fixation failure. However, in our study, preoperative chemotherapy, which was given to 50% of patients, affected neither postoperative mortality nor long-term survival. Therefore, in patients with preoperative MRI findings suggesting massive invasion of the TP

6 Ann Thorac Surg FADEL ET AL 2011;92: LUNG CANCER INVADING THE THORACIC INLET AND SPINE 1029 or VB and in those with bulky tumors we recommend platin-based induction chemotherapy, both to decrease the inflammatory response in the bone surrounding the tumor, and therefore the amount of abnormal tissue, and to identify patients with aggressive disease that will not benefit from surgical treatment. As most of the deaths in our study were due to distant metastases, postoperative chemotherapy seems warranted, particularly in patients with positive cervical or mediastinal lymph nodes. We recommend postoperative radiation in patients with positive margins or lymph nodes. Radiation therapy may induce local tissue fragility with potential spinal fixation destabilization, without affecting survival [5, 13]. Therefore, we no longer use postoperative radiation therapy in R0 N0 patients. Complete resection is the most important contributor to local tumor control. Two different strategies have been described to achieve resection of NSCLC invading the spine; namely, en bloc resection and intralesional resection. The main advantage of intralesional resection is decreased operative bleeding with a decreased risk of dilution coagulopathy and multiple blood transfusion, and, therefore, better postoperative outcomes [5 7]. The rates of complete resection after intralesional surgery ranged from 56% to 64% [5, 6]. However, tumor cells are probably disseminated in the operating field so that the term complete resection may be misleading. In 2 studies of multimodality treatment programs, the local recurrence rates were 31% and 38%, and 5-year survival rates were 27% and 26%, respectively [5, 6]. Survival beyond 5 years has not been reported in patients managed with intralesional surgery. En bloc resection is a radically different strategy that we and others recommend for the removal of NSCLC invading the spine [1 4]. In the present study, the largest case series to date, complete resection was achieved in 91% of patients, in keeping with the previously reported rates of 79% to 83% [1 4]. The local recurrence rate of 4% reported by Anraku and colleagues [2] and the 13% local recurrence rate in the current study with a longer follow-up period are lower than the rates reported with intralesional surgery, confirming the effectiveness of en bloc resection in providing local tumor control. Nevertheless, this very demanding surgical procedure is associated with high morbidity and mortality rates, ranging from 28% to 53% and from 0% to 9%, respectively [1 4]. Given this high complication rate, this complex multidisciplinary surgical procedure should be reserved for specialized centers. In highly selected patients with NSCLC invading the spine, the 3-year survival rates after en bloc surgery ranged from 58% to 68% [2, 4] in the most recent series. These rates are consistent with previous reports that margin status significantly influences not only local recurrence but also survival [5, 7]. In our study, we obtained the longest survivals reported to date, with a 10-year survival rate of 31% and 5 patients free of disease 10 years after surgery. These findings suggest that en bloc resection may provide better local control and long-term survival than intralesional surgery, in agreement with previously reported principles of oncologic surgery [8]. To resect tumors invading the spine, mainly when VB resection is planned, 2 different strategies have been described. The two-stage approach involves 2 separate surgical procedures performed at an interval of 1 or 2 weeks. The first procedure is done through the posterior midline approach [2, 5]. During the second procedure, the posterior approach is usually reopened to allow vertebral resection and spinal fixation. This strategy allows shorter operative times and less bleeding. However, after laminectomy and spinal fixation through the posterior midline approach, the intrapleural approach performed later on may show pleural or pericardial metastases or an unresectable tumor (eg, with invasion of the esophagus or the trachea) or even the absence of spinal invasion. In this situation, the first procedure was unnecessary. Furthermore, reopening the posterior approach after 1 to 2 weeks increases the risk of wound infection, which carries a risk of spread to the spinal instrumentation. Our one-stage mixed approach avoids these untoward situations. This single-center retrospective study has several limitations mainly due to the small number of tightly selected patients. However, despite these limitations, our case series is the largest reported to date and also has the longest follow-ups. We believe it contains several important messages. First, the mixed anterior transcervical and posterior midline approach is now a standardized technique that allows en bloc resection of NSCLC invading both the TI and the spine. We believe that this standardization in such a demanding procedure is crucial to improve the postoperative outcome. Second, in highly selected patients, this extensive resection can be performed with acceptable morbidity and mortality in specialized centers with interdisciplinary teams of thoracic and spine surgeons. Third, the complete en bloc resection technique, which complies with well-recognized oncologic surgical principles, provides a low local recurrence rate and excellent long-term survival. References 1. Fadel E, Missenard G, Chapelier A, et al. En bloc resection of non-small cell lung cancer invading the thoracic inlet and intervertebral foramina. J Thorac Cardiovasc Surg 2002;123: Anraku M, Waddell TK, de Perrot M, et al. Induction chemoradiotherapy facilitates radical resection of T4 nonsmall cell lung cancer invading the spine. J Thorac Cardiovasc Surg 2009;137: e1. 3. Grunenwald DH, Mazel C, Girard P, et al. Radical en bloc resection for lung cancer invading the spine. J Thorac Cardiovasc Surg 2002;123: Yokomise H, Gotoh M, Okamoto T, et al. En bloc partial vertebrectomy for lung cancer invading the spine after induction chemoradiotherapy. Eur J Cardiothorac Surg 2007; 31: Bolton WD, Rice DC, Goodyear A, et al. Superior sulcus tumors with vertebral body involvement: a multimodality approach. J Thorac Cardiovasc Surg 2009;137: Bilsky MH, Vitaz TW, Boland PJ, Bains MS, Rajaraman V, Rusch VW. Surgical treatment of superior sulcus tumors with spinal and brachial plexus involvement. J Neurosurg 2002;97(Suppl 3):301 9.

