DOI: https://doi.org/10.21203/rs.3.rs-2697044/v1
Objective Medullary carcinoma (MC) of the colon is a rare malignancy, and there is no survival prediction for this tumor. This study aimed to construct a nomogram to predict the overall survival (OS) of patients with MC
Methods We included 276 patients with a pathological diagnosis of MC between 2010 and 2015 from the Surveillance, Epidemiology, and End Results (SEER) database. The random forest method and multivariate Cox proportional risk model were used to identify independent prognostic factors for MC. The consistency index (C-index), the receiver operating characteristic (ROC) curve, and the calibration curve determined the nomogram's predictive accuracy and discriminative ability. Decision curve analysis (DCA) was used to evaluate the net clinical benefit of the nomogram.
Results The Cox regression analysis showed that age, N-stage, M-stage, tumor size, and chemotherapy were associated with OS of MC. Based on the identified independent factors, we constructed a nomogram for predicting OS in MC patients. The C-index value of the nomogram for predicting OS was superior to the TNM staging system (0.705 vs. 0.673). ROC and calibration curves showed the nomogram's good discriminatory and calibration ability. DCA showed that the nomogram had a more significant net clinical benefit than the TNM staging system.
Conclusion We developed a nomogram to accurately predict MC patients’ survival. The nomogram had excellent predictive efficacy and could help clinicians to assess the prognosis of MC patients.
Medullary carcinoma (MC) of the colon is an extremely rare tumor, accounting for 0.05% of all colon cancers[1], and is a poorly differentiated malignancy of the intestine first reported by Gibbs in 1977[2]. In the WHO (2010) classification of gastrointestinal tumors, intestinal MC was classified as a subtype of poorly differentiated or undifferentiated tumors. A growing number of studies also demonstrated that MC was an independent type of colon cancer[3–5]. Clinically, MC predominantly occurs in elderly, female, and right hemicolon individuals. Microsatellite instability (MSI) with B-RAF gene mutation is a common molecular feature of MC in immunohistochemical and molecular tests[6, 7]. MC showed a relatively good prognosis compared to other poorly differentiated and undifferentiated tumors of the colon[8]. So far, no predictive nomogram has been developed for this rare tumor.
Currently, the most common method for assessing the prognosis of colon cancer is the TNM staging system promulgated by the American Joint Committee on Cancer (AJCC). The TNM staging system effectively predicts the entire population of colon cancer patients, but it is less accurate in predicting the prognosis of individual patients. In addition, the system only considers tumor, lymphatic and metastatic conditions and does not consider other clinical variables important for predicting individual patients[9]. The TNM staging system has been reported to have limitations in accurately predicting patient prognosis, resulting in patients with advanced-stage colon cancer having a better prognosis than those with earlier-stage colon cancer[10, 11]. Therefore, we need to develop a predictive nomogram superior to the TNM staging system.
A nomogram is a statistical tool for predicting individual risk and is widely used for survival prediction, risk assessment, and clinical decision-making in patients with various cancers[12–15]. The nomogram is presented in a visualized form by including several important variables to predict the OS of patients with cancer[16, 17]. In addition, the nomogram usually has an advantage over the TNM staging system in predicting the individual risk of cancer patients[18, 19].
As far as we know, no predictive tool applies to individual patients with MC of the colon. The SEER database has collected a large amount of cancer incidence and survival data, which can provide a good source of data for studying rare tumors. Therefore, by incorporating the clinical and pathological characteristics of MC patients through SEER, we created the first nomogram for MC that could accurately predict OS. This nomogram could help physicians assess the prognosis of MC patients.
Patient data for this retrospective study cohort were filtered and downloaded using SEER*Stat version 8.4.0 (Incidence - SEER Research Plus Data, 17 Registries, Nov 2021 Sub (2000–2019)). MC patients eligible during 2010–2015 were included in this retrospective study. The following variables were selected when downloading data: age at diagnosis, sex, race, primary tumor site, degree of tumor differentiation, 7th edition AJCC stage, T-stage, N-stage, M-stage, tumor size, surgery, radiationtherapy, chemotherapy, months of survival, and survival status. Patient inclusion criteria: 1. Confirmed by pathological examination; 2. The first primary malignant tumor; 3. Undergoing surgical treatment. Patient exclusion criteria: 1. Non-first primary malignancy; 2. Unclear variable information or follow-up information. The detailed screening process for patients is shown in Fig. 1. In ethnicity, Asian or Pacific Islander, American Indian, and Alaska Native were classified as Others. The splenic region of the colon, descending colon, and sigmoid colon was defined as the left hemicolon, and the cecum, ascending colon, hepatic flexure of the colon, and transverse colon was defined as the right hemicolon. We used X-tile[20] software to identify the optimal cut-off point for tumor size (Fig. 2). The endpoint of this study was OS, defined as the time from the patient's diagnosis of MC to death.
