Intoduction

Endometrial carcinoma (EC) is the most common gynecologic cancer in developed countries [1,2,3,4]. EC is often detected in an early phase, when the tumor is limited to the uterus and lymph nodes are not involved, with a consequent excellent prognosis [5,6,7].

However, when EC metastasizes to lymph nodes, the 5-years survival rate dramatically decreases from 80 to 90% (FIGO stage I-II) to 51–58% (FIGO stage IIIC) [8]. Therefore, lymph nodes staging is a crucial step in determining the management of EC, and lymphadenectomy is recommended as standard intervention, except for stage FIGO Ia, low grade EC [9,10,11]. Nonetheless, lymphadenectomy requires surgical expertise, and it is associated with increased risk of perioperative and postoperative morbidity [12]. For this reason, the possibility of using sentinel lymph node (SLN) examination in EC has received great interest [13,14,15,16,17]. In fact, SLN examination allows the assessment of lymph node staging without the implications related to a wide lymphadenectomy [12]. Furthermore, the SLN examination may be performed intraoperatively, and its status may determine the radicality of the surgery [18]. In particular, The 2018 NCCN Guidelines indicate that SLN mapping may be considered in patients with apparent uterine-confined G1–2 ECs (clinical stage I disease). Moreover, recent evidence suggests that SLN mapping might be used even in high-risk histologies (serous carcinoma, clear cell carcinoma, and carcinosarcoma), although further studies are necessary to define the feasibility of such approach [9].

The optimal approach for SLN mapping appears to be pathologic ultrastaging, based on multiple sections stained with hematoxylin/eosin and immunohistochemistry. Such approach allows an accurate identification of metastases, resulting particularly useful for identifying low-volume metastases, which constitute approximately half of SLN metastases [19,20,21]. Unfortunately, ultrastaging takes much time for formalin fixation, paraffin embedment, tissue processing and immunohistochemical stainings, providing a delayed final diagnosis [22]. On the other hand, the intraoperative assessment of frozen sections appears too little sensitive for this purpose [23].

The one-step nucleic acid amplification (OSNA) method, based on Polymerase Chain Reaction (PCR), has been used to detect SLN metastases in breast cancer intraoperatively. OSNA method provides a fast and objective result by quantifying the copies of cytokeratin 19 mRNA in SLN [24,25,26].

OSNA has also been proposed in endometrial cancer [27], but the feasibility of such approach is still undefined.

In this study, we aimed to assess the diagnostic accuracy of OSNA in detecting SLN metastases in EC, by using pathologic ultrastaging as reference standard.

Materials and Methods

Study Protocol

This study followed methods of our previous reviews [28,29,30,31].

Methods for search strategy, study selection, data extraction, risk of bias assessment and data analysis were defined a priori. Two authors (AR, GFZ) independently performed all review steps. Disagreements were resolved by consensus among authors.

The study was reported following the Preferred Reporting Item for Systematic Reviews and Meta-analyses (PRISMA) statement [32] and the Synthesizing Evidence from Diagnostic Accuracy Tests (SEDATE) guideline [33].

Search Strategy

MEDLINE, Scopus, EMBASE, OVID, Web of Sciences, ClinicalTrial.gov, Cochrane Library and Google Scholar were used as electronic databases. Several researches were conducted by following different combinations of the following text words: “endometr*”; “cancer”; “carcinoma”; “lymph node*”; “sentinel”; “metasta*”; “on site”; “OSNA”; “cytokeratin 19”; “CK 19”, from the inception of each database to March 2019. References from relevant articles were also assessed for further eligible studies.

Study Selection

All peer-reviewed studies assessing lymph nodes from EC patients by using OSNA were included in the systematic review. Data not extractable, overlapping patient data, case reports and reviews were defined a priori as exclusion criteria. No language restrictions were planned.

Data Extraction

Data extraction followed methods of previous diagnostic accuracy studies [34,35,36,37,38].

Original data from each study were not modified during extraction. OSNA was considered as the index test. A threshold of 250 copies of cytokeratin 19 mRNA by μl was used to differentiate between negative test (mRNA copies <250/ μl) and positive test (mRNA copies ≥250/ μl). Histologic examination with ultrastaging (multiple histologic sections of lymph node plus immunohistochemical analysis) was considered as the reference standard. Two-by-two contingency table was prepared for each study, reporting true positives (OSNA+, histology+), false positives (OSNA+, histology-), false negatives (OSNA-, histology+) and true negatives (OSNA-, histology-).

Assessment of Risk of Bias within Studies

The revised Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) [39] were used to assess the risk of bias within studies. Four domains were evaluated: 1) Patient selection (i.e. if the patients were consecutively selected); 2) Index test (i.e. if methods for OSNA were unbiased); 3) Reference standard (i.e. if lymph node histologic assessment was correctly performed); 4) Flow and Timing (i.e. if all specimens were assessed with both the same index and the same reference standard). Authors’ judgments were categorized as “low risk,” “unclear risk” or “high risk of bias.”

