Risk factors for local and regional recurrence in patients with resected N0–N1 non-small-cell lung cancer, with implications for patient selection for adjuvant radiation therapy

Abstract

Background

The purpose of this study was to evaluate the actuarial risk of local and regional failure in patients with completely resected non-small-cell lung cancer (NSCLC), and to assess surgical and pathological factors affecting this risk.

Patients and methods

Between January 1998 and December 2009, 1402 consecutive stage I–III (N0–N1) NSCLC patients underwent complete resection without adjuvant radiation therapy. The median follow-up was 42 months.

Results

Local–regional recurrence was identified in 9% of patients, with local failure alone in 3% of patients, regional failure alone in 4% of patients, and both local and regional failure simultaneously in 2% of patients. Patients who had local failure were found to be at increased risk of mortality. By multivariate analyses, three variables were shown to be independently significant risk factors for local [surgical procedure (single/multiple wedges + segmentectomy versus lobectomy + bilobectomy + pneumonectomy), tumor size >2.7 cm, and visceral pleural invasion] and regional (pathologic N1 stage, visceral pleural invasion, and lymphovascular space invasion, LVI) recurrence, respectively.

Conclusion

Patients with N0–N1 disease have low rates of locoregional recurrence after surgical resection. However, several prognostic factors can be identified that increase this risk and identify patients who may benefit from adjuvant treatment.

introduction

Lung cancer is the leading cause of cancer-related death worldwide [1]. Patients with stage I, II, and selected stage III disease are considered potentially resectable for cure, with 5-year survival rates following complete resection of ∼60%–90%, 30%–70%, and 10%–30%, respectively [2–6]. The risk of locoregional recurrence (LRR) after surgery for non-small-cell lung cancer (NSCLC) is generally considered to be small in comparison with the risk of distant recurrence. Nevertheless, the reported rates of local–regional failure after surgery for NSCLC are variable, and have been reported to be as high as 40% [7–9]. While postoperative radiation therapy (PORT) is generally accepted in the case of adverse features for local failure such as positive margins [10, 11], the other risk factors for LRR [12] that would present an indication for PORT have not been well established. An accurate understanding of the patterns of failure after surgery is essential to guide appropriate adjuvant local therapy.

Furthermore, the optimal method of reporting patterns of failure is not clear. There is no standard definition for ‘local recurrence,’ which has a substantial impact on both the observed incidence of both local and distant recurrence and on which factors are associated significantly with these events [13]. For example, there have been varying definitions for recurrences that manifest as new ipsilateral parenchymal lung nodules, being reported as both regional and distant failure [13, 14]. In this report, we reviewed a large series of consecutive stage I–III (N0–N1) NSCLC patients who underwent complete resection. The purpose of this study was to investigate the risk factors for local and regional recurrence, which could in turn assist in selecting patients that may benefit from adjuvant local therapy.

patients and methods

patients

This institutional review board-approved retrospective study was carried out by searching the MD Anderson Cancer Center database for stage I–III (N0–N1) NSCLC patients with diagnoses based on the 6th edition TNM (T, tumor size; N, involvement of lymph nodes; M, distant metastasis) classification [15] who underwent surgery at MD Anderson between January 1999 and 2009. Patients who received neoadjuvant therapy (chemotherapy and/or radiation therapy), presented with synchronous primary lung tumors, had follow-up times of <12 months (for alive patients), had a prior history of lung cancer, or underwent PORT were excluded. All pathological factors including the tumor size and pleural and lymphatic invasion were documented from the pathologic analysis at the MD Anderson Cancer Center. The scoring of these items was made by different pathologists due to the 11-year time period of the study. The tumor size was defined as the greatest tumor dimension. Visceral and parietal pleural invasion was defined as the presence of tumor within the visceral and parietal pleura, respectively. Lymphatic invasion was defined as the presence of tumor within the lymphatic vessels.

patient follow-up

Chest X-rays or computed tomography (CT) scans of the chest were obtained every 3 months for the first 2 years after treatment, then every 6 months for 5 years, and then annually thereafter to check for evidence of recurrence. Further imaging and studies were obtained at the discretion of the treating physician.

