Reduced and Normalized Carbohydrate Antigen 19-9 Concentrations after Neoadjuvant Chemotherapy Have Comparable Prognostic Performance in Patients with Borderline Resectable and Locally Advanced Pancreatic Cancer

Background: The association between optimal carbohydrate antigen (CA) 19-9 concentration after neoadjuvant chemotherapy (NACT) and prognosis has not been confirmed in patients with borderline resectable (BRPC) and locally advanced pancreatic cancer (LAPC). Methods: This retrospective study included 122 patients with BRPC and 103 with LAPC who underwent surgery after NACT between 2012 and 2019 in a tertiary referral center. Prognostic models were established based on relative difference of the CA 19-9 (RDC), with their prognostic performance compared using C-index and Akaike information criterion (AIC). Results: CA 19-9 concentrations of 37–1000 U/mL before NACT showed prognostic significance in patients with BRPC and LAPC (hazard ratio [HR]: 0.262; 95% confidence interval [CI]: 0.092–0.748; p = 0.012). Prognostic models in this subgroup showed that RDC was independently prognostic of better overall survival (HR: 0.262; 95% CI: 0.093–0.739; p = 0.011) and recurrence free survival (HR: 0.299; 95% CI: 0.140–0.642; p = 0.002). The prognostic performances of RDC (C-index: 0.653; AIC: 227.243), normalization of CA 19-9 after NACT (C-index: 0.625; AIC: 230.897) and surgery (C-index: 0.613; AIC: 233.114) showed no significant differences. Conclusion: RDC was independently associated with better prognosis after NACT in patients with BRPC or LAPC. Decreased CA19-9 after NACT was a prognostic indicator of better survival and recurrence, as was normalization of CA 19-9 after both NACT and surgery.


Introduction
Pancreatic ductal adenocarcinoma (PDAC) is a rare gastrointestinal cancer, with patients having a dismal prognosis. Surgical treatment is the mainstay for curative treatment. However, only 15-20% of diagnosed patients have resectable disease, and only 30% have borderline resectable disease [1]. Radical surgery, including vascular reconstruction, has been reported as technically feasible, expanding surgical indications for PDAC [2]. Moreover, recent studies showed that patients with borderline resectable (BRPC) or locally advanced pancreatic cancer (LAPC) who underwent surgery after neoadjuvant chemotherapy (NACT) had better survival outcomes than those who underwent upfront surgery [3][4][5][6].
Although the optimal NACT regimen has not yet been determined, a recent meta-analysis found that FOLFIRINOX-based NACT yielded better oncologic outcomes than gemcitabine-based NACT, despite the former having greater toxicity [1,7]. The resection rate after NACT was 65.3%, with 57.4% of the patients who underwent surgery achieving R0 resection [7]. However, prognostic markers for responders to NACT have not yet been identified except circulating tumor cell or DNA [8].
Although several studies found that normalization of carbohydrate antigen (CA) 19-9 concentration is associated with better patient prognosis [9], 5-10% of patients with PDAC have normal CA  at diagnosis because of a Lewis-negative phenotype, and waiting until normalization of CA 19-9 is difficult in real-world practice [10,11]. This study investigated the ability of reduced CA 19-9 rather than normalized CA 19-9 after NACT to predict oncologic outcomes in patients with BRPC or LAPC. The present study also compared the prognostic ability of reduced and normalized CA 19-9 to evaluate response after NACT in patients with BRPC and LAPC.

Patients and Study Design
The present study included patients with BRPC and LAPC who underwent surgery following NACT at a tertiary referral center between July 2012 and August 2019. BRPC was defined as a tumor in contact with the common hepatic artery without extension to the celiac axis or hepatic artery bifurcation; a tumor in contact with ≤180 • of the circumference of the superior mesenteric artery; a tumor in contact with >180 • of the circumference of the superior mesenteric vein or portal vein; and a tumor in contact with ≤180 • of the circumference of either vein and with a contour irregularity or thrombosis of the vein but with possible reconstruction [12]. Patients who underwent upfront surgery were excluded.
NACT was administered based on each patient's general condition, and concurrent radiotherapy was not used routinely. The patients were evaluated by serial abdominal computed tomography (CT) and positron emission tomography (PET), and by measuring CA 19-9 concentrations during NACT. CT was evaluated using modified Response Evaluation Criteria in Solid Tumors [13]. Surgery after NACT was evaluated by a multidisciplinary team based on regressive or stable tumor with possibility of resectability of involved vessels. Pathologic response after surgery was reported using the College of American pathologist regression grading system [14]. After operation, the patients were administered chemotherapy except those with complete resolution of PDAC. Radiotherapy was used in the patients with R1 resection. The response to chemotherapy after surgery was evaluated every 3 months for 2 years by means of abdominal CT and tumor markers. Recurrence was diagnosed based on serial imaging studies with changing tumor markers and biopsy if possible. CA 19-9 concentrations were measured before and after NACT, and after surgery, with relative difference of the CA 19-9 (RDC) calculated as follows: [(CA19-9 after NACT) − (CA 19-9 before NACT)]/(CA 19-9 before NACT). The association between RDC and prognosis was investigated, and prognostic models were constructed for predicting overall survival (OS) and recurrence free survival (RFS). The abilities of normalized and reduced CA 19-9 concentrations to predict outcomes were compared.
Clinical data were obtained from patients' medical records. Recorded preoperative factors included age, sex, body mass index (BMI), American Society of Anesthesiologists (ASA) score, and imaging results before and after NACT, with tumor markers. Intraoperative factors included extent of resection, operation time, intraoperative transfusion, and estimated blood loss. Pathologic factors included tumor regression grade, node metastasis, the number of retrieved lymph nodes, and the presence of lympho-vascular or perineural invasion. Postoperative factors included length of hospital stay, postoperative complication based on Clavien-Dindo classification, 30-day mortality, recurrence, and survival. Informed consent was obtained from each patient before surgery. The study protocol was approved by the Institutional Review Board of Asan Medical Center (IRB No: 2018-1336).

