Introduction
Cervical neuroendocrine neoplasms (CNEN) are rare and aggressive cancers that account for only 1.4% of all cervical cancer cases (Tempfer et al.
2018). These neoplasms are subdivided into typical carcinoid, atypical carcinoid, small-cell neuroendocrine carcinoma (SCNEC), and large-cell neuroendocrine carcinoma (LCNEC) (Guadagno et al.
2016). SCNEC and LCNEC are high-grade neuroendocrine carcinomas (HGNEC) and are associated with poor outcomes, even when diagnosed at an early stage. Cervix is the most common primary site in the female genital tract of HGNEC. The 5-years survival of CHGNEC was reported at 36.8% in stage I–IIA and 8.9% in IIB–IV (Cohen et al.
2010). Its prognosis is inferior to that of squamous cell carcinoma, adenocarcinoma and adenosquamous carcinoma (Margolis et al.
2016). Patients with HGNEC are more likely to experience lymphatic and hematogenous spread, recurrence, and distant metastases due to the aggressive biological behavior of the disease (Gadducci et al.
2017). Despite the rarity of CNEN, stage (Bermudez et al.
2001; Boruta et al.
2001), tumor size (Chang et al.
1998; Yin et al.
2014), lymph node status (Sukpan et al.
2011), depth of invasion (Sukpan et al.
2011), LVSI (Sukpan et al.
2011), and margin status (Chan et al.
2003) have been identified as relevant prognostic variables. However, the prognostic characteristics of patients with CHGNEC remain controversial (Gadducci et al.
2017), and there is a lack of data regarding the biology, clinical behavior, and management of such aggressive tumors. The prognosis models developed for squamous cell carcinoma and adenocarcinoma are not applicable to CHGNEC. Furthermore, studies have primarily focused on SCNEC, and corresponding data on LCNEC are even scarcer. Therefore, the development of a new prognostic model to predict cancer-specific survival (CSS) for CHGNEC is both challenging and crucial. In this study, we aimed to construct a new prognostic model based on the Surveillance, Epidemiology, and End Results (SEER) database and validated it using clinical data from our hospital.
Methods
Data source and inclusion criteria
This study utilized data from the Surveillance, Epidemiology, and End Results (SEER) cancer registry database as the training dataset in accordance with the SEER data use agreement. The SEER*Stat software program (version 8.3.4) was used to extract data. The pathological diagnosis was based on the primary site following the International Classification of Diseases for Oncology, third edition (ICD-O-3). Our study included histology codes 8013 (large cell neuroendocrine carcinoma), 8041 (small cell neuroendocrine carcinoma), 8240 (neuroendocrine neoplasms), and 8246 (neuroendocrine carcinoma). We limited our study to patients with high-grade neuroendocrine tumors (small cell or large cell carcinoma) and included data for postoperative lymph node status and staging from 2004 to 2015. We excluded diagnostic surgeries and included only therapeutic excisions. We utilized the 7th American Joint Committee on Cancer (AJCC) staging system in this study.
Patient data and exclusion criteria
Sixteen patients were retrospectively studied as the validation dataset. The inclusion criteria were: (1) diagnosed with high-grade neuroendocrine tumors (small cell or large cell carcinoma) at our hospital; (2) received initial treatment between March 2007 and January 2017; and (3) diagnosed by two different pathologists according to the WHO classification of 2010. Exclusion criteria included: (1) incomplete survival data description; (2) incomplete description of metastatic status; or (3) presence of multiple primary tumors. The ethics committee of the Shanghai First Maternity and Infant Hospital, Tongji University School of Medicine, approved this retrospective study.
Demographic and clinical information, including age, grade, FIGO stage, and treatment strategies were extracted. Duration of follow-up and vital status, including the cause of death, were also included. The deadline for follow-up was December 31, 2020. Censored observations were recorded for patients alive at the last follow-up date. Survival time was defined as the duration from diagnosis to death, last contact, or December 31, 2020.
Predictor selection, model development, and validation
Cox proportional hazards risk regression was used to identify independent prognostic predictors. The least absolute shrinkage and selection operator (LASSO) regression analysis was used to identify potential risk factors for cancer-specific death (CSD) from the training dataset. LASSO regression analysis, through cross-validation, was used to penalize the absolute value of regression coefficients, prevent overfitting of variables from the training dataset, and only retain the most effective predictors in the model. We identified six variables with a non-zero coefficient value and corresponding lambda value and likelihood of deviance, which were then ascertained into the final model.
The prediction models were developed using Cox proportional hazards risk regression analysis and random survival forest (RSF) analysis. A nomogram was constructed and validated based on Cox regression analysis to visualize and quantify the effect of each selected variable on the estimated 3- and 5-years cancer-specific survival (CSS) probability. Internal validation was performed using a bootstrap resampling method, with replacement from the training dataset, and fitting the Cox regression model and random survival forest (RSF) model in 1000 bootstrapping replicates. Receiver operating characteristic curves (ROC) and calibration curves were depicted separately for 3- and 5-years CSS. Decision-curve analysis (DCA) was used to determine the clinical net benefit associated with established predictive models. Discrimination of predictive models was quantified with the area under the curve (AUC). The dataset from our hospital (n = 16) was used for external validation, and the performance of the model was further estimated using the AUC.
