Nomogram for Predicting Survival in Mucinous Adenocarcinoma of Prostate

Background: Few studies on predicting survival in patients with histology of mucinous adenocarcinoma of the prostate have been done. Objective: Nomogram is a mathematical model in which various important factors are combined to predict a specic end point. The purpose of this study was to evaluate the prognostic value of clinical factors in patients with mucinous adenocarcinoma and to construct nomogram to predict the survival rate. Design,Setting,and Participants: A nomogram was designed to predict the survival rate of 356 patients with invasive ductal carcinoma selected from the surveillance, epidemiology and nal results (SEER) database. Univariate and multivariate models of the cohort were done rstly.And then a model of nomogram that included age, race, grade, stage M and chemotherapy were constructed to predict the 3-year and 5-year survival of prostate cancer patients with histology of mucinous adenocarcinoma. Based on different time, we designed two nomograms. Results and limitations: After the discrimination and calibration, C-index was 0.8138. The specicity was 86.12%, sensitivity was 55.89% for 3-year nomogram . Its survival was 91.49% and AUC was 0.7467. The specicity was 82.42%, sensitivity was 54.83% for 5-year nomogram. Its Survival was 86.73% and AUC was 0.7555. Respectively, which indicated relative good discrimination of the nomogram. Conclusions: This clinical model shows a high ability to predict the survival rate of mucinous adenocarcinoma patients with histologically prostate cancer, making this model a novel and attractive model for predicting the survival rate of patients.


Background
Prostate mucinous adenocarcinoma is a rare variant whose prognosis is traditionally been considered to be worse than traditional adenocarcinoma [1] . The diagnosis of prostate mucinous adenocarcinoma is performed during prostatectomy. A diagnosis can be made when at least 25% of the excised tumors contain extracellular mucin [2,3] . The clinical signi cance of mucosal components of prostate cancer > 25% is unclear. Although some studies [4,5] have shown that these tumors show higher invasiveness than typical (non-mucinous) acinar adenocarcinomas, other studies [6,7] have shown that these two types of tumors have similar results. The incidence of prostate mucinous adenocarcinoma is approximately 0.2% [4,8−10] . However, according to the different data of isolated cases or small case series, the prognosis is still controversial compared with typical acinar cancer [6,11−15] . Therefore, it is necessary to establish a model to accurately assess the prognosis of patients with prostate mucinous adenocarcinoma.
Nomogram is a mathematical model in which various important factors are combined to predict a speci c end point [16] . By integrating these clinical and pathological factors, nomograms can provide individualized estimates of events over time, such as the individual probability of a patient's disease recurrence and death. Therefore, these algorithms can be used as reliable tools to predict clinical results and guide surgery, monitoring and auxiliary treatment decisions.
The construction of the nomogram is based on independent the relative factors of prognosis through Kaplan-Meier and multivariate Cox proportional hazard models. At present, nomograms have been frequently used to help surgeons formulate treatment schedule and assess the prognosis of cancer, for instance liver cancer [17] , gastric cancer [18] , nasopharyngeal cancer [19] and breast cancer [20] . Most important of all, it has been recorded in the NCCN Clinical Guidelines that prostate cancer can be detected early through nomograms [21] . Therefore, we attempted to evaluate the 3-year and 5-year operating systems based on serer database by establishing two nomograms to provide reference for surgeons.

Patients' general information
The study enrolled 356 patients with prostate mucinous adenocarcinoma from the SEER database from 2004 to 2015 [22] . Clinical pathological data included age, ethnicity, material status, follow-up time, tumor location, t-stage, n-stage, m-stage, RP, radiotherapy sequence, radiotherapy, chemotherapy, biopsy Gleason score, overall survival, age, and follow-up time. The inclusion of clinicopathological data started on Jan. 1st, 2011. Follow up starts on the date of radical prostatectomy operation, ends on the date of migration, death or December 31, 2015, whichever occurs rst. People have missing data were excluded.
After modeling, the information of 356 people built in the nomogram.
The mean age was 63. 13 year's old. The mean follow-up time was 67.07 months. Ethnic composition includes white, black and other races (Native American / AK, Asian / Paci c Islander). (Table 1) Survival analysis

Nomogram development
The Nearest Neighbor Estimation(NNE) method was used [23] . Generalization of the Somers DXY rank correlation of the c-index and the reviewed response variable: DXY = 2 (C-0.5). The inverse probability of the censored weight (IPCW) estimates of cumulative / dynamic time-dependent ROC curves.)

Nomogram validation
It is required to verify the accuracy of nomograph through 500 bootstrapping. The t was evaluated by a consistency index (C-index) and a calibration chart. There are two commands in R software, "rcorrcens" and "calibrate", through which you can obtain C-index and calibration. Statistical package R [25] and Empower Stats were used for all analyses. We used strati ed linear regression model for subgroup analysis. The modi cation and interaction of subgroups were examined by likelihood ratio test. All statistical tests were carried out from two aspects, P value less than 0.05 was considered statistically signi cant.