The optimal cut-off value for tumor size was calculated using X-Tile software. The random forest method and multivariate Cox proportional risk model were used to identify independent prognostic factors for the nomogram. The nomogram was used to estimate the probability of OS at 1, 3, and 5 years. Survival analysis of significant variables was performed using the Kaplan-Meier method. C-index, ROC curves, and calibration plots assessed nomograms' discriminative and calibration ability. DCA evaluated the clinical benefit of nomograms. All P values were two-tailed, and P < 0.05 was considered statistically significant. All statistical analyses were performed with the R software (version 4.2.1).
A total of 276 patients with MC eligible for the criteria were included in the SEER database. In the entire cohort, 88% of patients were white, 97% were elderly, 74% were female patients, 93% had a primary tumor site in the right hemicolectomy, 69% were poorly differentiated (grade III), and 29% were undifferentiated (grade IV). The proportions of stages I, II, III, and IV were 10%, 43%, 40%, and 10%, respectively. In addition, only 6 (2%) patients in the entire study population received radiotherapy, and 83 (30%) patients received chemotherapy. The longest follow-up time was 118 months, the median follow-up time was 55.5 months, and 130 patients died by the end of the follow-up time. Information on the clinical characteristics of the patients is shown in Fig. 3.
Random forest and COX regression analysis showed (Fig. 3): age (HR = 1.96, 95% CI: 1.05–3.68, P = 0.035), N-stage (HR = 1.83, 95% CI: 1.17–2.87, P = 0.008), M-stage (HR = 4.28, 95% CI: 2.33–7.87, P < 0.001), tumor size (HR = 2.46, 95% CI: 1.04–5.58, P = 0.041), and chemotherapy (HR = 0.29, 95% CI: 0.17–0.49, P < 0.001) were independent prognostic factors for OS of MC, with chemotherapy being a protective factor. K-M curves (Fig. 4) showed significant survival differences for each independent factor (P value < 0.05).
Based on the results of COX regression analysis, we constructed a nomogram using a total of five variables: age, tumor size, N stage, M stage, and chemotherapy to predict the OS for MC patients at one year, three years, and five years. As shown in Fig. 5, the patient's total score can be obtained by summing the respective scores for each variable. The total scores correspond to the 1-year, 3-year, and 5-year OS probabilities. The C-index of the nomogram was 0.70, and the C-index of the TNM staging system was 0.67. The ROC curves (Fig. 6) showed AUC values of 0.785, 0.722, and 0.702 for predicting 1-year, 3-year, and 5-year OS, respectively, indicating that the nomogram had a good discrimination ability. The calibration curves (Fig. 7) were close to the reference line, suggesting that the nomogram had an excellent calibration capability. DCA (Fig. 8) demonstrated that the nomogram would be clinically helpful and have a more significant net clinical benefit than the TNM staging system.
MC is a rare malignancy that has only been increasingly recognized and studied in the last 20 years. To our knowledge, no study has reported on the MC nomogram. Therefore, we constructed a nomogram to predict the prognosis of MC patients. Five variables (age, N stage, M stage, tumor size, and chemotherapy) were identified using a random forest method based on cox regression and incorporated into the nomogram. The validation of the nomogram showed that it had good discriminatory and calibration ability.
Some authors suggested that MSI was necessary to diagnose MC[21, 22]. In addition, MC patients with MSI had a better prognosis than other poorly differentiated adenocarcinomas[4]. Since there was no record of immunohistochemistry for MC patients in the SEER database, which prevented us from further investigating the effect of MSI on OS. Therefore, we did not include MSI in our nomogram. Of all the literature we reviewed on MC, there were few reports on the prognosis of MC patients, and the influence of clinicopathological features other than MSI on MC’s prognosis was unclear. MC is a subtype of colon cancer, and there is minimal report on its prognostic factors. The variables included in the current study were selected based on the predictive factors of colon cancer that had been identified in previous studies.
