Identification of Metastatic Nodal Disease in a Phase I Dose Escalation Trial ofIntraoperative Sentinel Lymph Node Mapping in NSCLC using Near-Infrared Imaging

Abstract

Objectives

Early stage non-small cell lung cancer (NSCLC) has a high recurrence rate and poor 5-yearsurvival, particularly if lymph nodes are involved. Our objective was to perform a dose escalationstudy to assess safety and feasibility of intraoperative near-infrared (NIR) fluorescence imaging toidentify the first tumor draining lymph nodes, i.e. sentinel lymph nodes (SLN) in NSCLCpatients.

Methods

A dose escalation Phase I clinical trial assessing real-time NIR imaging followingperitumoral injection of 3.8 – 2500 μg indocyanine green (ICG) was initiated inpatients with suspected stage I/II NSCLC. Visualization of lymphatic migration, SLN identification,and adverse events were recorded.

Results

Thirty eight patients underwent ICG injection and NIR imaging via thoracotomy(n=18) or thoracoscopic imaging (n=20). SLN identification increased with ICG dosewith <25% SLN detected in dose cohorts ≤600ug vs 89% success at≥1000 μg. Twenty six NIR⁺ SLNs were identified in fifteenpatients, with seven NIR⁺ SLNs (six patients) harboring metastatic disease onhistologic analysis. Metastatic nodal disease was never identified in patients with a histologicallynegative NIR⁺ SLN. No adverse reactions were noted.

Conclusion

NIR-guided SLN identification with ICG was safe and feasible in this initial doseescalation trial. ICG doses ≥ 1000ug yielded nearly 90% intrathoracic SLNvisualization, with presence or absence of metastatic disease in the SLN directly correlating withfinal nodal status of the lymphadenectomy specimen. Further studies are needed to optimize imagingparameters and confirm sensitivity and specificity of SLN mapping in NSCLC using this promisingimaging technique.

Introduction

Regional lymph node status is the most important prognostic indicator in early stage nonsmall cell lung cancer (NSCLC). While non-invasive pre-operative staging using CT and PET scans isuseful in determining metastatic disease in nodes greater than 1 cm, these modalities frequentlymiss regional lymphatic metastasis.¹ Surgicalresection and lymphadenectomy remain the standard, but routine complete lymphadenectomy is reportedin only 50% of patients.^2–4 Furthermore, the current practice of lymph node sampling can missmetastatic disease secondary to the propensity of NSCLC to “skip” to N2 mediastinalnodes and the lack of in-depth histologic analysis on multiple sampled nodes. Unfortunately,patients with missed metastatic disease are deemed stage I and do not receive adjuvant therapy,which likely contributes to the nearly 40% incidence of recurrent disease and50–60% five year survival rate reported for patients initially identified with earlystage disease^5–8

Compared with other early stage solid tumors, high recurrence rates and poor overallsurvival in early stage NSCLC suggests that patients are understaged and/or undertreated. Pooroutcomes may be attributed to inadequate nodal sampling, skip metastases, and/or inadequatehistologic processing for identification of occult metastases. ^5–7 Sentinel lymph node (SLN) mappingoffers the benefit of identifying nodes at greatest risk for metastatic spread, focusing in-depthimmunohistochemical analysis on “high-risk” nodes, and decreasing morbidity bylocalizing nodal dissection, and has become the standard of care in breast cancer andmelanoma.^9,10 However, SLN mapping has been unreliable in NSCLC due to difficult anatomic accessto hilar structures and poor signal-to-noise ratio, making tracer visualization in anthracotic nodesor identification of radioactive nodes suboptimal.^11–16 Near-infrared imaging (NIR) hassuccessfully been used for real-time intraoperative SLN identification in cancers of the skin,breast, and GI tract ^17–20 and is ideal for in vivo intraoperative imaging dueto low absorption, scatter, and tissue autofluorescence within the NIR spectrum (700 – 1000nm).²¹ As a result, NIR fluorescent lymphatictracers permits surgical dissection without distortion of critical anatomy.²² In addition, safety is increased since laser and radioactivity arenot required. Indocyanine green (ICG) is a common FDA-approved intravascular dye which alsofunctions as an NIR lymphatic fluorophore. Therefore, we hypothesized that NIR lymphatic imagingwith ICG may permit SLN identification in early stage NSCLC patients. The current manuscript reportsresults from the first Phase I dose-escalation clinical trial using ICG to assess the safety andfeasibility of NIR-guided SLN identification in early stage NSCLC patients via either an openthoracotomy imaging platform or thoracoscopic NIR system.

