Lymph node ratio and hematological parameters predict relapse-free survival in patients with high grade rectal neuroendocrine neoplasms after radical resection: a multicenter prognostic study

Background The prognostic nutritional index (PNI), alkaline phosphatase (ALP), and lymph node ratio (LNR) are reportedly related to prognosis. The aim of this study was to elucidate the clinical importance of the LNR and hematological parameters in patients with high grade rectal neuroendocrine neoplasms (HG-RNENs) who were undergoing radical resection. Methods We reviewed the medical records of patients with HG-RNENs from 17 large-scale medical centers in China (January 1, 2010–April 30, 2022). A nomogram was constructed by using a proportional hazard model. Bootstrap method was used to draw calibration plots to validate the reproducibility of the model. Concordance index (C-Index), decision curve analysis (DCA), and time-dependent area under the receiver operating characteristic curve (TD-AUC) analysis were used to compare the prognostic predictive power of the new model with American Joint Committee on Cancer (AJCC) TNM staging and European Neuroendocrine Tumor Society (ENETS) TNM staging. Results A total of 85 patients with HG-RNENs were enrolled in this study. In the 45 patients with HG-RNENs who underwent radical resection, PNI ≤ 49.13 (HR: 3.997, 95% CI: 1.379–11.581, P = 0.011), ALP > 100.0 U/L (HR: 3.051, 95% CI: 1.011–9.205, P = 0.048), and LNR > 0.40 (HR: 6.639, 95% CI: 2.224–19.817, P = 0.0007) were independent predictors of relapse-free survival. The calibration plots suggested that the nomogram constructed based on the three aforementioned factors had good reproducibility. The novel nomogram revealed a C-index superior to AJCC TNM staging (0.782 vs 0.712) and ENETS TNM staging (0.782 vs 0.657). Also, the new model performed better compared to AJCC TNM staging and ENETS TNM staging in DCA and TD-AUC analyses. Conclusions LNR, ALP, and PNI were independent prognostic factors in patients with HG-RNENs after radical resection, and the combined indicator had better predictive efficacy compared with AJCC TNM staging and ENETS TNM staging.


Introduction
Neuroendocrine neoplasms (NENs) are rare neoplasms arising from peptidergic neurons and neuroendocrine cells; they can demonstrate neuroendocrine differentiation and express neuroendocrine markers [1].These neoplasms are mostly found in the gastrointestinal tract, pancreas, and lung [2], with the rectum being the third most frequent site.The incidence of rectal neuroendocrine tumors has increased and was 1.3/100,000 from 2007 to 2016 [2,3].Rectal neuroendocrine neoplasms (R-NENs) are classified into rectal neuroendocrine tumor (R-NET), rectal neuroendocrine carcinoma (R-NEC), and rectal mixed adeno-neuroendocrine carcinoma (R-MANEC) based on differentiation degree, mitotic count, and Ki-67 index.Among these, rectal neuroendocrine carcinoma and mixed adeno-neuroendocrine carcinoma are also known as high grade rectal NENs (HG-RNENs), which have worse prognosis and are associated with higher metastatic risk than R-NET [4].
Surgical resection remains the preferred treatment option for primary localized HG-RNENs [4].However, despite curative resection, the prognosis of HG-RNENs is still poor [5].The American Joint Committee on Cancer (AJCC) and the European Association for Neuroendocrine Tumor (ENETS) TNM staging system are the most widely used prognostic evaluation method for HG-RNENs after radical resection.They divide HG-RNENs into stage I-IV based on T stage, number of positive lymph nodes (PLN), and distant metastasis [6,7].However, recent studies suggest that lymph node ratio (LNR) is a better prognostic predictor than PLN [8,9].Moreover, preoperative hematological parameters like prognostic nutritional index (PNI) and alkaline phosphatase (ALP) also demonstrate excellent prognostic efficacy [10][11][12][13].Currently, simple and effective prognostic model containing preoperative hematological parameters and pathological parameters is still lacking for HG-RNENs.Therefore, it is essential to establish an accurate prognostic prediction model to individualize treatment in HG-RNENs patients.
In this study, we retrospectively collected data from 17 large medical centers in China to construct a prognostic prediction model for patients with HG-RNENs.Furthermore, we compared the model with the traditional AJCC TNM staging and ENETS TNM staging to determine their prognostic predictive value.

Patients and data collection
We conducted a retrospective study of HG-RNENs patients in 17 large-scale medical centers in China, spanning from January 1, 2010, to April 30, 2022.Demographic, clinicopathologic, treatment, and outcome data were extracted from the electronic medical records of each hospital by surveyors with expertise in NENs, using standardized data collection templates.The inclusion criteria were as follows: (1) confirmation of neuroendocrine neoplasms by pathology, (2) tumor grade of R-NEC or R-MANEC, and (3) primary rectal neuroendocrine neoplasms.The exclusion criteria were as follows: (1) comorbidity with other malignancies and (2) incomplete clinical data or follow-up information.A total of 85 out of 1459 patients met our study criteria.The procedure for extracting eligible study subjects is shown in Fig. 1.

