Pathologic Nodal Staging Prior to SBRT for Early-Stage NSCLC Does Not Impact Overall Survival: A Propensity Score Matched NCDB Analysis

Abstract

Objective:

Stereotactic body radiation therapy (SBRT) for early-stage non-small cell carcinoma of the lung (NSCLC) is increasingly utilized. We sought to assess overall survival (OS) for early-stage NSCLC patients receiving SBRT depending on staging method.

Methods:

Early-stage NSCLC patients treated with definitive SBRT were identified in the National Cancer Database (NCDB), and OS was determined based on method of staging. Patient, disease, and treatment characteristics were also analyzed.

Results:

12,106 patients were included; 865 (7%) received invasive staging (nodal sampling, NS) and 11,241 (93%) had no nodal sampling (NNS). From this larger dataset, a propensity score matching (1:1 without replacement) was performed, which yielded 839 patients for each group (NNS and NS). With a median follow-up time of 3.12 years, median survival for all patients included in the matched dataset was 2.75 years (95% CI 2.55-2.93 years), with 2-year and 5-year OS estimated at 63.9% and 25.7% respectively. In a multivariable analysis on matched data, there was no difference in mortality risk between the NNS and NS groups (HR 1.08 (95% CI 0.94-1.24), p=0.25). Negative prognostic factors identified in the multivariable analysis of the matched data included: age >65, male sex, Charlson-Deyo Score ≥1, and tumor size ≥3cm.

Conclusion:

SBRT use in early-stage NSCLC steadily increased over the study period. Most patients proceeded to SBRT without nodal staging, conflicting with NCCN guidelines which recommend pathologic mediastinal lymph node evaluation for all early-stage NSCLC cases, except Stage IA. Our findings suggest similar OS in patients with early-stage NSCLC treated with SBRT irrespective of nodal staging. Furthermore, we highlight patient-, disease-, and treatment-related prognostic factors to consider when planning therapy for these patients.

Introduction

Carcinoma of the lung is the second most common cancer by incidence and the leading cause of cancer related death in both women and men in the United States, with an estimated 235,760 new cases in 2021 (roughly 12.5% of all new cancer cases).¹ Stage I and II non-small cell lung carcinoma (NSCLC) comprise roughly 30% of new NSCLC cases and roughly 5% of all new cancer diagnoses each year.² Stereotactic body radiation therapy (SBRT) has become an accepted alternative to surgical resection for those early-stage patients with T1-2N0 disease, with local control rates of approximately 90% at 2 years.³

The National Comprehensive Cancer Network (NCCN) recommends pathologic mediastinal staging as part of pre-treatment evaluation for all stage I and II NSCLC cases with the exception of peripheral T1 tumors and purely non-solid tumors measuring less than 3 cm (where the NCCN states that mediastinal staging is optional).⁴ Despite the recommendation for patients with a confirmed NSCLC to undergo invasive staging procedures such as endobronchial ultrasound (EBUS) or mediastinoscopy, there are instances when stage I-II NSCLC cases are treated with SBRT following PET or CT staging only or biopsy of the primary lesion without pathologic mediastinal staging. A prior analysis of National Cancer Database (NCDB) data from patients with early-stage NSCLC treated with SBRT found an increase over time in the proportion of patients treated without biopsy confirmation of malignancy.⁵

Numerous studies have been undertaken to elucidate the predictive value of staging modalities, reporting respective sensitivities and specificities of 80% and 90% for PET/CT,^6-8 89% and 100% for EBUS,⁸ and 78% and 100% for mediastinoscopy.⁸ Single-institution retrospective studies have also compared outcomes of invasive versus non-invasive staging with some reporting no significant differences in overall survival or loco-regional control,^9-11 and others showing higher rates of nodal failure in those patients with non-invasive staging.¹² Occult nodal disease (N2) in NSCLC with negative PET imaging occurs in ~16% of cases,¹³ and independent predictors of occult N2 disease include centrally-located tumors and right upper lobe tumors,^13,14 There is a possibility that forgoing invasive staging in these patients increases the risk of nodal recurrence, thereby decreasing survival.

