Bismuth’s classification was originally proposed to catalog postoperative strictures and to stratify their treatment. Therefore, the emphasis was placed on the length of healthy distal bile duct mucosa proximal to the injury [18, 19]. A level 6 was added in to the original 5 levels of stricture in 2001  to indicate isolated right duct or right branch strictures. Of note, the level of stricture (lower limit) does not always correspond to the level of injury  because of potential initial ischemic or thermal damage as well as the shortening that often accompanies upstream dilatation of the duct. This is why the initial site of the injury is indicated to define the injury level in the EAES classification. However, it is acknowledged that the length of stricture has direct therapeutic implications , and this parameter is found in the length of “loss of substance.” The Bismuth classification does not account for acute injuries (e.g., transection) or leaks; there is no mention of vasculobiliary involvement.
The classification of Strasberg et al.  was specifically devised for laparoscopic injuries. It is comprehensive in that it includes a wide spectrum of injuries, including intra- and extrahepatic injuries. Strasberg et al. distinguished between occlusions (types B and E), occlusions leading to biliary obstruction, and divisions without occlusion (types A, C, and D), leading to leaks and bilomas. They also classified isolated occlusion of the right hepatic duct (missing from the Bismuth classification). Although the authors made the distinction between complete (type C) and partial division (type D), partial occlusion is missing. The classification insists on mechanisms and technical errors. It also includes the location (integrating the Bismuth classification). However, partial division can be located on the CBD, common hepatic duct (CHD), left hepatic duct, or right hepatic duct (RHD), and this anatomic distinction cannot be categorized. Also, included in type A injuries are cystic duct leaks that occur as a result of physical injury as well as insecure clips or ligatures, leading to the question of whether this latter mechanism truly represents a bile duct injury or, as stated by Connor and Garden , a biliary complication. Of note, type B and C injuries occur mainly when an aberrant right duct is mistaken for the cystic duct when the latter runs directly into the RHD (rather than the CHD), an anomaly that involves between 2 %  and 28 %  of the population. Types A, B, and C can occur in almost any location along the biliary tree, not only in an aberrant RHD. Strasberg et al. suggested that it would be interesting to subclassify injury according to whether there was a loss of substance and the length of the loss of substance; according to injury with devascularization but without division; and according to concomitant injury to the right hepatic artery (RHA) (in types E1 and E2). However, they did not include these injury profiles in their classification. Theoretically, types B, C, and D involve the main bile duct but do not indicate the level of injury. The varieties e, d, and f in type E lesions are in fact redundant with types B and C. Last, type E categorizes occlusion but not division of the bile ducts above the superior confluence. Adequate and complete use of this classification should therefore include a letter A to D for the type plus, an E number, for the location, which is rarely seen in the literature. All these variations were included in our classification.
McMahon et al.  divided lesions according the extent of damage to the main bile duct, including laceration, transection, or excision, or, later, as stricture, leading to a simple two-class separation, minor and major. Minor includes <25 % of circumference of the CBD, laceration of the cystic CBD junction, or both. Major includes >25 % of circumference, going all the way to full transection, and involves the common bile or hepatic ducts. This classification has several negative aspects. For example, minor injuries are described for the CBD only, while major injury involves the CBD and CHD, without mention of injury including or above the superior confluence. Another negative aspect is that the cutoff of 25 % is arbitrary: minor injury can encompass a variety of injuries (Strasberg types A, B, C, and D) and ultimately can result in stricture (a major injury). Both a lateral laceration of less than 25 % of the circumference and the laceration of the cystic CBD junction are in the same category, minor, whereas the same type of latter injury can involve more than 25 % of the circumference and thus be a major injury. Moreover, although the distinction had certain therapeutic implications at the time the classification was derived, today, a 30 % to 35 % circumference injury might still be quite easily repaired surgically or by stent insertion , whereas the treatment options are not the same when the injury involves 90–99 % of the circumference. However, in this classification, both types of injury are classed in the same “major” group. These authors made no distinction between division (laceration and transection) and occlusion, and they did not mention longitudinal loss of substance or vasculobiliary involvement. Similar to Strasberg et al. , and McMahon et al.  included the Bismuth classification to designate the length of proximal healthy bile duct. However, they describe a class 0, which is not mentioned in any other publication on the topic and which does indicate how to integrate “major/minor” or the anatomic level into the two-category classification other than by a full description.