recurrence definition

Disease recurrence at the surgical resection margin was considered a local failure, while mediastinal, hilar, and supraclavicular fossa recurrence were assessed as regional failure. The recent lymph node map proposed by the International Association for the Study of Lung Cancer [16] was used for illustrating the regional recurrences. All other sites of failure were considered distant failures. On the basis of the American College of Chest Physicians recommendations [14], when the failure was detected in a different lobe of the lung (same histology) and there was N2,3 involvement or <2 years interval after the surgery, this recurrence was defined as a distant metastasis. If this particular type of failure did not meet one of the two criteria mentioned above, this occurrence was defined as second primary lung cancer. Nodal failures were defined as a new or enlarging lymph node ≥1 cm on short axis on CT or hypermetabolic on PET, which in the patient's subsequent clinical follow-up was consistent with a true failure. Local and regional failures were scored independently.

statistical analysis

Data were analyzed by using Stata/SE 11.1 software (Stata Corp LP, College Station, TX). The potential risk factors for recurrence were assessed in univariate analyses using the Cox proportional hazards model. Because of the possible confounding effect of clinical factors on survival, associations found to be significant in the univariate analysis were adjusted by patient (age, gender, and race), tumor (location, histology, pathologic size, T stage, N stage), treatment (surgical procedure), and pathological factors. Multivariate analyses were carried out using a logistic regression model, with a stepwise backward elimination procedure. Survival time was measured from the date of resection to the date of the first occurrence of the considered event [recurrence (local, regional, or distant), death] or last follow-up. A P value of ≤0.05 was considered to be statistically significant. Finally, we calculated the sensitivity and specificity of the resulting logistic regression model by constructing receiver operating characteristic (ROC) curves. Additionally, local and regional recurrence rates were reported [recurrence rate = (No. of patients with failures/No. of patients assessed) × 100]. We then calculated the area under the curve (AUC) statistic to estimate each model's ability to predict local or regional recurrence.

results

We identified 1402 patients who underwent surgery for T1–4 N0–1 NSCLC in the defined interval. The median follow-up was 42 months (range 1–156). Patient characteristics and surgical/pathological details are found in Table 1. The 5-year overall survival (OS) rate was 60% (Figure 1). Supplementary Figure S1, available at Annals of Oncology online, shows the Kaplan-Meier death estimates according to T stage. Local–regional recurrence was identified in 9% of patients (N = 123), with local failure alone in 3% of patients (N = 39), regional failure alone in 4% of patients (N = 59), and both local and regional failure simultaneously in 2% of patients (N = 25). There were 140 regional failures among the 84 patients who regionally recurred after surgery (supplementary Figure S2A, available at Annals of Oncology online). Regional recurrence restaging according to the 6th edition TNM classification was 15 patients N1, 44 patients N2, and 25 patients N3. Distant failure was identified in 220 patients (16%) and there were 224 distant failures (as the first site of failure) among these patients (supplementary Figure S2B, available at Annals of Oncology online). Lung was the most common site for distant recurrence (N = 88), with the presentation in the ipsilateral lung in 27 patients, in the contralateral lung in 41 patients, and bilaterally in 20 patients. Patients receiving adjuvant chemotherapy developed distant failure in 32 cases. Local–regional recurrences were confirmed by means of biopsy in 77%, CT in 17%, and PET in 6%. Distant recurrences were confirmed by biopsy in 66%, CT in 15%, and other imaging (PET, magnetic resonance imaging, or bone scan) in 19%.

Table 1.