This project aims to harness the NCDB dataset for this patient population in order to identify characteristics of patient, tumor, staging workup, and radiation dose and delivery technique that affect overall survival. We hypothesize that those patients receiving SBRT without full pathologic staging workup will have a poorer overall survival compared to patients receiving SBRT with a full pathologic staging workup, i.e., nodal sampling.

Patients and Methods

Study Design

Using data from the NCDB, a retrospective observational cohort study was performed investigating survival outcomes and prognostic indicators for early-stage NSCLC receiving definitive SBRT. The NCDB, one of the largest hospital-based cancer registries in the world with over 1500 accredited facilities, accounts for approximately 70% of newly diagnosed cancer cases in the Unites States. The NCDB is a joint project of the Commission on Cancer of the American College of Surgeons and the American Cancer Society. The database records numerous patient-, tumor-, and treatment-characteristics as well as overall survival. The data used in the study are derived from a de-identified NCDB file. The American College of Surgeons and the Commission on Cancer have not verified and are not responsible for the analytic or statistical methodology employed, or the conclusions drawn from these data by the investigator.

Patient Selection

The NCDB lung dataset utilized included over 1.5 million cases diagnosed from 2010 through 2015. Only those patients with a singular primary malignancy were included for analysis. Further parameters included early-stage NSCLC, which for the purposes of this study was defined as American Joint Committee on Cancer(AJCC) Stage I or II, node negative, with an NSCLC ICD-0-3 histology code including 8046 (non-small cell carcinoma, NOS), 807x (squamous cell carcinoma), 814x (adenocarcinoma), 856x (adenosquamous). Patients who did not receive any radiation were excluded from the cohort as well as those who received surgical resection or chemotherapy. Lastly, for those qualifying cases, radiation parameters defined as starting less than 180 days from diagnosis, receiving 1-5 fractions, with a total dose of 3000-7000 cGy, and total radiation time less than three weeks were applied leaving 12,106 patients meeting criteria for analysis (Figure 1). Available database variables known or presumed to be of clinical importance were also collected and analyzed for relationship with overall survival. Patient demographics such as age, race, Charlson-Deyo score, insurance status/type, income, education, and residential status were included. Disease factors such as tumor size and laterality were included as well as treatment characteristics including type of work-up, time to treatment, total dose of radiation, dose per fraction, and treatment modality. The NNS group was determined if the “Regional_Nodes_Examined” variable in the NCDB dataset was “00” with the NS group being the remainder. Patients were excluded if they did not have a code for the “Regional_Nodes_Examined” variable or if the variable was coded as unknown.

Figure 1.

Cohort selection diagram

Statistical Analysis

We calculated mean and standard deviation for continuous characteristics and count and percent for categorical characteristics as the descriptive statistics. We used the chi-square test and t-test to assess the univariable association between the staging approach and the categorical variables and continuous variables respectively. Variables associated with the nodal staging approach in the overall cohort were age, sex, race, Charlson-Deyo comorbidity index, histology, tumor size, calendar year of diagnosis, and month from diagnosis to radiation. In addition to these variables, based on the potential clinical significance, insurance(primary payer) and place of residence were also included in the logistic regression model to generate the propensity score for treatment assignment( nodal staging approach).Patients receiving either NS or NNS were matched 1:1 on the propensity score using a greedy nearest neighbor matching algorithm without replacement within the caliper size of 0.05 logit propensity score. The standardized difference was estimated after matching to assess the balance of covariates. The matching yielded 839 patients who received NS (and 839 patients who received NNS) balanced in all the covariates included in the model to calculate the propensity score. In the PS matching, 97% of the NS cases (839 out of 865) were matched with a control.

We used the Kaplan-Meier method to estimate the overall survival (OS), median survival duration, and the survival probability at 2 years and 5 years. Survival duration was defined as the time (in years) from cancer diagnosis to death censoring patients still living at date of last contact. The log-rank test was used to assess the difference in overall survival across the categories of diagnostic methods. We calculated the hazard ratio (HR) with 95% confidence intervals (CI) using unadjusted and adjusted Cox proportional-hazard models. Covariates that were significant in the unadjusted models were included in the adjusted models. Since, our primary exposure variable was staging approach (NS vs NNS), included the staging approach (NS vs NNS) regardless of its statistical significance. We assessed the two-way interaction among the staging approach and background characteristics using the maximum likelihood estimation and there was no interaction. An alpha of 0.05 was used. SAS software (version 9.4, SAS Institute Inc., Cary, NC) was used to analyze the data.

