Data from small cell neuroendocrine carcinoma of cervix : FIGO 2018 staging is more accurate than FIGO 2009


 Background: To evaluate whether the addition of stage IIIC to the revised 2018 International Federation of Gynecology and Obstetrics (FIGO) staging system of cervical cancer is reasonable based on small cell neuroendocrine carcinomas of cervix (SCNEC) data.Methods: We retrospectively reviewed 64 SCNEC patients with FIGO 2009 stage IB–IIA from January 2014 to December 2018 at the author’s hospital.Results: Univariate analysis showed that FIGO 2018 was related to more prognostic factors than FIGO 2009. Multivariate analysis showed the FIGO stage and the degree of lymph node metastasis (LNM) were significantly related to the prognosis of SCNEC patients. The 5-year overall survival rate (OS) was 78.5% and 22.2% in stage I and II, respectively (FIGO 2009). In FIGO 2018, the 5-year OS was 74.1%, 60.2%, and 0% in stage I/II, IIIC1, and IIIC2, respectively. After IIIC stage was subgrouped by T factor (T1, limited to cervix and vagina; T2, parametrial involvement; T3, pelvic and abdominal cavity involvement), the 5-year OS of IIICT1,IIICT2, and IIICT3 was 83.3%, 30.0%, and 0%, respectively, (P=0.010); while by N factor (N1a and N1b, metastasis to pelvic with the rate ≤0.20 and >0.20; N1c, metastasis to the para-aortic), the 5-year OS of IIICN1a, IIICN1b, and IIICN1c was 80.0%, 26.7%, and 0%, respectively (P=0.016).Conclusion: Compared with FIGO 2009, FIGO 2018 can evaluate the prognosis of SCNEC patients more accurately after subdivision of stage IIIC by combining tumor local invasion factors or the degree and location of LNM.


Introduction
Cervical cancer is a common gynecological malignancy and the fourth most common cancer in women [1]. Approximately 95% of cervical cancers are squamous cell carcinoma (SCC) and adenocarcinoma (AC), which have relatively good outcomes [1]. Small cell neuroendocrine cancer (SCNEC) is one of the remaining rare tumors with an extremely poor prognosis. Compared with common cervical cancer pathological types, SCNEC has a higher rate of lymph node metastasis (LNM). While approximately 40% of SCNEC patients are lymph node (LN)-positive, the rate is only 10%-15% for SCC and AC patients [3][4][5].
However, it is uncertain whether LNM affects the prognosis of patients. Most reports show that the prognostic factors include LNM in univariate analysis, but only staging is related to prognosis in multivariate analysis [3][4][5].
In 2018, the International Federation of Gynecology and Obstetrics (FIGO) revised the staging system for cervical cancer [8,9]. Compared with FIGO 2009, one of the main changes of FIGO 2018 is the division of stage IB stage from two subgroups into three subgroups (≤4, >4 cm vs ≤2, 2-4, >4 cm) according to tumor diameter to screen out patients with tumors ≤2cm who can consider preserving fertility. Another major change is to add stage IIIC according to LNM to predict patient prognosis. However, data from SCC and AC show that stage IIIC is not particularly accurate in predicting prognosis, and needs to be combined with tumor local invasion factors [10,11].
Therefore, we chose to analyze the data of SCNEC to address the following two questions: at the Obstetrics and Gynecology Hospital of Fudan University, which is the largest woman's hospital in China. Among them, 64 patients were diagnosed with SCNEC by pathology based on morphological criteria and immunohistochemical staining for neuron-speci c enolase, synaptophysin, chromogranin, and CD56. The morphological criteria revealed by hematoxylin-eosin staining included the presence of small cells with hyperchromatic nuclei and scant cytoplasm, absent or inconspicuous nucleoli, and numerous mitotic gures and extensive necrosis, and all tumors had to be positive for at least one neuroendocrine marker [6]. Tumors mixed with SCC or AC components were also included. All SCNEC patients underwent type III hysterectomy and pelvic lymphadenectomy, with or without para-aortic lymphadenectomy as primary treatment. The end point of follow-up was December 2019. For FIGO 2009, stage IB is de ned as tumors limited to the cervix. According to the maximum tumor diameter, it is divided into IB1 and IB2 (≤4 and >4cm). Stage IIA is de ned as tumors involving no more than the upper third of the vagina without parametrial involvement. If there is parametrial involvement, it is classi ed as stage IIB. For FIGO 2018, stage IB is also limited to the cervix, but according to the maximum diameter of the tumor, it is divided into IB1, IB2, and IB3 (≤2, 2-4, and >4 cm). Stages IIA and IIB are the same as FIGO 2009. However, as long as there is pelvic LNM, it is classi ed as IIIC1, and paraaortic LNM is classi ed as IIIC2, regardless of the local invasion of the tumor [8].