Patient clinicopathological data
All data are strictly ltered based on SEER database. This study analyzed 356 cases of mucinous adenocarcinoma of prostate. The average age was 63.13 months and the average follow-up time was 67.06 months. There were 238 (66.85%) married patients, 41 (11.52%) single patients, 36 (10.11%) divorced, widowed or separated patients and 41 (11.52%) unknown patients. A total of 283 (79.49%) white patients, 47 (13.20%) black patients, 20 patients of other races (5.62%) and 6 (1.69%) are unknown patients. The details of the validation queue are shown in the table1 Medium.)

Survival analysis and nomogram construction
For overall survival (OS), we conducted univariate and multivariate analysis. The results showed that age, race, material status, follow-up time, tumor area, T stage, N stage, M stage, RP, radiation sequence, radiation, chemotherapy, biopsy Gleason score, total survival time, age, follow-up time. (Table n) Through the univariate and multivariate analysis, the results showed that age, race, grade, stage M and chemotherapy were independent prognostic factors (P < 0.05). So, we consider all these factors by set up a nomogram. As shown in table 2.

Nomogram construction and validation
According to the nomograph, sum the total points, and then convert them to OS, because there are parallel lines below the graph, and their scales are linear with each other. We used the guide program to sample and calibrate the internal nomogram 500 times, and drew the graph with the appearance, deviation correction and ideal curve. All of this tells us that the nomogram is very consistent internally. Then, we draw the receiver's operating characteristics (ROC) inside and outside the training and veri cation set.
After the discrimination and calibration, C-index was 0.8138 (95% CI = 0.7483-0.8793,SD 0.0669). Based on different follow-up time, 3 years and 5 years, we developed two nomograms. (Figure 1, 2). In nomogram for 3 years, the sensitivity was 55.89% and the speci city was 86.12%. The survival was 91.49% and AUC was 0.7467. In nomogram for 5 years, the sensitivity was 54.83% and the speci city was 82.42%. The survival was 86.73% and AUC was 0.7555. (Figure 3, 4). The C-index value is greater than 0.7, and the calibration curve is in good coherence with 45 degree ideal line.

Discussion
We use Kaplan Meier method to calculate the estimated OS rate, which is showing no difference with the study reported by JAMA Oncol [27] . Through the subsist analysis of the surgical system, we found that the operating system of women was exceed that of men ( Fig. 1 and Table 2). In terms of age, we found that the operating system has declined for more than a year. Current results show that most patients are diagnosed after 50 years old [28] . Our results show that the operating system of patients in black is below that of other races, which is showing no difference with the study [29] .
The accuracy of the markers in the internal (modeling queue) and external (validation queue) is veri ed by using the C-index and calibration curve. 3-year and 5-year of operating systems internal validation of C-indexes was 0.8138. C-index value are greater than 0.7.There is excellent commonality between them.
The 3-year OS and 5-year OS for the prediction of prostate mucinous adenocarcinoma by the establishment of pictograms have many obvious advantages. In a word, it is simple and feasible to create an image-free graph to predict the process of 3-year and 5-year operating systems. Secondly, our patient base is large enough and selected from surveillance, epidemiology and end result (SEER) databases to make it more reliable. Secondly, our model has high precision and speci city for predicting 3-year and 5-year operating systems. This may be related to our large number of queues, which makes the relationship between variables and results quite exible.
Our research has advantages and some limitations. We based SEER database to nished a large sample retrospective research, and a more accurate marking model has been successfully established. A large number of studies have stated clearly that other correlative clinicopathological elements have an impact on the survival of oral cancer patients, for instance HPV [30] , lymph node involvement [31] , tumor thickness [32] , P53 [33] , EGFR [34] , cigarette and alcohol consumption [35] , and chemotherapy [36] . However, the SEER database does not contain the above elements. Similarly speaking, we cannot assess disease-free survival and regional control. Therefore, our scale cannot assess aforementioned factors. In order to test these indicators, we will conduct some prospective studies to make up for these limitations.
Finally, we carefully analyzed the survival and successfully established two accurate pictograms, which provided customized clinical treatment plan and personalized prognosis reference for surgeons.

Conclusion
We propose a nomogram which can predict the 3-year and 5-year survival rates. This clinical model shows a high ability to predict the survival rate of mucinous adenocarcinoma patients with histologically prostate cancer, making this model a novel and attractive model for predicting the survival rate of patients. In the subsequent validation, these promising results strengthen the potential of this prediction tool to predict 3-and 5-year survival rates.
Abbreviations SEER: surveillance, epidemiology and nal results; NNE = Nearest Neighbor Estimation; IPCW = inverse probability of the censored weight; OS = overall survival; ROC = receiver's operating characteristics; Cindex = consistency index Declarations Ethics approval and consent to participate All the information was downloaded from the SEER database via SEERStat software. The study was approved by the Ethics Committee of West China Hospital.

Consent for publication
Not applicable.

Availability of data and material
The datasets analyzed during the current study are available in the SEER database, Http://seer.cancer.gov.

Competing interests
The authors declare that they have no competing interests.  The clinical features were listed above.

SD=Standard Deviation
RP=Radical Prostatectomy OS=Overall Survival GS=Gleason Score