7 1030 FADEL ET AL Ann Thorac Surg LUNG CANCER INVADING THE THORACIC INLET AND SPINE 2011;92: Gandhi S, Walsh GL, Komaki R, et al. A multidisciplinary surgical approach to superior sulcus tumors with vertebral invasion. Ann Thorac Surg 1999;68: Ginsberg RJ, Martini N, Zaman M, et al. Influence of surgical resection and brachytherapy in the management of superior sulcus tumor. Ann Thorac Surg 1994;57: Dartevelle PG, Chapelier AR, Macchiarini P, et al. Anterior transcervical-thoracic approach for radical resection of lung tumors invading the thoracic inlet. J Thorac Cardiovasc Surg 1993;105: Farray D, Mirkovic N, Albain KS. Multimodality therapy for stage III non-small-cell lung cancer. J Clin Oncol 2005;23: Rusch VW, Giroux DJ, Kraut MJ, et al. Induction chemoradiation and surgical resection for superior sulcus non-smallcell lung carcinomas: long-term results of Southwest Oncology Group Trial 9416 (Intergroup Trial 0160). J Clin Oncol 2007;25: Kunitoh H, Kato H, Tsuboi M, et al. Phase II trial of preoperative chemoradiotherapy followed by surgical resection in patients with superior sulcus non-small-cell lung cancers: report of Japan Clinical Oncology Group trial J Clin Oncol 2008;26: DeMeester TR, Albertucci M, Dawson PJ, Montner SM. Management of tumor adherent to the vertebral column. J Thorac Cardiovasc Surg 1989;97: DISCUSSION DR MICHAEL M. LANUTI (Boston, MA): I have a quick question. I was concerned that the amount of N3 disease that you uncovered was 10 patients, thus 20% with N3 disease. Did you know that preoperatively, or did you intentionally resect patients with N3 disease? DR FADEL: As I told you, when we had clinical N3 disease, usually we didn t do this procedure. All of them were microscopic N3 disease on the definitive histology. DR LANUTI: And were they supraclavicular lymph nodes? DR FADEL: Supraclavicular, yes. DR CARLOS IBARRA-PEREZ (Mexico City, Mexico): I was wondering about the N2 patients. Did you know they were N2 before going into surgery or not? DR FADEL: As I said, patients with clinical N2 or N3 disease were not operated on. It was only those patients who had microscopic N2 or N3 disease; so, discovered on the final histological findings. DR IBARRA-PEREZ: I noted that only half of your patients had neoadjuvant therapy, if I understood well. Did you notice any difference in the survival of that group against the other group? DR FADEL: We didn t find any difference between this group and the patients that were not treated preoperatively in terms of survival, even in terms of recurrence, local or overall recurrence. DR IBARRA-PEREZ: That is interesting. Thank you very much. DR WALTER J. SCOTT (Philadelphia, PA): I wanted to ask about the lymph nodes again. Some people think that an ipsilateral supraclavicular lymph node is a local lymph node. Could you comment? DR FADEL: Yes, we believe it s local. However, with statistical analysis we found that the supraclavicular lymph node had a lower survival rate than an N1 node. So it s probably not only local. It s probably in between N1 and N2. DR NAZLY M. SHARIATI (Newark, NJ): That was a beautiful presentation. When you do an extensive vertebral resection, you explained that you do some form of a fixation. You also presented 2 of your patients who had even the root at the C8 level resected. Can you explain how you did the fixation in the two-field, what was the most cephalad limit of your fixation in those 2 cases, and was it an anterior and posterior approach or was it an anterior approach or a posterior? DR FADEL: Thank you for your questions. Actually, when we have enough clinical findings suggesting a C8 paralysis, we don t do this procedure. Usually those patients had microscopic invasion of C8 and we try not to operate these patients because of the postoperative complications associated with that. We always start the procedure with the anterior approach to assess resectability, and if we find massive invasion of the brachial plexus during the anterior approach, we stop the procedure. We do the anterior and the posterior approach during the same procedure. So we start by assessing resectability, and if we see that invasion is too high in the brachial plexus, we stop the procedure. DR STEVEN D. HERMAN (Brooklyn, NY): I want to congratulate you on an incredibly difficult clinical presentation and having such outstanding results. I m particularly impressed by your low rate of local recurrence along the spinal-vertebral area. In our experience, frankly, that has been the hardest area for control and that s the most vulnerable for recurrence, whether it s in the upper thorax or the lower thorax. We have had cancers reappear and grow into the vertebrae that we have resected. To help manage this issue, we have begun the use of brachytherapy seed implants along these bone resection margins to control local recurrence. Your diagrams were incredibly nice, but frankly, when we have been inside, the vertebral bone resection portion doesn t really look quite so clean with precise margins. How do you judge and do your actual vertebral resection in there to get such an outstanding local recurrence rate? DR FADEL: Thank you very much. It s an important question. I think to-date, the MRI [magnetic resonance imaging] is the best exam to assess invasion of the vertebrae and the spinal canal. Each time we try to maximize our predictive R0 resection, so probably we have a very high selection of those patients, because, as you see, it s a major surgery, a complex surgery associated with a high postoperative morbidity. Our strategy is to achieve en bloc resection, so we have a very high selection of patients and based mainly on preoperative MRI. As you see, patients who have bone invasion, when the bone cancellous is invaded, we try to give these patients preoperative chemotherapy to reduce the inflammatory response of the bone near the tumor and also to assess the aggressiveness of those tumors.