Introduction

One of the most elaborately studied surgical procedures that benefits from image guidance is the so-called sentinel lymph node (SLN) procedure. During this procedure, lymphatic flow from a primary tumor is visualized using lymphangiographic agents (e.g., methylene blue, fluorescein, or indocyanine green (ICG) [1,2,3]) and/or node-accumulating imaging agents, so-called SLN tracers (radiocolloids such as Technetium-99 m (99mTc)-nanocolloid) [4,5,6]. Lymphangiographic agents can be used to rapidly stain draining lymphatic templates due to their “small” size and rapid migration kinetics. On the other hand, node-accumulating agents allow careful classification of sequential drainage patterns [7, 8]. Despite these fundamental differences in tracer kinetics, both approaches are employed during SLN procedures.

Their proven ability to prevent extensive nodal dissections has made SLN procedures standard of care for clinically node-negative breast cancer and melanoma patients [9, 10]. The procedure is also being put forward as a valid treatment option in vulvar, cervical, and penile cancer [11,12,13,14] and has been extensively explored in a wide variety of cancers, among which prostate cancer (PCa). As early as 1999, Wawroschek et al. reported on the validity of radio-guided SLN procedures in PCa patients, who incorporated preoperative nuclear imaging with intraoperative gamma tracing [15]. Their pioneering work indicated that the procedure has also the potential to reduce perioperative morbidity and may increase the sensitivity of the detection of micrometastases. With the rise of robotic surgery, traditional gamma probes became difficult to implement and laparoscopic near-infrared (NIR) fluorescence imaging started to take over as leading intraoperative guidance modality [7].

The SLN procedures in PCa essentially evolved towards two approaches: (1) use of ICG-99mTc-nanocolloid (a fully integrated (hybrid) tracer entity whereby the SLNs are both radioactive and fluorescent) [7] and (2) the multi-step implementation of 99mTc-nanocolloid as SLN agent combined with free-ICG as lymphoscintigraphic agent (sequential tracer approach) [16]. In both cases, preoperative lymphoscintigraphy including single photon emission computed tomography with low-dose computed tomography (SPECT/CT) provides the operating surgeon with a roadmap, which is a valuable addition that indicates the anatomical location of the SLN and helps to plan the intervention accordingly [17]. A multitude of SLN studies in PCa, using either 99mTc-nanocolloid or ICG-99mTc-nanocolloid tracer, has indicated that the SLN procedure is a reliable diagnostic procedure for nodal staging with high sensitivity and specificity rates [18,19,20,21,22,23,24,25,26]. Logistical reasons, but also the fact that not every medical center has a highly skilled nuclear medicine department, lead to the suggestion that intraoperative fluorescence imaging can possibly replace the preoperatively radiocolloid-based method. Six groups reporting use of free-ICG (0.5–2 mg/patient) indicated that fluorescence detection proved highly sensitive but relatively nonspecific for disease [27,28,29,30,31,32,33]. While using ICG-99mTc-nanocolloid means a substantially lower amount of ICG is injected (0.25 mg/patient), thus making detection potentially more challenging, the procedure is claimed to yield an increased SLN specificity [26]. A systematic comparison between the use of ICG-99mTc-nanocolloid and free-ICG in PCa, however, has only taken place in pre-clinical studies [34].

To provide a head-to-head comparison between the use of ICG-99mTc-nanocolloid and a cocktail of separately injected 99mTc-nanocolloid and free-ICG approach, we conducted a phase II randomized trial in PCa patients. Here, we focused on determining the number of nodes harvested with fluorescence guidance and the positive predictive value for tumor-positive LNs.

Methods

Patients

Eligible patients with histopathologically proven PCa, that were clinically N0M0 on preoperative imaging (PET/CT and/or bone scan), had a risk of nodal metastases of  >5% according to the 2012 Briganti nomogram, and were scheduled for a robot-assisted radical prostatectomy (RARP) with extended pelvic lymph node dissection (ePLND), were included in this prospective randomized study. Exclusion criteria were history of iodine allergy, hyperthyroid or thyroidal adenoma, and kidney insufficiency. Patients were randomly allocated 1:1 to the hybrid tracer or sequential tracer group. ICG-99mTc-nanocolloid (250 µg ICG combined 200 MBq 99mTc-nanocolloid in 2 mL) or 99mTc-nanocolloid (200 MBq 99mTc-nanocolloid in 2 mL) was administered approximately 5 h prior to surgery in 4 deposits of 0.5 mL into the peripheral zone of the prostate using transrectal ultrasound (TRUS) guiding. Before incision and docking of the robot, patients allocated to the sequential group were separately administered 2 mL ICG (10 mg) in 2 deposits of 1 mL directed towards the peripheral zone of the prostate (Fig. 1).