Statistical Analyses
Continuous variables are reported as the mean (standard deviation) and are compared by Student's t-tests, and categorical variables are reported as numbers and percentages and are compared by χ 2 tests. Survival rates were estimated by the Kaplan-Meier method and compared by log-rank tests. A multivariable Cox proportional hazards model was used to identify factors prognostic of OS and RFS. These variables were selected based on their clinical significance and statistical significance in a univariate Cox model, with caution to avoid overfitting and to ensure generalizability. To compare three methods of parameterization of CA19-9 (RDC, normalization after NACT, and postoperative normalization of CA19-9), c-indices were calculated for the final Cox model: one with RDC, the same model with RDC replaced by post-NACT normalization of CA 19-9, and the same model with RDC replaced by postoperative normalization of CA19-9. To evaluate the statistical difference of these three c-indices and Akaike information criterion (AIC), their standard errors were determined using 500 bootstrap samples, and their p-values were calculated. All statistical analyses were performed using SPSS ® version 22.0 (SPSS Corp., Chicago, IL, USA) and R 3.5.1 (R Foundation for Statistical Computing, Vienna, Austria) software, with two-sided p-values <0.05 considered statistically significant.

Comparative Prognostic Performance of Reduced and Normalized CA 19-9 after NACT and after Surgery
We compared prognostic performance of model using factors related with prognosis in this study. Prognostic model for OS included adjacent vein resection, and intraoperative transfusion, Additionally, Model 1, 2, and 3 included RDC as a continuous variable, normalization of CA19-9 after NACT, and normalization of CA19-9 after surgery, respectively. Model 1, 2, and 3 had C-index values for OS of 0.653, 0.625, and 0.613, respectively. Although the C-index of model 1 was higher than those of models 2 and 3, the differences were not statistically significant (p = 0.904 and p = 0.680, respectively). The AIC values for OS of models 1, 2, and 3 were 227.243, 230.897, and 233.114, respectively, with no statistically significant differences between model 1 and models 2 (p = 0.896) and 3 (p = 0.912). Prognostic model for RFS included adjacent vein resection. Additionally, Model 1, 2, and 3 included RDC as a continuous variable, normalization of CA19-9 after NACT, and normalization of CA19-9 after surgery, respectively. The three models had C-index values for RFS of 0.604, 0.584, and 0.602, respectively, with no statistically significant differences between model 1 and models 2 (p = 0.812) and 3 (p = 0.592). The AIC values for RFS of models 1, 2, and 3 groups were 636.138, 640.246, and 638.247, respectively, with no statistically significant differences between model 1 and models 2 (p = 0.900) and 3 (p = 0.924). Thus, the prognostic performances of the three models for OS and RFS were similar (Table 5). Table 5. Prognostic performance of models that included decreases and normalization of carbohydrate antigen 19-9 concentration after neoadjuvant chemotherapy and surgery on overall survival and recurrence free survival.