Patient's characteristics

Characteristic	No. of patients (%) (N = 1402)
Age, years
Median	67
Range	26–90
Gender
Male	698 (50)
Female	704 (50)
Race
Caucasian	1267 (90)
African American	68 (5)
Hispanic	42 (3)
Asian	25 (2)
Locationa
Right upper lobe	518 (37)
Right middle lobe	115 (8)
Right lower lobe	243 (83)
Left upper lobe	410 (29)
Left lower lobe	234 (17)
Surgical procedure
Single wedge	55 (4)
Multiple wedges	15 (1)
Segmentectomy	85 (6)
Lobectomy	1123 (80)
Bilobectomy	24 (2)
Sleeve lobectomy	34 (2)
Pneumonectomy	66 (5)
Pathologic stage
IA	631 (45)
IB	445 (25)
IIA	67 (16)
IIB	179 (7)
IIIA	17 (2)
IIIB	63 (4)
Pathological T
T1	698 (50)
T2	579 (41)
T3	62 (4)
T4	63 (4)
Pathological N
N0	1171 (84)
N1	231 (16)
Tumor size, cm
Median	2.7
Range	0.2–14
Histology
Squamous cell	421 (30)
Adenocarcinoma	825 (59)
Adenosquamous	25 (2)
Bronchoalveolar	42 (3)
Large cell	34 (2)
NSCLC, NOS	55 (4)
Histologic differentiation
Well	175 (12)
Moderate	616 (44)
Poor	492 (35)
Unknown	119 (8)
Visceral pleural invasion
Yes	267 (19)
No	1135 (80)
Parietal pleural invasion
Yes	40 (3)
No	1362 (97)
Lymphovascular space invasion (LVI)
Yes	172 (12)
No	1230 (88)
Adjuvant chemotherapy
Yes	126 (9)
No	1276 (91)

NSCLC, NOS, non-small-cell lung cancer, not otherwise specified.

^aThe tumor was located in more than one lobe contiguously in 118 patients (80 cases in the right lung and 38 cases in the left lung).

Figure 1.

Survival curves for stage I–III (N0–N1) non-small-cell lung cancer patients (NSCLC).

The discriminative value of various cut-off points for the primary tumor size was analyzed in the context of local recurrence. A broad range of cut-offs between 1.5 and 4 cm showed significance, but 2.0 cm was the most discriminative in this particular series (P = 0.0006). Nevertheless, in order to avoid large differences in the sample size between the comparison groups, we used the median value (2.7 cm) for the analysis. Univariate Cox proportional hazards analyses of the data (Table 2) showed a higher risk of local failure in those patients who had advanced T stage (T3/T4; P = 0.047), had a tumor size of >2.7 cm (P = 0.001), and had visceral and parietal pleural invasion (P = 0.001 and P = 0.046, respectively). In addition, those patients who had N1 disease (P = 0.007) had visceral pleural invasion (P = 0.011), and had lymphovascular space invasion (LVI; P = 0.003) showed a higher risk of regional failure. Finally, we found that multiple factors, including those parameters that were associated with local and regional failure, were associated with an increased risk of mortality (Table 2). Moreover, we also observed that patients who had local, regional, or distant failure were found to be at increased risk of mortality (P < 0001).

Table 2.

Univariate analyses of associations between the patient, tumor, treatment, and pathologic characteristics and survival

Variables	Comparison	Overall survival		Disease-free survival		Local disease-free survival		Regional disease-free survival
		HR	P value	HR	P value	HR	P value	HR	P value
Age (years)	≤65	1.00		1.00		1.00		1.00
	>65	1.71	<0.0001	0.98	0.877	0.91	0.715	1.27	0.272
Gender	Female	1.00		1.00		1.00		1.00
	Male	1.66	<0.0001	1.43	0.003	1.36	0.217	1.47	0.079
Race	Caucasian	1.00		1.00		1.00		1.00
	Other	0.91	0.565	0.74	0.553	0.81	0.663	1.33	0.397
Location	1 lobe	1.00		1.00		1.00		1.00
	>1 lobe	1.8	<0.0001	1.45	0.075	0.95	0.928	1.16	0.711
Surgical procedure	Single/multiple wedges + segmentectomy	1.00		1.00		1.00		1.00
	Lobectomy + bilobectomy + pneumonectomy	0.63	<0.0001	0.88	0.513	0.61	0.153	0.81	0.530
Pathologic T stage	T1 + T2	1.00		1.00		1.00		1.00
	T3 + T4	1.62	<0.0001	1.92	<0.0001	2.04	0.047	1.34	0.424
Pathologic N stage	N0	1.00		1.00		1.00		1.00
	N1	1.63	<0.0001	1.84	<0.0001	0.91	0.793	1.95	0.007
Tumor size (cm)	≤2.7	1.00		1.00		1.00		1.00
	>2.7	1.90	<0.0001	2.03	<0.0001	2.27	0.001	1.52	0.054
Histology	Squamous cell	1.00		1.00		1.00		1.00
	Non-squamous cell	0.59	<0.0001	0.94	0.699	0.92	0.781	0.75	0.228
Grade	Well	1.00		1.00		1.00		1.00
	Moderate	1.71	0.0016	1.53	0.047	1.07	0.852	1.28	0.493
	Poor	1.95	<0.0001	1.62	0.027	1.09	0.833	1.40	0.362
Visceral pleural invasion	Yes	1.00		1.00		1.00		1.00
	No	0.79	0.036	0.69	0.009	0.43	0.001	0.53	0.011
Parietal pleural invasion	Yes	1.00		1.00		1.00		1.00
	No	0.46	<0.0001	0.52	0.028	0.35	0.046	0.64	0.452
Lymphovascular space invasion	Yes	1.00		1.00		1.00		1.00
	No	0.70	0.005	0.64	0.008	0.59	0.107	0.45	0.003
Adjuvant chemotherapy	Yes	1.00		1.00		1.00		1.00
	No	1.84	0.003	0.75	0.135	2.10	0.209	0.92	0.939
Local recurrence	No	1.00		NA		NA		NA
	Yes	2.21	<0.0001
Regional recurrence	No	1.00		NA		NA		NA
	Yes	2.13	<0.0001
Distant recurrence	No	1.00		NA		NA		NA
	Yes	2.81	<0.0001

