Professional Documents
Culture Documents
Chisholm 2018
Chisholm 2018
Background and Aims: Markedly increased liver chemistries in patients presenting with acute calculous chole-
cystitis (AC) often prompt an evaluation for concomitant choledocholithiasis (CDL). However, current guidelines
directing the workup for CDL fail to address this unique population. The aims of this study are to define the range
of presenting laboratory values and imaging findings in AC, develop a model to predict the presence of concur-
rent CDL, and develop a management algorithm that can be easily applied on presentation.
Methods: We conducted a retrospective review of patients presenting with AC to a large tertiary hospital over a
3.5-year period. CDL was defined as common bile duct (CBD) stone(s), sludge, or debris seen on any of the
following studies: US, CT, magnetic resonance imaging/MRCP, EUS, ERCP, or intraoperative cholangiogram. A
multivariable model to predict CDL was developed on 70% of the patients and validated on the remaining 30%.
Results: A total of 366 patients were identified and 65 (17.8%) had concurrent CDL. Univariable analysis was used
to predict CDL and demonstrated statistically significant odds ratios for transaminases >3 times the upper limit of
normal, alkaline phosphatase (AlkPhos) above normal, lipase >3 times the upper limit of normal, total
bilirubin 1.8 mg/dL, and CBD diameter >6 mm. In the validation cohort, an optimal model containing alanine
transaminase (ALT) >3 times the upper limit of normal, abnormal AlkPhos, and CBD diameter >6 mm was found
to have an area under the receiver operating curve of 0.91. When 0 or 1 risk factors were present, 98.6% of
patients did not have CDL. When all 3 risk factors were present, 77.8% were found to have CDL.
Conclusions: The prevalence of CDL is high among patients with AC. When a validated model is used, applica-
tion of cutoffs for ALT, AlkPhos, and CBD diameter can effectively triage patients with low and high likelihood for
CDL to surgery or ERCP, respectively. (Gastrointest Endosc 2018;-:1-7.)
investigations: EUS, MRCP, intraoperative cholangiogram considered in confirming or refuting the diagnosis of acute
(IOC), or ERCP.3–9 These tests can be expensive, invasive, cholecystitis.
associated with significant adverse events, and often delay Patients were identified as having CDL if a common
definitive care.10–13 bile duct (CBD) stone was detected on any of the above
Much of the motivation for this approach comes from imaging studies, including those found intraoperatively
American Society for Gastrointestinal Endoscopy (ASGE) at IOC. For the purposes of this study, we considered
guidelines directed at identifying patients who are at CBD sludge or debris seen on any imaging study as
high risk for CDL.4 However, these guidelines are CDL. All other patients were placed in a group labeled
derived largely from studies of individuals without “no CDL.”
cholecystitis, and thus likely represent a distinct and Given that patients may present to medical attention at
separate patient population from those with concomitant many time points during their presentation of AC and
AC. In addition, these guidelines have been limited by laboratory results may fluctuate in the disease course,
poor performance on retrospective validation studies and rather than using presenting values, we used the highest
may result in unnecessarily high referral rates for recorded value before surgical intervention for our anal-
diagnostic ERCP.14,15 Other studies to address this ques- ysis, or “peak preoperative” value.