The Amsterdam Academic Medical Center classification derives from two publications [17, 28]. Most of the diagnoses were made postoperatively, through endoscopic and/or percutaneous transhepatic cholangiography. Although mentioned and discussed, the classification itself does not indicate the location of the injury, except for A (cystic duct and aberrant or peripheral hepatic radicals); B I (main CBD); and B II (aberrant segmental extrahepatic branches). It is not clear what the authors mean by “minor” and “major”—whether this represents “not severe” and “severe,” injury to the NMBD and MBD, or partial (B) versus complete (D) divisions. Bile leaks from the cystic duct, aberrant or peripheral hepatic radicals, and minor bile duct lesions are lumped together in type A. Type D injuries include complete transection, but nothing in the classification allows us to determine whether there has been a loss of substance. The classification does not provide insight into the mechanism of injury. There is no distinction between complete transection with (the bile ducts will dilate) or without (the bile duct will leak) occlusion (ligature, clip). Last, there is no mention of vasculobiliary involvement.
The classification of Neuhaus et al.  was originally published in German; an English translation can be found in Schmidt et al. . This classification individualizes the nonmain BDI (type A), distinguishing between cystic duct (A1) and hepatic bed leaks (A2). However, it does not separate the physical cystic duct injuries from leaks due to slipped clips or ligatures. It incorporates vascular injuries, but these are not integrated into the figures, and there is no indication as to how to list them. The length of injury (type C) as well as loss of substance (“structural defect”) are included. However, the 5-mm cutoff value for length of injury is arbitrary, with no explanation of why this was chosen. The level of injury is not indicated for acute injuries, only for stenosis. The word stenosis is used to designate stricture, which might mean any narrowing, not necessarily one that is the result of a BDI. Last, the extent of circumferential damage is binominal (partial and complete division only, without any quantification).
The classification of Csendes et al. , which is based on the analysis of three varied circumstances of recruitment (retrospective multicentric, prospective monocentric, and referrals) as well as etiologic and anatomic considerations, is restrictive in that there is no indication of the anatomical level and it does not include injury above the superior confluence. Thus, it is difficult to distinguish between partial and complete type III injuries, and there is no distinction between occlusion and division. The classification brings forth some of the possible mechanisms, but energy-driven injuries are not separated from mechanical (scissors) injury, and there is no mention of vasculobiliary involvement. It takes into consideration the therapeutic consequences of longitudinal loss of substance (Davidoff injury) . Moreover, it implies that injury to the RHD can (always) be repaired during the index operation and that type I and II lesions can (always) be repaired by T-tube insertion.
The classification of Way et al.  was originally published in Annals of Surgery, but the Stewart et al. classification  is most often cited. Curiously, however, the two classifications differ somewhat because the subdivision of class III lesions is found only in the first article and is rarely used or cited. Subdivision of class III is based on the proximal extent of the injury as follows: class IIIa, remnant CHD; class IIIb, CHD transected at the bifurcation; class IIIc, bifurcation excised; and class IIId, proximal line of resection above the first bifurcation or of at least one of the lobar ducts. The classification is based on anatomy but also on the mechanism of the lesion. This classification also emphasizes the role played by vascular injury, but this is represented in class IV only, while this type of injury can occur in all four categories. The authors separate aberrant duct from RHD injuries. The subdivision of classes IIIa and IIIB differentiates between common BDI without and with loss of substance, but it does not describe the Davidoff lesion. Isolated injury (whether transection or resection) is described in the mechanism of injury but is not represented in the diagram. Moreover, this classification does not provide descriptors for simple bile leaks (from the cystic duct or the liver bed) or take into account the lesions evolving into bile duct leaks due to cautery or ischemic injuries. It does not describe late complications, such as strictures, or identify transections at or above the bifurcation (except the RHD in class IV), and it does not categorize injury to the right sectorial ducts. Last, the senior author published another classification in Wikisurgery  in which the author used the term type as opposed to class. Of note, types I and II are identical to classes I and II, but types III and IV are different from classes III and IV, which only adds to the confusion.
