A nomogram for predicting the risk of common bile duct stones based on preoperative laboratory tests and ultrasonography

Objective: The objective of this study was to determine the predictive factors for common bile duct (CBD) stone and establish a nomogram model based on the preoperative laboratory tests and imaging findings. Methods: A total of 1701 patients who underwent laparoscopic cholecystectomy (LC) combined with common bile duct exploration (CBDE) for suspected choledochlithiasis from November 2014 to October 2020 were eligible for this analysis. All patients were divided into the training set (from November 2014 to November 2019, n=1,453) and validation set (from November 2019 to October 2020, n=248) based on the admission time. The predictive factors for CBD stone were determined by the univariate and multivariate logistic regression model. A nomogram model for predicting the presence of CBD stone was developed based on significant variables, and receiver operating characteristic (ROC) curve, calibration plot and decision curve analysis (DCA) were used to assess the predictive performance of the nomogram. Results: The results of multivariate logistic regression analysis demonstrated that multiple gallbladder stones (OR: 7.463, 95%CI: 5.437-10.243, P<0.001), the maximal diameter of CBD stone measured by preoperative ultrasonography (OR for 0.8-1.5 cm: 4.756, 95%CI: 3.513-6.438, P<0.001; OR for 1.5-2.0 cm: 9.597, 95%CI: 4.621-19.931, P<0.001; OR for >2.0 cm: 24.473, 95%CI: 2.809-213.207, P<0.001), preoperative GGT level (OR for 90-225 U/L: 2.828, 95%CI: 1.898-4.214, P<0.001; OR for 225-450 U/L: 9.994, 95%CI: 4.668-21.403, P<0.001; OR for >450 U/L: 12.535, 95%CI: 4.452-35.292, P<0.001) and above-mentioned variables. ROC curve showed that the C-index of the nomogram model for the training set and validation set was 0.875 (95% CI: 0.857-0.893) and 0.834 (95% CI: 0.784-0.883), which were better than that of MRCP for preoperative diagnosis of CBD stone. The calibration curve and DCA curve demonstrated that the nomogram model had a good clinical utility for predicting the presence of CBD stone . Conclusion: The nomogram based on preoperative laboratory tests and ultrasonography had an excellent predictive power for CBD stone, and it might provide useful information for making treatment strategies.


Introduction
It has been reported that the prevalence of secondary common bile duct (CBD) stone was approximately 15-18% in patients with gallbladder stones [1]. CBD stone could further result in hepatic injury, pain, cholangitis, and even acute obstructive cholangitis or biliary pancreatitis [2]. In the past, CBD stone was mainly diagnosed by endoscopic retrograde cholangiopancreatography (ERCP), but the diagnostic method had a relatively high risk of acute pancreatitis and hemorrhage [3]. On the other hand, laparoscopic cholecystectomy (LC) combined with endoscopic retrograde cholangiopancreatography (ERCP) or laparoscopic common bile duct exploration (LCBDE) had been performed to determine the presence of CBD stone during the operation [4,5]. LCBDE largely preserved the integrity of the common bile duct, and avoided unnecessary choledochotomy and preserved the function of the Oddi sphincter [6,7]. With the advancement in the diagnostic technique and equipment, a variety of non-invasive imaging examinations such as abdominal ultrasonography (US), magnetic resonance cholangiopanreatography (MRCP), endoscopic ultrasonography (EUS), and abdominal CT scan were widely applied to identify the presence of CBD stone before surgery.
To date, abdominal ultrasound usually was considered as the primary imaging examination followed by MRCP [8]. It has been shown that the diagnostic sensitivity and specificity of MRCP for CBD stone was more than 95%, with a high degree of accuracy [9,10]. However, MRCP examination was more expensive and its diagnostic accuracy was easily affected by several subjective conditions. Therefore, it is urgent to develop a simple, non-invasive and efficient tool to precisely predict the presence of CBD stone before surgery.
Recently, the nomogram model, whihc was a graphical computational scale, was proposed by many studies to predict the possibility of interest events [11,12]. In the present study, patients who underwent surigcal treatment for suspected choledocholithiasis were reviewed and predictive factors for CBD stone were analyzed. We aimed to construct a nomogram model for predicting the presence of CBD stone according to the preoperative laboratory test and imaging findings in order to help clinicians make therapeutic strategies.

Patient
In this study, a total of 1701 consecutive patients who underwent surigcal treatment for suspected choledocholithiasis at the department of general surgery, Beijing Friendship Hospital from November 2014 to October 2020 were analysed retrospectively. Eligibility criteria required: (1) All patietnts suffering from cholangitis had definitive imaging examinations, including MRCP and abdominal ultrasonography, before surgery; (2) complete laboratory findings at the time of first presentation to the hospital. The exclusion criteria were: (1) patients without bile duct exploration; (2) contraindications to laparoscopic surgery; (3) clinical data were incomplete.