tion in patients with AC have had similar limitations.16–21 The relationship between CDL, preoperative labora-
It remains unclear if unique factors predict the presence tory results, and imaging results was then assessed using
of bile duct stones in this specific subset of patients and descriptive statistics with SAS 9.4 (Cary, NC). A model
when further testing should be performed for development cohort, consisting of a random 70% sample
intermediate-risk patients.20 We set out to both identify of our study population, was selected for univariable anal-
the distribution of presenting laboratory and imaging ysis of the laboratory and imaging data. For our analysis,
features and create a model to appropriately direct we used serum transaminase and lipase values greater
patients who would benefit from preoperative endoscopic than 3 times the upper limit of normal, and AlkPhos
intervention or imaging before cholecystectomy. values above the upper limit of normal for our reference
laboratory results. In addition, based on previous guide-
lines for CDL, we used a CBD diameter >6 mm, total
MATERIALS AND METHODS bilirubin 1.8 mg/dL, and total bilirubin >4 mg/dL as cut-
off values.4
We performed a retrospective review of all patients pre- Odds ratios (ORs) were calculated, and a threshold for
senting between January 1, 2013, and June 30, 2016, with a statistical significance was set at a P value <.05. We
preoperative clinical diagnosis of AC who underwent a cho- selected laboratory and imaging predictors found to be sig-
lecystectomy on the same admission to the University of nificant in univariable analysis from our model develop-
Michigan Medical Center, a large tertiary academic medical ment cohort for development of multivariable models
center. The study was approved by the University of Mich- predicting the presence of CDL in a logistic regression
igan Institutional Review Board (study ID, HUM00115511). model. The area under the receiver operating curve
To identify patients, we queried our electronic medical (AUROC) was calculated for each model. We used the
record database for ICD-9 and ICD-10 codes associated remaining 30% of patients not selected for our model
with acute cholecystitis (Supplementary Table 1, available development as our validation cohort to test each pro-
online at www.giejournal.org). The extracted records posed model. The management algorithm proposed was
were manually reviewed (P.R.C., A.H.P.). Patients were designed based on the model found to both maximize
excluded if the ICD coding did not match the the AUROC as well as the number of patients at either
preoperative clinical diagnosis as determined by the high or low risk for CDL.
surgeon performing the cholecystectomy. Patients were
also excluded if preoperative liver function test or
imaging results (US, CT, or magnetic resonance imaging RESULTS
[MRI]) were not available for review in the electronic
medical record. Patient population
Demographic data included age at the time of the pro- A total of 737 patients were identified in our query of the
cedure, sex, height, weight, medications, and comorbid- medical records and, of these individuals, 366 patients met
ities. Imaging data included reports from abdominal US, the inclusion criteria (Fig. 1). The most frequent reason for
CT, IOC, MRI, MRCP, EUS, and ERCP. Surgical operative exclusion was cholecystectomy done for reasons other than
notes were reviewed and approach, operative findings, AC (biliary colic, gallstone pancreatitis, acalculous
and any intraoperative adverse events were noted. Labora- cholecystitis, ascending cholangitis). Table 1 shows the
tory data between the time of presentation and surgical general characteristics for the study population. The mean
intervention were collected. As cholecystitis is primarily age of the patients was 52.0 years and 40.4% were male.
a clinical diagnosis, postoperative pathology was not Most of the surgeries (77.9%) were performed
TABLE 2. Comparison of select serum and imaging characteristics among patients with and without choledocholithiasis
Peak preoperative laboratory values (normal values) All (n [ 366), mean (SD) No CDL, mean (SD) CDL, mean (SD) P value
TABLE 3. Univariable predictors for the presence of choledocholithiasis in the model development cohort
without additional testing and 8.5% were recommended approach. In this study, we defined the range of presenting
for preoperative ERCP without MRCP or EUS beforehand. laboratory and imaging findings in patients with AC and strat-
The remaining patients (21.7%) would be referred for ified these values based on the presence or absence of CDL.
confirmatory tests via MRCP or EUS before definitive The risk factors identified by univariable modeling were
intervention. then successfully applied to achieve high-performing multi-
variable models.
As in other studies, we found that increases in levels of
DISCUSSION transaminases, total bilirubin, and AlkPhos frequently
occur in AC, with or without the presence of CDL.
