Outcomes of patients with stage III nonsmall cell lung cancer treated with chemotherapy and radiation with and without surgery




The objective of this study was to identify the factors associated with improved outcome after treatment for stage III nonsmall cell lung cancer (NSCLC).


A retrospective review of stage III NSCLC patients treated at who were treated at the Dana-Farber Cancer Institute/Brigham and Women's Cancer Center was done with institutional review board approval. Patients were followed for toxicity, local and distant failure, and overall survival. Multivariate Cox logistic regression analysis was used to determine the factors associated with treatment outcome.


Between August 2000 and November 2006, 144 patients received concurrent chemoradiation (CRT) for stage III NSCLC. Eighty of 144 patients were men (56%), and the median age was 61 years (range, 33-81 years). Sixty-two patients (43%) had stage IIIA NSCLC, and 82 patients (57%) had stage IIIB NSCLC. Radiotherapy (RT) was given concurrently with chemotherapy to all patients; 100 patients (69%) received CRT without surgery, and 44 patients (31%) received with neoadjuvant CRT followed by surgical resection. The median RT dose was 60 grays (Gy) (range, 46-70 Gy). The median follow-up was 15 months (range, 3-64 months), the median potential follow-up was 37 months (range, 12-84 months), and the median overall survival was 22 months (95% confidence interval, 15-28 months). The 1-year and 2-year survival rates were 68% and 47%, respectively. Among the 44 patients who underwent resection, the median survival was 61 months, and the 2-year survival rate was 73%. On multivariate analysis, stage at the time of treatment (stage IIIA vs stage IIIB) and use of surgery were the only factors associated with improved outcome (P = .01 and P = .001, respectively).


In this retrospective series, those patients who were able to undergo resection appeared to have improved outcome after induction CRT. Cancer 2009. © 2009 American Cancer Society.

Lung cancer is the leading cause of cancer death in both men and women.1 Nonsmall cell lung cancer (NSCLC) represents 85% of lung cancer cases, and approximately 33% of those patients present with stage III disease.2 Stage III NSCLC represents a very wide range of patients, from those with minimal disease, to those with a single N2 lymph node involved, to those with multiple, bulky mediastinal and supraclavicular lymph nodes.3 Consequently, multiple regimens have been developed to treat this disease. Among the most common treatment approaches are concurrent chemoradiation (CRT)4, 5 and trimodality therapy, which involves CRT followed by surgical resection.6, 7

The role of surgery in the treatment of stage III NSCLC remains unproven. A randomized phase 3 trial of CRT versus trimodality therapy produced an impressive 27% 5-year survival rate with trimodality therapy and a significant improvement in local control over CRT alone but did not demonstrate an advantage in overall survival (OS).8 A high surgical mortality rate in 12 of 54 patients (26%) who underwent pneumonectomy likely contributed to their shorter survival compared with patients who received CRT without surgery. In a second randomized trial in Europe, chemotherapy alone as induction therapy followed by surgery and postoperative radiotherapy was compared with chemotherapy followed by radiotherapy alone, and no benefit was observed from the addition of adding surgery (median survival, 16.4 months vs 17.5 months, respectively; P value not significant).9 The conclusions from those 2 studies are uncertain, because neither study met its primary endpoint for improved OS with surgery, yet the intergroup trial did produce a meaningful improvement in both local control and progression-free survival with surgery. Therefore, thoracic oncologists are left with several questions: Is surgery effective only when proceeded by both chemotherapy and radiotherapy, or should concurrent CRT alone should be the standard of therapy for stage III NSCLC?

The Dana-Farber/Brigham and Women's Cancer Center (DF/BWCC) has a long history of treating patients who have stage III NSCLC with trimodality therapy. Our experience has been reported from the early 1990s and has emphasized the importance of a pathologic complete response in the mediastinal lymph nodes as a predictor of outcome.7 Recurrence in the brain was reported in the late 1990s as a frequent problem after trimodality therapy.9 More recently, the Intergroup study reported a surgical mortality rate of 26% for patients who underwent pneumonectomy after CRT, which was greater than the 6% mortality rate observed in our single-institutional study of 73 patients who underwent pneumonectomy after CRT.10 However, the efficacy of trimodality therapy compared with CRT was not addressed. Therefore, we believed that it would be useful to review our most recent experience (since 2000) in patients with stage III NSCLC who were treated with definitive CRT as well as trimodality therapy at our institution. There is significant patient selection between these therapies; therefore, the objective of the current study was not to compare these groups directly but rather to report on overall outcomes (both local and distant), factors associated with local control, and patterns of recurrence to provide further information for planning future prospective studies.