Fig. 1
figure 1

Flow chart of the study

Full size image

This phase II randomized controlled trial was approved by the local Ethic Committee and registered as the M13PSN study with registration number NL46580.031.13. Written informed consent was obtained from all patients prior to the received treatment.

Preoperative SLN imaging

Following tracer injection, lymphatic mapping was performed using static planar lymphoscintigraphy of the pelvic area (15 min and 2 h post-injection) followed by SPECT/CT using a dual head gamma camera (Symbia T, Siemens, Erlangen, Germany), similar in both arms of the study. Planar images and SPECT, CT, and fused SPECT/CT including volume-rendering reconstructions were displayed using OsiriX medical imaging software (Pixmeo, Geneva, Switzerland). The images were interpreted and hotspots were assigned as SLNs or higher echelon nodes by an experienced nuclear medicine physician. An earlier study describes the protocol in detail [35].

Surgical procedure

Prostatectomy and nodal sampling were performed by three surgeons experienced in robot-assisted surgery using a Firefly fluorescence endoscope equipped da Vinci Si surgical system (Intuitive surgical, Sunnyvale, CA, USA). Fused axial SPECT/CT images were used as roadmap to indicate the anatomical localization of hotspots and were displayed in the operating room to enable easy consultation by the operating surgeon. During surgery, the nodal fluorescence was identified using the Firefly camera’s fluorescence imaging mode of the da Vinci Si surgical system (Fig. 2). Hereby, a SLN package was defined as a surgically resected specimen that can contain multiple LNs. After removal of the nodes, the radioactive counts in LNs were determined ex vivo at the operating table with a laparoscopic gamma probe (Europrobe 2; Eurorad, Eckbolsheim, France). In this way, SLNs that were not detected by fluorescence intraoperatively were still labeled as SLNs for pathological assessment based on their radioactivity. The removed nodes were correlated with location of the SLNs on SPECT/CT imaging.

Fig. 2
figure 2

Intraoperative view of nodal fluorescence using the Firefly camera’s fluorescence imaging mode of the da Vinci surgical system, showing a moderate fluorescent signal in accumulating lymph nodes with the hybrid approach (left side), compared to intense fluorescence in nodes and lymphatic vessels with the free-ICG tracer (right side)

Full size image

Following identification of fluorescent LNs and their subsequent removal, a complementary ePLND was performed, followed by a RARP, in that order. The ePLND template consisted of the common iliac artery, cranially confined by the ureteric crossing; the internal iliac vessels and the obturator fossa and external iliac vessels, caudally confined by the deep circumflex vein and femoral canal. The lateral border was the genitofemoral nerve and the medial border the peri-vesical fat.

Pathology

Excised fluorescent and radioactive LN packages were labeled as SLNs and as such fixed in formaldehyde and subsequently cut in 2-mm-thick slices perpendicular to the long axis. They were embedded in paraffin prior to serial sectioning in 3 slides at 200 µm intervals. All slides were stained with hematoxylin and eosin (H&E) staining and the middle slide was additionally stained with a pancytokeratin staining. Non-fluorescent and/or radioactive LNs were cut in 3-mm slices, usually longitudinal to the long axis, and only one H&E slide per block was made. Pathological assessment was performed by an experienced uropathologist specialized in PCa.

Indications where the image guidance procedure failed were defined as nodal metastases not residing in the resected SLN packages, but elsewhere in the ePLND templates.

Complications and follow-up

Intra- and postoperative complications (within 90 days after surgery) were scored using the Clavien-Dindo score [36]. Secondary endpoints were biochemical-free survival (BFS) and metastasis-free survival (MFS). Because MFS is considered a validated surrogate endpoint for overall survival in localized PCa [37], patients were evaluated for MFS of 2–8 year follow-up data. BFS was defined as the time between randomization and a PSA level  >0.2 ng/ml after RARP. MFS was defined as the time between randomization and the appearance of a metastatic recurrence (any N1 and/or M1) as suggested by bone scintigraphy, CT, or PSMA PET/CT.

Statistics

The rate in the number of tumor-positive SLNs was used to calculate the sample size. In an analysis from an earlier SLN study using ICG-99mTc-nanocolloid, we detected a median of 3 SLNs per patient [7]; a free-ICG study from the Salzburg group identified a median of 10 SLNs [28]. The number of tumor-positive SLNs was 12% vs. 15%, respectively. To test which tracer results in the highest yield of tumor-positive LNs, with an 80% power, statistics indicated we needed to include 69 patients in both study arms. For continuous variables, normality of distribution was verified using the Kolmogorov-Smirnov test. For variables that had no normal distribution, the nonparametric Mann-Whitney U test was used. For binomial variables that had normal distribution, the chi-square and Fisher’s exact test were used. Kaplan-Meier curves using log-rank rest were generated for BFS and MFS analyses. p-values  <0.05 were considered significant. Statistical analysis was performed using SPSS version 27 (IBM).

