Regional Lymph Node Uptake of [¹⁸F]Fluorodeoxyglucose After Definitive Chemoradiation Therapy Predicts Local-Regional Failure of Locally Advanced Non-Small Cell Lung Cancer: Results of ACRIN 6668/RTOG 0235

Abstract

Purpose/Objective(s)

ACRIN 6668/RTOG 0235 demonstrated that standardized uptake value (SUV) on post-treatment [¹⁸F]fluorodeoxyglucose positron emission tomography (FDG-PET) correlates with survival in locally advanced non-small cell lung cancer (NSCLC). This secondary analysis determines if SUV of regional lymph nodes (RLNs) on post-treatment FDG-PET correlates with patient outcomes.

Methods and Materials

Included for analysis were patients treated with concurrent chemoradiation therapy using radiation doses ≥60 Gy, with identifiable FDG-avid RLNs (distinct from primary tumor) on pre-treatment FDG-PET, and post-treatment FDG-PET data. ACRIN Core Laboratory SUV measurements were used. Event time was calculated from the date of post-treatment FDG-PET. Local-regional failure was defined as failure within the treated RT volume and reported by the treating institution. Statistical analyses included Wilcoxon signed-rank test, Kaplan-Meier curves (log rank test), and Cox proportional hazards regression modeling.

Results

Of 234 trial-eligible patients, 139 (59%) had uptake in both primary tumor and RLNs on pre-treatment FDG-PET, and had SUV data from post-treatment FDG-PET. Maximum SUV was greater for primary tumor than for RLNs before treatment (p<0.001), but not different post-treatment (p=0.320). Post-treatment SUV of RLNs was not associated with overall survival. However, elevated post-treatment SUV of RLNs, both the absolute value and the percent residual activity compared to the pre-treatment SUV, were associated with inferior local-regional control (p<0.001).

Conclusions

High residual metabolic activity in RLNs on post-treatment FDG-PET is associated with worse local-regional control. Based on these data, future trials evaluating a radiotherapy boost should consider inclusion of both primary tumor and FDG-avid RLNs in the boost volume to maximize local-regional control.

Introduction

Outcomes with concurrent chemotherapy and radiation therapy (CRT) for Stage III non-small cell lung cancer (NSCLC) remain poor, with the highest reported prospective median survival of 28.7 months after 60 Gy of radiation with concurrent chemotherapy(1). Current clinical trial efforts to improve survival include targeted systemic therapy for patients with actionable mutations and hypofractionated radiation dose escalation to residual [¹⁸F]fluorodeoxyglucose (FDG)-avid tumor sub-volumes, such as is being tested in Radiation Therapy Oncology Group (RTOG) trials 1308 and 1106, respectively.

Another approach is to apply stereotactic body radiation therapy (SBRT) as a boost following CRT for inoperable Stage III disease. One published retrospective series delivered SBRT as a boost to the primary tumor only, leaving involved nodes outside of the boost volume(2). Surgical data from several clinical trials involving induction CRT for advanced-stage NSCLC support the notion that pathologic clearance of regional lymph nodes (RLNs) is associated with significantly improved local-regional control (LRC) and improved overall survival (OS)(3). For patients who receive definitive CRT, it is not known if the response of the RLNs (defined as mediastinal and hilar nodes) differs compared to the response of the primary lung tumor, and if lack of complete response in RLNs correlates with worse outcomes. Knowing this information would help to formulate clinical trials using SBRT to target residual disease after CRT.

The American College of Radiology Imaging Network (ACRIN) (now ECOG-ACRIN) and RTOG (now NRG Oncology) performed a cooperative, prospective Phase II clinical trial (ACRIN 6668/RTOG 0235) in which patients with locally advanced NSCLC receiving definitive CRT underwent both pre-treatment and post-treatment positron emission tomography (PET) with FDG. The primary objective of the trial was to determine the relationship between post-treatment SUV of the primary tumor and overall survival(4). The initial results of this trial were reported previously and demonstrated that residual tumor maximum standardized uptake value (SUV_max) >5 on post-treatment FDG-PET was prognostic for decreased OS(4). Here, we present a subset analysis of ACRIN 6668/RTOG 0235, with the intent to compare the metabolic response of the RLNs versus the primary lung tumor as measured by FDG-PET, and to determine the relative prognostic value of RLN SUV_max as assessed by post-treatment FDG-PET.