Concerns about applicability were also evaluated for the domains 1, 2 and 3 (i.e. if the study methods did not fit the objective of our review, regardless of their correctness).

Data Analysis

Sensitivity, specificity, positive likelihood ratio (LR+), negative likelihood ratio (LR-) and diagnostic odds ratio (DOR) with 95% confidence interval (CI) were calculated for each study and as pooled estimates and reported graphically on forest plots. Statistical heterogeneity among studies was quantified by using the inconsistency index (I2): heterogeneity was categorized as null (I2 = 0%), minimal (0% < I2 < 25%), low (25 ≤ I2 < 50%), moderate (50 ≤ I2 < 75%) or high (I2 ≥ 75%), as previously described [40, 41]. A random effect model was used to pool data, as recommended by the SEDATE guidelines, since an actual heterogeneity is expected in meta-analyses of diagnostic accuracy [33].

Area under the curve (AUC) was calculated on summary receiver operating characteristic (SROC) curves. The diagnostic accuracy of OSNA was categorized as follows: absent for AUC ≤ 0.5, low for 0.5 < AUC ≤ 0.75, moderate for 0.75 < AUC ≤ 0.9, high for 0.9 < AUC ≤ 0.97, very high for AUC > 0.97. LR+, LR- and DOR were interpreted as previously described [59, 60]

The data analysis was performed using Meta-DiSc version 1.4 (Clinical Biostatistics Unit, Ramon y Cajal Hospital, Madrid, Spain) and Review Manager 5.3 (Copenhagen: The Nordic Cochrane Centre, Cochrane Collaboration, 2014).

Results

Selection and Characteristics of the Studies

Four studies were included in the systematic review [22, 27, 42, 43]. Two other studies that used PCR for detecting SLN metastases were excluded, because they did not adopt OSNA [44, 45]; furthermore, one of them used CK 20 instead of CK19 [44], while the other one assessed only macrometastases [45]. The whole process of study selection for the database PubMed is presented in Fig. 1.

Fig. 1
figure 1

Flow diagram of studies identified in the systematic review (Prisma template [Preferred Reporting Item for Systematic Reviews and Meta-analyses])

Full size image

Overall, 691 lymph nodes from 237 endometrial cancer patients were assessed. Histologic diagnoses included 202 endometrioid carcinomas, 21 clear cell or serous carcinomas, 6 carcinosarcomas, 4 mixed carcinomas, 2 adenosquamous carcinomas and 2 undifferentiated carcinomas. SLNs were identified by injecting indocyanine green or blue dye intracervically. Intraoperatively, each SLN was sectioned, and part of SLN underwent OSNA assay, while the remaining part underwent histologic examination with ultrastaging.

Characteristics of the included studies are reported in Table 1.

Table 1 Characteristics of the included studies
Full size table

Risk of Bias within Studies Assessment

For the “patient selection” domain, one study was considered at low risk of bias due to the inclusion of consecutive patients, while in the other 3 studies it was unclear whether patients were consecutive [22, 27, 42].

For the “index test” domain, all studies were considered at low risk of bias, since all used at least half of SLN for OSNA analysis; moreover, all studies considered 250 copies of cytokeratin 19 by μl as criterion for positivity, which has been shown to be the most accurate threshold [27].

For the “reference standard” domain, 3 studies were considered at low risk, since they performed multiple histologic sections and immunohistochemical staining to detect lymph node metastases; the other study was considered at unclear risk, because it did not report whether immunohistochemistry was used for ultrastaging [27].

For the “flow and timing” domain, all studies were considered at low risk, since all SLN underwent both the same index and the same reference standard.

No concerns about applicability were raised. Authors’ judgements about the risk of bias are shown in Fig. 2.

Fig. 2
figure 2

a Assessment of risk of bias. Summary of risk of bias for each study; Plus sign: low risk of bias; minus sign: high risk of bias; question mark: unclear risk of bias. b Risk of bias graph about each risk of bias item presented as percentages across all included studies

Full size image

Diagnostic Accuracy Analysis

Pooled sensitivity and specificity of OSNA in detecting lymph node metastases were 0.88 (95% CI 0.74–0.96) and 0.93 (95% CI, 0.90–0.95) respectively, with a pooled LR+ and a LR- of 17.95 (95% CI 4.72–68.33) and 0.15 (95% CI 0.07–0.32) respectively. Pooled DOR was 191.23 (95% CI 46.95–778.78). The heterogeneity was null for sensitivity and LR- (I2 = 0%), minimal for DOR (I2 = 15.6%) and high for specificity (I2 = 92%) and LR+ (I2 = 87.5%) (Fig. 3).

Fig. 3
figure 3

Sensitivity, specificity, positive and negative likelihood ratios (LR+ and LR-) and diagnostic odds ratio (DOR) of OSNA assay in detecting sentinel lymph node metastases in endometrial cancer

Full size image

SROC curves analysis showed high diagnostic accuracy (AUC = 0.959) (Fig. 4).