Bold values indicate statistical significance. HR, hazard ratio; NA, not applicable.

There were 616 patients with N1 and/or tumor size >3 cm (N1 and >3 cm, N = 141; N1 and ≤3 cm, N = 90; and N0 and >3 cm, N = 385). Ninety-two of these patients received adjuvant chemotherapy. We assessed the risk of developing local and regional recurrence in patients with N1 and/or tumor size >3 cm (N = 616) compared with the remainder of patients (N = 786). There was no significant increased risk of developing local (P = 0.40) and regional (P = 0.13) recurrence. Additionally, we evaluated the risk of developing local and regional recurrence in patients with N1 and/or tumor size >3 cm who received chemotherapy (N = 92) compared with (i) the rest of the group (N = 1310) and (ii) patients with N1 and/or tumor size >3 cm who did not receive chemotherapy (N = 524). There was no significant increased risk of developing local and regional recurrence in any comparison (P > 0.05).

After adjustment for other covariates by multivariate analysis (Table 3), three unique variables were shown to be independently significant risk factors for local (surgical procedure [single/multiple wedges + segmentectomy versus lobectomy + bilobectomy + pneumonectomy], tumor size >2.7 cm, and visceral pleural invasion) and regional (pathologic N1 stage, visceral pleural invasion, and LVI) recurrence, respectively (Figure 2). The local recurrence rates were relatively similar according to the surgical procedure (6.5% for single/multiple wedges + segmentectomy versus 4% for lobectomy + bilobectomy + pneumonectomy group). In contrast, we observed twice the local recurrence rate for patients with a larger tumor size (3% for those with ≤2.7 cm versus 6% for those with >2.7 cm) and visceral pleural invasion (4% for those without invasion versus 8% for those with invasion). Regional recurrence rates were 5% for N0 patients versus 9% for N1 patients; 5% for those without visceral invasion versus 9% for those with invasion; and 5% for those without LVI versus 10% for those with invasion. Additionally, we found that multiple factors, including age >65 years, male gender, tumor in >1 lobe, advanced T stage (T3/T4), and squamous cell histology were associated with an increased risk of mortality (Table 3). In contrast, adjuvant chemotherapy was associated with reduced mortality (P < 0001). Finally, local (supplementary Figure S3, available at Annals of Oncology online) and distant failure were found to be at increased risk of mortality after adjustment for other covariates (P = 0.001 and P < 0001, respectively).

Table 3.

Multivariate analyses of associations between the patient, tumor, treatment, and pathologic characteristics and survival

