Construction and Validation of the Predictive Model for Gallbladder Cancer Liver Metastasis Patients: A SEER-based Study

doi:10.21203/rs.3.rs-2484536/v1

Download PDF

Research Article

Construction and Validation of the Predictive Model for Gallbladder Cancer Liver Metastasis Patients: A SEER-based Study

https://doi.org/10.21203/rs.3.rs-2484536/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background

The purpose of this present research was to construct a nomograph model to predict prognosis in gallbladder cancer liver metastasis (GCLM) patients so as to provide a basis for clinical decision-making.

Methods

We surveyed patients diagnosed with gallbladder cancer liver metastasis (GCLM) in the Surveillance Epidemiology and the End Results database (SEER) between 2010 and 2019. They were randomized 6:4 into a training set and a validation set. In the training set, statistically meaningful prognostic factors were determined using univariate and multivariate Cox regression analyses, and an individualized nomogram prediction model was generated. The prediction model was evaluated by C-index and calibration curves from the training set and the validation set.

Results

727 confirmed cases were enrolled in the research, 426 in the training set and 301 in the validation set. Factors including bone-metastasis, surgery, chemotherapy and radiotherapy were independent prognostic factors for cancer specific survival (CSS) rates and were employed in the construction of the nomogram prediction model. The C-index for the training set and validation set were 0.685 and 0.709, respectively. Calibration curves indicated that the cancer specific survival (CSS) rates of the nomograph prediction model were consistent with the actual prognosis at 6-months-CSS, 1-year-CSS and 2-years-CSS in both the training set and validation set.

Conclusions

We have successfully constructed a nomogram prediction model that can predict cancer specific survival (CSS) rates in patients with gallbladder cancer liver metastasis (GCLM). This prediction model can help patients in counseling and guide clinicians in treatment decisions.

Oncology

GCLM

CSS

SEER

metastasis

nomograph

prognosis

Gallbladder cancer (GC) refers to the most common malignant neoplasms in the biliary system, which is characterized by insidious symptoms, rapid development, early metastasis and weak prognosis, and its incidence and death rate have been increasing slowly in recent years¹. At present, there is still a lack of early diagnosis means with good specificity and sensitivity for gallbladder cancer, and most of the gallbladder cancers found clinically are in the middle and late stages². Since the gallbladder is adjacent to the liver, gallbladder neoplasms often invade and metastasize to the liver, which makes prognosis prediction and treatment more difficult.

Currently, the 8th edition of the American Joint Committee on Cancer staging system (TNM staging system) is used the most broadly in order to assess the prognosis of confirmed cases with gallbladder cancer³. Nevertheless, there’s a limitation that the clinicopathological and therapeutic information is ignored in the TNM staging system⁴. If the liver metastasis has been diagnosed clinically, it’s usually regarded as a late stage of gallbladder cancer liver metastasis (GCLM). Also, it’s impossible to predict the prognosis by the conventional TNM staging. Therefore, there is an imperative need for more precise prognostic models to impart clinical decision to patients suffering from gallbladder cancer liver metastasis. Recently, the nomograph has been extensively used as a prediction approach for oncology, which has been proved to have better predictive ability than conventional TNM classification⁵. In the former study^6,7, the nomogram model showed good discrimination for patients with gallbladder cancer as well. However, to date, there haven’t been studies on the nomograph prediction model of gallbladder cancer liver metastasis (GCLM).

Through the retrospective analysis among confirmed cases from 2010 to 2019, the prediction model of gallbladder cancer liver metastasis is constructed on the basis of the Surveillance Epidemiology and the End Results database (SEER). And its reliability and feasibility have been assessed through an independent validation cohort.

Patient selection

For the gallbladder cancer liver metastasis in this study, the exclusion criteria are as follows: tumors of multiple origins; incomplete tumor type, differentiation and staging; incomplete follow-up information; patients lacking a clear pathological diagnosis; patients with incomplete information; patients whose first tumor was a non-primary gallbladder cancer; and patients aged ≥ 80 years or < 20 years. The selection criteria and screening process are illustrated in Fig. 1.

Data extraction

On the basis of the inclusion and exclusion criteria, the SEER*Stat 8.4.0 software was used to search for the information of confirmed cases in the SEER database, which includes the patient ID, year, gender, race, marital status, age at diagnosis, average household income, the type of pathology, whether the patient received timely treatment, whether the patient had pulmonary bone and brain metastasis, whether the patient was operated, whether the patient received radiotherapy, whether the patient received chemotherapy, and follow-up situations, etc. The extracted data were recoded. Gallbladder Cancer-Specific Survival (GCSS) refers to the patient's survival time from the diagnosis with gallbladder cancer to the death of gallbladder cancer. The survival situation is recorded in accordance with the latest follow-up. On the basis of the above inclusion and exclusion criteria, patients with gallbladder cancer liver metastasis who were registered in the SEER database between 2010 and 2019 were included in the study for retrospective analysis.