The presence of significantly abnormal liver indices in pa- Studies of liver histology in acute cholecystitis have shown
tients presenting with AC poses a significant diagnostic and multiple contributors to abnormal liver chemistries, most
therapeutic dilemma. Suspicion for concurrent CDL can delay commonly a non-specific reactive hepatitis.22 Additional
definitive cholecystectomy, sometimes inappropriately when studies have shown that this hepatocellular injury is
CDL is ultimately not identified. In addition, assessment of acute and transient with return of liver enzyme levels to
CBD stones in the background of concomitant AC likely rep- normal after cholecystectomy.23,24 These studies highlight
resents a subset of patients distinct from those addressed by the importance of developing other tools to differentiate
current management guidelines and thus requires a tailored increased liver function test results with uncomplicated
TABLE 4. Performance characteristics of multivariable model based on presence of cumulative risk factor cutoffs in the validation cohort
3 AlkPhos > ULN, 0.92 (0.89-0.95) 0 15 (100.0) 47 (51.7) 44 (74.6) 0 (0.0) 44.3
ALT > 3 ULN, 1 15 (100.0) 68 (74.7) 23 (60.5) 0 (0.0) 64.2
AST > 3 ULN
CBD >6 mm, 2 13 (86.7) 76 (83.5) 15 (53.6) 2 (2.6) 73.6
total bilirubin >1.8 3 11 (73.3) 81 (89.0) 10 (47.6) 4 (4.7) 80.2
4 4 (26.7) 89 (97.8) 2 (33.3) 11 (11.0) 94.3
AUROC, Area under the receiver operating curve; AlkPhos, serum alkaline phosphatase; ULN, upper limit of normal; ALT, serum alanine transaminase; CBD, common bile duct; AST,
serum aspartate transaminase.
*Predictor cutoff represents the number of predictors at that point. For example, for the first model, a cutoff of 2 represents a model that predicts choledocholithiasis at 3
predictors versus 0, 1, or 2 predictors present.
acute cholecystitis versus those caused by other factors On the basis of these data, we propose a management
such as CDL. algorithm for patients with acute cholecystitis for whom
Our study found that nearly 18% of patients with AC the presence of CDL is in question (Fig. 2). The goal of
have concurrent CDL. This high rate of CDL corresponds our algorithm was to limit (1) the number of patients
with other reports suggesting that CDL is more common referred directly for cholecystectomy with concurrent
in patients undergoing cholecystectomy for AC as opposed CDL (false negative), and (2) the number of patients who
to those having cholecystectomy for other indications.25 undergo unnecessary diagnostic ERCP when CDL is
This further illustrates that AC represents a population at ultimately not present (false positive). In previous
higher risk for CDL and may require an alternative CDL studies, CBD stones were found to complicate close to
risk stratification strategy other than those proposed by 5% of cholecystectomies for low-risk patients with symp-
current guidelines. tomatic cholelithiasis from any cause,26 suggesting a
In the present study, univariable predictors of CDL model that misses less than 5% of cases with CDL is
informed multivariable models based on cumulative risk appropriate.27 Similarly, although the significant risk
factors. As each model demonstrated similar performance, profile of ERCP demands it be used in the most
we selected model 1 as the basis of our proposed manage- appropriate settings, false-negative ERCP will occur despite
ment algorithm (Fig. 2), because it contained the fewest thorough preprocedural evaluations. Indeed, there is
variables and thus was the most convenient for rapid disagreement among experts and limited data regarding
clinical application. This algorithm uses only 3 variables an acceptable rate of false-negative ERCP. In an attempt
(increased AlkPhos, ALT >3 times the upper limit of to clarify this, a recent survey of gastroenterologists deter-
normal, and CBD diameter >6 mm) and differentiates mined that a negative ERCP rate of 25% was acceptable
patients into high, low, and intermediate risk of CDL. given the potential benefit of the procedure.27 Thus, our
We opted to use the peak preoperative laboratory model selection and subsequent proposed algorithm was
values in the statistical analysis to address the common derived with the guidance of these 2 thresholds.