Results

Patients

From August 2014 to March 2020, a total of 138 patients underwent randomization, 69 were assigned to the hybrid group and 69 to the sequential group. Demographic and disease characteristics were well balanced between the two groups (Table 1). Median follow-up was 42 months (IQR 26.8–54.3).

Table 1 Prostate cancer patient characteristics indicated as median values (Median) with interquartile range (IQR)
Full size table

Preoperative imaging findings

The median number of SLNs identified on SPECT/CT did not differ between the two groups (3 vs. 4, p = 0.511). Absence of drainage (i.e., bilateral non-visualization) occurred in 2 patients (2.9%) in the hybrid group and 3 patients (4.3%) in the sequential group (Table 2). Respectively 14 (20.3%) and 10 (14.5%) patients had unilateral visualization on SPECT/CT imaging.

Table 2 Preoperative imaging findings. Results are indicated as median values (Median) with interquartile range (IQR)
Full size table

Intraoperative findings

The time interval between radionuclide tracer injection and surgery was similar for both groups (5:15 h (hybrid group) vs. 5:17 h (sequential group), p = 0.941; Table 3). The median number of total removed (S)LN packages was comparable between the two groups (10 vs. 10, p = 0.139; Table 3).

Table 3 Intraoperative findings. Results are indicated as median values (Median) with interquartile range (IQR)
Full size table

Some SLNs on SPECT/CT imaging were surgically not pursued due to inaccessible locations, 55 (hybrid tracer) vs. 74 (sequential tracer; p = 0.472), resulting in the SPECT/CT directed resection of 177 (hybrid tracer) vs. 190 (sequential tracer; p = 0.442) SLN packages (Table 3). It is known that intraoperative fluorescence imaging during SLN procedures can be relatively insensitive, making confirmative back table fluorescence imaging and gamma probe measurements essential [35, 38]. Real-time intraoperative fluorescent nodal identification rates of SLN packages were 52% (hybrid group) and 76% (sequential group), respectively (p < 0.001). The other SLN packages were identified by back table specimen analyses. The ex vivo radioactive SLN packages on SPECT/CT imaging did not differ between the two groups (94% (hybrid group) vs. 92% (sequential group); Table 3). In both groups, the distribution of resected fluorescent LN packages—within or outside ePLND template—was similar (Table 4). Duration of surgery (2:29 h (hybrid tracer) vs. 2:20 h (sequential tracer), p = 0.654) was again similar (Table 3).

Table 4 Surgical location of fluorescent (S)LN packages
Full size table

Pathologic examination

The median number of resected LNs per patient (SLN+ePLND = 20 vs. 21; p = 0.727) and the total number of LN metastases (44 vs 33; p = 0.470) were comparable for both groups (Table 5). In the hybrid group, the amount of nodes resected by image guidance was halved (310 vs 665; p < 0.001; Table 5). Interestingly, 7.4% of the fluorescent nodes with the hybrid approach yielded metastases, whereas only 2.6% of the fluorescent nodes from the sequential group contained metastases (p = 0.002; Table 5). This indicates a threefold increase in positive predictive value for the former. The percentage fluorescent node-positive SLNs, correlated with the location of radioactive SLNs on SPECT/CT, was comparable for both groups (68% (hybrid group) vs. 61% (sequential group; p = 0.22). The same was true for indications where image guidance failed (1.4% vs. 1.4%, respectively).

Table 5 Pathological findings. Characteristics are indicated as median values (Median) with interquartile range (IQR)
Full size table

Complications and follow-up

Overall, complication rate within 90 days after surgery was 30.4%, of which 18.1% were Clavien-Dindo grade III/IV complications. No significant differences were found between both groups for incidence (29.0% vs. 31.9%, p = 0.78; Table 6) or Clavien-Dindo severity score of complications. Focusing on complications related to SLN and ePLND (lymphedema, lymphocele, thrombo-embolic events, iatrogenic ureter lesion), these complication rates were again not significantly different (8.7% vs. 11.6%, p = 0.57).

Table 6 Complications within 90 days after surgery, using the Clavien-Dindo score
Full size table

With a median follow-up of 42 months (IQR 26.8–54.3), the median BFS was 31.0 months in the hybrid group and 32.0 months in the sequential group (p = 0.73). Subsequently, the median MFS was 38.0 and 39.0 months in the hybrid- and sequential group, respectively (p = 0.38).