Criteria
We collected blood routine test records within 7 days before radical resection and calculated the PNI from serum albumin and peripheral lymphocyte counts.For patients who underwent radical resection with lymphadenectomy, we obtained lymph node-related information, with PLN defined as the number of positive lymph nodes and LNR defined as the ratio of the number of positive lymph nodes to the number of dissected lymph nodes.
Tumor size was determined based on the longest diameter of the tumor recorded in the pathology report.In cases where patients only underwent biopsy, tumor size was determined based on endoscopic findings before treatment.For patients who underwent palliative surgery, tumor size was assessed based on imaging reports.Tumor stage was classified according to the AJCC and ENETS TNM staging system, and tumor grade was classified based on the WHO 2010 classification.The mitosis count was expressed as the number of mitotic cells in ten high-power fields from hematoxylin and eosin stained slides examined with microscopy, and the Ki-67 index was calculated as the percentage of cells labeled by immunohistochemistry.

Follow-up
We defined cancer-specific survival (CSS) as the interval from diagnosis to death attributed to neuroendocrine neoplasms and relapse-free survival (RFS) for patients in whom neoplasms were removed as the interval from the date of intervention to the date of recurrence.All patients were followed up through telephone, outpatient, and inpatient means.Patients after complete resection were followed every 3~6 months and annually after 5 years, while patients who underwent palliative resection or biopsy were followed every 3 months.We conducted the last follow-up in July 2022.Follow-up exams included routine blood tests, chromogranin A tests, chest CT scans, and whole abdominal and pelvic contrastenhanced CT or MRI; PET-CT was performed if recurrence or metastasis was suspected.Loss to follow-up was defined as failure to contact either the patient or their family.

Statistical analysis
We presented continuous variables as the median with interquartile range and categorical data as numbers and percentages.CSS and RFS were analyzed using the Kaplan-Meier method, and variables were compared using the log-rank test in univariable analysis and Cox proportional hazard regression in multivariable analysis.
We used the R software (version 4.0.0)for statistical analysis and the X-tile software to determine the optimal cut-off value of continuous variables.Variables with a P < 0.10 in the univariate analysis were included in the multivariate analysis, and P < 0.05 was considered statistically significant.

Establishment and validation of nomogram
The independent prognostic factors identified by multivariate analysis are integrated to draw the nomogram by using the "rms" package of the R software (version 4.0.0).The average predicted survival rate was compared with the average actual survival rate, and calibration plot was used to estimate the precise predictions for 1-, 2-, and 3-year RFS.DCA analysis was performed to evaluate the clinical utility of the nomogram based on net benefits at different threshold probabilities.The time-dependent area under the receiver operating characteristic curve (TD-AUC) analysis was calculated and plotted by using the "timeROC" package of the R software (version 4.0.0).

Establishment and validation of nomogram
Based on the three independent predictors, we established a nomogram for the RFS of HG-RNENS patients.LNR had the largest weight in the nomogram, followed by PNI and ALP, as shown in Fig. 3.The accuracy of nomograms is verified by using bootstrap resample.The similarity between the actual survival rate and the predicted survival rate of nomograms was verified by the calibration plot (Fig. 4a-c).The actual survival rate and the survival rate predicted by nomogram were on the X-axis and Y-axis, respectively, using the Kaplan-Meier method.The calibration plot indicated that the nomogram predicted the 1-year, 2-year, and 3-year RFS rates of the patients with HG-RNENS.

Comparison of nomogram and TNM staging
The C-index of our novel nomogram was higher than that of ENETS TNM staging (0.782 vs 0.657) and AJCC TNM staging (0.782 vs 0.712) in predicting RFS.Predicted net benefit was compared between the new model and TNM staging using DCA analysis.The DCA curve indicated that the nomogram was feasible for making valuable and informed judgments of the prognosis.Our analysis suggested that if the threshold probability of recurrence in patients was approximately 0-70% predicted by this nomogram, the use of this nomogram to guide treatment measures in patients with HG-RNENS would provide more benefit than either the "treat all patients" or the "treat none patient" schemes.Also, the DCA decision curve suggests that the net yield of the new model is higher than that of AJCC TNM staging and ENETS TNM staging in the range of threshold 10-40% (Fig. 4d).Furthermore, the time-dependent ROC curve of the new model was found to be consistently more favorable in predicting RFS from one to nine years, as shown in Fig. 5.

