Biological risk based on preoperative serum CA19‐9 and histological grade predicts prognosis and improves accuracy of classification in patients with pancreatic ductal adenocarcinoma

Abstract Background Carbohydrate antigen (CA) 19‐9 and histological grade can serve as indicators of the biological characteristics of pancreatic ductal adenocarcinoma (PDAC). Aims The aim of this study was to investigate the combined impact of preoperative CA19‐9 levels and histological grade on prognosis and classification accuracy in PDAC patients. Methods and results A retrospective cohort study was conducted on 612 patients with PDAC who underwent curative pancreatectomy, and a biological risk model based on preoperative CA19‐9 levels and histology grade was established. The prognostic importance of the biological risk model was evaluated, and its validity was confirmed through a validation cohort of 218 patients. The survival of patients with PDAC was independently associated with preoperative CA19‐9 levels and histology grade, indicating a biological risk. This biological risk was incorporated into the eighth edition of the TNM staging system, leading to the development of a modified TNM (mTNM) staging system. Receiver operating characteristic (ROC) curves demonstrated that the mTNM staging system had a significantly larger area under the curve (AUC) than the TNM staging system. The discriminatory capacity of the mTNM staging system was further validated in an independent cohort. Conclusion Biological risk based on preoperative CA19‐9 and histological grade could predict the survival of patients with PDAC. The incorporation of biological risk into the TNM staging system has the potential to enhance the accuracy of patient classification in PDAC, predicting patient survival and enabling the development of individualized treatment plans.

The American Joint Committee on Cancer (AJCC) tumornode-metastasis (TNM) staging system is the most commonly used classification for prognostic evaluation and treatment decisions in PDAC (Table 1). 4 The TNM classification has continuously improved to accurately reflect the current understanding of disease extent. 5,68][9] In contrast, several other studies have demonstrated deficiencies in survival differences between substages of the TNM classification and proposed a modified TNM or a newly developed staging system. 10,11Therefore, it may be necessary to combine the AJCC TNM stage with tumor biological characteristics to improve the classification of PDAC.
Tumor biology has a substantial impact on the prognosis of patients with malignancy and has been incorporated into the staging system to improve the accuracy of prognostic prediction. 12,13Since both tumor burden and malignant degree could reflect the characteristics of tumor biology, 14,15 we hypothesize that combining factors representing tumor burden and malignant degree would provide a comprehensive understanding of tumor biology in predicting cancer prognosis.Incorporating factors associated with tumor biology into the staging system could enhance the accuracy of TNM staging in predicting prognosis.Carbohydrate antigen 19-9 (CA19-9) is associated with tumor burden, 16 and histology grade is associated with the degree of malignancy. 17,18Therefore, the combination of preoperative serum CA19-9 and histological grade can accurately reflect the tumor biology of PDAC.In the present study, we retrospectively analyzed the data of 612 patients with PDAC who underwent curative resections to explore the effect of tumor biology based on preoperative serum CA19-9 levels and histology grade on prognosis and the accuracy of patient classification in PDAC.

| MATERIALS AND METHODS
This study was approved by the Ethical Review Committees of Tianjin Medical University Cancer Institute and Hospital and was conducted in accordance with the ethical guidelines of the Declaration of Helsinki.Informed consent was waived due to the retrospective nature of the study.

| Patients
This study was conducted using a training cohort and an independent validation cohort.The training cohort consisted of 612 patients with PDAC who underwent curative resection at the Department of Pancreatic Cancer, Tianjin Medical University Cancer Institute and Hospital between January 1, 2011, and December 31, 2018.As this study is a retrospective study, patients of any age were enrolled.Therefore, the new staging system is suitable for all patients regardless of their age range.All patients underwent either a partial or total pancreatectomy, with superior mesenteric vein or portal vein resection being performed in cases where vascular infiltration was suspected.Standard lymph node dissection was routinely carried out, excluding patients who accepted palliative surgery or exploratory laparotomy without resection as well as those with missing values of preoperative CA19-9 and histology tumor grade.Therefore, the inclusion criteria for this study comprised (1) patients diagnosed with PDAC, (2) patients who underwent pancreatectomy with curative intent, (3) patients without distant metastasis, (4) patients with no history of other malignancies, (5) patients who were not lost to follow-up and (6) those who did not die during their initial hospital stay or within one month Stage III T1,T2,T3 N2 M0

T4
Any N M0

Stage IV Any T Any N M1
Note: T, primary tumor; T0, no evidence of primary tumor; Tis, Carcinoma in situ; T1, tumor ≤2 cm in greatest dimension; T2, tumor >2 cm and ≤4 cm in greatest dimension; T3, tumor >4 cm in greatest dimension; T4, Tumor involves the celliac axis, superior mesenteric artery, and/or common hepatic artery, regardless of size; N, regional lymph nodes; N0, no regional lymph node metastasis; N1, metastasis in one to three regional lymph nodes; N2, metastasis in four or more regional lymph nodes; M, distant metastasis; M0, no distant metastasis; M1, distant metastasis.