clinical dilemma of whether the proposed management In this algorithm, patients with 0 or 1 risk factor are
plan should change when there is a significant change in directed toward cholecystectomy, patients with 2 risk fac-
serial serum liver chemistries before proposed interven- tors are recommended to have MRCP or EUS to first
tion. To evaluate the impact of using peak preoperative exclude CDL, and patients with 3 risk factors should be
values for this study, we examined the performance charac- considered for direct referral for preoperative ERCP.
teristics of our selected multivariable model using the first Applied to our validation cohort, patients with 0 to 1
available presenting laboratory values (Supplementary CDL risk factor (approximately 70% of this cohort) would
Table 3, available online at www.giejournal.org) and be referred directly for cholecystectomy with only 1.4% un-
found no significant reduction in model performance dergoing cholecystectomy in the presence of CBD stones.
(AUROC, 0.86; range, 0.81-0.92). Patients with all 3 risk factors (approximately 10% of the
Figure 2. Management algorithm for patients presenting with acute calculous cholecystitis. ALT, Serum alanine transaminase.
validation cohort) had negative ERCP 22% of the time. Us- later pass stones spontaneously without symptoms,29 or
ing these 2 cutoffs, our model was successful in labeling may present in the future with symptomatic biliary stones
approximately 80% of patients as either low or high risk, remaining from their initial cholecystectomy. In addition,
leaving only 20% of our population in the intermediate- cases presenting to a tertiary care center can select for
risk category for which EUS or MRCP would be recommen- a population with higher comorbidities than at other
ded. For those endoscopists for whom this rate of institutions. This may explain why the average peak
diagnostic ERCP (22%) is unacceptably high, confirmatory presenting total bilirubin value in our study (1.6 mg/dL)
EUS or MRCP can still be applied to those patients with 2 was higher than the average presenting bilirubin value
or more risk factors, while avoiding such tests in those seen in other studies of AC.19,24,30 Third, we recognize that
with 1 or fewer risk factors. Similarly, at the very least, there is significant variation in practice among surgeons
the relative sensitivity and specificity at each cutoff may and endoscopists driven by physician preference, practice
be applied on an institution-specific basis to mitigate a po- patterns, and institutional guidelines. The relatively low
tential overutilization of diagnostic ERCP. rate (10.4%) of IOC performed in this study may reflect
There are many strengths to the present study. It is one the preference at our institution for CBD clearance before
of the largest retrospective studies evaluating CDL in this surgery, in part due to published data showing that the
population with an emphasis on using nearly universally presence of CDL at the time of cholecystectomy is an inde-
obtained laboratory testing and imaging for its prediction pendent predictor of CBD injury.31 A practice in which
model. The model developed had high predictive value IOC and intraoperative CBD exploration is more routinely
and was validated internally. performed may negate the need for preoperative
There are a few limitations to our study. First, inherent in evaluation for CDL in this population. However, the rate of
our retrospective study design is the reliance upon clinical CBD exploration in the United States has decreased
diagnosis and ICD coding for the identification of our study markedly and is associated with increasing rates of
population. As the aim of our study was to develop a preop- technical adverse events.32,33
erative decision tool, we opted to use the preoperative clin- In conclusion, in patients with acute cholecystitis, the
ical impression of the attending surgeon in the identification presence of concomitant CDL predictably increases the
of our study population, understanding that there is overlap serum levels of liver indices. Through application of a
in the presenting symptoms of AC, CDL, and other acute multivariable logistic regression, we were able to identify
biliary pathology. Second, retained CBD stones may present a simple algorithm to triage patients presenting with acute
several years after cholecystectomy.28 The design of our cholecystitis to minimize unnecessary testing and expedite
study presumed patients not found to have CDL within appropriate care. Prospective application and validation of
our follow-up period did not have CDL, but it is possible this algorithm, assessing its effects on length of stay,
that a few in this group may have asymptomatic CDL,6 resource utilization, and patient outcomes, are warranted.