Statistical analysis

The study was statistically analyzed in its entirety using R software. The confirmed cases were divided into 2 groups, the training set and the validation set, in a ratio of 6:4 to make sure the outcome event (dead or alive) was randomly distributed between the 2 data groups. A chi-square test was used for comparing the baseline data between the 2 groups (in Table 1 for details). For the training group data, univariate analysis was employed, and factors with P < 0.05 were further included in multifactor cox regression analysis to screen out the factor with P < 0.05 as predictors. Based on the predictors, R called the "rms" package to construct the COX proportional risk model function and applied the Nomogram tool to build the survival prediction model. Internal and external validation was performed using C-index and calibration curves in the training set and the validation set, separately, to test the distinction and consistency of this model. All statistical tests were bilateral, and P (< 0.05) was statistically significant.

Characteristics of the training set and validation set.

727 confirmed cases of GCLM diagnosed between 2010 and 2019 were screened out from the SEER database. After randomization of all patients in a 6:4 ratio, there were 426 confirmed cases in the training set and 301 confirmed cases in the validation set. No statistically significant differences were shown in demographics, pathological characteristics and treatment information between the 2 sets (p > 0.05), as detailed in Table 1, indicating no obvious difference between the 2 groups.

Screening out prognostic factors

Univariate analysis indicated that factors affecting CSS comprised delayed treatment, bone metastasis, PTR, chemotherapy and radiotherapy (all P < 0.05). Based on Cox regression, multivariate analysis included the remarkable factors from the univariate analysis. Bone metastasis, PTR, chemotherapy and radiotherapy (P < 0.05). Table 2 indicates the outcomes of univariate versus multivariate analyses. On the basis of the Kaplan-Meier approach as well as Cox regression model, the 4 factors were selected for model construction: bone metastasis, PTR, chemotherapy and radiotherapy.

Construction of the nomogram predicting model in the training set

In the nomogram model, four independent prognostic factors were selected as variables. and we constructed a prediction model using R software for predicting the median survival time, 6-months CSS, 1-year CSS, and 2-years CSS (results are shown in Fig. 2). The results showed bone metastasis (yes score 7.7, no/unknown score 0), PTR (yes score 0, no/unknown score 7.3), chemotherapy (yes score 0, no/unknown score 10) and radiotherapy (yes score 0, no/unknown score 6.2). As is seen in the nomograph, the higher the total score is, the lower the CSS and the median survival time are. We got the total score by summing the scores of each factor and then referring to the prediction model to get the more accurate median survival time and CSS rate (Fig. 2).

Calibration and Validation of the Nomogram

The nomograph model was further calibrated and validated. The C-index in training and validation cohorts were 0.685 and 0.709, respectively. This showed that the nomograph model matched well with the actual situation, which then suggested that the model had good discrimination. Figures 3 and Figs. 4 shows the calibration curves of the nomogram at 6 months CSS, 1-year CSS and 2-years CSS, which confirms the high consistency between the predicted CSS and actual CSS in training and validation groups. In conclusion, the outcome indicated that our constructed nomograms have identification and calibration abilities.

Summary of main results

When gallbladder tumors locally infiltrated into the liver, they were often combined with vascular invasion and extensive metastasis of regional lymph nodes, or even distant metastasis, with a high degree of malignancy and a rather poor prognosis. Improving the prognosis and prolonging the survival time of the confirmed cases had been a long-standing goal of biliary surgeons. The bleak prognosis of GCLM was mainly due to the inability to diagnose and manage it as early as possible. Although gallbladder cancer liver metastasis was classified as stage M1 in TNM staging, the prognosis of patients with tumors remains heterogeneous because of distinctions in age, metastatic organs, and cures. Nevertheless, to date, there’s no available prognostic nomograph for M1 stage in GCLM. In this research, we developed and validated a prognostic nomogram for confirmed cases with GCLM, using a large population data as the study population. The factors in the nomogram are available from clinical data. In addition, we verified the good performance of the nomograph through various validations. More importantly, it can also be greatly convenient to patients and clinicians.

In the research, the clinicopathological data of 727confirmed cases with gallbladder cancer liver metastasis from the US SEER database from 2010 to 2019 were retrospectively analyzed to four independent prognostic factors, including bone metastasis, surgery, chemo derive therapy, and radiotherapy, which affected the survival prognosis of patients with GCLM from a large sample and multicenter perspective. Based on these independent prognostic factors, the survival model of gallbladder cancer liver metastasis was established using R software, and the unique visualization of the model could help biliary surgeons visualize the contribution of every factor to the prognostic survival of confirmed cases. In addition, based on the model, the mean survival time, 6-month CSS, 1-year CSS and 2-year CSS could be calculated easily.