Results

Use of SBRT Over Time

The utilization of SBRT increased 2.5-fold during the time of the study from 1181 in 2010 to 3057 cases in 2015. The percentage with nodal sampling also increased through time from 4.6% in 2010 to 8.4% in 2015.

Patient Characteristics

Characteristics for the entire cohort are described in Table 1. The final cohort consisted of 12,106 patients with 865 (7%) having received nodal sampling (NS, considered to be standard of care) as part of their staging workup and 11,241 (93%) having no nodal sampling (NNS) as a part of their staging work up. Figure 2 shows the survival curve for the entire cohort. As reported in Table 1, the NNS and NS groups are not evenly matched in terms of patient, tumor, or treatment characteristics. Therefore, a propensity score matched (PSM) dataset was created which resulted in 839 patients for each of the NNS and NS groups. With a median follow-up time of 3.12 years, the median survival for all patients included in the matched dataset was 2.75 years (95% CI: 2.55-2.93 y), with 2- and 5-year OS estimated at 63.9% and 25.7%, respectively (Fig. 2).

Table 1.

Patient, Tumor, and Treatment Characteristics by Nodal Staging-Entire Cohort

Factor	Total Cohort	NNS(N= 11241)	NS(N=865)
	Count (%)	Count (%)	Count (%)	P-value
Patient Characteristics
Age
<65 years	1871 (15	1723 (15)	148 (17)	<0.001
65-74 years	4070 (34)	3725 (33)	345 (40)	.
>=75 years	6165 (51)	5793 (52)	372 (43)	.
Sex
Male	5453 (45)	5012 (45)	441 (51)	<0.001
Female	6653 (55)	6229 (55)	424 (49)	.
Race
White	10714 (88)	9976 (89)	738 (85)	0.002
Black	1069 (9)	964 (9)	105 (12)	.
Others/unknown	323 (3)	301 (3)	22 (3)	.
Charlson-Deyo Score
0	6390 (53)	5981 (53)	409 (47)	0.007
1	3291 (27)	3037 (27)	254 (29)	.
2	1633 (13)	1497 (13)	136 (16)	.
>=3	792 (7)	726 (6)	66 (8)	.
Tumor Characteristics
Histology
Squamous cell carcinomas	4407 (36)	4027 (36)	380 (44)	<0.001
Adenocarcinomas	5737 (48)	5362 (48)	375 (43)	.
Adenosquamous carcinomas	114 (1)	99 (1)	15 (2)	.
Non-small cell carcinomas, NOS	1848 (15)	1753 (16)	95 (11)	.
Tumor Size
<1-1 cm	503 (4)	472 (4)	31 (4)	<0.001
1-2 cm	4599 (38)	4347 (39)	252 (29)	.
2-3 cm	4109 (34)	3803 (34)	306 (35)	.
3-4 cm	1819 (15)	1651 (15)	168 (19)	.
4-5 cm	642 (5)	574 (5)	68 (8)	.
>5 cm	236 (2)	203 (2)	33 (4)	.
Unknown, size not stated	197 (2)	190 (2)	7 (1)	.
Staging Approach
Nodal Staging	11241 (93)
No Nodal Staging	865 (7)
Treatment Characteristics
Radiation Modality
External beam, NOS	311 (3)	293 (3)	18 (2)	0.051
Proton	1710 (14)	1560 (14)	150 (17)	.
IMRT	948 (8)	873 (8)	75 (9)	.
3D/Conformal	289 (2)	270 (2)	19 (2)	.
Radiosurgery	8811 (73)	8212 (73)	599 (69)	.
Other, NOS	37 (<1)	33 (<1)	4 (<1)	.
Number of Fractions
1	101 (1)	92 (1)	9 (1)	0.068
2	14 (<1)	14 (<1)	0 (0)	.
3	2964 (25)	2785 (25)	179 (21)	.
4	3298 (27)	3050 (27)	248 (29)	.
5	5729 (47)	5300 (47)	429 (50)	.
BED
BED <100 Gy	421 (4)	388 (3)	33 (4)	0.067
BED 100-140 Gy	8731 (72)	8082 (72)	649 (75)	.
BED >140 Gy	2954 (24)	2771 (25)	183 (21)	.
Year of Diagnosis
2010	1181 (10)	1126 (10)	55 (6)	<0.001
2011	1498 (12)	1414 (13	84 (10)	-
2012	1788 (15)	1670 (15)	118 (14)	-
2013	2172 (18)	2013 (18)	159 (18)	-
2014	2410 (20)	2220 (20)	190 (22)	-
2015	3057 (25)	2798 (25)	259 (30)	-
Month from Diagnosis to Radiation
≤1	1576 (13)	1507 (13)	69 (8)	<0.001
2-4	9557 (79)	8877 (79)	680 (79)	.
≥5	915 (8)	802 (7)	113 (13)	.