| Evaluation of clinicopathological variables and survival
The clinicopathological variables studied included 14 factors: sex, age at surgery, preoperative serum levels of CA19-9, carcinoembryonic antigen (CEA), and CA242, tumor location, type of surgery, TNM stage, T stage, N stage, histology grade, lymphovascular invasion, perineural invasion and postoperative adjuvant chemotherapy.
Preoperative serum tumor markers (CA19-9, CEA, and CA242) were detected within one week prior to surgery.The normal upper limits of serum tumor markers were adopted as follows: CA19-9 (37.0 U/mL), CEA (5.0 ng/mL) and CA242 (20 U/mL).In this study, all patients with a total bilirubin level ≥ 250 mmol/L were treated with either percutaneous transhepatic biliary drainage or endoscopic retrograde biliary drainage.
After biliary drainage for approximately 1-2 weeks, the operation was performed if the total bilirubin level was less than 100 mmol/L, and the preoperative serum CA19-9 level was rechecked along with bilirubin.poorly differentiated or undifferentiated adenocarcinoma (Figure 1).
Lymphovascular infiltration (LVI) refers to the invasion of blood vessels and lymphatic channels, while perineural invasion is the infiltration of peripheral nerves by the primary tumor in close proximity.
Patients who received more than three cycles of postoperative adjuvant chemotherapy were classified as the chemotherapy group.

| Follow-up
The patients were reviewed monthly for 6 months following the surgery.Subsequently, they underwent follow-up evaluations every 3 months for up to 2 years post-surgery, then every 6 months for up to

| Statistical analysis
The receiver operating characteristic (ROC) curves were used to identify potential cut-off values of preoperative serum CA19-9 for Examples of differentiation at different grades in pancreatic ductal adenocarcinoma (PDAC) stained with H&E.
predicting survival.The ROC curves were also used to compare the area under the curves (AUC) of the modified TNM staging system and the eighth edition of the TNM staging system, demonstrating their respective advantages in predicting survival.Discriminant function analysis was used to calculate the coefficients for predicting biologic risk based on preoperative serum CA19-9 levels and histology grade.
The OS curves were calculated using the Kaplan-Meier method based on the time elapsed between primary surgical treatment and final follow-up or death.The log-rank test was used to assess significant differences between curves.Independent prognostic factors were identified by the COX proportional hazard regression model, which was used to measure homogeneity and discriminatory ability.A The optimal cut-off value for predicting the survival of pancreatic cancer was determined to be 112 U/mL, considering both maximum sensitivity and specificity.
bilateral p-value of less than 0.050 was considered statistically significant.The statistical analysis was performed using IBM SPSS 23.0 and MedCalc v.19.6.1.

| Clinicopathological characteristics of the whole cohort
The baseline characteristics of the training cohort are listed in

| Survival analysis of patients with PDAC
The ROC curves demonstrated that preoperative serum CA19-9 could predict the prognosis of patients with PDAC (AUC = 0.576, 95% CI: 0.535-0.615,p = 0.003) (Figure 2).The optimal cutoff point for preoperative serum CA19-9 to predict survival was determined to be 112 U/mL.Then, the patients were categorized into two groups based on the optimal cutoff value of preoperative serum CA19-9: Group 1, in which the level of preoperative serum CA19-9 was ≤112 U/mL; and Group 2, in which the level of preoperative serum CA19-9 was >112 U/mL.
The results of the univariate and multivariate survival analyses are presented in Table 3.In the univariate analysis, the following 8 factors were evaluated and found to have a significant effect on survival: age at surgery (<70 years vs. ≥70 years), preoperative CA19-9 (≤112 U/mL vs. >112 U/mL), preoperative serum CEA (≤5 ng/mL vs. >5 ng/mL), postoperative adjuvant chemotherapy, TNM stage, histology grade (low grade vs. high grade), perineural invasion, and lymphovascular invasion.
In the multivariate analysis, preoperative serum CA19-9 (with a hazard ratio of 1.355 for CA19-9 levels >112 U/mL and p = 0.005) and histology grade (with a hazard ratio of 1.545 for high grade and p < 0.001) were identified as independent prognostic factors for overall survival, along with age at surgery (≥70 years), TNM stage, and postoperative adjuvant chemotherapy.