Variables	Comparison	Overall survival		Disease-free survival		Local disease-free survival		Regional disease-free survival
		HR	P value	HR	P value	HR	P value	HR	P value
Age (years)	≤65	1.00
	>65	1.59	<0.0001		–		–		–
Gender	Female	1.00
	Male	1.45	<0.0001		–		–		–
Location	1 lobe	1.00
	>1 lobe	1.45	0.013		–		–		–
Histology	Squamous cell	1.00
	Non-squamous cell	0.76	0.006		–		–		–
Surgical procedure	Single/multiple wedges + Segmentectomy	1.00				1.00
	Lobectomy + bilobectomy + pneumonectomy	0.56	<0.0001		–	0.43	0.021		–
Pathologic T stage	T1 + T2	0.22		1.00
	T3 + T4	1.82	<0.0001	1.74	0.002		–		–
Pathologic N stage	N0			1.00				1.00
	N1		–	1.61	0.001		–	1.76	0.026
Tumor size (cm)	≤2.7	1.00		1.00		1.00
	>2.7	1.68	<0.0001	1.70	<0.0001	2.38	0.001		–
Visceral pleural invasion	Yes					1.00		1.00
	No		–		–	0.44	0.003	0.60	0.044
Lymphovascular space invasion	Yes							1.00
	No		–		–		–	0.53	0.028
Adjuvant chemotherapy	Yes	1.00
	No	2.12	<0.0001		–		–		–
Local recurrence	No	1.00
	Yes	1.75	0.001		–		–		–
Distant recurrence	No	1.00
	Yes	2.58	<0.0001		–		–		–

Bold values indicate statistical significance HR, hazard ratio.

Figure 2.

Freedom from local and regional recurrence based on the (A) tumor size, (B) visceral pleural invasion, (C) nodal status, and (D) Lymphovascular invasion.

On ROC analysis, we found that the tumor size was the most important predictor for local recurrence, with an AUC of 0.62. By adding the other significant variables (surgical procedure and visceral pleural invasion), the predictive power of the local recurrence risk model significantly improved, enhancing the AUC by 0.08 (AUC 0.70, P < 0.0001). For the regional recurrence risk, the three significant variables found on multivariate analysis had similar values for AUC (pathologic N1 stage, AUC 55; visceral pleural invasion, AUC 53; and LVI, AUC 56). By combining the three variables, the predictive power of the model improved with a marginal significance (AUC 0.60; P = 0.0665).

We then attempted to find the highest risk groups for local and regional failure that were clinically relevant. To this end, we selected only those subsets of high risk patients who represented at least 10% of the whole series and whose recurrence rates of local/regional failure were at least 5% above the comparison group. We found that patients with tumor size >2.7 cm and visceral invasion (N = 155; 11% of the whole series) had 10% local failure compared with 4% in those with a lower tumor size and without visceral invasion [hazard ratio (HR): 3.05; P = 0.001]. In terms of regional recurrence, only those patients with LVI (N = 172; 12% of the whole series) met the criteria outlined above.

discussion

The risk of local and regional recurrence after surgery for resected NSCLC has not been well defined, but is generally felt to be low and overshadowed by the risk of distant recurrence. Nevertheless, as systemic therapy improves [17, 18], achieving local control will likely assume greater importance. Our pertinent findings can be summarized as follows. First, we found that the extent of surgical resection (single/multiple wedges + segmentectomy versus lobectomy + bilobectomy + pneumonectomy), tumor size >2.7 cm, and visceral pleural invasion were shown to be independently significant risk factors for local failure. Second, we found that pathologic N1 stage, visceral pleural invasion, and LVI were risk factors for regional failure. Finally, we found that multiple factors, including the presence of local failure, were associated with an increased risk of mortality. This information could be used to help identify patients with a high risk of recurrence who would be candidates for more aggressive postoperative treatment, and if validated could be potentially utilized to further refine future staging systems.

There is no clear consensus as to which patients with N0–N1 disease and negative margins could derive the most benefit from adjuvant radiation therapy [10, 19, 20]. For instance, Kelsey et al. [10] reported that >50% of recurrences after surgery for early-stage NSCLC involve local sites, with local recurrence being defined as at the surgical resection margin, ipsilateral hilum, and mediastinum. Surgical procedures (wedge/segmentectomy versus lobectomy) and LVI were found to be risk factors for local failure. Jonnalagadda et al. [19] found that the number of lymph nodes in N1 disease was a risk factor for disease failure. And in a recent study, Varlotto et al. [20] examined 60 consecutive unirradiated patients with N1 disease and found high failure rates in both the hilum and mediastinum (46% local failure at 5 years). The authors proposed a PORT portal encompassing the bronchial stump, the anastomosis, the hilum, the subcarinal area, and the ipsilateral and contralateral mediastinum for stage II NSCLC and N1 nodal disease patients. It should be noted that this reported local failure rate is much higher than that which has been previously reported in the literature for this subset of patients. Nonetheless, in the current study with over 1400 patients, we found similar factors for LRR as have been published in prior studies (i.e. N1 versus N0, LVI, surgical procedure).