In our study, MC occurred mainly in patients > 60 years old, which may be related to hMLH1 hypermethylation with an accompanying lack of protein expression[23]. Multivariate COX regression analysis showed that > 60 years was a risk factor for MC prognosis, consistent with previous studies reporting the effect of age on colon cancer prognosis[24]. Previous studies have demonstrated that lymph node metastasis was also an important prognostic factor for patients with colon cancer[24]. Previous studies had demonstrated that lymph node metastasis was also an important prognostic factor for patients with colon cancer[25, 26]. Our findings showed that a more positive number of lymph nodes was associated with a worse prognosis of MC, and survival curves suggested significant survival differences between patients with different numbers of lymph node metastases. According to the 7th edition AJCC criteria, M0 represented no distant organ metastasis, and M1 represented a single distant organ metastasis. The prognosis of patients with M1 was significantly worse than that of patients with M0. The predictive results of the N-stage and M-stage variables in our nomogram were the same as those of the N-stage and M-stage in the TNM staging system. Other studies revealed that tumor size was also an independent prognostic factor for colon cancer and was negatively associated with patient survival[27, 28]. We used the X-Tile software to determine the optimal cut-off values for tumor size (3–37 mm, 38–97 mm, and 98–162 mm, respectively), and the survival curves for the three groups of tumor sizes were significantly different. The results of COX regression analysis showed that the larger the tumor (98–162 mm), the poorer the prognosis of patients, which was consistent with the results of previous studies on the association between tumor size and the prediction of colon cancer.
NCCN guidelines recommend surgery and postoperative adjuvant chemotherapy as the standard treatment principles for non-metastatic colon cancer[29]. However, because MC is a rare disease, most previous studies were about pathomorphological features, and few were about therapeutic methods. Until now, there are no treatment guidelines and consensus on MC. A total of 276 patients treated with surgery were included in this study; only 6 received radiotherapy, and 83 received chemotherapy. The results of our COX regression analysis suggested that chemotherapy was a protective factor for the prognosis of MC patients. We could not find any reports on chemotherapy and the prognosis of patients with MC. However, many reports confirmed that chemotherapy could significantly improve the survival of patients with colon cancer[30, 31]. Similarly, our findings suggested that chemotherapy improves OS in MC patients. Based on the above reports and the results of our analysis, we incorporated age, N stage, M stage, tumor size, and chemotherapy to construct a nomogram for predicting OS in patients with MC.
A nomogram is a tool commonly used to estimate the survival rate of individual cancer patients, and it can calculate the cumulative effect by integrating all prognostic factors. There have been many reports on constructing prognostic nomograms for colon cancer. However, no prognostic nomograms have been developed for patients with MC. The SEER database covers approximately 30% of the U.S. population and can provide sufficient patient data for rare tumors. In this study, we created prognostic nomograms of MC from the SEER database to predict the incidence of 1-year, 3-year, and 5-year OS. Internal validation showed good efficacy of the nomogram, and DCA showed that our nomogram had more significant advantages than the TNM staging system.
There are several limitations to our study. First, as a retrospective study, selection bias is inevitable. Second, immunohistochemical information (e.g., MSI) is unavailable in the SEER database, and MSI is generally associated with prognosis in patients with MC. Finally, because MC is a rare disease, it was tough for us to collect enough clinical data for external validation. The nomogram we developed still requires further verification and improvement in future clinical practice to make it more convincing.
Our study is the first to develop a nomogram to predict the prognosis of MC patients based on the SEER database. This nomogram included five independent prognostic factors that influence the OS of MC: age, tumor size, N stage, M stage, and chemotherapy. The validation showed that the nomogram performed well and could be helpful for clinicians to assess the prognosis of MC patients and answer the consultation.
Ethical approval
This study does not include any animal or human experiments performed by the authors. The SEER database does not record patients' names and does not involve patients' privacy.
Competing interests
All authors guarantee that they have no conflict of interest.
Authors' contributions
All authors contributed to the conception and design. Data collection and manuscript preparation were performed by Huabin Zhou and Jiayi Chen. Statistical analysis and data analysis were performed by Yulan Liu and Chao Zheng. The first draft of the manuscript was written by Huabin Zhou and reviewed by Min Li. All authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.
Funding
This research did not receive any specific grants from funding agencies in the public, commercial, or not-for-profit sectors.
Availability of data and materials
Te primary data used to support the fndings of this study are available from the corresponding author upon request.
Acknowledgments
We sincerely thank the SEER project for making this database public.