Patient population

We retrospectively reviewed the records of patients diagnosed with stage III NSCLC between January 2000 and October 2006 who were treated with CRT at DF/BWCC. Patients were identified from an institutional database of patients who received chest radiotherapy for NSCLC. The database was created with institutional review board approval. Patients were included if they had stage III NSCLC that was treated with concurrent CRT or neoadjuvant chemotherapy followed by CRT. The patients were evaluated initially by a medical oncologist, a radiation oncologist, and a general thoracic surgeon in a multidisciplinary clinic. All patients underwent pretreatment staging, which included chest computed tomography (CT) scans, bone scans or [18F]2-fluoro-2-deoxy-D-glucose positron emission tomography scans, and central nervous system imaging with either contrast-enhanced CT or magnetic resonance imaging. Surgical staging, including fiberoptic bronchoscopy and mediastinoscopy, pulmonary function testing, and split-lung ventilation/perfusion testing, was done routinely in patients who were considered for pulmonary resection.

Patients who had pleural effusions with cytologic involvement of NSCLC, those who had recurrent lung cancer, and those who underwent surgery as the initial modality of treatment were excluded from this analysis. All patients had their performance status graded from 0 to 4 on the Eastern Cooperative Oncology Group scale, and histology was assessed according to World Health Organization criteria.11 Race was identified based on documentation at the time of initial consultation.


The regimens of systemic therapy that were given concurrently with radiation therapy varied during the course of the current review, as explained below. In the initial years of the study (2000-2004), patients typically were treated with induction carboplatin (area under the curve [AUC], 6) plus paclitaxel (200 mg/m2) for 2 cycles followed by weekly carboplatin (AUC, 2) plus paclitaxel (50 mg/m2) with chest radiotherapy.12, 13 After presentation/publication of several clinical trials suggesting that patients who received etoposide/cisplatin had a better outcome,8, 14, 15 the patients who were deemed fit to receive the cisplatin-based regimen were treated with every-4-week cisplatin and etoposide (50 mg/m2), and the patients who were considered candidates for that regimen were given weekly carboplatin and paclitaxel. These same trials, in addition to others16, 17 that demonstrated the lack of benefit of induction chemotherapy before CRT, influenced our group also to offer initial, concurrent CRT as the standard of care. Cisplatin/etoposide and weekly carboplatin/paclitaxel also were used concurrently with radiation therapy in the neoadjuvant setting. When concurrent chemotherapy and radiation therapy was not followed by surgical resection, consolidation chemotherapy with the same drugs or with docetaxel was administered.15 Patients who received sequential chemotherapy followed by radiation alone or patients in whom the chemotherapy regimen varied from our standard practice, as defined above, were excluded from the analysis. In total, 18 patients were treated with other regimens: 2 patients received vinorelbine/cisplatin, 4 patients received vinorelbine alone concurrent with radiation, 1 patient received cisplatin/gemcitabine, and 11 patients received sequential chemotherapy followed by radiation alone.


Three-dimensional conformal radiation therapy with CT-based treatment planning was used in all cases. The patients typically were simulated with Vac-loc bags (Med-Tec, Orange City, Iowa) and T-Bar immobilization. CT simulation was performed using slices from 2.5 mm to 3 mm in thickness. Treatment planning was done with the ECLIPSE planning system (Varian Medical Systems, Palo Alto, Calif). The primary target volumes included the gross primary tumor volume observed on CT and/or positron emission tomography scans and the involved lymph node stations. Although the use of elective nodal irradiation (ENI) differed according to the treating physician, the overall trend was to offer ENI less as time progressed, consistent with reports in the literature during that period.18, 19 After 2004, lung heterogeneity corrections were used using the “Modified Batho” method. The radiation dose was prescribed to cover at least 95% of the planning tumor volume (PTV). Patients received radiation daily in fractions from 1.8 Gy to 2 Gy. The total dose varied (range, 46-70 Gy) according to the intent of treatment (preoperative or definitive). Patients who were to undergo potential surgical resection typically received up to 54 Gy. Patients who did not undergo surgery typically received up to 60 Gy before the implementation of lung heterogeneity corrections (in 2004) and from 66 Gy to 68 Gy when lung heterogeneity corrections were used.