Methods and Materials

Patients and Trial Design

ACRIN 6668/RTOG 0235 was a phase II prospective study that accrued 250 patients with inoperable Stage IIB/III NSCLC. Details of CRT, FDG-PET technique and PET image assessment were described previously(4). Briefly, treatment consisted of definitive, concurrent, platinum-based doublet CRT with radiation to doses ≥ 60 Gy. Patients underwent pre-treatment FDG-PET (which could have been up to 6 weeks prior to enrollment), and post-treatment FDG-PET at approximately 14 weeks (range 12 to 16 weeks) after radiotherapy (and at least 4 weeks after the completion of adjuvant chemotherapy, if applicable). Each participating institution obtained institutional review board approval prior to enrolling patients, and all patients provided written, study-specific informed consent. All participating centers performed FDG-PET on ACRIN-qualified scanners, using pre-specified protocols, which have been previously described(4).

FDG-PET Image Evaluation

In ACRIN 6668/RTOG 0235, both local interpretations performed at the participating institution and central interpretations performed by an expert nuclear medicine physician at the ACRIN Core Laboratory were collected. For purposes of our analysis, the core laboratory SUV measurements specific to the primary lung tumor and RLNs were used. For each region, the maximum SUV (SUV_max) within the region of interest (ROI) was reported. Peak SUV (SUV_peak) measurements are not reported because the primary trial results showed that SUV_max and SUV_peak were highly correlated and yielded similar primary endpoint findings(4).

The reporting schema of the central review from the ACRIN Core Laboratory was such that pre-treatment FDG-PET scans where RLN involvement could not be distinguished from the primary tumor, or where RLNs were judged to be “definitely not tumor” or “probably not tumor” did not have RLN SUV recorded. Therefore, because the goal of this analysis was to evaluate RLN metabolic activity, quantified using SUV, only those patients with identifiable FDG-avid RLNs, distinct from the primary tumor, on pre-treatment FDG-PET were included for analysis.

Patient Outcomes

The outcomes tested in this analysis are overall survival (OS) and local-regional control (LRC). In ACRIN 6668/RTOG 0235, LRC was determined by the treating institution. A local-regional failure was defined as a failure within the irradiated volume. For the purposes of this analysis, outcome information was correlated with the post-treatment RLN FDG PET results.

Statistical Analysis

Distributional summaries of primary tumor and RLN SUV_max were prepared for pre-treatment, post-treatment, and percent (%) residual SUV_max. Percent residual SUV_max was defined as the post-treatment value divided by the pre-treatment value multiplied by 100. Comparisons of these SUV measures between the primary tumor and RLNs were conducted using the non-parametric Wilcoxon rank sum test.

Kaplan-Meier (KM) curves (using the log-rank test) and Cox proportional hazards regression models were used to examine the association between RLN SUV_max and patient outcomes, with separate results reported by SUV measure (post-treatment vs. % residual) and outcome (OS vs. LRC). Of note, post-treatment RLN SUV_max was the primary PET parameter of interest and % residual RLN SUV_max was the secondary parameter of interest. For both outcomes, event time was calculated from the date of the post-treatment FDG-PET scan. Both univariate and multivariate models were fit. In the multivariate models, potential confounders were adjusted for, including age, sex, baseline Zubrod performance status, baseline clinical stage, chemotherapy regimen and radiation dose (Gy). Model diagnostics related to the Cox model were assessed, and appropriate interaction terms with time were included in the model for variables that violated the proportional hazards assumption(5).

In addition to continuous post-treatment RLN SUV_max, binary thresholds of 3.5, 5 and 7 were used in order to align with results reported previously for the trial(4). For both post-treatment RLN SUV_max and % residual RLN SUV_max, an optimal binary threshold was also identified by means of recursive partitioning in a conditional inference framework(6).

