Diagnostic value of four neuroendocrine markers in small cell neuroendocrine carcinomas of the cervix: a meta-analysis

Small cell neuroendocrine carcinoma of the cervix (SCNECC) is a highly invasive cervical cancer. The immunohistochemical criteria is an important aspect for assistant diagnosis of SCNECC. However, which markers can be appropriate selection for diagnosing SCNECC were not determined. The aim was to systematically evaluate expression levels of four neuroendocrine markers (containing synaptophysin (Syn), neural cell adhesion molecules (CD56), neuron-specific enolase (NSE) and chromograninA (CgA)) and to find out the appropriate selection for diagnosing SCNECC. Four English and three Chinese libraries were retrieved between 1984 and 2020. 23 studies about NSE, 36 studies about Syn, 23 studies about CD56 and 36 studies about CgA (all studies containing 581 patients) were eligible for meta-analyses. The pooled positive expression percentages (95% CI; I2) were as follows: 84.84% (79.41–90.27%; 76.7%) for Syn, 84.53% (79.43–89.96%; 37.5%) for CD56, 77.94% (69.13–86.76%; 83.5%) for NSE, and 72.90% (67.40–78.86%; 59.7%) for CgA. The positive proportions (95% CI; I2) ranked top three of simultaneous expressions of two markers were 87.75% (82.03–93.87%, 33.3%) for Syn and CD56, 70.92% (50.50–87.68%, 82.7%) for Syn and NSE, 65.65% (53.33–76.98%, 73.5%) for Syn and CgA. This confirms that Syn and CD56 are reliable indicators for diagnosing SCNECC.

and clinical data of patients was analyzed in each study.The clinical data were consisted of age, tumor stage, histopathological type, IHC results, treatment and prognosis. We searched the definition of IHC positive staining in each study and found three methods of description. The first method only mentioned the positive expression without specific description. The second defined positive staining by positive staining percentage of tumor cells. When the staining rate of neuroendocrine markers was more than 5% or 10%, the expression of neuroendocrine markers was positive. The third was definition of positive staining by a four-point scale. Staining was graded as 0, 1 + (less than 5% or 10% tumor cells), 2 + (5% or 10-50% tumor cells), 3 + (more than 50% tumor cells) respectively. The positive expression of neuroendocrine markers by any of the above three methods was regarded as positive expression.
The positive expression rate of individual neuroendocrine marker in one study was defined as follows: number of positive expression SCNECC cases/number of tested SCNECC cases (percentage). In addition, there are 6 combinations of the two markers, which are Syn and CD56, Syn and NSE, Syn and CgA, CD56 and NSE, CD56 and CgA, NSE and CgA. Similarly, the positive expression rate of two markers can be calculated as followed: number of simultaneous positive expression cases/number of tested cases in one study (percentage). If the study included patients with non-SCNECC, we tried to calculate the sensitivity and specificity of neuroendocrine markers expression.
Statistical analysis. All raw data extracted from the predetermined studies were managed using MetaProp function in statistical software R 3.5.0. The detailed approach of meta-analysis was shown in Supplementary  Fig. 1. The first step was to determine if a transformation of raw rate is needed. The converted rate was calculated with optimum one of the four proportion transform methods (log, logit, arcsine, and dsrsine) if the raw rate does not satisfy the normal distribution. Then the pooled expression proportions with 95% confidence intervals (CIs) of one marker or two markers could be determined using the appropriate transformation to give the effective value. A random effect model was selected when p ≤ 0.05 and a fixed effect model was used when p > 0.05.
Additionally, we also measured the effect of heterogeneity between the included studies using I 2 = 100% × (Q − df)/Q. I 2 value of 25%, 50%, and 75% were considered as low, moderate, and high degrees of heterogeneity respectively. The results of meta-analysis were presented in forest maps. Funnel plot asymmetry was assessed by the Egger's linear regression test. The results provided publication bias results for this meta-analysis. p < 0.05 was considered significant bias.

Results
Study searches and characteristics. A total of 369 literatures about Syn, 282 literatures about CD56, 300 literatures about NSE, and 278 literatures about CgA were identified as potentially eligible for inclusion. A flow diagram of the study selection is shown in Fig. 1. Of the remaining 42 studies warranting furthering review, 41 studies were case series and 1 study was a case report. 42 studies were harmonized for inclusion criteria, comprising 29 from Asia (China, Japan, Indian, and Thailand), 4 from Europe (England, Germany, Poland), 8 from the United States of America and 1 from Canada. All included studies were retrospective and a total of 581 patients were enrolled in the stratified meta-analyses. Characteristics of studies used in the research are enumerated in the Table 1.