Patients were evaluated for surgical resection before the start of treatment and after CRT. Patients were considered for resection if their disease was deemed anatomically resectable (without invasion of vertebral bodies, heart, or great vessels) and if they were deemed medically fit to undergo resection. The patients who were considered for surgical resection were reimaged during Weeks 4 and 5 of CRT if their resection status was in doubt or after CRT if resection was planned from the beginning. Patients who had tumors that were deemed unresectable continued chest radiotherapy in doses of 60 Gy to 68 Gy.

A complete pathologic response was defined as the absence of viable tumor cells in the pathologic specimen, both in the primary tumor and in the lymph nodes. Mediastinal downstaging was defined as the absence of tumor on pathologic examination of resected lymph nodes that previously had been involved (N2-N1or N0).20 Acute and late toxicity were scored according to the Common Toxicity Criteria (version 3.0).21

Patient Follow-Up

Because all patients in this cohort received all of their therapy within the DF/BWCC, they underwent close medical and radiation follow-up and, as indicated, surgical oncology follow-up. In general the initial follow-up consisted of scans obtained at 6 to 8 weeks after therapy. Thereafter, follow-up continued every 6 to 10 weeks as indicated. All patients were followed to death or to the last known follow-up. Follow-up documentation, including toxicity, recurrences, and survival, was obtained retrospectively from the electronic medical records of the DF/BWCC.

Statistical Analysis

The primary endpoint of this study was the relation between the treatment modalities and OS. Secondary endpoints were locoregional control (LRC) and distant control (DC). Survival and all other endpoints were calculated from the start of treatment. LRC was defined as freedom from any recurrence involving the primary tumor location or any regional lymph nodes22 All other recurrences were considered distant metastases. We retrospectively reviewed the scans that documented the local recurrences and compared the location of the recurrence with the PTV at the time of simulation to determine whether the local recurrences were in or outside the fields of treatment.

The 1-year, 2-year, and 3-year OS, LRC, and DC rates were calculated using the Kaplan-Meier method. Because of the retrospective nature of this analysis, known clinical variables that hypothetically were associated with the outcome variables were examined in a stepwise multivariate analysis using the Cox proportional hazards regression method. All covariates that were considered in the regression model were examined as categorical variables and included age, sex, race, Eastern Cooperative Oncology Group performance status, TNM classification, use of induction chemotherapy (yes or no), type of concurrent chemotherapy(cisplatin/etoposide vs carboplatin/paclitaxcel), consolidation chemotherapy, surgery (yes or no) and thyroid transcription factor 1 (TTF-1) status (yes or no). Hazard ratios (HRs) for each of the variables of interest were calculated with corresponding 95% CIs and P values. All P values were 2-sided, and P values ≤.05 were considered significant. Analyses were performed using R (R Foundation for Statistical Computing, Vienna, Austria).23


Patient Characteristics

In total, 144 patients received CRT for stage III NSCLC between January 2000 and October 2006. Sixty-two patients (43%) had stage IIIA disease, and 82 patients (57%) had stage IIIB disease; of these 82 patients, 15 (18%) had T4N0 or N1 disease. Nearly all patients (95%) had an excellent performance status (0 or 1). The majority of patients were men (56%). The patient characteristics are shown in Table 1.

Table 1. Patient Characteristics
Patient CharacteristicNo. of Patients (%)
Entire Group, N=144Surgery, n=44No Surgery, n=100
  1. WHO indicates World Health Organization; NSCLC NOS, nonsmall cell lung cancer not otherwise specified; FDG-PET, [18F]2-fluoro-2-deoxy-D-glucose positron emission tomography.