As two patient outcomes (OS and LRC) and two SUV measures (post-treatment RLN SUV_max and % residual RLN SUV_max) were examined, time to event analyses were adjusted for multiple comparisons by means of the Bonferroni correction in order to control the family-wise error rate. P-values below the adjusted threshold of 0.05/4=0.0125 were considered statistically significant. Data were analyzed using SAS v9.4 (SAS Institute, Cary, NC) and R v3.1.0 (R project, http://www.r-project.org/).

Results

Patients

250 patients were accrued to ACRIN 6668/RTOG 0235. Of these, 233 eligible participants underwent pre-treatment FDG-PET, 176 (76%) of whom were identified as having distinct FDG uptake within the primary tumor and identifiable RLNs. Those patients whose nodal tumor volumes were not distinguishable from primary tumor volumes on pre-treatment FDG-PET (i.e., central primary tumors), such that only one FDG-avid volume was identifiable, were excluded from analysis (Figure 1). Among patients with identifiable FDG-avid primary tumor and RLNs, 139 (79%) also underwent post-treatment FDG-PET and had available SUV data. Reasons for not having evaluable post-treatment FDG-PET can be found within appendix materials of the initial publication, with the most common reason being patient death (4). These 139 patients comprise the current analysis set (Figure 1). Note that 6 patients (4%) were excluded from LRC analyses because of unavailable LRC outcome (n=4) or because local-regional failure occurred before post-treatment PET (n=2). Patient demographics and treatment details are presented in Table 1.

Figure 1.

Analysis flowchart of patients included in the current study.

Table 1.

Participant demographics and clinical characteristics, both for all eligible participants and for the subset of analyzed participants.

Demographic or clinical characteristic	Eligible cases (N=234)		Analysis set (N=139)
Age (years)
Median	65		65
Range	(36–85)		(36–82)
	N	%	N	%
Gender
Male	150	64.10	87	62.59
Female	84	35.90	52	37.41
Ethnicity
Hispanic or Latino	7	2.99	4	2.88
Not Hispanic or Latino	217	92.74	132	94.96
Unknown	10	4.27	3	2.16
Race 1
White	171	73.08	110	79.14
African American	27	11.54	10	7.19
Asian	31	13.25	18	12.95
Other	9	3.85	4	2.88
Clinical stage 2
IIB	9	3.85	3	2.16
IIIA	118	50.43	81	58.27
IIIB	107	45.73	55	39.57
Performance status
0 (fully active)	102	43.59	71	51.08
1 (ambulatory, capable of light work)	132	56.41	68	48.92
Chemotherapy regimen
Carboplatin/paclitaxel	95	40.60	54	38.85
Cisplatin/etoposide	35	14.96	26	18.71
Other	89	38.03	59	42.45
Data not available	15	6.41	0	0.00
Radiation dose, Gy
<50	9	3.85	0	0.00
50–60 Gy	11	4.70	5	3.60
60–70 Gy	146	62.39	104	74.82
>=70 Gy	52	22.22	28	20.14
Data not available	16	6.84	2	1.44

¹Multiple races could be endorsed by a single participant, such that the total over all options may sum to greater than 100%.

²One participant with clinical stage recorded only as stage III was grouped into the stage IIIB row.

SUV Measurements

Pre-treatment SUV_max was significantly higher in primary tumors compared to the RLNs (median of 12.5 vs. 9.8, p<0.001) (Supplementary Table 1). In contrast, there was no difference in the post-treatment SUV_max measurements (median of 2.9 vs. 2.8, p=0.320). SUV_max decreased significantly for both the primary tumor and the RLNs after definitive CRT, with a lower % residual SUV_max observed in the primary tumor (Supplementary Table 1).