Figure 2.

Overall survival for early-stage NSCLC-PSM

Univariable and Multivariable Analysis of Overall Cohort and PSM Data

On univariable analysis of the overall cohort, older age, male sex, white race, higher Charlson-Deyo score, adenocarcinoma or squamous cell histology (compared to NSCLC NOS), tumor size ≥1 cm, BED <100 Gy, certain dose per fraction regimens, and Medicare as primary payer were all found to be associated with higher risk of mortality (Table 2). On multivariable analysis, age ≥75 years, male sex, white race, increasing Charlson-Deyo score, adenocarcinoma histology, tumor size ≥2 cm, BED <100 Gy, and Medicare as primary payer increased a patient’s hazard of death (Table 2). For example, the hazard of death is 1.28 times greater for males as compared females when adjusting for other variables in the model (HR 1.28, 95% CI 1.21-1.34). The PSM analysis differed somewhat from the overall cohort. In the multivariate analysis, only age ≥65 years (65-74 years HR 1.276, 95% CI 1.037-1.570, p=0.215; ≥75 years HR 1.369, 95% CI 1.118-1.677, p=0.0024), male sex (HR 1.214, 95% CI 1.061-1.389, p=0.0047), and increasing Charlson-Deyo score (compared to Charlson-Deyo 0, Charlson-Deyo 1: HR 1.289, 95% CI 1.104-1.506, p=0.0014; Charlson-Deyo 2: HR 1.239, 95% CI 1.013-1.514, p=0.0367; Charlson-Deyo ≥3: HR 1.745, 95% CI 1.342-2.270, p<0.0001) and tumor size (compared to <1 cm, 3-4 cm: HR 1.775, 95% CI 1.105-2.854, p=0.0178; >5 cm: HR 2.780, 95% CI 1.606-4.813, p=0.0003; size unknown/not stated: HR 3.512, 95% CI 1.571-7.850, p=0.0022) remained as significant predictors of mortality. Dose, fractionation and biologically effective dose (BED) results are discussed further in another section.

Table 2:

Univariable and Multivariable Cox Proportional Hazards Regression Analysis for Overall Survival for Overall Cohort (N= 12,106)

