Preoperative CA19-9 and GGT ratio as a prognostic indicator in ampullary carcinoma

This study retrospectively analyzed 162 patients with ampullary carcinoma who underwent radical pancreaticoduodenectomy in the Fourth Affiliated Hospital of Hebei Medical University from January 2011 to April 2021. Inclusion criteria: (1) Preoperative evaluation reached the indication of radical resection, and radical pancreaticoduodenectomy was performed; (2) Postoperative pathological diagnosis was ampullary carcinoma; exclusion criteria: (i) Patients died within one month after operation; (ii) Patients who died of non-ampullary cancer; (iii) the clinical and follow-up data of the patient were missing. According to electronic medical records, the general characteristics, perioperative characteristics, preoperative laboratory indexes, postoperative complications and pathological examination of patients were recorded, and many predictive parameters that may affect long-term survival were counted, such as CA19-9/GGT, PLR, NLR, Albumin to Alkaline Phosphatase Ratio (AAPR), Glucose to Lymphocyte Ratio (GLR), Gamma-glutamyltransferase to Platelet Ratio (GPR) and albumin to globulin ratio (AGR). At the same time, we use the AJCC 8th edition TNM staging system as a reference for clinical pathological staging [21].

Patients were followed up regularly after surgery and Postoperative follow-up included clinical and laboratory examinations. Regular follow-up once a month within 6 months after operation; If there is no recurrence within 6 months, follow-up every 3 months; Patients without recurrence in 3 years were followed up every 6 months. The time interval between the operation date and the first recurrence date or the last follow-up date was defined as RFS. The OS interval is from the date of surgery to the date of death or the last follow-up. The end point of follow-up was May 2022. All patients or their families were informed consent.

Continuous variables are expressed as mean ± standard deviation; The receiver operating characteristic (ROC) curve was used to determine the optimal cut-off value of routine blood indexes and prediction parameters; Chi-square test, Fisher's exact test and Mann–Whitney U test were used to compare the clinicopathological features of the two groups. In the comparison between groups of measurement data, independent sample t test is used for normal distribution, and rank sum test is used for non-normal distribution. χ² test or Mann–Whitney U test was used for the comparison between groups of enumeration data. Kaplan–Meier method was used to calculate the overall survival curve and non-recurrence survival curve; comparison of OS and RFS between CA19-9/GGT subgroups by log-rank test; The association of relevant variables with OS and RFS was assessed using Cox regression model; P < 0.05 indicated that the difference was statistically significant; the predictive ability of prognostic indicators was evaluated and compared by C-index. Nomogram was drawn by combining AJCC 8th TNM stage with CA19-9/GGT in R. C-index, calibration curve and Decision Curve Analysis (DCA) were used to compare and evaluate Nomograms with the simple AJCC 8th edition TNM staging system. Statistical analysis were performed by the software statistical package for social sciences version 26.0 (SPSS, Chicago, IL), R project version 4.0.5 (http://​www.​r-project.​org/​) and Medcalc software (version 15.2.2.0; Ostend, Belgium). The optimal cut-off values for prognostic indicators and inflammation-based scores were determined using X-tile3.6.1 software (Y ale University, New Haven, CT, USA). Time-dependent ROC was depicted using KM method via the survival ROC package in R. The nomogram was computed with the rms package in R.

During the study period, a total of 174 patients with AC underwent radical pancreatoduodenectomy, including 2 patients diagnosed with ampullary neuroendocrine tumors, 3 patients died of accidents, 1 patient died within 30 days after surgery, and 6 patients were excluded due to incomplete information or lost to follow-up, so a total of 162 patients with ampullary cancer were enrolled. Among 162 patients, 138 patients were found to have jaundice symptoms after the first examination after hospitalization. The chief complaints during interrogation included 67 cases of jaundice, 53 cases of upper abdominal discomfort, 11 cases of fever, 11 cases of loss of appetite, 8 cases of abnormal urine, 4 cases of nausea and anorexia, 5 cases of tumors found by examination, 1 case of waist and back discomfort, 1 case of skin pruritus, 1 case of emaciation. Among 138 jaundice patients, 38 patients received preoperative jaundice reduction treatment, including 12 patients received Endoscopic Retrograde Cholangiao-Pancreatography, Endoscopic sphincterotomy, Endoscopic Nasal Bile Drainage (ERCP + EST + ENBD) jaundice reduction treatment, and 26 patients received Percutaneous Transhepatic Cholangial Drainage (PTCD) jaundice reduction treatment. 87 patients (53.7%) received postoperative adjuvant therapy, but the specific therapeutic regimen was not known. The median follow-up time was 40 months [Inter-Quartile Range (IQR) = 21–58 months]. The 1-, 3-, 5-year OS and RFS were 87%, 60.5%, 44.1% and 67.9%, 54.3%, 40.6%, respectively. The specific clinicopathological features of the patients are detailed in Table 1.

