Interpretation
In this study we found that many pathologists examined only one H&E-stained frozen section for each SLN in order to reserve more tissue for DPE.25-29 This consideration may be more reasonable for assessing SLN of breast cancer, in which the axillary lymphadenectomy can be performed asynchronously and usually replaced by chemoradiotherapy.56 Actually, the use of FSE in SLNB for breast cancer has significantly decreased during the past years.57 In early-stage cervical cancer, however, recent viewpoints have begun to emphasize the avoidance of combination of surgery and radiotherapy, since there was a significantly increased morbidity.14 So the management will be challenging once the FSE result was found to be false. Therefore, best efforts are required in intraoperative diagnosis and the FSE protocol should be given enough attention, especially in fertility-preserving surgeries.58
Generally, FN results can be caused by technical errors in sectioning processes or judgment errors in reviewing processes. Gortzak-Uzan and colleagues49 reported a technical error on MAM with diameter of 4-mm, which was not observed in the frozen sections but hided in the remaining tissue. Only four MAM were missed among the 13 studies using L-protocol. However, such FN results seemed more common in the studies using E-protocol since there were 20 MAM omitted in total. In the study by Slama and colleagues29 one-level section was examined for each node and nine of 48 MAM were missed by FSE. The median diameter of these FN-MAM was 3.94 mm and the largest one reached 8.4 mm, which could hardly be neglected in reviewing processes. So, it is reasonable to speculated that most of these FNs were technical errors and could have been avoided by taking sections at short intervals.
Some may doubt that the high FN rates were due to strict ultrastaging in which more occult metastases might be revealed. This explanation also seems reasonable. However, in this meta-analysis, most of the studies used both serial sectioning and IHC examination for ultrastaging. We classified these techniques using a recommended criterion by previous pathological studies.38 Yet only the FSE protocol was found to be a source of heterogeneity in meta-regression, whereas the DPE protocol showed minimal impact on sensitivity. This observation was further confirmed by sensitivity analyses (supplementary). A more reasonable explanation is that, in E-protocol more lymph tissue was reserved for DPE, which inevitably carried higher opportunity to have metastasis within, regardless of the method for detection.
The clinical significance of MIM/ITC in SLN remains to be clarified.59,60 Okamoto and colleagues found that non-SLN were seldom involved if SLN harbored merely MIM/ITC.61 In the SENTICOL study, only one recurrence was observed among 16 patients having MIM/ITC in SLN.10 Besides, three included studies showed favorable oncological outcomes despite that PLND were omitted in FN-FSE cases.6,15,49 Taken together, these evidences suggested that MIM/ITC only represented the very beginning of lymphatic spread, and their impacts might be negligible provided that metastatic SLNs were removed. This inference was encouraged by the findings from a breast cancer study (IBCSG 23-01) supporting the exemption of axillary lymphadenectomy in patients presenting only MIM/ITC in SLN.5 If MIM/ITC was not considered, the pooled sensitivity for L-protocol would reach 0.97 (95%CI 0.89–0.99), which is high enough for intraoperative decision-making.
In the E-protocol subgroup there remained moderate heterogeneity in sensitivities, which may due to the remaining methodological difference. Since our aim was to determine the optimal protocol, the heterogeneity in this subgroup was less important. The sectioning intervals were 2–5 mm in L-protocol and the pooled sensitivity further increased when we restricted the criterion (supplementary p18). However, shortened sectioning intervals may increase the pressure upon pathologists and result in loss of tissue for DPE.18 Yamashita and colleagues examined 3 to 5 sections for each SLN and reported that the diagnosis usually finished within 30 minutes.30 This may be a rational workload.
The survival data of patients whose PLND were exempted for negative FSE-SLN is still insufficient. Only three observational studies6,15,49 and one randomized controlled trial (SENTICOL II=NCT01639820)7 had addressed this issue and the outcomes were generally good. However, the sample sizes of these studies are relatively small. High-quality evidence should be expected from several ongoing multicenter trials (SENTIX=NCT02494063, CSEM010=NCT02642471, SENTICOL III=NCT03386734, supplementary p28) in which patients with FSE-negative SLN are exempted from further PLND,37 or intraoperatively randomized into arms with or without PLND.36 The suggestion by this meta-analysis is to adopt L-protocol to reduce the risk of inadequate treatment and ensure the applicability of future findings.