Clinical information
The positive findings of imaging examination were defined as visible stones diagnosed by preoperative magnetic resonance cholangiopancreatograph (MRCP) or ultrasonography (US).
In our laboratory, ALT>40U/L, AST>35U/L, ALP>100U/L, GGT>45U/L, TB>17.1umol/L, DB>6.84umol/L were considered as abnormal value, respectively. The patients were divided into two groups according to the presence of CBD stone.

Statistical analysis
Statistical analysis was conducted with SPSS software (version 26) and R software (version 4.0.2). Continuous variables were presented as mean±SD and the differences between research groups were compared using a Student's t-test if they followed a normal distribution; otherwise, they were recorded as the medium (P25, P75) and were compared using Mann-Whitney U test.
Categorical variables were shown as frequency or percentage and were compared using chisquare test or Fisher's exact test. The cut-offpoint for CBDD measured by preoperative ultrasonography was determined by receiver operating characteristics (ROC) curve.
Univariate and multivariate logistic regression were used to determine the predictive factors for CBD stone, and relevant results were reported as odds ratio (OR) and its 95% confidence intervals (CIs).
In the present study, the nomogram model for predicting the presence of CBD stone before surgery was developed according to the regression coefficients estimated by multivariate logistic regression model.
The predictive performance of the nomogram model was assessed by the ROC curve.
Calibration curves were plotted to evaluate the association between predicted probabilities and actual observations. A two-sided P-value of <0.05 indicated a significantly statistical significance.

Univariate and multivariate analysis of predictive factor for CBD stone
Univariate analysis of predictive factor for CBD stone was performed in the training set, and the results revealed that the presence of CBD stone was significantly associated with multiple gallbladder stones (P<0.001), elevated levels of TB (P<0.001), DB (P<0.001), cholestatic liver enzymes (ALP and GGT, P<0.001) and hepatocellular liver enzymes (ALT and AST, P<0.001).
In addition, the presence of CBD stone was more likely to be detected by preoperative ultrasonography (P<0.001) and was associated with larger diameter of CBD stone measured by ultrasonography (P<0.001) (

Construction and validation of nomogram for the prediction of CBD stone
Next, we developed a nomogram model for the prediction of CBD stone according to the results of multivariate logistic regression analysis in the training set (Table 3 and Table 4). As shown in the Figure   and 57.6%, which was better than that of MRCP for preoperative diagnosis of CBD stone ( Fig.   2A). Meanwhile, the calibration curves showed a good consistency between predicted and observed outcomes in the training set and validation set ( Fig. 2B and C). In addition, the decision curve analyses (DCA) indicated that the clinical applicability of the nomogram model was comparable to that of MRCP for the diagnosis of CBD stone (Fig. 2D).

Discussion
In 2010 [14]. However, several studies showed unsatisfactory diagnostic accuracy of ASGE guideline for predicting CBD stones [15,16]. MRCP is widely accepted as a non-invasive diagnostic method with high sensitivity and specificity, but it is time-consuming, relatively expensive, and its diagnostic accuracy mainly depends on the image quality and radiologist's experience.
Therefore, to develop a simple, non-invasive and efficient model for predicting the presence of CBD stone was useful to make a rapid diagnosis and guide therapeutic strategies before surgery.
The nomogram enables the assessment of the probability of clinical events through individual patient characteristics, and its visual and iconic results provide a more intuitive representation for risk assessment of disease than other predictive models. In this study, we used the univariate and multivariate logistic regression analysis to determine the predictive factors for CBD stone. Our results revealed that multiple gallbladder stones, the maximal diameter of CBD stone measured by preoperative ultrasonography, preoperative GGT level and DB/TB ratio were independent predictors of CBD stone, which had been reported by previous studies [17][18][19].

Ethical approval
The ethical approval for the present study was obtained from the Beijing Friendship Hospital ethics committee, and the informed consent was obtained from all participants.

Availability of data and materials
The datasets used during the current study are available from the corresponding author on reasonable request

Consent for publication
Not applicable

Conflict of interest
The authors declare no conflict of interest        Fig. 2 (A)ROC curve based on the nomogram compared with MRCP. Calibration curve based on the nomogram model in the training set(B) and the validation set(C), the dashed line represents the predictive performance of the nomogram, and the 45 straight lines represent the perfect match between the actual (y-axis) and the predicted probability of the nomogram (x-axis). The closer the distance between the two curves, the higher the prediction accuracy of the nomogram; (D)Decision curve analysis for MRCP-reported and the prediction nomogram in predicting CBD stone. The y-axis measures the net benefit. The net benefit is calculated by subtracting the proportion of all false positives and adding the proportion of true positives to weight the relative harms of forgoing treatment against the harms of unnecessary treatment [9]. The gray line represents the assumption that all patients have CBD stone. The black line represents the assumption that no patients have CBD stone. Using the prediction nomogram (red curve) to predict CBD stone adds more benefit for patients than predicting by MRCP.