Selection of the flexible choledochoscope was made between one of the medium size (10 Fr, 3.3 mm) and the other of the large size (15 Fr, 5.5 mm). The small size choledochoscopes are not proper for this method because of inefficient vision, inadequate lavage facilities and the need for large baskets for sizable stone pick up.

The rest of the equipment included 30o laparoscope, curved pair of scissors, knife, a Semm forceps, irrigation-suction device, and a pair of laparoscopic needle holders (alternatively an endostitch apparatus). CBD suturing was made by 4/0 coated vicryl or 4/0 polysorb threads.

In addition to the classical sites of port placement in laparoscopic cholecystectomy, a fifth 11 mm port was put at right of the midline (in the immediate proximity) to the xiphoid for the insertion of the choledochoscope and cholan-giography was performed.

It was important that the continuity of the cystic duct with both the CBD and the gall bladder was maintained since this enables fixation of the CBD by lateral and upward traction of the gall bladder. After that, we were able to proceed to the anterior dissection of the supraduodenal segment of the CBD in about a 2-cm long and 1-cm wide area, while the duodenal bulb was mobilized downward. Mobilization of hepatic artery or its right branch is necessary if these entities cross the bile duct in this region.

Choledochotomy procedure: the opening should not exceed 1.0 cm at first instance. External massage and manipulations of the lower bile duct is surprisingly efficacious and should always be tried as the first measure. After that, a biliary balloon catheter was inserted through the subxiphoid port into the bile duct. Blind biliary balloon stone dislodgment was successful in some of our patients. Smaller stones can also be removed using the sucker tip.

Visually guided extraction with the flexible choledochoscope by wire baskets or long balloon catheters of Fogarty type was successful in the vast majority of our patients. Upon completion of the procedure, the entire biliary tract (proximal and distal) was inspected and shown to be free of calculi and debris and a cholangiogram was also done to ensure complete duct clearance.

Drainage of the extrahepatic biliary tree was done through a T-tube insertion or through the cystic duct stump with primary closure of the bile duct. Closure is accomplished with interrupted 4/0 absorbable sutures. After cholecystectomy, a subhepatic drain was inserted.

Cholangiography was performed on the 8^th postoperative day and the T-tube or the cystic duct cannula was removed 2 d later.

Flexible choledochoscope must have an outside diameter of no larger than 3 mm.

Not infrequently, the duct was cleared before the scope had been completely prepared by noncholedochoscopic maneuvers such as pharmacologic relaxation of the sphincter of Oddi, administration of 1-2 mg of intravenous glucagon, flushing of the cholangiocatheters with saline, balloon and helical or straight basket techniques.

Three tools are available to dilate the cystic duct for choledochoscope entry: curved forceps, mechanical graduated dilators and pneumatic high pressure balloon dilators. The choledochoscope was inserted through the midclavicular port into the cystic duct. Choledochoscopic stone retrieval was most commonly achieved using a basket inserted through the channel of the scope.

Balloon-tipped catheters were also used. These catheters generally do not fit in the working channel of the newer small-diameter scopes, so they must be inserted adjacent to the scope. In cases that scope or stone passage was not feasible, the cystic duct wall was cut longitudinally towards the common duct. If the incision included or abut the cystic duct-CBD junction, some stitches were sewn. Most often this was accomplished without the placement of a T-tube unless there was a good reason for doing so.

Patients were followed every 3 mo by phone and every 6 mo by personal conference and if mandatory, laboratory tests and ultrasonography were prescribed. There were no violations of the protocol.

The corresponding proportion and total number of patients were expressed for each clinical predictor. When distributed normally, continuous variables were described using mean and standard deviation, if not they were expressed as median and range. All parameters were analyzed as categorical variables.

A sensitivity analysis of all possible predictors was conducted through an ST Neural Network. The prognostic estimation of a biochemical test alone to differentiate between the absence and presence of choledocholithiasis was assessed using receiver operating characteristics (ROC) curve analysis. For each test, one cut-off value was chosen by determining the point closest to an ideal test with 100% specificity and sensitivity (the upper most left corner of the graph).

Discriminant forward stepwise analysis, using the ideal cut-offs determined at ROC curve analysis, was employed to retrospective data to establish the model that could best predict the presence of a CBD stone. Positive and negative predictive values were determined for all predictors. Prospective validation of the model that predicts choledocholithiasis was done by VGSR. In addition, the parameter coefficients of the initial model were compared with those obtained from the data of the prospective study.

Statistical analysis was made using the Statistica (StatSoft) software. Results with P values less than 0.05 were considered to be statistically significant.

Eight potential prognostic factors (positive history, past and present; abnormal values of LDH; amylase; SGOT; dBil; ALP and CBD diameter on ultrasound) were considered for evaluation as predictors of choledocholithiasis. A sensitivity analysis was conducted through an ST Neural Network. Variables pruned because of training, and verification ratios below the threshold were: past history, amylase and lactate dyhydrogenase, and were removed from the underdevelopment model of predictors. The “rank” on sensitivity datasheet was CBD diameter = 1, ALP =2 , dBil = 3, SGOT = 4. The best network found had good performance: correct classification rate 0.894737 and area under ROC curve 0.946328. ROC curve was used to select an optimum decision threshold (value that equalizes the probability of this classification of either class of false positive and false negative). Diameter of the supraduodenal CBD documented by ultrasonography was categorized into two cut-off values: >6 and ≥10 cm. The ideal cut-off was 10 mm. History was categorized as positive (presence of fever, jaundice, chills, pancreatitis) or negative. Categorical classification of liver enzymes was done by two values: simply elevated and equal or above twice normal. The ideal cut-off value was twice as normal.