	Unadjusted Analysis				Adjusted Analysis
Characteristics	HR	Lower CI	Upper CI	p- value	HR	Lower CI	Upper CI	p- value
Age
<65 years	Reference				Reference
65-74 years	1.13	1.04	1.22	0.0033	1.00	0.91	1.09	0.9471
≥75 years	1.27	1.18	1.38	<.0001	1.12	1.02	1.22	0.013
Sex
Female	Reference				Reference
Male	1.32	1.25	1.39	<.0001	1.28	1.21	1.34	<.0001
Race
White	Reference				Reference
Black	0.84	0.77	0.92	0.0002	0.87	0.79	0.95	0.0026
Others/unknown	0.83	0.70	0.98	0.0239	0.83	0.70	0.98	0.0262
Charlson- Deyo Score
0	Reference				Reference
1	1.11	1.05	1.18	0.0007	1.13	1.06	1.20	0.0001
2	1.25	1.16	1.35	<.0001	1.24	1.15	1.33	<.0001
≥3	1.58	1.43	1.74	<.0001	1.56	1.42	1.72	<.0001
Histology
Non-small cell carcinomas, NOS	Reference				Reference
Adenocarcinomas	0.81	0.75	0.87	<.0001	0.79	0.74	0.86	<.0001
Adenosquamous carcinomas	1.16	0.90	1.49	0.2489	1.06	0.82	1.37	0.6473
Squamous cell carcinomas	1.13	1.05	1.21	0.0015	1.03	0.96	1.11	0.4286
Tumor size
<1 cm	Reference				Reference
1-2 cm	1.17	1.01	1.35	0.0328	1.15	1.00	1.33	0.0586
2-3 cm	1.44	1.25	1.67	<.0001	1.40	1.21	1.62	<.0001
3-4 cm	1.69	1.45	1.97	<.0001	1.59	1.37	1.86	<.0001
4-5 cm	2.05	1.73	2.43	<.0001	1.90	1.60	2.25	<.0001
>5 cm	2.52	2.05	3.09	<.0001	2.29	1.86	2.82	<.0001
Unknown, size not stated	1.47	1.17	1.85	0.001	1.41	1.11	1.78	0.0043
Nodal examination
Nodal Staging (NS)	Reference				Reference
No Nodal Staging (NNS)	0.93	0.85	1.03	0.1495	1.00	0.90	1.10	0.9402
Treatment Modality
Radiosurgery	Reference				Reference
3D/Conformal	1.10	0.94	1.29	0.2367	1.11	0.95	1.30	0.1983
External beam, NOS	1.04	0.89	1.22	0.6176	1.01	0.86	1.18	0.9354
IMRT	1.10	1.00	1.20	0.0544	1.07	0.98	1.18	0.1438
Other, NOS	0.95	0.62	1.46	0.806	0.96	0.62	1.47	0.8409
Proton	1.03	0.96	1.11	0.4156	1.03	0.96	1.11	0.4071
BED
BED 100-140 Gy	Reference				Reference
BED >140 Gy	0.92	0.87	0.97	0.0036	0.98	0.80	1.20	0.8376
BED <100 Gy	1.37	1.21	1.55	<.0001	2.12	1.21	3.72	0.0091
Fraction # and Dose
5Fx (≥50 Gy)	Reference				Reference
3Fx (48-53.99 Gy)	1.06	0.55	2.04	0.8631	1.02	0.52	2.03	0.949
	0.88	0.71	1.09	0.2393	0.86	0.69	1.07	0.1643
3Fx (≥54 Gy)	0.88	0.83	0.94	0.0001	0.97	0.78	1.20	0.7608
3Fx (30-47.99 Gy)	1.31	0.97	1.76	0.0759	1.00	0.67	1.49	0.993
4Fx (30-47.99 Gy)	1.48	1.18	1.86	0.0007	0.73	0.40	1.32	0.2942
4Fx (≥48 Gy)	0.91	0.86	0.97	0.0028	0.94	0.88	1.00	0.056
5Fx (30-49.99 Gy)	1.19	1.02	1.39	0.0252	0.53	0.30	0.95	0.0334
Single Fx (≥30 Gy)	0.84	0.64	1.10	0.2065	0.88	0.65	1.19	0.394
Diagnosis to radiation
< 2months	Reference				Reference
2-4 months	1.07	0.99	1.15	0.0897	1.06	0.99	1.15	0.1102
>5 months	0.92	0.82	1.03	0.1636	0.94	0.83	1.05	0.2685
Year of Diagnosis
2015	Reference				Reference
2010	1.01	0.92	1.12	0.817	1.02	0.92	1.12	0.7469
2011	1.07	0.98	1.18	0.1364	1.07	0.98	1.18	0.1345
2012	1.02	0.93	1.12	0.6232	1.02	0.93	1.12	0.6328
2013	1.04	0.95	1.14	0.383	1.02	0.94	1.12	0.6022
2014	0.98	0.89	1.07	0.6258	0.97	0.89	1.07	0.5579
Primary Payer
Private	Reference				Reference
Insurance Status Unknown	1.22	0.97	1.52	0.0897	1.15	0.91	1.44	0.2455
Medicaid	1.01	0.87	1.17	0.9024	1.07	0.92	1.25	0.3899
Medicare	1.17	1.08	1.27	<.0001	1.11	1.02	1.21	0.0123
Uninsured	0.75	0.54	1.03	0.0755	0.76	0.54	1.05	0.0936
Other Government	0.99	0.85	1.16	0.9199	0.86	0.74	1.01	0.0622
Urban/Rural
Metro County	Reference				Reference
Not Available	0.79	0.66	0.95	0.0121	0.82	0.68	0.99	0.0387
Rural County	0.94	0.79	1.11	0.436	0.92	0.78	1.09	0.3312
Urban County	1.03	0.96	1.10	0.4334	1.00	0.93	1.07	0.9783