Median age [range], y61 [33-81]59.5 [33-74]62 [45-81]
 Women64 (44)21 (48)43 (43)
 Men80 (56)23 (52)57 (57)
 White122 (85)36 (81)87 (87)
 African American16 (11)7 (16)9 (9)
 Asian2 (1)1 (1.5)1 (1)
 Hispanic2 (1)1 (1.5)1 (1)
 Unknown2 (1)1 (1.5)2 (2)
WHO performance status
 018 (12)8 (18)10 (10)
 1119 (83)35 (80)84 (84)
 27 (5)1 (2)6 (6)
 NSCLC NOS55 (38)12 (27)43 (43)
 Squamous39 (27)10 (23)29 (29)
 Adenocarcinoma44 (31)22 (50)22 (22)
 Large cell6 (4)0 (0)6 (6)
 IIIA62 (43)32 (73)30 (30)
 IIIB82 (57)12 (27)70 (70)
Tumor classification
 TX2 (1)0 (0)2 (2)
 T116 (11)4 (9)12 (12)
 T272 (50)25 (57)47 (47)
 T313 (9)5 (11)8 (8)
 T441 (29)10 (23)31 (31)
Lymph node status
 N013 (9)5 (11)8 (8)
 N14 (3)2 (5)2 (2)
 N274 (51)34 (77)40 (40)
 N353 (37)3 (7)50 (50)
FDG-PET staging
 Yes99 (69)28 (64)71 (71)
 No45 (31)16 (36)29 (29)
 Yes99 (69)37 (84)62 (62)
 No45 (31)7 (16)38 (38)

Treatment Characteristics

Treatment characteristics are summarized in Table 2.

Table 2. Treatment Characteristics
Treatment characteristicNo. of Patients (%)
Entire Group, N=144Surgery, n=44No Surgery, n=100
  1. Gy indicates gray.

 Induction and concurrent56 (39)17 (39)39 (39)
 Concurrent51 (35)17 (39)34 (34)
 Concurrent and consolidation37 (26)10 (22)27 (27)
Concurrent chemotherapy
 Weekly carboplatin and paclitaxel105 (73)28 (63)77 (77)
 Cisplatin and etoposide every 4 wk39 (27)16 (37)23 (23)
 Lobectomy and bilobectomy26 (18)26 (59) 
 Pneumonectomy12 (8)12 (27) 
 Wedge resection6 (4)6 (14) 
 No surgery100 (70)  
Radiotherapy dose, Gy
 Median [range]60 [46-70]54 [54-70]60 [46-70]
 <541 (0.6)0 (0)1 (1)
 5433 (23)26 (59)7 (7)
 54-596 (4)0 (0)6 (6)
 6045 (31)7 (16)38 (38)
 61-6511 (8)1 (2)10 (10)
 6620 (14)2 (4)18 (18)
 6821 (14)7 (16)14 (14)
 707 (5)1 (2)6 (6)


There was some variability in the chemotherapy regimens used in this group. Fifty-six patients (39%) received induction carboplatin/paclitaxel chemotherapy before CRT. Fifty-three patients received 2 cycles of induction chemotherapy, including 1 patient who received 1 cycle and 2 patients who received 3 cycles. CRT consisted of weekly carboplatin/paclitaxel for 105 of 144 patients (73%) and cisplatin plus etoposide every 4 weeks for 39 of 144 patients (27%).


The median radiotherapy dose for the whole population was 60 Gy (range, 46-70 Gy), and only 1 patient received <54 Gy (Table 2). The breakdown of the radiotherapy doses was as follows: 46 Gy (1 patient), 54 Gy (33 patients), 58 Gy (3 patients), 60 Gy (44 patients), 61 Gy to 64 Gy (11 patients), 66 Gy (20 patients), 68 Gy (20 patients), and 70 Gy (7 patients). For the patients who received CRT without surgery, the median radiotherapy dose was 60 Gy (range, 46-70 Gy). Fourteen patients in that group received <60 Gy, and 5 of those patients received neoadjuvant CRT but did not undergo surgery because of disease progression (1 patient), treatment-related toxicity (pneumonitis; 2 patients), decline in performance status (1 patient), and stroke (1 patient). The remaining 9 patients had been planned to receive 60 Gy but missed 1 or 2 fractions because of toxicity at the end or because of patient-related scheduling concerns. All patients who received neoadjuvant CRT before surgery received ≥54 Gy (range, 54-70 Gy).