Post-treatment RLN SUV and Patient Outcomes

Among this set of patients with identifiable pre-treatment FDG-avid primary tumors and RLNs, who also had available post-treatment FDG-PET, a statistically significant association was identified between post-treatment RLN SUV_max and LRC, even after adjusting for potentially important confounders (Table 2). % residual RLN SUV_max was also significant in the multivariate model using the adjusted significance threshold (Table 2).

Table 2.

Summary of the results of univariate and multivariate Cox regression models.

Patient Outcome	Post-treatment RLN SUV measure	N	Hazard ratio 1 (95% CI)	P-value
Overall Survival	SUVmax (continuous) 2	139	1.051 (0.962, 1.148)	0.273
		137	1.053 (0.964, 1.150)	0.249
	% Residual SUVmax 3 (continuous)	139	1.052 (0.981, 1.129)	0.153
		137	1.061 (0.984, 1.144)	0.121
Local-Regional Control	SUVmax (continuous) 2	133	1.145 (1.048, 1.252)	0.003 †
		131	1.214 (1.094, 1.346)	<0.001 †
	% Residual SUVmax 3 (continuous)	133	1.122 (1.016, 1.238)	0.023
		131	1.149 (1.037, 1.275)	0.008†

¹In each cell, univariate hazard ratios are presented above and multivariate hazard ratios are presented below. The multivariate model adjusted for the following covariates: age, sex, baseline performance status, baseline clinical stage, chemotherapy regimen and radiation dose.

²Hazard ratios correspond to a 1-unit increase in RLN SUV_max.

³% Residual RLN SUV_max is defined as (post/pre)*100. Hazard ratios correspond to a 10% increase in the % Residual RLN SUV_max.

^†Analyses were adjusted for multiple comparisons. P-value is below the adjusted significance threshold of 0.0125.

Recursive partitioning analysis identified the optimal threshold for post-treatment RLN SUV_max as 5.2. Patients with post-treatment RLN SUV_max above this threshold exhibited significantly worse LRC compared with patients with lower values (p=0.006) (Figure 2A). Using the thresholds identified as significant for survival that were published in the original ACRIN6668/RTOG 0235 report(4), we found that RLN SUV_max greater than 5 (p=0.019), and greater than 7 (p=0.050) were of marginal significance for poorer LRC at the adjusted significance threshold (Supplemental Figure 2A, B). Using a threshold of 3.5, we found no significant association with LRC (p=0.193), and thus these data are not shown.

Figure 2.

Local-regional control (LRC) as a function of post-treatment RLN SUV_max, using varying cut-offs. A) Post-treatment RLN SUV_max with a cut-off at the optimal threshold, and B) % Residual RLN SUV_max with a cut-off at the optimal threshold. Solid lines correspond to the respective Kaplan-Meier (KM) curves, and dashed lines correspond to 95% confidence intervals of the respective curves.

For % residual RLN SUV_max, the optimal threshold identified on recursive partitioning analysis was 25.84%. Patients with % residual RLN SUV_max above this threshold exhibited significantly worse LRC compared with patients who had a lower % residual RLN SUV_max (p=0.010, Figure 2B). Interestingly, while patients with post-treatment RLN SUV_max >5.2 had the worst local-regional control, we found that patients with RLN SUV_max ≤5.2 could be further stratified by % residual RLN SUV_max above or below 25.84% (p = 0.004, Supplementary Figure 1).

Dichotomized post-treatment RLN SUV_max remained a significant predictor of LRC after adjustment for potential confounders in a multivariate Cox model (Table 3). Patients with RLN SUV_max above the optimal threshold of 5.2 had a 2.7-fold increase in the risk of local-regional failure over the observed follow-up period compared to those with SUV_max below the threshold (p=0.005). A similar multivariate Cox model using % residual RLN SUV_max is shown in Supplementary Table 2, and demonstrates a 2.8-fold increase in the risk of local-regional failure for patients with % residual RLN SUVmax >25.84%, compared to those with % residual RLN SUVmax ≤25.84% (p = 0.004).

Table 3.

Summary for local-regional control (LRC) of the results of a multivariate Cox regression model with post-treatment RLN SUV_max dichotomized at the optimal threshold.