Methods

ELIGIBILITY

After informed consent, a total of forty six patients scheduled to undergo lung resectionfor known or suspected lung cancer were enrolled between February 2009 and February 2013, in thisdose-escalation SLN trial approved by the Dana Farber Cancer Institute IRB. Exclusion criteriaincluded age <18 years of age, pregnancy, lactation, allergy to iodine, shellfish or indocyaninegreen (ICG). Eight patients were subsequently ineligible due to subsequent data ruling out stageI/II NSCLC, withdrawal of consent prior to surgery, or equipment repair (n=1). Therefore, 38patients received intra-operative peritumoral injection of ICG and underwent NIR-mediated SLNmapping. All study patients underwent pre-operative chest CT and PET/CT scanning, with cervicalmediastinoscopy or EBUS generally performed in patients with tumors > 2cm, mediastinallymphadenopathy (>1cm lymph nodes), PET avid nodes or more central tumors. In the current study,where all patients were required to have a negative PET/CT scan or prior negative c-med forenrollment, 13 of the 38 study patients (34 %) also underwent additional surgical stagingvia mediastinoscopy or EBUS prior to this study.

ICG Preparation

ICG was purchased from Akorn, Inc (IC-GREEN^™, Decatur, IL) or directlyfrom Novadaq Technologies (Bonita Springs, Florida), re-suspended in 10mL sterile saline to yield astock solution of 3.2 mM, and appropriately diluted in 50 cc of fresh frozen plasma (FFP) or25% human serum albumin (HSA; Baxter, Deerfield, IL) to yield ICG:FFP or ICG:HSAconcentrations of 10 μM to 3.2 mM. FFP was used in the initial 18 patients as a couplingagent for ICG until NIR mapping in other organ systems demonstrated technical equivalence in imagequality with HSA, which was used thereafter to avoid potential infectious risks with FFP exposure.¹⁷

NIR Fluorescence Imaging Systems

Real-time NIR fluorescence images were obtained using either theFLARE^™ open surgery imaging system²¹ or a 10 mm thoracoscopic NIR camera supplied by Novadaq Technologies. Each systemcaptures both white light and NIR fluorescence images and simultaneously displays these images inreal time. Video output contains three images of the surgical field: white light image (i.e. colorvideo as seen by the human eye), pure NIR image (black and white image) and a merged NIRpseudocolored (green) image superimposed on the white light video image. This approach avoidsobscuring the surgical field with visible contrast agents while providing the ability to correlateNIR images with surgical anatomy.

Intraoperative NIR Imaging Technique in Patients with Suspected Lung Cancer

The basic technique for intrathoracic NIR imaging has been previously described,although success with low dose ICG has been limited.²³ Based on pre-clinical trials in large animals and these earlier studies ²⁴, initial ICG dose was 3.8 μg and increased to 2500μg over the course of this dose-escalation trial. Briefly, following induction of anesthesiaand surgical incision(s), ICG:FFP or ICG:HSA formulations were injected using NIR-guidance into thelung parenchyma using a 25 gauge needle in four sites surrounding the suspected lung cancer. Priorto tumor excision, the lung was ventilated for five minutes to facilitate ICG migration within thelymphatic channels to the SLN. Visualization of lymphatic pathways and SLN in situwas attempted using NIR imaging via the open platform, when patients underwent surgical resectionvia a thoracotomy, or the minimally invasive NIR thoracoscope, for VATS procedures. The location ofthe visualized SLN was noted and the known or suspected lung cancer was resected in the standardfashion as determined by the surgeon. If benign, no further surgery was performed and the SLN wasleft in situ, although the location was recorded. If the tumor was malignant, the NIR-identifiedSLN(s) was resected, followed by lymphadenectomy, as determined by the surgeon, with subsequentimaging of the lymphadenectomy specimen and the surgical field to assure that all SLN wereidentified and removed. SLN identification rate, time to identification, number of SLNs identified,SLN station and pathology were recorded. Post-resection images of all specimens and the operativesite were stored for subsequent review to confirm NIR status. SLNs were fixed in formalin andembedded in paraffin for standard haematoxylin and eosin (H&E) analysis. If sufficient tissuewas available after routine clinical analysis, in-depth analysis of the NIR-identified SLNs formicrometastatic disease was performed which involved transverse sectioning of the paraffin-embeddedSLN at 1.5 mm intervals with H&E analysis of superficial, middle and deep aspects of the SLN andimmunohistochemistry (IHC) of the step sections for AE1/AE3 cytokeratin, cytokeratin 7 and thyroidtranscription factor 1 (TTF-1). Micrometastases were defined as tumor aggregates of 0.2 to 2 mm insize and isolated tumor cells or clusters if those aggregates measured less than 0.2 mm.

