Nomogram to predict overall survival of patients with pseudomyxoma peritonei of appendiceal origin: A retrospective cohort study

Pseudomyxoma peritonei (PMP) is a rare disease, with the rate of overall survival (OS) influenced by many factors. The present study aimed to define independent predictors and establish a nomogram for individual risk prediction in PMP patients.

with hyperthermic intraperitoneal chemotherapy (HIPEC) has been recommended as the optimal treatment for PMP 4 ; as a result, the recurrence rate has decreased and the overall survival (OS) has improved greatly.
Although the long-term outcomes after treatment are impressive in patients with PMP, there is still a significant rate of recurrence of the disease. 5 It has been confirmed that many factors-for instance, male sex, 6 high tumor marker levels, 7,8 high-grade mucinous carcinoma peritonei (especially with signet ring cells), 9,10 debulking surgery, 11 and even KRAS mutation 12 -were all independently associated with poor prognosis in PMP. A recent study constructed a nomogram to predict the OS for PMP patients based on large-scale population data from the Surveillance, Epidemiology and End Results (SEER) program database, where the C-index was 0.757 in the training cohort and 0.746 in the validation cohort. 13 To our knowledge, this is the first established prognostic nomogram for PMP; however, because this nomogram was constructed using a population-based cancer database, the availability of predictive factors is limited and some bioinformatics factors, such as tumor markers, that have been proven to be predictive for PMP patients were not included in the study.
In the present study, we intended to evaluate more comprehensive candidate prognostic factors to predict the OS for PMP patients, which were selected a priori based on either prior research (e.g., sex, carcinoembryonic antigen [CEA], carbohydrate antigen [CA] 125, CA 19-9, degree of radical surgery, and histological grade) or sound clinical reasoning [e.g., age, Barthel Index Score, hemoglobin, albumin, D-dimer, CA 724, CA 242, and peritoneal cancer index (PCI)]. Using this foundation, we sought to create and internally validate a nomogram to predict the individual risk of OS for PMP patients, which may be helpful for physicians involved in the prognosis judgment and treatment of PMP patients.

| Patients
The institutional review board (IRB) of Aerospace Central Hospital approved the present study (ethics no. 20200113-LCYJ-01), and all of the patients signed informed consent before CRS and consented to be followed up with after surgery.
This was a retrospective study. We retrieved cases with a diagnosis of PMP in the special follow-up database from Aerospace Central Hospital between June 1, 2013, and November 22, 2019. The PMP diagnosis was finally confirmed based on histological results of resected specimens, which were interpreted by two experienced pathologists according to the Peritoneal Surface Oncology Group International (PSOGI) criteria. 14 The study inclusion criteria were patients who received CRS and HIPEC treatment at our center for the first time (n = 160). The exclusion criteria were as follows: (1)

| Follow-up protocol
All patients were routinely followed up with every 3-6 months after having undergone CRS and HIPEC for the first time, and tumor markers and computed tomography (CT) imaging of the abdominopelvic region were routinely examined. Most patients routinely came to our center for follow-up; however, if the patient did not visit our center for follow-up, we assessed them by telephone to determine the patient's survival status.
F I G U R E 1 Study schematic. A total of 160 patients with PMP who underwent CRS and HIPEC at our center for the first time were included. Among the 160 identified cases, one with incomplete medical records, eight with disease of non-appendiceal origin (colorectal [n = 5], urachal [n = 1], biliary pancreatic [n = 1], and ovary [n = 1]), one combined with breast and thyroid cancer, and three lost to follow-up were all excluded. Ultimately, 147 subjects with PMP were included in the present study. CRS, cytoreductive surgery; HIPEC, hyperthermic intraperitoneal chemotherapy; PMP, pseudomyxoma peritonei

| Endpoint event determination
The time of CRS was taken as the starting time for follow-up. Allcause mortality served as an endpoint event in the present study. The final date of follow-up was November 8, 2020.