Parameter	Estimate (SE)	Hazard ratio (95% CI)	P-value
Age (years): continuous	0.015 (0.019)	1.015 (0.978, 1.053)	0.434
Sex: female (vs. male)	0.009 (0.301)	1.009 (0.560, 1.821)	0.975
Baseline performance status: ambulatory, capable of light work (vs. fully active)	0.841 (0.460)	- 1	0.068
Baseline performance status * Time (months)	−0.095 (0.047)	- 1	0.045
Baseline clinical stage: IIIB (vs. IIB/IIIA)	1.351 (0.596)	- 1	0.023
Baseline clinical stage * sqrt(Time) (months)	−0.582 (0.217)	- 1	0.007
Radiotherapy dose (Gy): continuous	0.096 (0.065)	- 1	0.141
Radiotherapy dose (Gy) * sqrt(Time) (months)	−0.057 (0.027)	- 1	0.034
Chemotherapy regimen: Cisplatin + Etoposide (vs. Carboplatin + Paclitaxel)	−0.442 (0.540)	0.643 (0.223, 1.852)	0.413
Chemotherapy regimen: Other (vs. Carboplatin + Paclitaxel)	−0.398 (0.594)	- 1	0.503
Chemotherapy regimen: Other * sqrt(Time) (months)	0.351 (0.198)	- 1	0.076
Post-treatment RLN SUVmax: > 5.2 (vs. ≤ 5.2)	0.994 (0.353)	2.702 (1.352, 5.402)	0.005 †

¹A single hazard ratio is not reported as the specified covariate is time-varying, thus implying that the hazard ratio varies over time.

^†Analyses were adjusted for multiple comparisons. P-value is below the adjusted significance threshold of 0.0125.

sqrt = Square root transformation

No statistically significant association was observed between post-treatment RLN SUV_max and OS. This held for RLN SUV_max as a continuous variable (Table 2), as well as across the variously defined thresholds of 3.5 (not shown), 5 and 7 (Supplemental Figure 2C, D). There was also no association observed between % residual RLN SUV_max and overall survival (Table 2). As no threshold for post-treatment SUV_max or % residual RLN SUV_max existed that provided significant separation in overall survival (and thus the optimal threshold by recursive binary partitioning is undefined), the same optimal thresholds identified for the LRC outcome (RLN SUV_max of 5.2 and % residual RLN SUV_max of 25.84%) were used for OS for comparison (Figure 3A, B).

Figure 3.

Overall survival (OS) as a function of post-treatment RLN SUV_max, using various cut-offs. A) Post-treatment RLN SUV_max with a cut-off at the same optimal threshold derived for local-regional control, and B) % Residual RLN SUV_max with a cut-off at the same optimal threshold derived for local-regional control. Solid lines correspond to the respective Kaplan-Meier (KM) curves, and dashed lines correspond to 95% confidence intervals of the respective curves.

Discussion

ACRIN 6668/RTOG 0235 was a non-randomized Phase II study in which patients with locally advanced NSCLC received pre-treatment FDG-PET, definitive concurrent CRT with RT doses of ≥60 Gy and then underwent FDG-PET between 12 and 16 weeks post therapy. The initial report of this trial confirmed the prognostic value of elevated post-therapy tumor FDG uptake in predicting overall survival. However, the initial analysis considered the primary lung tumor SUV and response, and did not evaluate pre- or post-treatment SUV of the RLNs. The current study presents prospective evidence that metabolic response—specifically of involved RLNs—is an important risk factor for local-regional failure after definitive CRT in patients with locally advanced NSCLC.

In this secondary analysis, we found that high post-treatment RLN SUV_max was associated with worse LRC but not with overall survival. Although recursive partitioning analysis revealed specific optimal cut-offs of post-treatment RLN SUV_max of 5.2 and % residual RLN SUV_max of 25.84%, we caution against the use of these exact cut-offs in clinical decision making until they are prospectively validated. Based on our results, thresholds of 5 for post-treatment RLN SUV_max and 25% for %residual RLN SUV_max would be sensible values to use in confirmatory studies. In contrast to the original report, which showed that primary tumor SUV_max on post-treatment FDG-PET is associated with overall survival, the current analysis included a smaller number of patients who had both pre- and post-treatment primary lung tumor and RLNs volumes reported, and therefore may not have had sufficient statistical power to detect a difference in overall survival.