| Development of biological risk model
Initially, all patients were categorized into four groups based on their There was no significant difference in survival between Group 2 and Group 3 (5-year OS: 23.0% vs. 18.6%, p = 0.805) (Figure 4A).
As the survival coefficients of À1 and 1 indicate no difference in survival prediction, we divided the coefficient range into three intervals: <À1,  4B).In the multivariate analysis, the BR model was found to be an independent prognostic factor for OS (HR was 1.418 for the middle BR group, p = 0.016; and HR was 2.221 for the high BR group, p < 0.001).(Table 4).

| Incorporation of biological risk into the eighth edition of the AJCC TNM staging system
With the TNM-stratified analysis, there was no significant difference in survival between the low BR group and middle BR group in stage I, while the overall survival of these two groups was significantly better than that of the high BR group in stage I (Figure 5A).Among the patients in stage II, there were significant differences in survival between each of the two groups (Figure 5B).There was no significant difference in survival among the three groups of patients in stage III (Figure 5C).The OS of patients in the low BR and middle BR groups, who were at stage I, was similar to that of patients in the low BR group who were at stage II.The OS of patients in the high BR group, who were at stage I, was similar to that of patients in the middle BR group, who were at stage II.The OS of patients in the high BR group, who were at stage II, was similar to that of patients in stage III (Figure 5D).
Based on the results of the strata analysis, we incorporated BR into the eighth edition of the TNM staging system and introduced our modified TNM (mTNM) staging system (Table 5).In the mTNM staging system, the 5-year OS rates of stages mI, mII and mIII were 31.5%,17.2% and 7.2%, respectively (χ 2 = 78.603,p < 0.001).In the TNM staging system, the 5-year OS rates of stages I, II and III were 28.2%, 14.9% and 7.7%, respectively (χ 2 = 37.794, p < 0.001) (

| Predictive performance of the mTNM staging system
The prognostic prediction differences between the eighth edition of the TNM staging system and the mTNM classification system were directly compared.The À2 log likelihood of the mTNM stage was 4688.361, which was lower than that of the TNM staging system (4694.876),indicating a better performance in survival prediction (Table 6).The ROC curves showed that the AUC for mTNM staging was significantly higher than that for TNM staging [AUC: 0.646 (95% CI: 0.606-0.683)vs. 0.587 (0.547-0.626), p = 0.0007] (Figure 7).

| External validation
Ultimately, the proposed staging system underwent external validation.The characteristics of the validation cohort are presented in Table 7.Both the mTNM and TNM staging systems were successfully used to stratify OS into distinct subgroups, as illustrated in Figure 8.
However, survival differences between the subgroups of mTNM were more pronounced than those observed with the TNM subgroups.The 5-year overall survival rates were 35.6%, 9.0%, and 5.0% for stages mI, mII, and mIII, respectively ( p < 0.001 for mI vs. mII, p = 0.006 for mII vs. mIII).The corresponding rates for stages I, II, and III were 21.4%, 14.1%, and 4.8% ( p = 0.041 for I vs. II, p = 0.038 for II vs. III).

| DISCUSSION
Both tumor burden and malignant degree can reflect the biological characteristics of a tumor, and they are also associated with the survival of patients with malignancy.The AJCC TNM classification, which reflects tumor burden, is the most widely used staging system for malignancy.However, it is a macroscopic classification mainly based on anatomy and does not incorporate factors that reflect microscopic tumor burden (which cannot be detected by imaging), 16 nor does it consider factors that reflect the malignant degree of a tumor.As a result, there are always accuracy biases when using the TNM stage to predict prognosis. 10,11In the present study, we aimed to incorporate preoperative serum CA19-9, which could compensate for the deficiency of TNM stage in evaluating microscopic tumor burden, and histology grade, which could make up for the defect of TNM stage in reflecting malignant degree, into the TNM staging system to reduce biases.
In  reported that both pretreatment and post-treatment levels of CA19-9, as well as their changes after treatment, have good prognostic value in determining the survival of pancreatic cancer patients. 19rrently, there is no established cutoff value for CA19-9 to predict the prognosis of PDAC [20][21][22] as its levels fluctuate widely.In our study, CA19-9 levels ranged from 0.6 to 178 850 U/mL.4][25][26] However, multiple studies have confirmed the superiority of CA19-9 over other serum tumor markers in predicting the prognosis of PDAC [27][28][29] which is consistent with our findings.
Histology grade is defined as the level of differentiation exhibited by a tumor, which refers to the degree of morphological and The impact of mTNM and TNM on the prognostic value of staging.functional similarity between a tumor cell and a normal cell from the same tissue.Malignant neoplasms typically progress from low grade to high grade.During the process of malignancy, tumors become more aggressive, and the tumor burden increases as the tumor grade rises. 30,31Our study also found that histology grade was significantly associated with the survival of patients with PDAC and could be used to modify the TNM staging system as demonstrated in many previous studies. 17,18 the best of our knowledge, this study is the first to categorize patients with PDAC according to their biological risk based on preoperative CA19-9 levels and histology grade and assess the impact of tumor biology on survival.In our opinion, the combination of preoperative serum CA19-9 levels and histology grade provides unique prognostic information that reflects microscopic tumor burden and malignancy degree, which are not included in the TNM staging system.Incorporating biological risk into the TNM staging system may classify patients with the same TNM stage into different subgroups, and it may also facilitate the design of individualized treatment regimens.
There are several limitations to the study.First, it was a retrospective study conducted at a single center.Second, patients included in the study were from 2011 to 2018; therefore, current neoadjuvant therapy was not routinely used.Additionally, we could not exclude patients who were negative for Lewis antigen, which may have an uncertain impact on accuracy.As the constraints listed above may introduce bias in the results, further validation is necessary through large-scale and prospective multicenter studies to confirm its superiority and applicability.F I G U R E 9 The receiver operating characteristic curve was used to compare the area under curves between mTNM and TNM in the validation cohort.AUC(mTNM) was 0.681, AUC(TNM) was 0.585, with a p value of 0.0041.
postsurgery.The validation cohort included 218 patients from the First Affiliated Hospital of Hainan Medical University and Shanxi Provincial People's Hospital between January 1, 2015, and December 31, 2018, who met the same criteria.
The pathological diagnosis was established by two professional pathologists.Tumors were staged according to the 8th edition of AJCC TNM staging of pancreatic cancer.Independent of tumor stage, histological tumor grade is an important predictor of disease outcome with higher grade tumors behaving more aggressively.Grade is usually based on microscopic features, including nuclear features.The more closely the tumor resembles normal tissue, the lower the tumor grade and the less aggressive it will behave.Lower-grade tumors generally grow more slowly and are less likely to spread and metastasize than higher-grade tumors.The histology grade was dichotomized based on the degree of tumor differentiation: low grade, including well or moderately differentiated adenocarcinoma, and high grade, including