Outcome
Prognostic

Discussion
This study showed that a decrease in CA19-9 concentration after NACT was an indicator of better prognosis in patients with BRPC or LAPC. Furthermore, comparisons of three prognostic models of reduced and normalized CA 19-9 after NACT, and of normalized CA 19-9 after surgery, showed that the three models were similarly predictive of OS and RFS.
CA19-9 is a Lewis blood group oligosaccharide, also called sialyl Lewis A antigen, which is synthesized by exocrine epithelial cells. It has shown a 70-90% predictive value for diagnosing pancreatic cancer [16]. However, elevated CA 19-9 has also been associated with other gastrointestinal tumors, as well as with biliary tract inflammation. Moreover, 5-10% of patients with PDAC are Lewis antigen negative, with normal CA 19-9 concentrations [11]. CA 19-9 concentration after NACT may be a biologic marker in patients with BRPC and LAPC because normalized or reduced CA 19-9 concentration after NACT has been reported to be an important prognostic marker of better OS and RFS [17]. Compared with patients with RDC ≤ 0.5, those with RDC > 0.5 experienced better survival and higher resectability after NACT, suggesting that early surgery may benefit rapid responders [18]. However, 83% of responders had RDC > 0.5 after NACT; of these, 24% had resectable PDAC, suggesting they were biologically good responders. Although normalization of CA19-9 after NACT was found to be more prognostic of survival outcomes than reduced CA 19-9, that study included patients with CA 19-9 >1000 U/mL, with this subgroup showing higher CA19-9 and a lower normalization rate after NACT than patients with CA19-9 <1000 U/mL [9]. In addition, the evaluation of the relationship between RDC and OS in that study also included patients with high CA 19-9 concentrations. The present study found that RDC after NACT affected patient prognosis. High preoperative CA 19-9 was shown to be associated with early recurrence and lower resectability rates [16,19,20]. Patients with CA19-9 >1000 U/mL were classified as having BRPC, with NACT recommended even in patients with resectable tumors [15]. In the present study, only 14% of patients with high CA 19-9 before NACT achieved normalization after NACT, with survival outcomes being poorer than in patients with CA19-9 <1000 U/mL before NACT, although the differences were not statistically significant. Other markers are required to evaluate tumor response in this subgroup. By contrast, evaluation of response to NACT using CA19-9 is inadequate for patients with CA19-9 <37 U/mL. These patients may have a Lewis-negative phenotype, suggesting that other markers, such as CEA and CA125, are needed to check their biologic status [21]. However, CA 19-9 concentrations were found to be elevated in patients with pancreatic cancer, despite 27.4% of these patients being Lewis antigen negative, suggesting that CA19-9 may be helpful in diagnosing pancreatic cancer in Lewis-negative patients, except in those with extremely low CA19-9 ≤5 U/mL [10,21].
This study found that RDC was an independent prognostic factor and that survival outcomes were better in good responders. Similarly, a previous study showed that RDC > 0.5 was an independent predictor of OS and that RDC > 0.9 was associated with pathologic complete regression [18]. The present study also showed similar prognostic performances of reduced and normalized CA 19-9 after NACT. That is, prognosis was similar in patients with higher RDC after NACT and in patients with normalized CA 19-9 after NACT or surgery.
The present study also found that the change of CA 19-9 was unable to predict the need for resection of adjacent vessels, R0 resection, or tumor regression grade. RDC was not a biologic marker predictive of curative resection after NACT. Similarly, normalization of CA 19-9 was not associated with a histopathologic response, with a negative predictive value of 28% [22]. Furthermore, radiologic response was not related to histologic response [23][24][25]. Additional studies are needed to identify biomarkers of resectability after NACT [26,27].
This study had several limitations, including its retrospective design, which may have resulted in potential selection bias. Furthermore, the relatively small number of patients was another limitation. However, this disease entity is rare, indicating a need for multicenter studies to evaluate larger patient populations.

Conclusions
RDC was independently prognostic of better OS and RFS rates in patients with CA19-9 concentrations of 37-1000 U/mL prior to NACT. Although normalization of CA19-9 after NACT is an indicator of good patient prognosis, its prognostic performance was comparable to a decrease in CA 19-9 during NACT.
Supplementary Materials: The following are available online at http://www.mdpi.com/2077-0383/9/5/1477/s1, Table S1: Characteristics of patients with relative difference of CA19-9 < 0 and ≥ 0, Table S2: Prognostic factors associated with overall survival in patients with borderline resectable pancreatic cancer and carbohydrate antigen 19-9 concentrations of 37-1000 U/ml before neoadjuvant chemotherapy (N = 73), Table S3: Prognostic factors associated with recurrence free survival in patients with borderline resectable pancreatic cancer and carbohydrate antigen 19-9 concentrations of 37-1000 U/ml before neoadjuvant chemotherapy (N = 73), Table S4: Prognostic performance of models that included decreases and normalization of carbohydrate antigen 19-9 concentration after neoadjuvant chemotherapy and surgery on overall survival and recurrence free survival for borderline resectable pancreatic cancer, Table S5: Prognostic factors associated with overall survival in patients with locally advanced pancreatic cancer and carbohydrate antigen 19-9 concentrations of 37-1000 U/ml before neoadjuvant chemotherapy (N = 60), Table S6: Prognostic factors associated with recurrence free survival in patients with locally advanced pancreatic cancer and carbohydrate antigen 19-9 concentrations of 37-1000 U/ml before neoadjuvant chemotherapy (N = 60), Table S7. Prognostic performance of models that included decreases and normalization of carbohydrate antigen 19-9 concentration after neoadjuvant chemotherapy and surgery on overall survival and recurrence free survival for locally advanced pancreatic cancer.