Fig. 4
figure 4

Area under the curve (AUC) calculated on summary receiver operating characteristic (SROC) curve for the diagnostic accuracy of OSNA assay in detecting sentinel lymph node metastases in endometrial cancer

Full size image

Discussion

According to our results, OSNA appears a highly accurate method to detect metastases in SLN in EC. To our knowledge, this is the first meta-analysis assessing the diagnostic accuracy of OSNA method in EC.

The use of SNL mapping in EC has been giving more and more interest given the clear advantages for patients and surgeons [12,13,14,15,16,17,18]. In fact, pelvic and/or paraaortic lymphadenectomy is a procedure which require time and surgical technique, and may cause perioperative morbidity and lower arts lymphedema in a considerable percentage of cases [12, 22]. On the other hand, not assessing lymph nodes entails a risk of understaging and undertreatment [18, 46]. In this background, SLN assessment appears as a good compromise [9, 10].

Over time, some uncertainties regarding methods for SLN mapping in EC have being resolved, in particular regarding the type of dye to be used and the optimal site of injection. In fact, several studies have shown superiority of indocyanine green over methylene blue in identifying SLN [47], and uterine cervix seems to be a reliable injection site [48]. To date, SLN mapping for EC is still experimental, but even guidelines recognize the increasing evidence about its usefulness [9, 10].

The use of frozen section is widespread and is necessary to perform an intraoperative histologic examination of SLN. In fact, frozen sections allow avoiding the time required for the standard processing of pathologic tissues [49]. Nonetheless, histologic diagnosis performed on frozen section may be little accurate, with a sensitivity as low as 0.50 [43]. Furthemore, frozen sections examination cannot be supported by ancillary techniques such as immunohistochemistry. On the other hand, pathologic ultrastaging, which is the gold standard, cannot be performed intraoperatively [19, 20] and no universal protocol for SLN ultrastaging has been defined [50].

As showed in breast cancer, the OSNA method may overcome these issues [24,25,26, 51]. Indeed, OSNA might be as fast as frozen section and as accurate as pathologic ultrastaging, eliminating subjectivity from the diagnosis. Our results appear to confirm the reliability of OSNA, showing a high accuracy in detecting metastases (AUC = 0.96). Such accuracy appears adequate to support the use of OSNA for EC. These results are strengthened by the consistency in the methods among the included studies. In fact, in all studies, every SLN underwent both OSNA (with the same threshold of 250 CK 19 mRNA copies/μl) and pathologic ultrastaging (which is regarded as the gold standard).

Interestingly, despite the high accuracy found, a small discrepancy still existed between OSNA results and ultrastaging results. It would be interesting to assess the causes for such a discordance. It cannot be excluded that variations in sensitivity and specificity may derive from the sampling method of SLN. Indeed, in order to assess each SLN with both OSNA and pathologic ultrastaging, part of the lymph node was used for the former, and another part for the latter. Thus, it is possible that small metastases were present only in the sections that underwent pathologic ultrastaging, and not in those used for OSNA, or vice versa. Another cause for the non-perfect specificity may lie in the possibility of benign epithelial inclusions within SLN, which has been described in the literature [22].

Limitations of OSNA are the impossibility of assessing the morphologic features of the metastasis (e.g. focal or diffuse lymph node involvement, extranodal involvement) and the possible low reliability in some histotypes, such as carcinosarcoma, undifferentiated carcinoma and dedifferentiated carcinoma, in which a loss of CK 19 expression might be hypothesized. To date, these limitations should not affect the clinical value of OSNA, since the current staging system does not consider the morphologic features of metastasis, and the current guidelines do not recommend SLN mapping in high-risk histologies [9, 52].

An argument against the use of OSNA may be its cost, which is high if compared to frozen section and even to pathologic ultrastaging (about 10 times more expensive). However, it has been suggested that these additional costs may be widely recouped by reducing costs of inadequate treatments. In fact, in breast cancer the cost/benefit ratio seems to clearly favor the use of OSNA [22, 53, 54].

Moreover, OSNA is also able to discriminate lymph node metastases according to their volume. In fact, a number of CK mRNA copies ranging from 250 to 4999/μl is considered indicative for a micrometastasis, while a value of 5000/μl or more is considered indicative for a macrometastasis [22, 27, 42, 43]. The prognostic significance of micrometastases and isolated tumor cells is still undefined. In fact, some authors suggested that the prognostic value of lymph node metastases may depend on their volume [55,56,57]. This idea is well accepted in breast cancer, where the staging system include the size of lymph node metastases [58]. Similarly, the 8th edition of the AJCC Staging Manual incorporates pN0 (i+) and pN1mi sub-categories for lymph node metastases in EC [52]. In this regard, OSNA might also allow an accurate risk stratification for a more tailored management. Further studies are necessary in this field.

Conclusion

OSNA appears as a quick, objective and reliable method to intraoperatively detect SLN metastases in endometrial cancer. Its high accuracy may support its introduction in the common practice.