REFERENCES 16. Abboud PA, Malet PF, Berlin JA, et al. Predictors of common bile duct
stones prior to cholecystectomy: a meta-analysis. Gastrointest Endosc
1. Strasberg SM. Clinical practice. Acute calculous cholecystitis. N Engl J 1996;44:450-5.
Med 2008;358:2804-11. 17. Taylor TV, Armstrong CP, Rimmer S, et al. Prediction of choledocholi-
2. Trowbridge RL, Rutkowski NK, Shojania KG. Does this patient have thiasis using a pocket microcomputer. Br J Surg 1988;75:138-40.
acute cholecystitis? JAMA 2003;289:80-6. 18. Huguier M, Bornet P, Charpak Y, et al. Selective contraindications
3. Al-Jiffry BO. Non-invasive assessment of choledocholithiasis in patients based on multivariate analysis for operative cholangiography in biliary
with gallstones and abnormal liver function. World J Gastroenterol lithiasis. Surg Gynecol Obstet 1991;172:470-4.
2013;19:5877. 19. Peng WK, Sheikh Z, Paterson-Brown S, et al. Role of liver function tests
4. ASGE Standards of Practice Committee; Maple JT, Ben-Menachem T, in predicting common bile duct stones in acute calculous cholecystitis.
Anderson MA, et al. The role of endoscopy in the evaluation of sus- Br J Surg 2005;92:1241-7.
pected choledocholithiasis. Gastrointest Endosc 2010;71:1-9. 20. Yang D, Yachimski P. When does assessment for bile duct stones need
5. Song SH, Kwon CI, Jin SM, et al. Clinical characteristics of acute chole- to be performed prior to cholecystectomy for calculus gallbladder dis-
cystitis with elevated liver enzymes not associated with choledocholi- ease? Clin Gastroenterol Hepatol 2018;16:331-2.
thiasis. Eur J Gastroenterol Hepatol 2014;26:452-7. 21. Gurusamy KS, Giljaca V, Takwoingi Y, et al. Ultrasound versus liver func-
6. Rosseland AR, Glomsaker TB. Asymptomatic common bile duct stones. tion tests for diagnosis of common bile duct stones. Cochrane Data-
Eur J Gastroenterol Hepatol 200;12:1171-3. base Syst Rev 2015. https://doi.org/10.1002/14651858.CD011548.
7. Tse F, Barkun JS, Barkun AN. The elective evaluation of patients with 22. Geraghty JM, Goldin RD. Liver changes associated with cholecystitis.
suspected choledocholithiasis undergoing laparoscopic cholecystec- J Clin Pathol 1994;47:457-60.
tomy. Gastrointest Endosc 2004;60:437-48. 23. Chang C-W. Acute transient hepatocellular injury in cholelithiasis and
8. Sirinek KR, Schwesinger WH. Has intraoperative cholangiography cholecystitis without evidence of choledocholithiasis. World J Gastro-
during laparoscopic cholecystectomy become obsolete in the era of enterol 2009;15:3788.
preoperative endoscopic retrograde and magnetic resonance cholan- 24. Fikry A. Elevated liver enzymes in patients with cholecystitis. J Surg
giopancreatography? J Am Coll Surg 2015;220:522-8. 2014;2:38.
9. Giljaca V, Gurusamy KS, Takwoingi Y, et al. Endoscopic ultrasound 25. Lee D, Ahn YJ, Lee HW, et al. Prevalence and characteristics of clinically sig-
versus magnetic resonance cholangiopancreatography for common nificant retained common bile duct stones after laparoscopic cholecystec-
bile duct stones. Cochrane Database Syst Rev 2015. tomy for symptomatic cholelithiasis. Ann Surg Treat Res 2016;91:239.
https://doi.org/10.1002/14651858.CD011549. 26. Iranmanesh P, Tobler O, De Sousa S, et al. Prospective validation of an
10. Ward WH, Fluke LM, Hoagland BD, et al. The role of magnetic reso- initial cholecystectomy strategy for patients at intermediate-risk of
nance cholangiopancreatography in the diagnosis of choledocholithia- common bile duct stone. Gastrointest Endosc 2017;85:794-802.