Impact of Omitting Nodal Sampling

Table 3 illustrates well balanced groups after propensity score matching. The only variable that differs significantly between the two groups is time from diagnosis to radiation, which is longer for the NS group (median 3 months) than the NNS group (median 2 months). Figure 3 shows the survival curves for the NS and NNS groups. Median survival was 2.83 years (95%CI 2.53-3.1 years) and 2.72 years (95% CI 2.47-2.94 years) for the NS and NNS groups respectively. On univariable analysis, mortality risk was not significantly different between NNS versus NS groups (HR 1.06 (95% CI 0.93-1.21), p=0.39).

Table 3:

Patient, Tumor, and Treatment Characteristics by Nodal Staging – Propensity-Score Matched Data

		Nodal Staging Status
Covariate	Total Cohort (N=1678)(%)	No Nodal Staging (NNS) (N=839)(%)	Nodal Staging (NS) (N=839)(%)	P-value
Age
<65 years	286 (17)	150 (18)	136 (16)	0.259
65-74 years	646 (38	307 (37)	339 (40)
≥75 years	746 (45)	382 (45)	364 (44)
Sex
Female	830 (49)	404 (48)	426 (51)	0.283
Male	848 (51)	435 (52)	413 (49)
Race
White	1454 (87)	735 (88)	719 (86)	0.318
Black	188 (11)	90 (11)	98 (12)
Others/unknown	36 (2)	14 (1)	22 (2)
Charlson-Deyo Score
0	806 (48)	403 (48)	403 (48)	1
1	502 (30)	251 (30)	251 (30)
2	254 (15)	127 (15)	127 (15)
>=3	116 (7)	58 (7)	58 (7)
Histology
Non-small cell carcinomas, NOS	748 (45)	383 (46)	365 (44)	0.634
Adenocarcinomas	737 (44)	367 (44)	370 (44)
Adenosquamous carcinomas	21 (1)	9 (1)	12 (1)
Squamous cell carcinomas	172 (10)	80 (9)	92 (11)
Tumor Size
<1 cm	52 (3)	26 (3)	26 (3)	1
1-2 cm	504 (30)	252 (30)	252 (30)
2-3 cm	606 (36)	303 (36)	303 (36)
3-4 cm	322 (19)	161 (19)	161 (19)
4-5 cm	130 (8)	65 (8)	65 (8)
>5 cm	54 (3)	27 (3)	27 (3)
Unknown, size not stated	10 (1)	5 (1)	5 (1)
Radiation Modality
External Beam, NOS	36 (2)	18 (2)	18 (2)	0.411
Proton	257 (15)	114 (14)	143 (17)
IMRT	144 (9)	69 (8)	75 (9)
3D/Conformal	36 (2)	17 (2)	19 (2)
Radiosurgery	1200 (72)	619 (74)	581 (69)
Other, NOS	5 (<1)	2 (<1)	3 (<1)
Number of Fractions
1	18 (1)	9 (1)	9 (1)	0.997
3	353 (21)	176 (21)	177 (21)
4	478 (29)	241 (29)	237 (28)
5	829 (49)	413 (49)	416 (50)
BED
<100	71 (4)	42 (5)	29 (4)	0.284
100-140	1247 (74)	617 (74)	630 (75)
>140	360 (22)	180 (21)	180 (21)
Fraction # and Dose
Single Fraction (≥30 Gy)	18 (1)	9 (1)	9 (1)	0.885
3 Fractions (30-47.99Gy)	8 (<1)	5 (1)	3 (<1)
3 Fractions (48-53.99 Gy)	18 (1)	10 (1)	8 (1)
3 Fractions (≥54 Gy)	327 (19)	161 (19)	166 (20)
4 Fractions (30-47.99 Gy)	14 (1)	9 (1)	5 (1)
4 Fractions (≥48 Gy)	464 (28)	232 (28)	232 (28)
5 Fractions (30-49.99 Gy)	53 (3)	30 (4)	23 (3)
5 Fractions (≥50Gy)	776 (46)	383 (46)	393 (47)
Year Of Diagnosis
2010	107 (6)	54 (6)	53 (6)	0.932
2011	174 (10)	93 (11)	81 (10)
2012	229 (14)	115 (14)	114 (14)
2013	316 (19)	158 (19)	158 (19)
2014	351 (21)	169 (20)	182 (22)
2015	501 (30)	250 (30)	251 (30)
Primary Payer
Uninsured	10 (1)	4 (<1)	6 (1)	0.846
Private Insurance	218 (13)	115 (14)	103 (12)
Medicaid	70 (4)	32 (4)	38 (5)
Medicare	1300 (78)	650 (77)	650 (77)
Other Govt	65 (4)	30 (4)	35 (4)
Unknown	15 (1)	8 (1)	7 (1)
Urban Rural
Metro Counties	1313 (78)	659 (79)	654 (78)	0.64
Urban Counties	304 (18)	152 (18)	152 (18)
Rural Counties	46 (3)	23 (3)	23 (3)
Not Available	15 (1)	5 (1)	10 (1)
Time from Diagnosis to RT (Months)
N	1678	838	836	<0.001
Mean	2.81	2.58	3.05
Median	3	2	3
≤1 month	163 (10)	95 (11)	68 (8)	<0.001
2-4 months	1341 (80)	684 (82)	657 (79)
≥5 months	170 (10)	59 (7)	111 (13)