Because we were interested in ascertaining the importance of RLN response to chemoradiation as determined by FDG-PET, the current analysis included only patients who had SUV values recorded for distinct RLN volumes on pre-treatment FDG-PET. The current analysis was thus limited in patient number by having to exclude those patients with only one FDG-avid tumor volume identifiable on pre-treatment FDG-PET (i.e., central primary tumors).

Complete pathologic tumor response and LRC after definitive CRT for locally advanced NSCLC are known to be associated with improved overall survival(7–10). LRC with current doses employed for definitive treatment of locally advanced NSCLC remains suboptimal, with 25–35% of patients experiencing local-regional failure as a part of disease recurrence(1, 11). In the current study, local-regional recurrence was reported as an aggregate, and did not specify whether failure was in the primary tumor or RLNs. Therefore, we were not able to determine the association between post-treatment RLN SUV and control of local disease, regional nodal disease, or both. Nevertheless, because post-treatment RLN SUV remained significantly associated with LRC on multivariate analysis, these data show that better metabolic response of the previously involved RLNs is an important outcome for definitive CRT. These data support those in a retrospective report by Guerra et al demonstrating a similar association between high post-RT SUV_max specifically of the RLNs and higher risk of death and recurrence(12). Supplemental Table 3 shows the primary findings of prior published studies evaluating the prognostic value of FDG-PET in patients with locally advanced NSCLC treated with definitive CRT(4, 12–24). Studies published prior to 2000 are nicely summarized in a review by deGeus-Oei et al (25).

Efforts to improve LRC and therefore overall survival include radiation dose-escalation. RTOG 0617, a phase III randomized trial, which tested the efficacy of dose-escalated 74 Gy versus standard 60 Gy with concurrent chemotherapy with or without cetuximab for patients with Stage III NSCLC, paradoxically showed lower disease-free survival, overall survival, and LRC in the experimental higher-dose arm(1). The explanation for this wholly unexpected result is being meticulously sought. It may relate to toxicity associated with large radiation volumes resulting in inadvertently high doses of radiation to normal organs including the heart, in a unique population of patients who, because of the same risk factors predisposing many of them to develop NSCLC (most notably smoking), also have significant cardiovascular and pulmonary comorbidities.

Knowing that radiation dose escalation using conventional fractions is not beneficial, current trials in development include other methods of escalating radiation dose. One approach is SBRT, a technique that delivers high doses of radiation to well-defined volumes, with millimeter precision and steep gradients, allowing rapid fall-off of dose away from the tumor. SBRT has revolutionized the use of radiotherapy for early stage NSCLC, with local control rates as high as 95% and low toxicity rates(26–28).

Given the success of SBRT in patients with early-stage lung cancers, many investigators are currently considering the use of an SBRT boost following an initial course of conventionally fractionated radiation therapy as a means to improve LRC in locally-advanced disease. Feddock et al. reported results of a prospective trial of 26 patients with Stage II or III NSCLC who received a median of 59.4 Gy to the primary tumor and involved nodes followed by an SBRT boost to the residual primary tumor volume(2). The intent of this prospective trial was to deliver biological equivalent doses of 100 Gy₁₀ and to assess toxicity. The results indicate that SBRT can be safely delivered to peripheral primary tumors following CRT, with hilar tumors requiring more careful fractionation schemes to avoid excess toxicity. Local control with short follow-up was 82.9%, though the authors caution that follow up data are pre-mature. A phase I institutional trial was recently launched at Emory University, delivering 44 Gy in 2 Gy daily fractions with concurrent platinum-based chemotherapy, to be followed by an SBRT boost to both the primary tumor and RLNs(29). This trial is in its early phases.