Clinical and pathological variables
Analyzed clinical and pathological variables included patient age, symptoms, human papillomavirus (HPV type, pathological diagnosis (pre-and post-surgery), FIGO stage (2009/2018), tumor size, LNM, and total number of LNs, depth of stromal invasion, lymph vascular space invasion (LVSI), parametrial extension, lower segment involvement, surgical margin, vaginal involvement, and treatment modalities. The primary end point was any cancer-related death. All end points were calculated from the date of radical hysterectomy to death or were censored at last follow-up. The date of death was obtained from our hospital follow-up department.

Statistical analysis
Relationships between stage and clinicopathological features were analyzed by Pearson x 2 test. The Kaplan-Meier method was used to construct survival curves, and the log-rank test was used to examine the statistical difference between curves. The Cox proportional hazards model was used to estimate the independent factors prognostic for overall survival (OS). All analyses were performed using SPSS 22 software (SPSS, Chicago, IL, USA).

Compared with FIGO 2009, FIGO 2018 is related to more prognostic factors
The total number of early surgically treated cervical cancer patients was 5544, of which 64 (1.15%) were SCNEC. The mean age of SCNEC patients was 44.48 ± 10.63 years (median, 43.50 years; range, 22-76 years). The total 5-year OS was 67.9%. The main symptom was abnormal vaginal bleeding (71.9%), and other complaints included abnormal physical examination (14.1%) and abnormal vaginal discharge (10.9%). Approximately half of the patients who underwent an HPV test were HPV18-positive (56.7%); only 3.3% were HPV16-positive, and 33.3% were not classi ed. The remaining 6.7% patients were HPVnegative. These results indicated that more than 90% of SCNEC patients were HPV-positive, and most were HPV18-positive. However, before surgery, the rate of pathological diagnosis for SCNEC was only 48.3% because some SCNEC cases were mixed type.
Other clinicopathological parameters are detailed in Supplementary Table 1, and were compared between the two staging systems. Because FIGO 2009 is a clinical staging system, it is not accurate in determining vaginal involvement, parametrial in ltration, and tumor size. Therefore, it was observed that a small number of stage I patients had pathological indications of vaginal involvement, and a small number of stage IIA patients had parametrial in ltration, while some tumor sizes did not match the stage standard. However, the FIGO 2009 staging system is also related to LNM (P=0.023), parametrial in ltration (P<0.001) and lower uterine involvement (P=0.014), which may be related to prognosis [12,13] suggesting that the FIGO 2009 staging system also has a role in predicting prognosis. Compared with the old system, FIGO 2018 staging based on pathological staging is more accurate regarding pathological parameters such as vaginal involvement, parametrial involvement, tumor size, and LNM, and is related to more prognostic factors (two medium risk factors in Sedlis standard: tumor invasion depth and LVSI). Therefore, it can be theoretically inferred that the FIGO 2018 staging system is more accurate for prognosis.
FIGO stage (2009/2018) and the degree of lymph node metastasis were signi cantly related to patient prognosis There was no difference in the 5-year OS between IB1 and IB2 (78.2% vs 80%, P=0.723), as well as between IIA1 and IIA2 (20.0% vs 25.0%, P=0.463) of FIGO 2009. The same was true in stage I and stage II groups of FIGO 2018 (Table 1). We conducted univariate analysis by not subdividing substage; the results are shown in Table 2  It has been reported that the ratio of LNM is more helpful than the presence of LNM in determining prognosis [16]. Therefore, we calculated the rate of LNM; the mean was 0.201 ± 0.156. According to the receiver operating characteristic curve (ROC), we con rmed that rate of LNM was more effective than the presence LNM to predict survival (AUC 0.631 vs 0.604), and calculated the optimal threshold value of 0.20 (Fig 1A). According to this threshold value, patients were divided into different degrees of LNM: non LNM group (ratio = 0), low LNM group (ratio ≤0. 20 (Fig 1B). There was also no difference in 5-year OS between FIGO 2018 stage I and II (P=0.761), and the 5-year OS for FIGO 2018 stage I/II, IIIC1, and IIIC2 was 74.1%, 60.2%, and 0%, respectively (P=0.003; Fig 1C). However, in FIGO 2018 staging, as long as there is LNM, it will be classi ed as stage IIIC, which ignores an important prognostic factor, tumor local invasion [16]. Therefore, we further divided stage IIIC into IIICT1, IIICT2, and IIICT3 according to pathological ndings (T1, limited to the cervix and vagina, and no parametrial in ltration; T2, parametrial involvement; and T3, pelvic and abdominal cavity involvement). The 5-year OS of IIICT1, IIICT2, and IIICT3 was 83.3%, 30.0%, and 0%, respectively (P=0.010; Fig 1D). Another interesting nding was that when IIIC was divided into IIICN1a, IIICN1b, and IIICN1c according to the degree and location of LNM (N1a, limited to the pelvic cavity with a metastasis rate ≤0.20; N1b, limited to the pelvic cavity with a metastasis rate >0.20; N1c, metastasis to the para-aortic LNs), the 5-year OS was 80.0%, 26.7%, and 0%, respectively (P=0.016; Fig1E).
Taken together, the prognosis of stage II (FIGO 2009) was underestimated, and stage IIIC (FIGO 2018) was ambiguous. Therefore, the combination of the FIGO 2018 staging system and tumor local invasion factors can more individualize and accurately evaluate the prognosis of SCNEC patients. It is also a reasonable method to combine the FIGO 2018 staging system with the factors of LNM for SCNEC.