Discussion
Precise diagnosis is very crucial for SCNECC treatment. This disease requires specialized management recommendations depending on its unique biological behavior. Our results confirmed that the positive expression percentage of Syn was the highest among four classic neuroendocrine markers, and the positive rates of combination (Syn and CD56) were the highest among six combinations (Fig. 5). This is the first meta-analysis of the expression levels of neuroendocrine markers in SCNECC studies with the largest sample size. Therefore, the results of quantitative evaluation will help us select suitable markers for assisting diagnosing SCNECC. The cancer cells of SCNECC have the similarity of neuroendocrine characteristics 44 . This is also the basis for distinguishing other morphologically similar tumors from SCNECC. Our study showed Syn had the highest expression rate, followed by CD56 in single marker expression. The combination Syn and CD56 have the highest positive expression rate in double marker expression simultaneously. The differences in the expression levels of four neuroendocrine markers are related to their molecular characteristics. Syn is a calcium binding protein located on the membrane of synapse vesicles, which diffusely expresses in the cytoplasm of neuroendocrine cells 4 . The molecular biological characteristic may explain why it is expressed with high degree. The expression level of CD56 is second only to Syn. And the heterogeneity of CD56 expression is smaller than those of other three markers. CD56 is a glycoprotein on the surface of cell membrane and also a member of cell adhesion molecule, which plays an important role in infiltration and metastasis of tumor cells 45 . The high expression level of CD56 corresponds with the aggressive properties of SCNECC. Moreover, CD56 has its own unique advantages in terms of stable expression detected by IHC method.
The positive expression rates of NSE and CgA markers are relatively low in our study, especially the expression level of CgA is the lowest. CgA and NSE are valuable markers for diagnosing neuroendocrine cancer, and their expressions are relevant to the patient's prognosis 6,12 .Our study did not reach a similar conclusion. The potential possibilities are as follows. Firstly, SCNECC may have a decline in the expression of some neuroendocrine cell     www.nature.com/scientificreports/ characteristics for high degree of malignancy and poor differentiation. Secondly, expression rates of neuroendocrine markers may be affected by detection technology. The expressions of CgA and NSE can be detected by a serological assay, which were not included in our studies. The 5-year overall survival rates of SCNECC range from 20 to 46.6%, and the prognosis of patients with advanced stage was very poor regardless of therapy 46 . However, SCNECC patients with early stage have the potential to receive multimodality therapy and have long term survival 47 . This difference emphasises the importance of early accurate diagnosis of SCNECC. There are some controversies in diagnostic criteria of SCNECC focusing on the necessity of neuroendocrine markers in the diagnosis. Some researchers pointed out that SCNECC was a morphologic diagnosis and the IHC evidence of neuroendocrine differentiation was not a requirement for diagnosis 25 . But actually, many studies have confirmed that accurate diagnosis of SNCECC require IHC staining of neuroendocrine markers which have been performed in clinical work too 1,3,12,19,22,29,48,49 . These findings highlighted the expression of two or more markers was a necessary criteria for diagnosing SCNECC.
The evidences reveal that differential diagnosis by neuroendocrine markers is particularly important in two situations. One is to differentiate SCNECC from other tumors with small cell morphological characteristics, and the other is to determine whether cervical adenocarcinoma or squamous carcinoma coexist with SCNECC 1,3,18 .  To obtain more valuable results, we also analyzed the combined expression of two markers. We found that the combination (Syn and CD56) had the highest expression rate, which was consistent with the level of positive expression rate for single markers. Syn and CD56 are sensitive indicators for diagnosing SCNECC. However, the expression levels of both markers are highly variable.
Our study managed to collect almost all the related studies. However, the quantity, quality, and type of these studies still limited the level of evidence of this meta-analysis. All the included studies were retrospective types with small sample sizes. Heterogeneity of some studies existed in this meta-analysis. There were not adequate data and studies for the meta-analysis of prognosis. Since the data including IHC expressions of the four neuroendocrine markers in non-SCNECC were too small, it was not possible to compare the diagnostic specificity of these four markers. Thus, more studies including patients with non-small cell neuroendocrine cancer or clinical trials with a larger sample size are expected in the future.

conclusion
The positive expression percentage of Syn was the highest among four neuroendocrine markers, and the positive rates of combination (Syn and CD56) were the highest among six combinations. It is confirmed that Syn and CD56 are reliable indicators for diagnosing SCNECC. The sample size of the combination of CD56 + NSE is too small to evaluate the publication bias. One requirement of asymmetric analysis is that the case number of included studies is greater than nine. So we eliminated the analysis of studies of CD56 and NSE for whose sample size did not meet analytical criteria.