In 42 patients with choledocholithiasis, positive histories were present in 36%, CBD diameter ≥10 mm on ultrasound in 95% and more or equal twice normal values of SGOT in 59%, dBil in 42%, and ALP in 50%. In the remaining 238 patients without evidence of CBD lithiasis in intraoperative cholangiography, positive histories were present in 5%, CBD diameter ≥10 mm in 15% and more or equal twice normal values of SGOT in 11%, dBil in 7% and ALP in 7%.

Discriminant forward stepwise analysis disclosed that high values (≥2×normal) of SGOT, ALP, conjugated bilirubin and CBD diameter on ultrasound ≥10 mm were all prognostic factors of CBD lithiasis in univariate and multivariate analysis at a level of P<0.01. History was not included in the model.

Positive predictive value (PPV), when considering all predictors was 93%, while the negative predictive value (NPV) was 95%.

From April 1997 to August 2000, 97 patients fulfilled the criteria of enrollment in the protocol as dictated by the preliminary retrospective study.

The following 19 patients were excluded from the study:

·Physical status ≥ASA III (seven patients)

·Did not sign the informed consent (six patients)

·Past procedures which should exclude laparoscopy (two patients)

·Recurrence of CBD lithiasis (one patient)

·Huge esophageal hernia (one patient)

·Von Willebrand disease (one patient)

·Choledochal cyst (one patient).

There were no statistically significant differences regarding demographic and clinicopathological characteristics, ultraso-nography and liver function tests among patients of both groups (Mann-Whitney U test was applied). Postoperative follow-up was made for 7-36 mo (median 22.36 mo).

Group A: Twenty-eight of thirty-six patients of group A had CBD lithiasis, 12 patients had direct CBD exploration and, the trancystic duct approach was applied in the remaining 16.

The application of direct CBD exploration was unsuccessful in three patients: there was conversion to open in one patient and the rest two patients were referred for endoscopic stone extraction. We had only one failure in the application of the trancystic duct approach and the patient was also referred for endoscopy.

The median duration of laparoscopic CBD exploration was 90 min (range 70-310 min). Median hospital stay was 7.4 d (5 d for trancystic duct approach and 11 d for direct CBD exploration).

Group B: Thirty-two of forty-two patients of group B had CBD lithiasis. In five patients, endoscopic stone extraction was unsuccessful and all were referred for open choledochotomy. Median hospital stay was 9 d. Duration of ERCP and ES in addition to the time spent for LC was 60-255 min (median 105 min).

The short-and long-term results from the application of the above-mentioned laparoscopic and endoscopic procedures are shown in Table 4.

Complications encountered among patients of both groups are depicted in Table 5.

Rank sum test disclosed that there was not a statistically significant difference among patients of both groups concerning morbidity (P = 0.87), mortality (P = 0.70) and hospital stay (P = 0.07). On the contrary, there was statistically significant difference regarding hospital stay in favor of patients that underwent the trancystic duct approach (P<0.01).

To compare long-term results, we applied the Gehan’s Wilcoxon test but a statistically significant difference was not found (P = 0.70).

In order to validate prospectively the multivariate model, we used the VGSR module to build a model for our design with multiple degrees of freedom effects for categorical predictor variables.

The best model for predicting bile duct stones included the following independent predictors: CBD diameter ≥10 mm (dichotomized, using a cut-off value of 10 mm) and elevated values (≥2 normal) of SGOT, dBil, ALP (dichotomized using a cut-off value of equal to twice normal).

All the above were predictors on univariate analysis even after using the Bonferroni-Holm method to adjust for multiple comparisons. Level of significance in univariate analysis for CBD diameter ≥ 10 mm, and elevated values (≥2 normal) of SGOT, dBil, ALP was P<0.01 for each of them. Table 6 shows the performance of each possible predictor of CBD lithiasis analyzed including positive and NPVs. Table 7 depicts the model building via forward stepwise regression.

There were no significant differences between observed and predicted probabilities in the subgroups of patients when broken down according to the different possible predictor combinations.

The respective coefficients for each of the predictors were: CBD diameter ≥10 mm (0.314982024), SGOT (0.227198578), dBil (0.244231286), ALP (0.207177395).

Prediction equation for choledocholithiasis was: CBD lithiasis = -0.03912693+0.314982024×CBD ≥10 mm+ 0.227198578×SGOT+0.244231286×BIL+0.207177395×ALP in which CBD diameter ≥ 10 = 1 if yes or = 2 and if not, values of SGOT, dBil, ALP = 1 if ≥2 normal or 2 if lower. Presence of CBD stone in a given case, is to be expected if the result of the equation is ≤1.5, the opposite is true if the result is >1.5. Consequently, parameter coefficients of the model did not vary significantly when retrospective and prospective data were combined. PPV, when considering all these predictors, was 93.3%, while the NPV was 88.8%. Sensitivity of the model was 96.5% and specificity 80%.