On the basis of the initial assessment, 50 of 144 patients (35%) had tumors that were deemed potentially resectable after CRT. Forty-one of those 50 patients had stage IIIA disease, and the remaining 9 patients had stage IIIB disease. Of the remaining 94 patients who were not considered for a neoadjuvant approach, the majority (74 of 94 patients) had stage IIIB disease. Of the patients who were selected for a neoadjuvant approach, 36 of 50 patients (72%) underwent surgery, and they all had a complete resection. Of 14 patients initially were deemed candidates for resection who did not undergo surgery, 6 patients did not have a radiographic response in the mediastinum, and 3 other patients developed clinically evident distant metastases at the time of restaging before surgery. Five additional patients had tumors that were not deemed resectable based on anatomic criteria, including 2 patients who had grade 2 radiation pneumonitis, 1 patient who had a decline in performance status, 1 patient who had a stroke, and 1 patient who developed Guillen-Barre syndrome. Eight patients received CRT as their definitive approach and then underwent re-evaluation by the multimodality team: An attempt at surgical resection was deemed appropriate because of their radiographic response to CRT. Two of these patients had initial stage IIIA disease (T2N2), and the remaining 6 patients had initial stage IIIB disease (4 patients had T2/T3 N3 disease, and 2 patients had T4N2 disease). Among the 44 patients who underwent surgery after CRT, the operations consisted of lobectomy (24 patients; 14 left, 10 right), bilobectomy (2 patients), pneumonectomy (12 patients; 11 left, 1 right), and wedge resection (6 patients). The median time between the end of CRT and surgery was 2 months (range, 1-4 months). The patient and treatment characteristics in the 2 groups (surgery vs no surgery) are detailed in Tables 1 and 2.

Neoadjuvant Therapy Pathologic Response

Thirty-two of 44 patients (73%) who had documented mediastinal (N2 or N3) lymph node involvement and underwent surgery had pathologic lymph node downstaging (from N3/N2 to N1 or N0). A pathologic complete response was achieved by 7 of 44 patients (16%) who underwent surgery.

Survival Analysis

At the time of the current analysis, 81 of 144 patients (56%) had died, and the remaining 63 patients remained alive. The median follow-up for the entire group was 15 months (range, 3-64 months), and the median potential follow-up was 37 months (range, 12-84 months). The median OS for the entire group was 22 months (95% confidence interval [CI], 15-28 months). The rates of OS, LRC, and DC for the entire group of 144 patients were 68%, 71%, and 51% at 1 year, respectively; 47%, 46%, and 38% at 2 years, respectively; and 36%, 42%, and 31% at 3 years, respectively (Fig. 1).

Figure 1.

This chart illustrates overall survival for all (n = 144) patients in the study.

The OS rate for the patients who received CRT followed by surgical resection was 93% at 1 year, 73% at 2 years, and 58% at 3 years. Their median survival was in excess of 3 years. It is noteworthy that the 30-day postoperative mortality rate in this group was only 1 of 44 patients (3%). The 1 patient who died of acute respiratory distress syndrome/multisystem organ failure 10 days after undergoing a left upper lobectomy.

The remaining 100 patients received CRT alone (median dose, 60 Gy; range, 46-70 Gy). The 1-year and 2-year OS rates for this group were 59% and 37%, respectively, and the median OS was 15 months.