RESULTS

Patient demographics

Twenty five female and thirteen male patients, for a total of thirty eight patients withsuspected Stage I/II NSCLC, underwent peritumoral injection of ICG at the time of surgicalresection. The median age was 64.9 ± 10.3 years. There was a predominance of right and leftupper lobe tumors, although tumors in all five anatomic lobes were represented. Lung resectionsincluded wedge, segmentectomy, lobectomy or pneumonectomy as determined by the attending surgeon,with a majority of tumors removed via lobectomy or segmentectomy. Tumor location, extent of surgicalresection performed, and pathology of the resected lung nodule for each patient is listed in Table 1. A total of thirty-three patients were found to have NSCLC,with adenocarcinoma representing over eighty percent of all tumors resected regardless of the lobeof origin.

Table 1.

Dose-Escalation of ICG and Successful Intra-operative Sentinel Lymph Node Mapping in Patientswith NSCLC

Patient #	Location of Tumor	Tumor Size (cm)	Type of Surgical Resection	Tumor Pathology	ICG Dose (ug)	SLN Found?	SLN Node Stationa
1	RUL	1.2	Segmentectomy	Adenocarcinoma	3.8	No	--
2	RUL	1.0	Wedge Resection	Adenocarcinoma	3.8	No	--
3	LUL	1.3	Wedge Resection	Metastatic prostate cancer	3.8	Nob	--
4	LUL	1.1	Wedge Resection	Adenocarcinoma	3.8	No	--
5	RUL	7.0	Lobectomy	Adenocarcinoma	7.75	No	--
6	LUL	6.0	Lobectomy	Adenocarcinoma	7.75	Yes	4L
7	LLL	2.0	Lobectomy	Adenocarcinoma	7.75	No	--
8	RML	3.0	Lobectomy	Adenocarcinoma	7.75	No	--
9	RLL	2.2	Lobectomy	Adenocarcinoma	7.75	No	--
10	RLL	3.8	Segmentectomy	Adenocarcinoma	15.5	No	--
11	LUL	2.1	Lobectomy	Adenocarcinoma	15.5	No	--
12	L hilum	3.7	Pneumonectomy	Squamous Cell	100	No	--
13	RUL	1.3	Segmentectomy	SquamousCell	100	Yes	10.1R, 10.2R,10.3R
14	RUL	1.7	Lobectomy	Large Cell	100	Yes	11R
15	LLL	1.5	Segmentectomy	Metastatic melanoma	200	Nob	--
16	RUL	2.3	Lobectomy	Adenocarcinoma	200	Yes	10.1R
17	LUL	2.4	Wedge Resection	Adenocarcinoma	200	No	--
18	RLL	3.0	Lobectomy	Squamous Cell	400	No	--
19	RLL	1.4	Wedge Resection	Nocardiosis	400	Nob	--
20	LUL	0.9	Lobectomy	Adenocarcinoma	400	No	--
21	RML	0.3	Segmentectomy	Adenocarcinoma	600	No	--
22	LUL	0.6	Wedge Resection	Adenocarcinoma	600	No	--
23	LUL	1.6	Segmentectomy	Adenocarcinoma	600	Yes	10L, 11L
24	RLL	4.0	WedgeResection	Adenocarcinoma	800	Yes	9R
25	RUL	3.2	Lobectomy	Adenocarcinoma	800	Yes	10.1R, 11R
26	RUL	2.0	Wedge Resection	Adenocarcinoma	800	No	--
27	RLL	1.1	Wedge Resection	Adenocarcinoma	800	No	--
28	RUL	--	Wedge Resection	Necrotizing granuloma	800	Nob	--
29	RUL	1.2	WedgeResection	SarcomatoidCarcinoma	1000	Yes	4R
30	LUL	2.6	Lobectomy	Adenocarcinoma	1000	Yes	9.1L,10.1L
31	RUL	0.9	Wedge Resection	Lymphoma	1000	Technical Failurec	--
32	LUL	6.5	Lobectomy	Adenocarcinoma	1000	Yes	5, 10.1L
33	RML	1.5	Lobectomy	Adenocarcinoma	1000	No	--
34	LLL	1.0	Segmentectomy	Adenocarcinoma	1000	Yes	10.1L
35	RUL	1.1	WedgeResection	Adenocarcinoma	2500	Yes	4.1R,10.1R
36	RML	2.4	Lobectomy	Adenocarcinoma	2500	Yes	4R, 10.1R,10.2R
37	LUL	2.0	Lobectomy	SquamousCell	2500	Yes	10.1L,10.2L
38	RLL	2.9	Lobectomy	Adenocarcinoma	2500	Yes	11.1R,11.2R