sis: do benefits outweigh the costs? Am Surg 2015;81:720-5. 27. Almoner BJ, Debenedet AT, Volk ML, et al. Clinical yield of diagnostic
11. Kaltenthaler E, Vergel YB, Chilcott J, et al. A systematic review and eco- endoscopic retrograde cholangiopancreatography in orthotopic liver
nomic evaluation of magnetic resonance cholangiopancreatography transplant recipients with suspected biliary complications. Liver
compared with diagnostic endoscopic retrograde cholangiopancrea- Transpl 2012;18:1479-84.
tography. Health Technol Assess 2004;8(iii):1-89. 28. Cox MR, Budge JPO, Eslick GD. Timing and nature of presentation of
12. Morris S, Gurusamy KS, Sheringham J, et al. Cost-effectiveness analysis unsuspected retained common bile duct stones after laparoscopic
of endoscopic ultrasound versus magnetic resonance cholangiopan- cholecystectomy: a retrospective study. Surg Endosc 2015;29:2033-8.
creatography in patients with suspected common bile duct stones. 29. Collins C, Maguire D, Ireland A, et al. A prospective study of common bile
PLoS ONE 2015;10:e0121699. duct calculi in patients undergoing laparoscopic cholecystectomy: nat-
13. Netinatsunton N, Attasaranya S, Sottisuporn J, et al. Comparing cost- ural history of choledocholithiasis revisited. Ann Surg 2004;239:28-33.
effectiveness between endoscopic ultrasound and endoscopic retro- 30. Padda MS, Singh S, Tang SJ, et al. Liver test patterns in patients with
grade cholangiopancreatography in diagnosis of common bile duct acute calculous cholecystitis and/or choledocholithiasis. Aliment Phar-
stone in patients with predefined risks: a study from a developing makoi Ther 2009;29:1011-8.
country. Endosc Ultrasound 2016;5:165-72. 31. Kohn JF, Trenk A, Kuchma K, et al. Characterization of common bile
14. Adams MA, Hosmer AE, Wamsteker EJ, et al. Predicting the likelihood duct injury after laparoscopic cholecystectomy in a high-volume hos-
of a persistent bile duct stone in patients with suspected choledocho- pital system. Surg Endosc 2018;32:1184-91.
lithiasis: accuracy of existing guidelines and the impact of laboratory 32. Wandling MW, Hungness ES, Pavey ES, et al. Nationwide assessment of
trends. Gastrointest Endosc 2015;82:88-93. trends in choledocholithiasis management in the United States from
15. Prachayakul V, Aswakul P, Bhunthumkomol P, et al. Diagnostic yield of 1998 to 2013. JAMA Surg 2016;151:1125-30.
endoscopic ultrasonography in patients with intermediate or high like- 33. Livingston EH, Rege RV. Technical complications are rising as
lihood of choledocholithiasis: a retrospective study from one common duct exploration is becoming rare. J Am Coll Surg
university-based endoscopy center. BMC Gastroenterol 2014;14:165. 2005;201:426-33.
SUPPLEMENTARY TABLE 2. Demographic and laboratory characteristics of model development cohort and validation cohort
SUPPLEMENTARY TABLE 3. Performance characteristics of multivariable model 1 using the first presenting laboratory value in the validation
cohort
First AlkPhos > ULN, 0.86 (0.81-0.92) 0 13 (86.7) 56 (61.5) 35 (72.9) 2 (3.5) 54.7
first ALT > 33 ULN, 1 12 (80.0) 78 (85.7) 13 (52.0) 3 (3.7) 76.4
CBD >6 mm
3 6 (40.0) 89 (97.8) 2 (25.0) 9 (9.2) 92.5
AUROC, Area under the receiver operating curve; AlkPhos, serum alkaline phosphatase; ULN, upper limit of normal; ALT, serum alanine transaminase; CBD, common bile duct.