To better understand the factors associated with improved outcome, univariate and multivariate analyses were performed using relevant clinical parameters. Univariate analysis revealed that stage at diagnosis (IIIA vs IIIB) and surgical resection were the only factors associated with prolonged survival (P = .01 and P = .001, respectively). Age, sex, type of chemotherapy, and race were not predictive of survival. Surgical resection was the most important factor associated with longer survival in the multivariate Cox regression analysis, with an HR of 0.27 in favor of patients who underwent resection (95% CI, 0.14-0.049; P<.001). In the cohort of patients that underwent resection (n = 44), 2 clinical factors were significant predictors of improved outcome. Patients who received initial concurrent CRT before surgery (n = 27) fared significantly better than those who underwent induction chemotherapy before CRT (n = 17) in preparation for surgery (HR, 0.20; P = .016). The survival difference between these groups at 2 years was 46% (94% vs 48%) (Fig. 2). In addition, patients who underwent positron emission tomography staging before surgery (n = 28) had a better outcome than those who only had CT-based staging (n = 16), with an HR of 0.30 (P = .05) and a significant difference in 2-year survival (92% vs 58%). It is noteworthy that neither of these factors was significant in the larger group of patients who received CRT alone.

Figure 2.

This chart illustrates survival difference between the groups who did and did not undergo surgery.

No statistically significant differences were observed in OS, LRC, or DC rates, depending on the type of chemotherapy delivered. The results did not demonstrate that the dose of radiotherapy influenced OS, local recurrence, or DM.

Patterns of Recurrence

At the time of this analysis, 90 of 144 patients (63%) had experienced either distant recurrence (DR), local recurrence, or both. The median time to progression was 9 months (range, 2-57 months). Among the 90 patients who experienced recurrence or progression, the site of first recurrence had a component of local recurrence in 33 patients (37%) and a component of DR in 57 patients (63%). Among the 33 patients who had locoregional recurrence (LRR) as a component of the first site of cancer progression, 11 patients had LRR only, and none of these patients had undergone surgical resection. Of the 57 patients who had some component of DR as their first site of recurrence, 33 had DR only. Ten of the 33 patients who had DR only were patients who underwent surgery, and the other 23 of 33 patients had received CRT alone. Forty-six patients had first recurrences that were both LRR and DR. Of the remaining 54 patients who did not develop recurrent disease, 27 had received trimodality therapy, and 27 patients received CRT only (Table 3)

Table 3. Recurrence Patterns in the Surgery and Nonsurgery Groups (Any Site)
 No RecurrenceLRR OnlyLRR and DRDR OnlyTotal
  1. LRR indicates locoregional recurrence; DR, distant recurrence.

Surgery (n=44)27071044
Nonsurgery (n=100)27113923100
Total (N=144)54114633144

LRR at any time was observed in 57 of 144 patients (40%), and it was the first site of recurrence (or progression) in 33 of those patients. The median time from diagnosis to LRR was 11 months (range, 3-44 months) for all patients and 10.5 months (range, 3-44 months) for patients who had LRR as their first site of recurrence. The specific locations of LRR were at the primary site of disease only for 18 of 57 patients (32%), lymph nodes only for 15 of 57 patients (26%), and the remaining patients (24 of 57; 42%) developed recurrent disease in the primary site and lymph nodes simultaneously.

Among the 57 patients who had LRR, 31 patients (54%) recurred within the radiation field, 10 patients (17%) recurred beyond the radiation field, and 16 patients (29%) recurred both within and outside of the chest radiation field. Of the 57 patients who developed LRR, only 7 of 44 patients (%) had undergone surgery, whereas 50 of 100 patients (50%) received CRT alone. In 32 of these 50 patients, the LRR was their first site of recurrence.

DR at any time was observed in 79 of 144 patients (55%) and, in 57 of 79 patients (72%) was the first site of recurrence. The DR rate was 39% (17 of 44 patients) after trimodality therapy; in all but 1 of these patients, DR was the first site of recurrence. The DR rate was 62% (62 of 100 patients) in the group that received CRT only, and 65% of those represented the first site of recurrence. The most common sites of DR were brain (n = 33), contralateral lung (n = 17), and adrenal (n = 13).

In total, 33 patients (23%) developed brain metastases at any time during follow-up. The brain was the first site of recurrence in 20 of 33 patients. The median time to the development of brain metastasis was 9.5 months (range, 3-60 months), and it was 7.5 months (range, 3-20 months) in patients who experienced brain metastasis as their initial site of recurrence. The rate of brain metastasis was 16% (7 of 44 patients) after trimodality therapy and 26% (26 of 100 patients) after CRT alone. The median time to brain metastasis was 11 months in the trimodality group and 8.5 months in the CRT-only group.