In the era of precision medicine treatment, the model is a practical and easy clinical prediction tool that fills the gap in the prognosis of confirmed cases suffering from gallbladder cancer liver metastasis. For specific clinical application, when the prognosis of a patient is evaluated according to the model, if the CSS of that patient is low and the prognosis is poor, as a clinician, a closer follow-up plan can be made for the patient and more aggressive treatment such as radiotherapy and chemotherapy can be administered.

Potential limitations in the study

(1) In the SEER database, there aren’t detailed information on documenting adjuvant chemotherapy and neoadjuvant therapy (including treatment regimens, cycles, etc.), which is likely to influence the precision of assessing prognosis.

(2) In the SEER database, the surgical procedure, the surgeon, and the pathologist for surgical specimen detection are not described in detail. This is because radical versus palliative surgery can have a significant influence on the survival prognosis of confirmed cases.

(3) In the SEER database, there isn’t information on postoperative treatments for confirmed cases, such as retreatment after tumor recurrence and metastasis, targeted therapy, etc., which are key factors affecting the survival prognosis of confirmed cases.

(4) The research is a retrospective study, and there may be some selection bias.

(5) The data used in the research were gotten from the SEER database with a North American population, and there may be bias when applied to the Asian population.

In conclusion, the research aimed to construct a prognostic prediction model for gallbladder cancer liver metastasis, and a more accurate and effective prediction nomograph was constructed by means of retrospectively analyzing the SEER database. Despite certain shortcomings, the study will provide research ideas to achieve more accurate prediction of prognosis and provide a basis for implementing accurate individualized cure and implementing prospective clinical studies. The research is a retrospective study with some bias, so a large sample, prospective, multicenter randomized controlled trial is needed to offer more precise data to improve and update this model.

In conclusion, we successfully established a nomograph model to predict the median survival time and CSS of GCLM patients. Our nomogram model has high discrimination and good concordance, and has excellent clinical application value.

Acknowledgments

This work was supported by grants from the Fujian Provincial Clinical Medical Research Center for First Aid and Rehabilitation in Orthopaedic Trauma（2020Y2014）.

Conflict of the interest statement

We declare that we do not have any commercial or associative interest that represents a conflict of interest in connection with the work submitted.

Huang J, Patel HK, Boakye D, et al. Worldwide distribution, associated factors, and trends of gallbladder cancer: A global country-level analysis. Cancer Letters. 2021;521:238–251. doi:10.1016/j.canlet.2021.09.004
Sharma A, Sharma KL, Gupta A, Yadav A, Kumar A. Gallbladder cancer epidemiology, pathogenesis and molecular genetics: Recent update. WJG. 2017;23(22):3978. doi:10.3748/wjg.v23.i22.3978
Giannis D, Cerullo M, Moris D, et al. Validation of the 8th Edition American Joint Commission on Cancer (AJCC) Gallbladder Cancer Staging System: Prognostic Discrimination and Identification of Key Predictive Factors. Cancers. 2021;13(3):547. doi:10.3390/cancers13030547
Wang J, Bo X, Shi X, et al. Modified staging classification of gallbladder carcinoma on the basis of the 8th edition of the American Joint Commission on Cancer (AJCC) staging system. European Journal of Surgical Oncology. 2020;46(4):527–533. doi:10.1016/j.ejso.2019.10.015
Cai YL, Lin YX, Jiang LS, Ye H, Li FY, Cheng NS. A Novel Nomogram Predicting Distant Metastasis in T1 and T2 Gallbladder Cancer: A SEER-based Study. Int J Med Sci. 2020;17(12):1704–1712. doi:10.7150/ijms.47073
Xu X, He M, Wang H, Zhan M, Yang L. Development and validation of a prognostic nomogram for gallbladder cancer patients after surgery. BMC Gastroenterol. 2022;22(1):200. doi:10.1186/s12876-022-02281-2
Chen M, Li S, Topatana W, et al. Development and Validation of a Nomogram for Predicting Survival in Gallbladder Cancer Patients with Recurrence After Surgery. Front Oncol. 2021;10:537789. doi:10.3389/fonc.2020.537789

Tables 1 and 2 are available in the Supplementary Files section.

Table1.pdf
Table 1 Characteristics of the training set and validation set.
Table2.pdf
Table 2 Univariate and Multivariate analysis of prognosis in patients

Download PDF

Version 1

posted

You are reading this latest preprint version

Construction and Validation of the Predictive Model for Gallbladder Cancer Liver Metastasis Patients: A SEER-based Study

Status:

Version 1

Abstract

Background

Methods

Results

Conclusions

Figures

Introduction

Materials And Methods

Patient selection

Data extraction

Statistical analysis

Results

Screening out prognostic factors

Construction of the nomogram predicting model in the training set

Calibration and Validation of the Nomogram

Discussion

Summary of main results

Potential limitations in the study

Conclusions

Declarations

Acknowledgments

Conflict of the interest statement

References

tables

Supplementary Files

Status:

Version 1