Figure 3:

Overall Survival Comparison of NS and NNS Groups -PSM

Effect of BED on Overall Survival

BED was calculated using the linear quadratic equation, assuming an α/β of 10. Patients were then binned into three groups based on BED (<100 Gy, 100-140 Gy, and >140 Gy). These bin levels were selected based on reported tumor control probabilities for Stage I NSCLC in Brown et al.¹⁵ and Onishi et al.¹⁶ Table 2 reports the associated hazard ratio with BED 100-140 Gy as the reference group. On univariable analysis, BED >140 was associated with improved survival while BED <100 Gy was associated with inferior survival. On multivariable analysis, BED <100 Gy remained significant for decreased survival (HR 2.12, 95% CI 1.21-3.72, p=0.0091). These differences did not remain significant in the PSM analysis. The calculated hazard ratios were 1.11 (95% CI 0.8-1.53, p = 0.53) for BED <100 and 0.92 (95% CI 0.78-1.08, p = 0.29) for BED >140 Gy when compared to the 100-140 Gy group. Table 2 also displays specific fractionation schemes compared to a reference (5 fractions regimen to≥50 Gy). On multivariable analysis, only 5 fraction regimens with total dose from 30-49.99 Gy were significantly associated with improved survival, but there was no significant difference between regimens in the PSM analysis.

Discussion and Conclusion

The use of SBRT to treat early-stage NSCLC has continued to increase over the time period of this study, and the vast majority proceeded to SBRT without nodal staging. The findings of this study suggest similar overall survival for those patients with early-stage NSCLC treated with SBRT who underwent mediastinal nodal staging compared to those without nodal staging. Therefore, the hypothesis that omitting nodal sampling would compromise survival is not supported per the current study parameters.

With a median follow-up time of 3.12 years, the median survival for all patients included in the matched dataset was 2.75 years (95% CI 2.55-2.93 years), with 2-year and 5-year OS estimated at 63.9% and 25.7% respectively (Figure 2). Median survival was 2.83 years (95%CI 2.53-3.1 years) and 2.72 years (95% CI 2.47-2.94 years) for the NS and NNS groups respectively. On univariable analysis of the PSM cohort, the risk of mortality was not significantly different between NNS versus NS groups (HR 1.06 (95% CI 0.93-1.21), p=0.39). Furthermore, this study highlights numerous patient-, disease-, and treatment-related prognostic variables to consider when counseling and planning therapy for this patient subset.