F I G U R E 4
Overall survival curves were plotted after curative resection.(A) The patients were divided into 4 groups based on the status of CA19-9 and histology grade: group 1 included patients with CA19-9 ≤ 112 U/mL and low grade; group 2 included patients with CA19-9 ≤ 112 U/mL and high grade; group 3 included patients with CA19-9 > 112 U/mL and low grade; group 4 included patients with CA19-9 > 112 U/mL and high grade.(B) The patients were categorized into 3 groups according to the biological risk: the low-risk group consisted of patients with CA19-9 ≤ 112 U/mL and low grade; the middle-risk group consisted of patients with either CA19-9 ≤ 112 U/mL + high grade or CA19-9 > 112 U/mL + low grade; the high-risk group consisted of patients with both CA19-9 > 112 U/mL + high-grade.F I G U R E 3 Overall survival curves were analyzed after curative resection.(A) Patients were grouped according to their CA19-9 levels ( p < 0.001, log-rank test).(B) Patients were also grouped based on histology grade ( p < 0.001), with low grade indicating well and moderate differentiation, and high grade indicating poor differentiation and undifferentiation.

F I G U R E 5
Overall survival curves stratified by TNM stage showed that the survival differences were only observed in patients with stage I and stage II pancreatic cancer.(A) Stage I. (B) Stage II.(C) Stage III.(D) Overall survival curves of pancreatic cancer patients according to TNM stage and biological risk.T A B L E 5 PDAC patients were divided into three groups according to the biological risk of PDAC: Low, CA199 ≤ 112 U/mL + low grade; Middle, CA19-9 ≤ 112 U/mL + high grade or CA19-9 > 112 U/mL + low grade; High, CA19-9 > 112 U/mL + high grade.The biological risk was incorporated into the eighth edition of TNM staging system and composed the newly modified TNM (mTNM) stage.
is an independent prognostic factor for PDAC.After incorporating the BR model into the eighth edition of the TNM staging system, we established the mTNM staging system, which was confirmed to be more accurate in predicting postoperative prognosis for PDAC patients than the eighth edition of the AJCC TNM staging system.CA19-9 is the most widely used serum tumor marker.Kang et al.

F I G U R E 6
There were significant differences in overall survival (OS) curves between patients with different mTNM or TNM stages of pancreatic cancer in the training cohort.(A) mTNM stage.(B) TNM stage.

Table 2 .
Out of the 264 patients with low histology grade, 23 were well differentiated and 241 were moderately differentiated.The remaining 348 poorly differentiated patients were classified as having a high histology grade.The median overall survival (OS) of the training cohort was 18 months (range: 6-126 months), with survival rates of 71.3%, T A B L E 3 Univariate and multivariate survival analysis of all the patients with PDAC.

Table 6 ,
Figure Multivariate survival analyses of all PDAC patients after including biological risk.