^aNodes retrieved from the same nodal station (x) are labelled in sequential order as x.1, x.2 etcto maintain individual identity

^bNo nodes were removed given finding of Non-lung cancer pathology

^cFollowing injection, there was a camera malfunction with loss of video display prior toimaging

Intraoperative NIR imaging with ICG Dose Escalation

NIR imaging using ICG was performed in a total of 38 patients over an ICG dose range of3.8μg to 2500 μg. Eighteen patients underwent intraoperative NIR imaging using anopen imaging system at the time of thoractomy and twenty patients underwent videoscopic explorationand intraoperative NIR imaging using a minimally invasive thoracoscope. For doses less than 600μg, unreliable lymphatic migration and SLN identification were demonstrated as only 4 of 20patients (21%) demonstrated positive lymphatic tracking and SLN identification. However, ineach successive cohort of 600 μg, 800 μg, 1000 μg, and 2500 μglymphatic migration and SLN identification improved with the rate of SLN identification increasingfrom 33%, to 40%, 80%, and 100% respectively.(Figure 1)

Figure 1. SLN Identification.

A dose dependent increase in SLN identification is demonstrated as a function of ICG dose from3.8 μg to 2500 μg. The SLN identification rate was 100% in the highestcohort.

NIR-guided SLN Identification

Of the 38 patients enrolled in the study, five were found to not have NSCLC and thus noSLN dissection or subsequent lymphadenectomy was performed. In the thirty three patients with NSCLC,and therefore available for lymph node analysis, a total of 26 sentinel lymph nodes were identifiedintra-operatively in 15 patients (Table 1). Figure 2 is an intraoperative image of lymphatic migration visualizedin situ with identification of the SLN in green. Time from ICG injection to SLNidentification and subsequent removal using NIR imaging ranged from 3 to 125 minutes (when resectionof the primary tumor was required to establish NSCLC diagnosis before SLN removal could beperformed), with a mean time of 30 minutes. All resected SLN were imaged ex-vivo to confirm NIRpositivity. Figure 3 is an ex vivo image of aSLN in normal white light, pure NIR and a merged pseudocolored (white light-NIR) image highlightingthe appearance and the localization of the NIR signal within the SLN.

Figure 2. Intraoperative Real-time Identification of SLN using NIR Fluorescent Imaging.

A. The lymphatic migration of ICG is noted in green and progresses from the site of injectiontowards the hilum (arrow). B The SLN is identified using NIR fluorescent imaging and is shown ingreen (arrow).

Figure 3. Ex-vivo SLN Imaging using NIR Fluorescent Imaging.

Following intraoperative identification and excision of the SLN, the SLN is imaged exvivo using NIR imaging to confirm fluorescence. A. White light image of the SLN. B. PureNIR image demonstrating NIR fluorescence in white (arrow head). C. Merged pseudocolored image of theSLN with NIR fluorescence within the SLN shown in green.

Following NIR imaging, patients with the diagnosis of NSCLC underwent a radicallymphadenectomy or lymph node sampling at the surgeons discretion and all specimens were imaged forany additional NIR⁺ SLN, of which none were identified in this study. The averagenumber of lymph node stations sampled per patient within the resectedsurgical specimen was 2.9 ± 2 (i.e. hilar, level 4 and 7) and were independent from priorstations sampled via mediastinoscopy, EBUS or SLN mapping. In addition, the averagenumber of additional nodes harvested per patient, was a minimum of 6.5± 5.3 nodes, as multiple nodes from the same station where often not counted separately onthe final pathology report.