Treatment-related toxicity is summarized in Table 4. Nonhematologic toxicities included grade 3/4 nausea and vomiting for 8 of 144 patients (5%) and acute esophagitis for 38 of 144 patients (26%). Of the patients who had acute esophagitis, 6 developed late chronic esophageal stricture, which required dilatation. The doses for these 6 patients were 54 Gy (1 patient), 60 Gy (2 patients, 66 Gy (1 patient), and 68 Gy (2 patients), respectively. The rates of grade 2 and 3 combined-modality pneumonitis were 12% and 1%, respectively. Two patients developed fatal combined-modality pneumonitis. One of those patients died 7 months after the completion of CRT and already had developed distant metastases. The second patient died 6 weeks after the completion of CRT. The volume of lung that received ≥20 Gy in these 2 patients was 34% and 33.5%, respectively. Both patients received CRT and did not undergo surgery. These were the only deaths from nonsurgical toxicity in the entire cohort of 144 patients.

Table 4. Treatment-Related Toxicity
ToxicityNo. of Patients (%)
Grade 2Grade 3Grade 4Grade 5
Anemia33 (23)1 (0.7)0 (0)0 (0)
Leukopenia32 (21)26 (18)1 (0.7)0 (0)
Thrombocytopenia9 (6)2 (1)1 (0.7)0 (0)
Nausea/vomiting17 (12)7 (5)0 (0)0 (0)
Esophagitis71 (49)37 (26)1 (0.7)0 (0)
Pneumonitis17 (12)2 (1)0 (0)2 (1)


Currently, the most prevalent approach to treating stage III NSCLC in the United States is concurrent chemotherapy plus chest radiotherapy (CRT).24 Typically, this approach produces a median survival from 17 months to 20 months and a 3-year survival rate of 23% to 27%. The best current results, from a phase 2 cooperative group trial by the Southwest Oncology Group (SWOG 9504), produced a 29% 5-year survival rate.25-27 The questions arise: Where does this approach fail, and how can it be improved? We reviewed our single-institution experience in patients with stage III NSCLC to determine the patterns of recurrence in patients who received CRT with and without surgery and compared the results with those reported in the literature.

Most striking in our cohort was the outcome for the subset of 44 patients who received CRT followed by surgical resection. These patients had a median survival in excess of 3 years and had 2-year and 3-year survival rates of 73% and 58%, respectively. These results compare favorably to our experience from the 1990s with trimodality therapy in which the 2-year and 3-year survival rates were 54% and 44%, respectively.9 Explanations for the apparent improvement in outcome for the more recent patients may be improved staging or better selection of candidates.

Clearly, the patients who were selected for surgery in our cohort represented a more favorable subgroup than those who received CRT alone. However, on multivariate analysis, when we corrected for prognostic factors like disease stage, age, and performance status, surgery remained the most important prognostic factor for survival. To determine the reason for this, we examined the patterns of recurrence in our patients. Patients who received CRT followed by resection had a significantly lower rate of local recurrence (7%) than patients who received CRT alone (50%). Similarly, there was a lower distant metastasis rate for the trimodality group compared with the CRT-only group (39% vs 62%). The distant metastasis rate of 39% for trimodality treatment is similar to that reported by Albain et al. in the Intergroup 0139 trial for the surgical arm (37%), whereas our rate of distant metastasis in the CRT arm far exceeded that reported in the Intergroup trial (42%).8

Excellent results with trimodality therapy also have been reported in other institutions. Takeda et al. reported 100 patients who received trimodality therapy in Japan. In their study, they used a combined chemotherapy regimen of mitomycin C, vindesine, and cisplatin together with a somewhat lower dose of radiation (median, 41 Gy) and demonstrated a median survival of 40 months.28 The treatment-related mortality rate in that study was 2.5%. Similar results also were reported by Uy and colleagues, who reported on 40 patients who received cisplatin/etoposide and thoracic radiation (median dose, 45 Gy); those authors also reported a 40-month median survival and a treatment-related mortality rate of 7%.29 Finally, researchers at the University of Maryland reported a study of 40 patients in which, in addition to giving concurrent cisplatin-based chemotherapy with thoracic radiation, they increased the dose of chest radiation to at least 59 Gy. Those authors reported an increase in median survival to 53 months and a 0% treatment-related mortality rate.30 To put our results into context with these other single-institution results, we observed a median survival in excess of 3 years in our 44 surgically treated patients and a 2% treatment-related mortality rate. In our cohort, we typically delivered higher doses of radiation preoperatively (median, 54 Gy), similar to the approach and results from the Maryland group.