A near 2.5-fold increase in the number of SBRT cases was seen from 2010 to 2015. This is likely related to greater physician comfort with the technique as well as increased utilization in central tumors.³ The percentage of patients receiving nodal sampling actually increased over the years analyzed in this study, but still only represented a small portion of the total population receiving SBRT (4.4% received nodal staging in 2010 versus 7.8% of total SBRT cases in 2015). Year of diagnosis was not associated with survival.

While NCCN guidelines recommend nodal staging for a large proportion of early-stage NSCLC patients, studies such as RTOG 0813 and 0915 did not require nodal staging except in the setting of enlarged, PET positive nodes. RTOG 0813 included patients who were medically inoperable, performance status 0-2, with T1-T2 N0 M0 disease, and central tumors (theoretically the most at risk for nodal metastases). CT and PET were required within 8 weeks of registration, and any enlarged or PET-positive nodes had to be sampled to confirm N0 status, but otherwise mediastinal staging was not required.³ RTOG 0915 had similar eligibility criteria for patients with peripheral tumors, staged by CT and PET with no requirement for mediastinal staging.¹⁸

Mullins et al. looked at the effect of nodal staging in a single-institution retrospective study.²⁰ In this study of 158 patients, 149 received PET/CT staging (94%), all underwent tumor directed biopsy. 79 patients had nodal staging and 79 had no nodal staging. There was no statistically significant association between receipt of nodal staging and time-to-event recurrence or survival.²⁰ These findings agree with the current study which also did not show a difference in OS in a propensity score matched dataset.

Histology has been reported as a significant prognostic factor in previous NCDB studies of this population. Abel et al. reported that squamous cell carcinoma histology was an independent predictor of worse survival on propensity score matched multivariable comparison with median survival of 44 months (adenocarcinoma) versus 33 months (squamous cell) (p<0.0001).¹⁹ In the PSM data from the current study, patients with adenocarcinoma had an HR of 0.94 for death,, however, this did not reach statistical significance.

By employing the propensity score matching statistical technique we were able to have two well balanced groups in terms of patient, tumor, and treatment characteristics, however, there was still a difference in time from diagnosis to radiation. Time from diagnosis to radiation was longer in those receiving nodal staging with median time being 2 months for the NNS group versus 3 months for the NS group (p = <0.001). This difference is most likely due to the extra time necessary to schedule, perform, and await final pathology results for nodal staging before treatment initiation. On multivariable analysis, the interval from diagnosis to receipt of radiation was not a statistically significant factor on OS. Notably, patients with interval from diagnosis to receipt of SBRT greater than or equal to 180 days were excluded from this analysis.

Biologically effective dose (BED) has been shown to be a significant factor in determining local control in this population. In the PSM analysis, treatment with a BED of less than 100 Gy had a HR of 1.11 (95% CI 0.8-1.53, p = 0.53) when compared to the cohort that was treated with a BED of 100-140 Gy and treatment with a BED beyond 140 Gy had a HR of 0.92 (95% CI 0.78-1.078, p = 0.29), neither reaching statistical significance. Although this did not reach statistical significance, this is consistent with previous studies reporting greater local control with BED greater than 130 or 140 Gy.^16,17 While a 5-fraction regimen with total dose <50 Gy (BED <100 Gy) was associated with improved survival in the overall cohort, this did not hold true in the PSM analysis.

Certain limitations of this study include lack of information regarding the location of tumors which may impact decision-making regarding total dose, fractionation, staging, and outcome. There is also no coding within the NCDB to elucidate how the dose was prescribed, for example to the isocenter, to an isodose line, or to a volume. Additionally, there is no information within this dataset regarding the planning technique. An inherent limitation of retrospective studies of large databases is the potential for coding errors. Although we can reduce some bias by using a PSM dataset – we can only control for the factors included in the PSM, therefore not all confounding factors can be accounted for due to some non-available patient characteristics in the NCDB.