Discussion

By performing a randomized trial, we have been able to unravel the clinical impact of relying on the intraoperative fluorescence guidance provided by the hybrid tracer ICG-99mTc-nanocolloid (a SLN specific agent) versus the fluorescence guidance provided by free-ICG (a lymphangiographic agent). Since the procedures were performed combined with an ePLND, there were no differences in the total number of resected (S)LNs or histologically positive (S)LNs. Looking at the LNs resected under fluorescence guidance, the hybrid tracer decreased the fluorescence-based harvesting twofold, while increasing the chance of identifying tumor-bearing LNs using fluorescence threefold.

As the radiocolloid concentration and amount of radioactive activity were similar for both groups, no difference was seen on SPECT/CT. Both the total number of SLNs and the SLNs with aberrant drainage patterns were similar. This corresponds to the findings of Brouwer et al., who concluded that the lymphatic drainage pattern of ICG-99mTc-nanocolloid is identical to that of 99mTc-nanocolloid on SPECT/CT [39]. Several studies comparing free-ICG with 99mTc-nanocolloid, e.g., in early stage cervical and vulvar cancer [40,41,42], underscore discrepancies in behavior between the two agents. In our study, use of free-ICG meant a 40 times higher quantity of fluorescent dye was injected as compared to that used in ICG-99mTc-nanocolloid (10 mg vs 250 μg), whereby free-ICG can be considered lymphangiographic agent and ICG-99mTc-nanocolloid a SLN specific agent. We observed that the sequential group yielded more than twofold increase in the number of fluorescent packages resected. The general assumption is that increasing the amount of resected LNs converts to increasing the chance of harvesting more histologically positive LNs (“more is better"). Our current finding shed doubt on the validity of this approach and perhaps even indicates “less is more.” While the overall tumor recovery rates were similar, intraoperative ICG fluorescence in the sequential group was about 3 times less likely to indicate the presence of tumor than in the hybrid group, thus making the hybrid approach the more tumor specific of the two and with that the more accurate image guidance approach.

In fluorescence-guided surgery, signal intensity and in particular the signal to background ratio drive surgical decision-making [43]. In our study, the complex pelvic anatomy and the fact that a less sensitive laparoscopic fluorescence camera was used [44] meant that real-time in vivo fluorescence detection percentage were relatively low for both groups, 53% (hybrid group) and 76% (sequential group; p < 0.001), percentages that are substantially lower than our previous reports [35, 45], underlining how critical back table analyses are for these procedures. In addition, gamma tracing of the radioactive signal provided a valuable back-up for lesions with low fluorescence intensity [46]. The fact that real-time fluorescence imaging can already be challenging in SLN procedures raises concerns for the pursuit of receptor-target image-guided surgery procedures for, e.g., the prostate-specific membrane antigen (PSMA) [47] and other future fluorescence-guided surgery applications wherein the amount of tracer accumulated in target lesions will be relatively low [38], a setting where hybrid approaches that allow for the back-up by gamma tracing could even be more critical.

This study comes with some limitations. To accurately compare the impact of ICG-99mTc-nanocolloid and free-ICG on the surgical resection, we send all intraoperative fluorescent and/or radioactive nodes to pathology as SLNs. While this facilitates the comparison between the groups, it deviates from the concept that SLNs are only the nodes designated as SLNs in the SPECT/CT images [48]. Critics may state that the multi-step use of 99mTc-nanocolloid and free-ICG, while greatly over sampling the SLNs, yielded similar pathological findings without adding complications. This argument, however, fails to consider the use of free-ICG doubled the pathological workload; SLNs are more meticulously examined. In general, the SLN concept is poised as a diagnostic tool, with the aim to reduce complications by minimizing the nodal harvesting, while retaining diagnostic accuracy [26]. This only seems to be the case for the hybrid approach, but as we performed the SLN concept in conjunction with an ePLND, the benefit of having the ePLND as control meant we were not able to reduce nodal harvesting with either technology. As a result, the amount of healthy LNs that has been resected during this study was relatively high (97.1% (hybrid group) vs. 97.8% (sequential group)). The limited utility of the rigid laparoscopic gamma probe in the robotic setting hampering its application in small-pelvis robot surgery meant radioguidance was not regularly used in vivo. The recent availability of the robot-tailored DROP-IN gamma probe may change this for future studies [49], highlighting the importance of multidisciplinary collaboration to achieve the best outcomes for PCa patients [50]. Finally, recent literature indicates that the positive predictive value of the SLN procedure can be increased by applying intratumoral, rather than peripheral hybrid tracer administration [51].

Conclusion

This study clearly indicates that the composition wherein ICG is used greatly impacts the fluorescence guidance aspect of SLN procedures. The hybrid ICG-99mTc-nanocolloid approach halved the number of LNs harvested under fluorescence guidance, while increasing the positive predictive value of the procedure by a factor of three. As a result, the hybrid tracer ICG-99mTc-nanocolloid enables the most reliable and minimal invasive method for LN staging in PCa patients.