Balanced against these excellent results is the concern that CRT followed by surgery may be too toxic for some patients. In the randomized Intergroup trial, a benefit of adding surgery to CRT was not demonstrated for the entire group. One hypothesis for the absence of benefit from surgical resection was the high rate of perioperative mortality (14 of 54 patients; 26%) after pneumonectomy.8 However in different centers (as discussed above), including our own, the mortality with CRT followed by pneumonectomy has been modest (range, 0%-13%).11, 10, 30, 31 This suggests that trimodality therapy likely will continue to have a role for select patients with stage III NSCLC.

Despite the success of trimodality therapy, the majority of our patients (69%) were not deemed candidates for surgery and received definitive CRT alone. Because only 20% to 30% of patients with stage III disease are likely candidates for surgery, what can be done to improve the outcome of patients who are treated with CRT alone? Consistent with other reported data, we observed a median survival of 15 months in this group; and, as expected, there was a higher percentage of patients with stage IIIB disease (70%).16, 32 The analysis of patterns of recurrence indicated that 50% (50 of 100 patients) developed local progression after CRT alone. This high rate of local recurrence is consistent with multiple studies that have used radiation doses in the 60-Gy range.33-35 These data could argue for testing more aggressive local approaches to this disease. Evidence exists in a cooperative group setting2, 36 that radiation dose escalation to 74 Gy can be achieved in patients with stage III NSCLC; and, in a single-institutional setting, doses even beyond 80 Gy have been deliverred.37 Two retrospective analyses from Memorial Sloan-Kettering and the University of Michigan have produced improvements in OS with increased doses of radiation.38, 39 Whether these efforts, in turn, will lead to improved clinical outcome in stage III NSCLC remains to be seen and is the subject of an ongoing phase 3 randomized trial by the Radiation Therapy Oncology Group examining 74 Gy versus 60 Gy with concurrent chemotherapy. However, in our data set, we did not observe any improvement in outcome when the dose of radiation was examined as a prognostic factor.

The main impediment to improved outcome, however, is the high DR rate. We observed a 62% DR rate after CRT alone. Although 50% of patients developed a local recurrence in this series, only 11 of 100 patients (10%) developed an isolated local recurrence. This low level of isolated local recurrence, compared with DR, is consistent with other studies of CRT. In a recent Cancer and Leukemia Group B phase 3 trial, the isolated local recurrence rate was 20% compared with a DR rate of 80%.16 In SWOG 9504, there was a 36% local recurrence rate compared with a DR rate of 64%.15 These data, along with the frequent occurrence of a short interval between DR and death among patients with NSCLC, indicate that more effective systemic therapy is necessary as the primary method for improving the overall outcome in patients with this disease.

The brain was the predominant DR site in this series, in that 23% of all patients developed brain metastases. This high incidence of brain metastases is consistent with the rate observed in our previous experience in patients with stage III NSCLC patients.9 More significantly, the time to development of brain metastasis as a first site of failure was 7.5 months. This indicates the significant effect of brain metastases on OS in these patients. One option, which must be seriously considered in light of these results, is the use of prophylactic cranial irradiation. That approach is currently under study in the cooperative group setting, and more data on this issue may be forthcoming.

The current large cohort of patients who were treated in the era of concurrent CRT provides some observations. First, patients who are able to undergo surgical resection after CRT have an excellent outcome, and surgery remains a therapeutic option for properly selected patients who are treated at experienced centers. Second, the patterns of failure after trimodality therapy differ from those after CRT alone. Finally, despite the use of concurrent CRT, significant room for improvement in systemic therapy is needed to limit DR and significantly improve OS in patients with stage III NSCLC.

Conflict of Interest Disclosures

The authors made no disclosures.