Discussion
Our study showed that SCNEC accounted for 1.15% of cervical cancer cases with a 5-year OS of approximately 67.9%, similar to another report [17]. Before 2018, most studies reported that staging (FIGO 2009) was the most important prognostic factor for surgically treated SCNEC [17]. For example, Wang et al. analyzed 116 cases of surgically SCNEC. Multivariate analysis showed that only stage (FIGO 2009) was an independent prognostic factor, while in univariate analysis, LNM was also related to prognosis [7], consistent with our initial results in the current study. Several studies have suggested several other prognostic indicators, including age, ethnic background, tumor size, and depth of invasion [4,18,19].
However, only stage (2009) and LN status are the most consistent prognostic factors in retrospective studies [15,20]. Coupled with the decision of FIGO to introduce LNM into staging in 2018, we speculated that LNM was a de nite prognostic factor. Therefore, we attempted to change the analysis strategy to prove this.
Many literatures reported that quantitative analysis of LNM can better predict patient prognosis than qualitative analysis (LNM rate vs LNM or not) [16,21,22]. First, we compared the e ciency of the LNM rate and the presence of LNM with prognosis by ROC curve. The results showed that the former was better (AUC 0.631 vs 0.604). Then we divided the patients into different degrees of LNM according to the rate of LNM. Multivariate regression analysis showed that staging (FIGO 2009/2018) and the degree of LNM were signi cant prognostic factors. This result addresses the rst question in the Introduction and con rms that the basis of stage IIIC is correct. But it also raises another question as to whether we need to optimize stage IIIC with the rate of LNM. In fact, the rate of LNM cannot fully represent the role of LNM in tumor prognosis. Therefore, we divided stage IIIC into three groups according to the LNM rate and location and found that the prognosis information was more detailed (IIICN1a, IIICN1b, and IIICN1c: 80.0%, 26.7%, and 0%, respectively; P=0.016). Thus, for SCNEC, a special type of cervical cancer, stage IIIC could be divided into three groups: IIICN1a, IIICN1b, and IIICN1c.
For our second question, our data showed that FIGO 2018 was signi cantly better than the old version. FIGO 2009 underestimated the prognosis of most stage II patients, which might bring a negative psychological burden to the patients and is not conducive to treatment. Compared with the old staging system, the new one was associated with more prognostic factors and more detailed prognosis evaluation. However, it has also been highlighted that in AC and SCC, stage IIIC only refers to the LN factor and ignores the local invasion factor, and thus its role in predicting prognosis is not so accurate [11]. For this reason, stage IIIC was further divided into IIICT1, IIICT2, and IIICT3 with 5-year OS rates of 83.3%, 30.0%, and 0%, respectively. Thus, this strategy is also applicable to SCNEC, and is expected to be extended to all pathological types of cervical cancer.
What is special about our study is that SCNEC, in which LNM easily occurs, was selected to validate the effectiveness of FIGO 2018, especially for stage IIIC. However, our study also had some limitations. First, the number of cases was small because of the rarity of this disease. Second, our data did not include patients of advanced stage, who are not suitable for surgery.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.