The location of the 26 identified NIR⁺ SLN in relation to the primarytumor and injection site are illustrated in Figure 4. Nineteenof these lymph nodes were identified in the hilar (n=14) or interlobar (n=5) N1stations. Of the seven lymph nodes in the N2 stations, two lymph nodes were identified in thepulmonary ligament (station 9), one in the A-P window (station 5) and four in the superiormediastinum (station 4). SLN from the LUL seemed to be particularly diverse in their location.

Figure 4. Lymphatic Map of SLN Stations Identified via NIR-image Guidance.

A total of 26 NIR⁺ SLN were identified in 15 patients. The location of theidentified SLN in relation to the site of the primary tumor (and injection site) aredemonstrated.

Histologic Analysis of NIR⁺ SLN

Of the twenty six nodes identified with NIR guidance in 15 patients, seven SLN (sixpatients) demonstrated evidence of previously unknown metastatic disease on subsequent H+Eanalysis (Table 2). All seven histologically positiveNIR⁺ SLN were located within level 10 or 11 nodal stations. Of note, althoughmetastatic disease was identified in some patients in additional lymph nodes within the surgicallymphadenectomy specimen, it is significant that these nodes were generally at the same or moreproximal lymph node stations thus signifying lymphatic spread of metastatic disease beyond the SLN.Metastatic nodal disease was never identified in patients with a histologically negativeNIR⁺ SLN. In fact, in two of the six patients the only histologically positivenode, including those within the surgical lymphadenectomy specimen, was the single SLN identifiedvia NIR imaging. Interestingly, in-depth immunohistochemical analysis on sevenNIR⁺ SLN deemed negative by standard H&E did not reveal evidence of occultmetastatic disease.

Table 2.

SLN vs non-SLN Pathology for all Patients with NIR⁺ SLNs

Patient #	Tumor Location	Tumor Size (cm)	Tumor Pathology	SLN Node Stationa	SLN pathology	IHC b	# Node Stations Sampled c	# Nodes Harvested d	non-SLN pathology
									Positive	Negative
6	LUL	6.0	Adenocarcinoma	4L	Negative	Negative	4	9	None	5, 10Ls, 11L
13	RUL	1.3	Squamous Cell	10.1R 10.2R 10.3R	Negative	Negative	2	6	None	10R × 2, Intraparenchymal LNs
14	RUL	1.7	Large Cell	11R	Negative	Negative	2	3	None	4R × 2
16	RUL	2.3	Adenocarcinoma	10.1R	Negative	Negative	1	3	None	10R × 2
23	LUL	1.6	Adenocarcinoma	10L, 11L	Negative	Negative	2	7	None	10L × 2, 11L × 3
24	RLL	4.0	Adenocarcinoma	9R	Negative	Negative	5	12	None	7, 8, 9 × 2, 10R × 5, 11R × 2
25	RUL	3.2	Adenocarcinoma	10.1R 11R	Negative	Negative	3	6	None	4R, 10R × 3
29	RUL	1.2	Sarcomatoid Carcinoma	4R	Negative	Insufficient tissue	1	2	None	4R × 2
30	LUL	2.6	Adenocarcinoma	9.1L 10.1L	10.1L Positive		5	14	10L 11L × 3	5 × 3, 7, 9L × 3,11L
32	LUL	6.5	Adenocarcinoma	5 10.1L	10.1L Positive		4	20	2/13 Hilar nodes	4L × 3, 10L × 2, 11/13 Hilar nodes
34	LLL	1.0	Adenocarcinoma	10.1L	10.1L Positive		1	2	None	10L
35	RUL	1.1	Adenocarcinoma	4.1R 10.1R	Negative	Insufficient tissue	2	5	None	4R, 10R
36	RML	2.4	Adenocarcinoma	4R 10.1R 10.2R	10.1R & 10.2R Positive		3	6	7 10R	10R
37	LUL	2.0	Squamous Cell	10.1L 10.2L	10.1L Positive		4	8	5 × 2 and 7 10L, 11L	11L
38	RLL	2.9	Adenocarcinoma	11.1R 11.2R	11.2R Positive		6	14	None	4R × 2, 7 × 2, 8, 9, 10R × 5, 11R

^aNodes retrieved from the same nodal station are labeled in sequential order as *.1, *.2 etc to maintain individual specimen identity with multiple nodes shown as × 2 or × 3 etc.