CA19-9/GGT was obtained by the patient's first laboratory examination after hospitalization, and its optimal cut-off value was obtained using the receiver operating characteristic (ROC) curve, which was approximately 0.14 [area under the curve (AUC) = 0.839, 95% CI 0.779–0.899]. The corresponding sensitivity was 87.7%, and the specificity was 68.5%. Patients were divided into two groups by the optimal cut-off value of CA19-9 / GGT: low-risk group (CA19-9/GGT ≤ 0.14, n = 69) and high-risk group (CA19-9/GGT > 0.14, n = 93) (Table 2). Compared with the low-risk group, the high-risk group patients (CA19-9/GGT > 0.14) were presented with more postoperative adjuvant therapy (P = 0.025), lymph node metastasis (P = 0.036), and nerve invasion (P = 0.002); lower differentiation grade (P = 0.001); higher levels of total bilirubin (P = 0.011) and CA19-9 (P < 0.001) and more advanced TNM stage (P < 0.001).

Receiver operating characteristic (ROC) curve was used to calculate optimal cutoff values for predictors. Cox regression model was used for survival analysis of risk factors (Table 1). In univariate analysis, gender (P = 0.023), tumor size (P = 0.023), differentiation extent (P = 0.023), lymph node metastasis (P = 0.004), nerve invasion (P = 0.001), AJCC 8th edition TNM stage (P < 0.001), total bilirubin (P < 0.001), CA19-9 (P < 0.001), PLR (P = 0.027), GPR (P = 0.008), GLR (P = 0.004), AGR (P = 0.003), CA19-9/GGT (P < 0.001) were identified as risk factors for OS (Table 1); gender (P = 0.027), tumor size (P = 0.002), differentiation extent (P < 0.001), lymph node metastasis (P < 0.001), nerve invasion (P = 0.004), AJCC 8th edition TNM staging (P < 0.001), total bilirubin (P < 0.001), CA19-9 (P < 0.001), GPR (P = 0.011), GLR (P = 0.003), AGR (P = 0.005), CA19-9/GGT (P < 0.001) affected postoperative recurrence-free time (RFS) (Table 1).

In multivariate analysis, the independent risk factors affecting OS were CA19-9/GGT (P = 0.001, HR 2.459, 95% CI 1.450–4.167), medium/low differentiation (P < 0.001, HR 6.372, 95% CI 1.470–27.610), AGR (P < 0.001, HR 6.474, 95% CI 2.972–13.850), CA19-9 (P < 0.001, HR 3.401, 95% CI 1.953–5.922); the independent risk factors affecting RFS included CA19-9/GGT (P = 0.002, HR 2.333, 95% CI 1.371–3.971), medium / low differentiation (P < 0.001, HR 4.760, 95% CI 1.077–21.030), lymph node metastasis (P = 0.022, HR 1.762, 95% CI 1.089–2.849), CA19-9 (P < 0.001, HR 3.477, 95% CI 1.991–6.073), AGR (P < 0.001, HR 5.582, 95% CI 2.576–12.093). CA19-9/GGT > 0.14 was significantly correlated with poor prognosis of OS (P < 0.001) and RFS (P < 0.001) (Fig. 1).

CA19-9/GGT, other inflammation-based scores, routine laboratory parameters and AJCC 8th staging system were used to evaluate the predictive ability by the concordance index (C-index) (Table 3). Compared with other prediction parameters, CA19-9/GGT had the highest C-index for OS and RFS prediction, which was 0.753 ( 95% CI 0.748–0.838) and 0.745 ( 95% CI 0.737–0.827), respectively. In terms of prediction accuracy, CA19-9/GGT was significantly superior to other inflammatory scores and CA19-9 in OS and RFS prediction.

A new Nomogram prediction model was constructed by combining the AJCC 8th TNM staging system with CA19-9/GGT (Fig. 2). The C-index derived from the traditional AJCC 8th staging system and the Nomogram prediction model has been entered in Table 3. Compared with AJCC 8th TNM staging system (OS: C-index = 0.759, 95%CI = 0.632–0.785; RFS: C-index = 0.772, 95%CI = 0.659–0.805), the C-index of Nomogram prediction model in OS (C-index = 0.787, 95%CI = 0.685–0.712) and RFS (C-index = 0.795, 95%CI = 0.696–0.713)elevated.

The calibration curve intuitively reflected that the 1-year, 3-year, and 5-year OS and RFS of the actual observations were in good consistency with the 1-year, 3-year, 5-year OS and RFS calculated by the Nomogram prediction model (Fig. 3). Decision curve analysis (DCA), as an assessment method reflecting the clinical net benefit of the predictive model, intuitively confirms that the Nomogram predictive model involved a wider range of threshold probabilities than the traditional AJCC 8th staging system, yielded better net benefit, thereby provided higher prediction accuracy (Fig. 4, Table 4).