| Tumor markers determination
All tumor markers were determined within seven days before CRS. All measurements were performed according to manufacturer's instructions.
CEA (ng/ml), CA 125 (U/ml), and CA 19-9 (U/ml) were measured by chemiluminescence immunoassay (CMIA) (Abbott), the detection method of CA 724 (U/ml) was same as the former three tumor markers (Autobio, Zhengzhou, China), while CA 242 (kU/L) was measured by flow fluorescent technology (Luminex). Internal quality control (IQC) and external quality assessment (EQC) were all performed for all five tumor markers.

| Peritoneal cancer index
The PCI scoring system divides the abdomen into nine anatomical areas, with four further subareas of the small bowel delineated. Tumors were assessed in each area, and a score of 0-3 points was awarded for each of the 13 areas (i.e., 0 points for no tumor, 1 point for nodules <0.5 cm, 2 points for nodules between 0.5 and 5 cm, and 3 points for nodules >5 cm). The total score was then calculated by adding all the scores and ranges from zero to 39 points. 15

| CRS and HIPEC
The CRS of PMP was performed consistent with standard operation method. Complete removal of all visible disease was scored as CC 0 cytoreduction, while residual disease less than 0.25 cm was scored as CC 1; both CC 0 and CC 1 are considered as having undergone complete CRS (CCRS). If the patient could not achieve complete cytoreduction, debulking treatment was performed, and any residual tumor deposit between 0.25 and 2.5 cm was scored as a CC 2 cytoreduction, while a residual tumor deposit greater than 2.5 cm was scored as CC 3 cytoreduction, and CC 2 and CC 3 were considered as having undergone maximum tumor debulking (MTD). 16 Once cytoreduction was complete, intraoperative hyperthermic chemotherapy was delivered, wherein 5-fluorouracil (1000 mg) together with cisplatin (80 mg) was heated to 43°C and continuously infused using a HIPEC machine for 1 h.  OS was estimated from the initial date of CRS and HIPEC to death or censored at the last follow-up. 17 The Kaplan-Meier method was used to describe the survival data and plot survival curves. All candidate continuous variables were processed by using the X-tile program to obtain the best cut-off points because it is easier to explain the results in clinical practice when continuous variables have been transformed into categorical variables. OS rate differences between groups were compared using the univariate log-rank test. In the present study, we performed the purposeful selection of variables to establish our prediction model for PMP patients. First, we used univariate Cox regression to screen the candidate influencing factors, and only variables with p values less than .10 were then included in multivariate Cox proportional hazard models. We defined confounding as a change in any remaining parameter estimate of more than 20% relative to the full model. A change in a parameter estimate above the specified level indicates that the excluded variable was important in the sense of providing a needed adjustment for one or more of the variables remaining in the model. 18 After the independent predictors were established, the interactions between the variables were assessed in the model.

| Statistical analysis
After multivariate analyses, a nomogram was constructed using the RMS package in R to visually predict the one-, two-, and 3-year OS rates for PMP. The maximum score of each variable was set as 100 points. The performance of the established nomogram was measured using the Harrel concordance index (C-index) and calibration plotting. The C-index was used to assess the discrimination ability of the predicting model, and calibration plotting was used to determine whether the predicted survival was in concordance with the actual survival. The nomogram was subjected to 1000 bootstrap resamples for internal validation by discrimination and calibration.
Two-sided p values less than .05 indicated a statistically significant difference.

| Patient characteristics
Among the 160 patients who underwent CRS and HIPEC at our center for the first time, 147 subjects were finally included for survival analysis, while the remaining 13 subjects were excluded. The statistical results indicated that there was no significant difference in several key clinicopathological characteristics of the 147 subjects in comparison with the total 160 subjects (eTable in the Supplement).
The clinicopathological features of the included patients with PMP are presented in Table 1

| Influence of independent prognostic variables on survival
The best cut-off points of continuous variables were calculated using the X-Tile software, and details are listed in