^bImmunohistochemistry (IHC) only available if sufficient tissue available after all clinically necessary analysis performed for patient care.

^cNode stations sampled in addition to stations sampled via c-med, EBUS and SLN mapping.

^dMinimum # of nodes harvested since multiple nodes within a given station often grouped together.

Discussion

Sentinel lymph node biopsy is a well-established prognostic indicator and importantcriteria in the staging of malignant disease for many cancers. However, standard approaches withblue dye and radioisotopes have not translated into successful identification of the SLN(s) in NSCLCpatients. Little et al first attempted SLN biopsy in lung cancer in 1999 with only a 47%success rate.¹¹ In 2000, Liptay et al successfullyutilized technetium for SLN identification in 87% of patients with lung cancer however theseresults were not reproducible in a phase II multi-center CALGB trial which noted SLN identificationin only 51% of patients.^12–13 Variations of techniques have been attempted but resultshave been unreliable with poor sensitivity and/or specificity.^11–16 Suboptimal results have beenattributed to anthracotic nodes in the thorax, “shine through” effect (where highradioactive signal from the injection site produces false positive results due to increasedbackground signal in nearby tissues), and distortion of critical anatomy by blue dye.

NIR imaging has specific advantages over current SLN mapping techniques and has proveneffective in multiple cancers including breast, cervical, gastrointestinal and skincancers.^17–20 The current study confirms that NIR imaging can be performed intra-operatively, inreal time during lung surgery, for visualization of lymph nodes without distortion of criticalintrathoracic structures as seen with blue dyes or the radioactive risk associated withradioisotopes. Although only a single institutional proof of concept study, at least one SLN wasidentified in all patients within the 2500 μg ICG dose cohort. While this earlydose-escalation trial is not able to determine the role of NIR imaging in the staging of patientswith NSCLC, we have identified a dose response for SLN identification and demonstrated safety andfeasibility of minimally invasive NIR-guided SLN mapping in patients with early stage non-small celllung cancer. Furthermore, although the primary objective of this trial was to establish safety anddose-response, we have shown that SLN identification can lead to detection of nodal metastaticdisease and permit increased histologic analysis via immunohistochemistry or other modalities, withthe goal of improving NSCLC staging.

Although the ICG dose required for intrathoracic SLN visualization is higher than thatrequired for clinical SLN imaging in breast cancer and melanoma patients, this study reports successat significantly lower doses than utilized for cardiac output measurements or other lungimaging.^{17, 20,25} Surprisingly, despite sequential doubling of thesuccessful ICG dose used for lung SLN mapping in swine (i.e. 10ug), SLN were not identified inhumans at these low doses. Similar findings in translation of NIR technology to human breast cancerpatients resulted in a new dose escalation scheme starting at 100 μg, suggesting thatdespite the similar size of humans and large animal models, lymphatic mapping may fundamentallydiffer between species and organ system. Since commencement of our trial in 2009, our dose of ICGhas increased significantly from 3.8 μg to 2500 μg. The current dose of 2500ug isabove the 488 ug used in NIR imaging for both melanoma and breast cancer patients ^17,20 but far lessthan the 5–10 mg doses previously reported in Japanese studies.²⁵ Given the reported risk, albeit rare, of anaphylaxis and distortionof visual anatomy using high doses of ICG, our goals were to identify a safe, optimal dose forinjection that did not alter surgical vision, yet resulted in identification of lymphatic migrationin lung cancer patients. We have demonstrated the dose-dependent identification of SLN with successat doses of 1000ug to 2500ug. Although the sample size is small, the detection of SLN in allpatients within the 2500ug cohort suggests that although metastatic disease can reach thecontralateral lymph node basins, localization of the first draining lymph node (i.e. SLN) to thesecontralateral nodes is likely rare. Given the absence of NIR+ nodes within the surgicallymphadenectomy specimen at any dose, failure to identify the SLN in 20% of patients at adose of 1000 μg is unlikely to be the result of SLN being present elsewhere in themediastinum but certainly could be the result of inadequate ICG dose or secondary to ICG leakagefrom the intraparenchymal injection site. We are currently working on methods to improve ourtechnique to maximize SLN detection.

