Linxweiler J, Sprenk J, Cascetta K et al: Robotic salvage lymph node dissection in recurrent prostate cancer: lessons learned from 68 cases and implications for future clinical management. J Urol 2021; 206: 88.

While the management of nonmetastatic prostate cancer (PCa) has markedly improved during recent years, many patients still develop biochemical recurrence (BCR) after local, curative-intended therapy.¹ For those, salvage (s) extracorporeal beam radiotherapy (EBRT) and/or palliative androgen deprivation therapy (ADT) was the standard of care for many decades. However, with the advent of innovative imaging techniques such as prostate-specific membrane antigen (PSMA)-positron emission tomography/computerized tomography (PET/CT), a new clinical presentation of PCa, often termed “oligometastatic” or “oligorecurrent” accrued. This raised the question of whether metastasis-directed therapy by means of targeted salvage EBRT or salvage lymph node (LN) dissection (sLND) might be beneficial.² While there is 1 published prospective, randomized phase II trial comparing metastasis-directed radiotherapy to surveillance in oligometastatic recurrent PCa,³ no results of prospective studies comparing sLND to the standard of care are available to date. However, several retrospective studies were performed investigating the safety and oncological effectiveness of sLND in patients with LN metastases detected by PET/CT at BCR after RP.⁴ The majority of these studies describe an open surgical approach, while there are only a few case series reporting on minimally invasive sLND.

In this study, we aimed at reporting perioperative and oncological outcomes of robotic sLNDs performed at our institution over the last 8 years.⁵ To determine why PSMA-PET/CT often underestimates the true number of LN metastases,⁶ we selected patients with a mismatch between preoperative imaging results and the number of positive LNs in pathology reports, performed PSMA immunohistochemistry on the removed LNs and systematically compared these staining results with preoperative imaging data.

We identified 68 patients with a median age of 67.5 years who underwent robotic sLND. The median time between RP and sLND was 4.6 years, the median preoperative prostate specific antigen (PSA) was 1.57 ng/ml. The median number of suspicious LNs in preoperative imaging (choline-PET/CT in 11 patients, PSMA-PET/CT in 57 patients) was 1, and suspicious LNs were mostly located around the internal and external iliac vessels. During robotic sLND, a median number of 8 LNs were removed of which a median number of 1 was histologically tumor-positive. The median operation time was 126 minutes with a blood loss of 50 ml, and a hospital stay of 4 days. No Clavien–Dindo grade IV or V complications were seen. Vascular injury (most often of the iliac vein) occurred in 7 cases, the management of which required conversion to open surgery in 1 case. PSA followup was available for 62 patients and clinical followup for 63 patients. The median time to further therapy after robotic sLND was 12.4 months (fig. 1). Overall, 45/62 patients (73%) showed any PSA decline after robotic sLND, 33/62 patients (53%) showed a PSA decline >50%, and 16/62 patients (26%) showed a PSA decline >90% (fig. 2, A). We found that 23/62 patients (37%) reached a cBCR (postoperative nadir <0.2 ng/ml), and the majority of these (16/23; 70%) did not receive any further salvage therapy for 12 months (fig. 2, B). A completely undetectable PSA was observed in 7/63 patients (11%) and was maintained for more than 1 year in 3 of them.

Figure 1. Therapy-free survival after robotic sLND.

PSA responses were more frequent and more pronounced after preoperative PSMA-PET/CT compared to preoperative choline-PET/CT (fig. 2, A and B; median PSA-change —59% vs +10% [p=0.018]; complete biochemical response [cBCR] in 41% vs 18% [p=0.19]). A preoperative PSA <1.57 ng/ml (p=0.004), a time since RP of more than 4.6 years (p=0.004), a preoperative PSMA-PET/CT (p=0.003) and a postoperative cBCR (p=0.001) predicted significantly longer therapy-free survival after robotic sLND (fig. 3, parts A through D). In contrast, the number of removed LNs (above vs below the median), Gleason score at RP (<8 vs ≥8), N-status at RP (pN1 vs pN0), the sLND dissection template and whether another kind of salvage therapy (mostly sEBRT) had been performed between RP and sLND (yes vs no) had no significant impact on therapy-free survival.

Figure 2. Oncologic results after robotic sLND. A, waterfall plot demonstrating relative postoperative PSA changes. Patients with preoperative choline-PET/CT are shown in green, and patients with preoperative PSMA-PET/CT are shown in blue. B, Swimmer plot illustrating postoperative course of patients who experienced complete biochemical response after sLND. Again, patients with preoperative choline-PET/CT are shown in green, and patients with preoperative PSMA-PET/CT in blue. Black dot indicates initiation of further therapy, and red dot means that further therapy has not yet been started.

Figure 3. Kaplan–Meier analyses to identify factors predictive of longer therapy-free survival after robotic sLND. PSA at time of sLND (A), time between RP and sLND (B) and kind of preoperative PET/CT (C) were identified as factors predicting significantly longer therapy-free survival after robotic sLND. Variables preoperative PSA (A) and time between RP and sLND (B) were dichotomized into 2 groups divided by mean.

To explore the limitations of and the reasons for false-negative preoperative PSMA-PET/CTs, we performed PSMA immunohistochemistry (IHC) on 48 removed LNs from 10 patients with a mismatch between their preoperative imaging results and the results of histopathological LN workup. According to hematoxylin and eosin (H&E) histology, 38 LNs were tumor-positive and 10 LNs were tumor-negative. PSMA IHC showed that all LN metastases expressed PSMA (fig. 4, A), while all tumor-free LNs did not (fig. 4, B).

Figure 4. H&E histology (left column) and PSMA immunohistochemistry (right column) in lymph nodes removed during robotic sLND. Representative microphotographs of H&E stained, as well as PMSA-stained, sections of lymph nodes removed during robotic sLND. A, lymph node completely infiltrated by prostate cancer cells, which was detected by preoperative PSMA-PET/CT. B, tumor-free lymph node, which showed negative signal on preoperative PSMA-PET/CT. C, micrometastasis (dark brown cell cluster marked with asterisk) in otherwise tumor-free lymph node that was not detected by preoperative PSMA-PET/CT. Scale bar=500 μm.

To compare PSMA IHC results with preoperative PSMA-PET/CT imaging, all 48 LNs were assigned to 24 distinct anatomical regions according to surgical reports. We observed a concordance of PSMA IHC and PSMA-PET/CT in 14 regions and a discordance in 10 regions. In 9 cases of discordance, PSMA-PET/CT was negative despite a PSMA-expressing LN metastasis (false-negative), while in 1 case PSMA-PET/CT was positive despite a tumor-free LN (false-positive). In the 9 cases of false-negative PSMA-PET/CT, PSMA-PET/CT tended to especially miss small metastatic foci <5 mm within a LN.

After implementation of PSMA-PET/CT imaging for biochemical recurrence, sLND has emerged as a new treatment option for PCa patients with LN only recurrences after RP in recent years. Since then, several large retrospective studies could provide evidence for the safety and efficacy of open sLND.^4,7,8 In contrast, this study represents the largest published series on robotic sLND in PCa to date. Robotic sLND proved to be a safe procedure with clinical benefit in the majority of patients regarding PSA response and thus a delay of palliative systemic therapy. Furthermore, we defined predictive factors for improved therapy survival after robotic sLND and showed that false-negative PSMA-PET/CT signals are caused by small PSMA positive tumor foci (<5 mm) within the LN rather than by a lack of PSMA-expression of tumor cells.