| DISCUSSION
Reliable prognostication in any cancer after surgical resection is critical to patients and treating physicians alike for the purpose of making good decisions. 19 The present study identified five independent prognostic factors of PMP patients and further built a nomogram to predict the OS effectively and visually. After been underwent bootstrap internal validation, the nomogram exhibited a reasonable level of predictive performance.
The particular strength of the present study is that the missing follow-up rate was very low, with only three patients lost to follow-up because, at the beginning of the establishment of our follow-up database, we had arranged for a physician to be responsible for follow-up; as such, meanwhile, the present research established a nomogram with a C-index of 0.825, which seems to offer a better predictive ability than that of the former study. The main reason for this difference may be that the variables included in the prediction model were different between these two studies. The former investigation 13 was based on data from the SEER database, and some positive clinicopathological characteristics associated with prognosis were not available in the SEER database and thus could not be incorporated into the study analysis. However, our nomogram was developed based on our single-center follow-up database; so, the candidate predictors were relatively more comprehensive. In addition, the TNM staging system is not commonly used in PMP patients at our center, while PCI is commonly used for reflecting the tumor burden instead of TNM staging system.
An elevated pretreatment D-dimer level has been confirmed to be markedly associated with poor OS in patients with solid tumors, 20 such as those with colorectal cancer, 21 pancreatic carcinoma, 22 and lung cancer. 23,24 Interestingly, we also discovered that a high D-dimer level was associated with a poor survival which is also expressed in peritoneal malignancy and can be elevated in patients with any source of peritoneal irritation. 1 CA 125 may also help to judge the tumor burden of peritoneal carcinoma. In the present study, CA 125 seems as well to be a useful marker for prediction survival of PMP patients, with a high CA 125 level denoting a poor prognosis of PMP.
While the univariate analysis revealed that elevated CEA, CA 724, and CA 242 levels were all associated with poor survival in PMP, the inclusion of these parameters in the multivariate analysis did not yield a degree of statistical significance. A former study also confirmed that CEA boasts a low value in the prediction survival of PMP patients. 7 Due to the extremely low incidence rate of PMP, few cases were included in the present study; therefore, studies with larger sample sizes and longer follow-up times are needed to assess the predictive value of tumor markers in PMP patients.
The completeness of cytoreduction is one of the most important prognostic factors for PMP, 1 and the present study revealed that the MTD group had an obvious reduced rate of survival relative to the CCRS group. Notably, this result is consistent with those of previous studies. 6,29 However, a large proportion of participants in the former study reached CCRS, while, in our study, the majority of patients could only undergo major debulking surgery; we speculate that most PMP patients in China  31 have confirmed that PCI is a risk factor for postoperative morbidity in univariate analysis; however, no statistically significant correlation was found during the multivariate analysis. The present study also found that PCI was not an independent prognostic factor for survival prediction in PMP. In clinical practice, we found that some patients with high tumor loads can also achieve CCRS; therefore, the degree of surgery may boast a higher predictive value for prognosis than that of PCI in PMP patients.
Some researchers reckon that the prognosis of PMP correlates closely with histopathological classification. 32 By contrast, others have suggested that PMP is unlike other tumors, and histopathology does not reliably predict tumor behavior. 33 The present study revealed that histological grade is an independent prognostic factor for PMP patients,  [22][23][24][25][26][27][28][29][30][31][32][33]; Z = −2.062; p = .039). Physicians at our center often choose patients with a heavy tumor burden to undergo preoperative systemic chemotherapy, and patients with a heavy tumor burden often have a poorer prognosis; therefore, this seemingly contradictory phenomenon was also found in the present study.
There were several limitations to the present study. First, due to the limitations of a retrospective study design, records of some data in the database were missed. Second, the predictive accuracy and discriminative ability of the established nomogram in the present study only underwent internal validation; an external validation assessment with a separate cohort remains to be undertaken in future.
Finally, some other prognostic factors were not included in the prediction model, such as KRAS mutations, which have been proven to be independently prognostic for progression-free survival in PMP patients. 12 Future work will further verify this hypothesis.

| CONCLUSION
To conclude, the present study comprehensively identified five independent prognostic factors for PMP patients, including D-dimer, CA 125, CA 19-9, degree of radical surgery, and histological grade.
Based on these variables, the established nomogram possesses a reasonable degree of prediction performance, and we could calculate an accurate individual survival probability. More research is needed to verify and improve the prediction model, particularly prospective large-sample studies for external validation.