Although lymphatic mapping may be affected by the bulk of metastatic nodal disease, thepatients in this study were specifically chosen to clinically have no evidence of nodal disease asevident by negative PET/CT and/or cervical mediastinoscopy.²⁶ Nonetheless, tumors of various sizes and lobar locations were included in both highand low dose ICG cohorts, given that tumor size has been shown to correlate with the incidence ofmetastatic nodal disease. Similar to previously published data for SLN biopsy in breast cancerpatients,^27,28 the results of the current study did not demonstrate a relationship betweenpositive identification of SLNs and tumor size or location. Instead, ICG dose was shown to be theprimary factor determining the likelihood of SLN identification in this study. However, it wasevident that there is significant variability in the lymphatic drainage patterns from the lung tumorto SLN. Although nearly three-quarters of the 26 NIR⁺ SLN identified were locatedin the nearby N1 stations, a surprising 27% of patients had SLN that had“skipped” to an N2 station, placing them at risk for having occult Stage III diseasedespite a pre-operative assessment of clinical stage I disease. This is consistent with the20–30% incidence of skip metastases reported by others using technetium 99 anddemonstrates the importance of accurately identifying the SLN in these patients ^{29, 30}.

Only three SLN nodes were identified in the intralobar position (station 11) with no SLNidentified in the more distal sublobar and segmental positions. We hypothesize that this is theresult of high fluorescence at the injection site and surrounding lung parenchyma which can obscureour ability to identify a discrete separate NIR signal above background within the intraparenchymallymph node stations (segmental and subsegmental), causing the equivalent of a “shineeffect”. We have been able to decrease the impact of this by positioning the injection siteout of the visual field while looking at the hilum and mediastinum but this may be still beimpacting our ability to imaging nodes closer to the tumor itself. Secondly, imaging is particularlydifficult if ICG leaks from the injection site onto the nearby lung which may have distorted ourability to identify discreet nodes in these lymph node stations and suggests that individual imagingof segmental nodes dissected ex vivo from the resected specimen may have identified additional SLNnodes in these stations. Third, although NIR imaging is reported to be feasible for up to 1 cm indepth²² this does not appear to be the case inair-filled lung or within the mediastinum around the bronchial structures. Having said this, finalpathology has not identified metastatic nodes within the more distal segmental positions even inthose patients with metastatic SLN identified in the level 10 or 11 nodes, indicating that theidentified NIR+ SLN are sufficient for identification of metastatic nodal disease.

The high incidence of mediastinal (N2) SLN and the variability of SLN location betweenpatients highlights the difficulty in predicting the site of metastatic nodal disease in patientswith NSCLC. In the current trial, despite a negative pre-operative workup via PET and possiblemediastinoscopy, seven lymph nodes were positive for metastatic disease on H+E staining.Therefore, although our numbers are small, SLN analysis resulted in the upstaging of six of the 15NSCLC patients in whom SLN were identified in this study. This suggests that 40% of patientswere understaged pre-operatively, which is not significantly different than the 27.5% ofstage I NSCLC patients found to have node positive disease at the time of resection, despite anegative pre-operative workup, in CALGB 9761.³¹

Identification of the sentinel lymph node in patients with NSCLC may allow for detailedhistologic analysis via immunohistochemistry or PCR and thereby leading to more accurate staging innon small cell lung cancer. Although immunohistochemical assessment of a small number of the SLNfound to be negative on H & E in this study have not yet identified occult metastases, we expectthat additional analysis, including PCR, in a larger population will potentially identifymicrometastatic disease in a subset of these H& E negative SLN. This may result inidentification of patients that may benefit from adjuvant therapy for previously undetected occultdisease, with the goal of improving survival. Our current phase I trial has identified an ICG doseresponse for SLN identification and has demonstrated the safety and feasibility of NIR imaging forSLN identification in patients undergoing resection for NSCLC via a minimally invasive,intraoperative approach. Although we have demonstrated the detection of metastatic disease withinNIR+ SLN, additional studies are needed to improve SLN imaging and dose delivery, determinesensitivity and specificity of this approach and to ultimately assess the impact of SLN analysis onadjuvant therapy and overall outcomes in patients with early stage NSCLC.