The Hanover classification [16, 27] has tried to combine several items included in other classifications and comes close to the goal of being all-inclusive, but some information is still missing. In particular, this classification does not distinguish between intraoperative and postoperative detection of the injury, or the mechanism of injury (except for the clip mechanism, type B). The authors used the term stenosis to designate an occluded duct; this might lead to confusion with stricture. Again, the 5-mm length of loss of substance was used [9, 29] without explanation of where this cutoff value came from. The letters used to designate the vascular injury are based on Latin and therefore might not be integrated easily—for example, d for dextra (right), s for sinistra (left), p for propria hepatic artery, and DHC for ductus hepatocholedochus. It is not clear what the authors mean by defect in type D, particularly whether this represents loss of substance; if this is so, then there is no clear difference between the two types of injury when they occur above the bifurcation. The level of injury is not indicated for acute injuries, only for “stenosis,” which once again is a possible source of confusion with “stricture.”
The classification of Lau et al. [5, 15], taking into account anatomic, etiologic, and vascular problems, mentions loss of substance but does not include the short-term cautery or ischemic injuries or the long-term septic consequences. It attempts to class BDI according to ascending order of severity as well as the date of appearance of the complication (early and late). However, it is not always true that vascular injury to the right hepatic artery (type V) is more severe than type IV. Moreover, vascular injuries are lumped together in one class. It does not distinguish between sectorial and main BDI. The level of acute injury is indicated as CBD, CHD, right/left hepatic, or sectorial duct.
Four other simple classifications were found, one by Siewert et al.  (reported in English by Weber et al. ) and another by Cannon et al. , both essentially based on economic considerations and severity; one by Sandha et al.  based on endoscopic findings; and one by Kapoor .
The Siewert classification [10, 30] divides acute BDI into two anatomic categories, peripheral and central (probably nonmain biliary tract and main biliary tract). Siewert and colleagues [10, 30] described four types of injury, noting that there was an increasing order of severity; surprisingly, late bile duct strictures, classified as type II, are found between the immediate biliary fistula group (type I) and the central lesions (types III and IV). This classification includes vascular injury, without any specific description for types III and IV. Derived from surgical series, it was used essentially to class injury before endoscopic-only treatment. Lesions that healed without consequences were apparently not recorded; this is perhaps one of the reasons that the incidence of BDI is probably higher than previously reported.
The classification of Cannon et al.  provides three grades of injury, I to III. Grade I consists of leaks from the cystic stump, duct of Luschka, or accessory right hepatic ducts; grade II includes all other levels of injury lumped together, without any clear distinction of level, from the CBD to the intrahepatic ducts; and grade III embraces all combined vascular and biliary injuries. Severity was indicated by an increased financial burden, referral decisions, and mortality (0, 1.4, and 15 %, respectively) as the grade went from I to III.
The goal of the endoscopic classification of Sandha et al.  was to distinguish between leaks discovered early or late in the postoperative period and according to their intensity. Low-grade (LG) leaks are those identified after opacification of intrahepatic radicals (extravasation of contrast material requires hyperpressure), and high-grade (HG) leaks are those detected before radicular opacification (spontaneous extravasation of contrast material). Although this classification was the result of identification by endoscopic retrograde cholangiopancreatography, it can also be applied to cholangiograms, with HG corresponding the spontaneous bile leaks and LG to those detected after opacification takes place. Other than this distinction, however, the classification is largely insufficient, including the lack of anatomic location, the absence of nonleaking lesions, the absence of vascular involvement, and bile volume dependence (drainage is not mandatory). It was therefore not included in the tables we present here.
Last, Kapoor , in a letter to the editor, published a classification similar to our own, in that letters pertaining to the type of injury were used (nominal), rather than a categorical sequence (ordinal or cardinal). However, the abbreviations Kapoor used did not always correspond to semantics; for instance, “By” was used for bile leak (“y” for “yes”), “Bn” for no bile leak, corresponding to ligation or clip, circumference involved (“Cf” for full circumference [transection or excision] or “Cp” for partial circumference [clip, cautery, hole, excision]) and duct injured (“Ds” for significant duct [CBD, CHD, RHD, right sectoral or segmental duct] and “Di” for insignificant duct [cystic duct, subsegmental duct, subvesical duct]). Missing is full-circumference occlusion. Vascular injury was included: the letter V is added when there is associated vascular injury. There was no clear indication as to how to describe the level of injury.
Two other partial classifications have their importance: one by Li et al. , which separates segmental from Luschka duct lesions, and the other by Connor and Garden , which added an E6 injury to the Strasberg classification to describe complete excision of the extrahepatic confluence.