Statistical analysis
Continuous variables were described as mean ± standard deviation (SD) and median with interquartile range (IQR) values, while categorical variables were displayed with numbers and percentages per group. The Chi-squared, Fisher exact and Wilcoxon rank-sum tests were used to compare frequency distribution among categorical and numerical variables, respectively.
All statistical analyses were performed using R version 4.0.3 (
http://www.r-project.org), with
p < 0.05 considered statistically significant for all analyses.
Discussion
In the current study, we developed a prognosis prediction model using SEER database and further validated externally the model using real cases from our hospital. This study analyzed 306 patients diagnosed with CHGNEC, revealing that small cell neuroendocrine carcinoma is the most common subtype. The majority of patients were white, had insurance, and were diagnosed at stage IV. The primary treatments included chemotherapy, radiation therapy, and cancer-directed surgery. The study found that older age, advanced tumor stages, higher T stage, lymph node metastasis, and distant organ metastasis were associated with increased risk of CHGNEC-specific death, while surgery, chemotherapy, and radiation therapy were protective factors. Six variables including age at diagnosis, stage 1 status, stage 4 status, T1 status, N0 status, and surgery of the primary site were included in the final predictive model. The RSF model outperformed the Cox regression, with AUCs of 0.81 and 0.83 at 3- and 5-years, respectively. A nomogram was developed to estimate 3- and 5-years survival. The study also reported clinical data from our own institution and external validation. Overall, this study provides valuable insights into CHGNEC and highlights the importance of surgery, chemotherapy, and radiation therapy in the treatment of this disease.
Neuroendocrine tumors (NETs) are a group of rare tumors that arise from cells of the neuroendocrine system, which produces hormones and controls various physiological functions. NETs can occur in various parts of the body, including the gastrointestinal tract, lungs, pancreas, and other organs. Compared to other neuroendocrine cancers, CNEC is relatively rare. Small cell lung cancer (SCLC) is the most common subtype of neuroendocrine cancer (Meerbeeck et al.
2011), accounting for approximately 15% of all lung cancers. Gastroenteropancreatic neuroendocrine tumors (GEP-NETs) are another common subtype of neuroendocrine cancer, accounting for approximately 70% of all NETs (Cives and Strosberg
2018). These tumors arise from neuroendocrine cells in the gastrointestinal tract and pancreas. In terms of treatment, the management of neuroendocrine tumors depends on the location and extent of the tumor (Oronsky et al.
2017). Surgery is often the first-line treatment for localized tumors, followed by adjuvant therapy, such as chemotherapy or radiation therapy. For metastatic disease, systemic therapy is often used, including somatostatin analogs, targeted therapies, and immunotherapy (Mangano et al.
2016). However, the optimal treatment for CNEC is not well established, and current treatment strategies often involve a multimodal approach, including surgery, chemotherapy, and radiation therapy (Kunz et al.
2013).
Unlike squamous and adenocarcinoma subtypes, which spread primarily by local extension, cervical high-grade neuroendocrine tumors have a high rate of lymphatic and hematogenous metastasis even when disease is clinically limited to the cervix (Salvo et al.
2019). Therefore, for newly diagnosed patients, we suggest a diagnostic imaging work-up to rule out bone, liver, brain, and bone marrow metastases. The NCCN guideline for cervical cancer highly recommended a PET/CT scan for initial radiologic staging (Abu-Rustum et al.
2020).
Early prevention and screening are crucial for the effective management of high-grade neuroendocrine cervical cancer (HGNEC) due to its early hematogenous metastasis characteristic and poor prognosis. However, there is no recognized precursor for intervention prior to becoming invasive cancer. Therefore, cervical cancer prevention requires a multipronged approach involving primary, secondary, and tertiary prevention Aggarwal and (Aggarwal
2014). In terms of primary prevention, almost all HGNEC patients were infected with high-risk HPV, primarily HPV18 and HPV16. These findings are consistent with previous studies showing that most small cell neuroendocrine carcinomas (SCNC) and large cell neuroendocrine carcinomas (LCNC) are caused by HPV (Castle et al.
2018), mainly HPV18 and HPV16. HPV vaccines are effective in preventing HPV-related cancers. However, with respect to secondary prevention, cytology-based screening tests are not effective in identifying HGNEC patients. Many patients with HGNEC have normal pap smear results (Chiang et al.
2017). HPV screening strategies may be better than cytology-based screening for HGNEC, and a biopsy is recommended for patients who test positive for HPV16 and/or HPV18.
The present study has the following limitations. First, the nomogram was based on retrospective analysis, which may have caused biases due to the lack of random assignment, patient selection, and some missing values. Second, information on some potential independent prognostic variables, such as parametrial involvement, margin status, stromal invasion, and LVSI were unavailable from the SEER database, which might also increase the performance index of the model. Third, although the prediction model has been internally validated with the SEER database and externally validated using data from SFMIH, it should be further validated using data from more institutions before it is applied to the general population.
In conclusion, high-grade neuroendocrine cervical cancer is rare but vicious, more likely to suffer hematogenous metastasis and with poor prognosis. HPV test might be helpful in screening, and out nomogram is helpful in prognosis evaluation as well as personized therapy.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.