Discussion
HG-RNENS is a highly heterogeneous group of neoplasms that can present either as low-grade malignant or as highly aggressive [14].The incidence of HG-RNENS has increased in recent years, mainly due to increased health awareness and advances in endoscopic techniques [15].Most HG-RNENS are nonfunctional neoplasms, which means that these patients have no symptoms related to hormone secretion and lack specific clinical manifestations; thus, HG-RNENS is mostly diagnosed at the time of advanced stage [16].In this study, more than half of the patients developed lymph node metastasis at the time of diagnosis, and about a quarter developed distant metastasis, with a 5-year CSS of only 59.0% in the whole group.This suggests that the overall prognosis of HG-RNENS is poor.Thus, in addition to conventional contrast-enhanced imaging, clinicians should perform EUS, PET-CT, pathological biopsy, and other examinations to accurately assess tumor pathological grade, invasion condition, lymph node, and distant metastasis, which may be essential for choosing the most appropriate treatment.In the further analysis, we constructed a prognostic predictive model based on clinicopathologic parameters that significantly impact the prognosis of patients with HG-RNENS.To our knowledge, this is the first prognostic model study for HG-RNENS.
Recent studies have shown that some preoperative hematological parameters are helpful to predict the prognosis of patients with digestive tract tumors.Tokunaga et al. retrospectively analyzed 468 patients with colorectal cancer who underwent radical resection, and the results showed that using 45.0 as the cutoff value of PNI can predict OS and RFS effectively [10].Also, Tominaga et al. suggested that distinguishing PNI by 42.4 could effectively predict the prognosis of elderly patients with colorectal cancer [11].These studies suggested that PNI could be used as an effective predictor of the prognosis of patients with colorectal cancer, and low PNI was often associated with adverse prognostic events.Hu et al. included 61 patients with gastric cancer, and the results showed that the prognosis of patients with ALP ≤ 225.0 U/L was significantly better than that of patients with ALP > 225.0 U/L [12].Chau I et al. also obtained similar conclusions in the study of esophagogastric junction cancer; their study showed that using 100.0U/L as the cutoff value of ALP could effectively predict the prognosis of patients with esophagogastric junction cancer [13].In Lymph node metastasis is a risk factor for the prognosis of RNENs patients, and PLN-based TNM staging system is currently the most widely used staging system [6,7].The PLN staging system depends only on the number of positive nodes and does not consider the influence of the number of dissected nodes/numbers of negative nodes.Studies have found that even in patients with the same number of positive lymph nodes, insufficient number of dissected lymph nodes might indicate a poor prognosis [8].In contrast to PLN, LNR is less susceptible to the number of dissected lymph nodes during surgery and is thus more stable.As a novel pathological factor, LNR has been demonstrated to be associated with the prognosis of diseases such as gastric cancer, colorectal cancer, and esophageal squamous cell carcinoma [17][18][19].In this study, we found that LNR > 0.40 was associated with adverse prognostic events for HG-RNENS.Our new model based on LNR, ALP, and PNI showed better predictive power than AJCC TNM staging and ENETS TNM staging in C-index and DCA and TD-AUC analyses and demonstrated better predictive power.
Predictive models, including scoring model and nomogram, have been widely studied and applied in the clinical practice of NENs, but such models are still lacking for the diagnosis and treatment of HG-RNENS [20][21][22].In this study, we constructed an accurate and effective prognostic prediction model for RFS in HG-RNENS patients based on the three aforementioned independent predictors.In the establishment of RFS prognostic model, the novel nomogram showed good predictive efficacy with C-index of 0.782.This suggests that the new model based on LNR, ALP, and PNI constructed in this study has a high predictive value for the prognosis of HG-RNENS patients.Given its simplicity and effectiveness, the new model can help clinicians select treatment modalities and even help screen patients suitable for clinical research.
This study has some limitations.First, although we collected multicenter data to reduce potential bias, the nature of the retrospective study still impacts the statistical power of this study and reduces clinical value.Second, limited by the incidence of HG-RNENS, the effectiveness of the predictive model was studied only through a small sample.Although we have verified and compared the models, the clinical effect of this model still needs to be further verified by international multicenter studies with larger sample size.However, to our knowledge, this is currently the study with the largest sample size for HG-RNENS and can provide some guidance for the diagnosis and treatment selection of HG-RNENS.

Conclusions
In summary, LNR, ALP, and PNI were found to be independent predictors of prognosis in HG-RNENS in this study.In addition, we constructed a predictive model for RFS in patients with HG-RNENS after radical resection based on data from multicenter data, which showed better predictive power than AJCC TNM staging and ENETS TNM staging.The new model can effectively predict RFS and can be used to guide the development of treatment options for patients with HG-RNENS.

Fig. 2 a
Fig. 2 a Cancer specific survival for the whole group.b Relapse-free survival stratified by PNI.c Relapse-free survival stratified by ALP.d Relapse-free survival stratified by LNR

Fig. 4 a
Fig. 4 a Calibration plot of 1-year relapse-free survival.b Calibration plot of 2-year relapse-free survival.c Calibration plot of 3-year relapse-free survival.d The decision curve analysis (DCA)

Table 1
Clinicopathological data of patients with high grade rectal neuroendocrine neoplasms

Table 2
Univariable and multivariable analysis of factors associated with relapse-free survival in patients with high grade rectal neuroendocrine neoplasms Analysis for patients who underwent radical resection + lymphadenectomy with available lymph node information and did not receive neoadjuvant therapy a HR Hazard ratio b CI Confidence intervalClinicopathological