Maeda et al. [21] reported a statistically significant difference in OS between adenocarcinoma and squamous cell carcinoma (5-year OS rate: 79.7% and 63.8%, respectively) in resected NSCLC patients. Consistently, in our study, there was a significant increase in the risk of mortality for those patients with squamous cell carcinoma. In addition, age and gender were also found to be prognostic factors for OS after multivariate analysis. Our findings are concordant with those of other studies in NSCLC in which OS was better in women than men [22–24].

Our study also showed that adjuvant chemotherapy was associated with a decreased risk of mortality. This conclusion is consistent with the Lung Adjuvant Cisplatin Evaluation meta-analysis [25] of 4584 NSCLC patients, which showed that adjuvant chemotherapy increased survival from 64% to 67% for stage IB, from 39% up to 49% for stage II and from 26% up to 39% for stage III NSCLC at a median follow-up of 5.1 years. On the basis of this and other recent meta-analyses [25–27], the 2010 European Society for Medical Oncology Clinical Practice Guidelines [28] recommend the use of adjuvant chemotherapy in the stage II–III radically resected NSCLC.

In contrast to most prior studies, the current analysis is novel in distinguishing local and regional recurrence. These findings have clinical implications for the use of radiation therapy. Specifically, our data suggest that with larger tumors with visceral pleural invasion, regardless of margin status, a radiation field encompassing the primary site could reduce local failure, while for patients with N1 disease, visceral pleural invasion, and/or LVI, a mediastinal field could decrease the incidence of regional recurrences. Potential next steps for future analyses would be to incorporate the role of genetic factors into clinical and pathological features, and then to collectively analyze these aggregate findings into a prospective analysis assessing the benefit of further individualizing the criteria for PORT.

We acknowledge that there are limitations to our study. First, a systematic method of follow-up, including imaging, would be optimal to evaluate the patterns of failure after surgery. We acknowledge that given the retrospective nature of this analysis and the 11-year time period which it covers, optimal imaging with modern techniques (e.g. at least a CT scan of the chest) did not occur in a proportion of patients. However, at MD Anderson, chest CT was incorporated routinely in the follow-up of NSCLC patients in approximately 2003. Therefore, virtually all patients followed up after this time (∼2/3 of patients of our series) underwent follow-up with this imaging modality. Second, although several subsets of patients showed significant differences in terms of local and regional failure, some of them had relatively low and similar recurrence rates (i.e. local failure: 6.5% for the single/multiple wedges + segmentectomy group versus 4.3% for the lobectomy + bilobectomy + pneumonectomy group). Therefore, we would recommend the consideration of PORT when, in addition to having a significant risk of local or regional recurrence, the recurrence rates were also high enough to be clinically relevant. We found two subsets of patients who met these criteria. First, those patients with tumor size >2.7 cm and visceral invasion represented 11% of our series and had 10% local failure, compared with 4% in those with lower tumor size and without visceral invasion. In addition, patients with LVI represented 12% of the current series and had 10% regional failure, compared with 5% in those without invasion. Of course, a determination as to whether 10% is thought to be a ‘meaningful’ risk of locoregional failure is at the treating physician's discretion. Finally, we should emphasize that while prior studies [29] have shown that postoperative radiation significantly reduces local relapses in selected NSCLC patients, an extrapolation to improvements in OS is not certain. However, given that we found a correlation between local recurrence and OS, we believe that this finding provides an important foundation for further assessing the role of PORT in specific clinical scenarios as outlined above.

In one of the largest published series of surgical stage I–III (N0–N1) NSCLC patients, we demonstrated a high rate of locoregional control, but were able to identify distinct clinical and pathologic risk factors for both local and regional recurrence. We also found that improved local control was associated with increased OS. This information can be used to further refine the criteria for PORT in the setting of resected N0–N1 disease.

funding

This work was supported in part by Cancer Center Support (Core) Grant CA016672 from the National Institutes of Health to The University of Texas MD Anderson Cancer Center.

disclosure

The authors have declared no conflicts of interest.