RTOG 1106 is an ongoing Phase II randomized trial comparing standard definitive CRT versus adapted dose-escalated hypofractionated radiotherapy using mid-course FDG-PET to adapt radiotherapy volumes(30). The investigational arm of this trial delivers at least 60 Gy to the initially involved nodal clinical target volume, and dose-escalated radiation to as high as 80.4 Gy to persistently metabolically active nodal regions on mid-therapy FDG-PET. This trial is designed to demonstrate the utility, or lack thereof, of adapting radiotherapy based on response to therapy using mid-treatment FDG-PET.

Evidence exists that pathological complete response in the RLNs is important for overall survival. Emami et al. showed improved in-field control when nodal regions with gross disease were adequately covered with radiation doses >60 Gy(31). Results of the Intergroup 0139 trial in which patients with stage IIIA lymph node-positive NSCLC were randomized to receive either induction CRT with 45 Gy prior to surgical resection, or definitive CRT to 60 Gy demonstrated that patients in the induction therapy arm with pathologic clearance of the lymph nodes at the time of surgery had significantly better survival than those with persistent nodal disease. RTOG 0229, in which patients received full-dose radiation (60.2 Gy) to areas of gross disease and chemotherapy prior to surgery demonstrated similar findings with improved survival in patients with nodal clearance(3). This trial, in which involved lymph node regions received the higher dose of radiation, had an increased rate of pathologic nodal clearance (63%) compared to Intergroup 0139 (40%), supporting the notion that higher doses to involved lymph nodes may result in better nodal clearance rates(3, 11). Based on these data and this secondary analysis of ACRIN 6668/RTOG 0235, eradicating tumor within regional lymph nodes is important for LRC and, most likely, for overall survival as well.

Conclusions

These data from a prospective Phase II multi-institutional trial with central imaging review suggest that persistent regional nodal FDG uptake on post-treatment PET is a marker of poor LRC. Clinical trials to escalate radiation dose in locally advanced NSCLC are currently being defined, and these data support the inclusion of FDG-avid regional nodes.

Supplementary Material

1

Supplementary Figure 1. Local-regional control (LRC) as a function of both post-treatment RLN SUV_max, and percent residual RLN SUV max, dichotomized at ≤ or > 5.2, and ≤ or > 25.84%, respectively. Solid lines correspond to the respective Kaplan-Meier (KM) curves, and dashed lines correspond to 95% confidence intervals of the respective curves.

2

Supplemental Figure 2. (A and B) Local-regional control (LRC) as a function of post-treatment RLN SUV_max, using varying cut-offs. A) Post-treatment RLN SUV_max with a cut-off of 5, B) Post-treatment RLN SUV_max with a cut-off of 7; and (C and D) overall survival (OS) as a function of post-treatment RLN SUV_max, using various cut-offs. C) Post-treatment RLN SUV_max with a cut-off of 5, D) Post-treatment RLN SUV_max with a cut-off of 7. Solid lines correspond to the respective Kaplan-Meier (KM) curves, and dashed lines correspond to 95% confidence intervals of the respective curves.

3

Supplementary Table 1. Distributional summary of SUV data based on timing of PET for both the primary tumor and RLNs.

4

Supplementary Table 2. Summary for local-regional control (LRC) of the results of a multivariate Cox regression model with percent (%) residual RLN SUV_max dichotomized at the optimal threshold.

5

Supplementary Table 3. Prior published studies evaluating the prognostic value of FDG-PET in patients with locally advanced NSCLC undergoing definitive CRT.

Summary.

Patients with locally advanced NSCLC continue to have poor outcomes. ACRIN 6668/RTOG 0235 is a prospective non-randomized trial in which patients with locally advanced NSCLC received pre-treatment FDG-PET, definitive chemoradiation and post-treatment FDG-PET. In this secondary analysis, we found that high SUV in the regional lymph nodes on post-treatment FDG-PET was independently associated with poor local-regional control. These data support future efforts to improve metabolic response of the RLNs as one means of improving LRC.