Does near-infrared fluorescent cholangiography with indocyanine green reduce bile duct injuries and conversions to open surgery during laparoscopic or robotic cholecystectomy? — A meta-analysis

Abstract

Background

Bile duct injury and conversion-to-open–surgery rates remain unacceptably high during laparoscopic and robotic cholecystectomy. In a recently published randomized clinical trial, using near-infrared fluorescent cholangiography with indocyanine green intraoperatively markedly enhanced biliary-structure visualization. Our systematic literature review compares bile duct injury and conversion-to-open–surgery rates in patients undergoing laparoscopic or robotic cholecystectomy with versus without near-infrared fluorescent cholangiography.

Methods

A thorough PubMed search was conducted to identify randomized clinical trials and nonrandomized clinical trials with ≥100 patients. Because all near-infrared fluorescent cholangiography studies were published since 2013, only studies without near-infrared fluorescent cholangiography published since 2013 were included for comparison. Incidence estimates, weighted and unweighted for study size, were adjusted for acute versus chronic cholecystitis, and for robotic versus laparoscopic cholecystectomy and are reported as events/10,000 patients. All studies were assessed for bias risk and high-risk studies excluded.

Results

In total, 4,990 abstracts were reviewed, identifying 5 near-infrared fluorescent cholangiography studies (3 laparoscopic cholecystectomy/2 robotic cholecystectomy; n = 1,603) and 11 not near-infrared fluorescent cholangiography studies (5 laparoscopic cholecystectomy/4 robotic cholecystectomy/2 both; n = 5,070) for analysis. Overall weighted rates for bile duct injury and conversion were 6 and 16/10,000 in near-infrared fluorescent cholangiography patients versus 25 and 271/10,000 in patients without near-infrared fluorescent cholangiography. Among patients undergoing laparoscopic cholecystectomy, bile duct injuries, and conversion rates among near-infrared fluorescent cholangiography versus patients without near-infrared fluorescent cholangiography were 0 and 23/10,000 versus 32 and 255/10,000, respectively. Bile duct injury rates were low with robotic cholecystectomy with and without near-infrared fluorescent cholangiography (12 and 8/10,000), but there was a marked reduction in conversions with near-infrared fluorescent cholangiography (12 vs 322/10,000).

Conclusion

Although large comparative trials remain necessary, preliminary analysis suggests that using near-infrared fluorescent cholangiography with indocyanine green intraoperatively sizably decreases bile duct injury and conversion-to-open–surgery rates relative to cholecystectomy under white light alone.

Introduction

Laparoscopic cholecystectomies (LC) are among the most common surgical procedures performed worldwide, accounting for between 600,000 and 900,000 procedures annually in the United States alone.^1,2 Starting in the mid-1980s, a shift from open to laparoscopic cholecystectomies (LC) occurred, for reasons that included markedly less scarring, shortened hospital stays and recovery times, and reduced operative times and costs.³ This transition has not been without problems; however, as bile duct injuries (BDI), the number one concern of surgeons performing LC,4, 5, 6 appeared to increase in frequency with the emergence of LC, from 1 to 2 BDI per 1,000 procedures⁷ to as many as 1 per 100.8, 9, 10, 11, 12 This increased incidence has persisted over time,^13,14 despite the adoption of imaging techniques such as intraoperative cholangiography (IOC) and ultrasound. Such BDI may significantly prolong hospital stays, increase the need for further surgery, markedly elevate hospital costs, and result in chronic morbidity and significantly increased mortality.15, 16, 17 Many BDI patients experience reduced quality of life for years,18, 19, 20 and roughly 4% of BDI patients had premature mortality directly attributed to their BDI.^21,22

Another problem that arises during LC is the need for conversion from laparoscopic to open surgery, which occurs in an estimated 6.2% of patients.²³ Such conversion also markedly increases patient morbidity and mortality, duration of hospital stay and convalescence, and healthcare costs.^24,25 In a 151-study meta-analysis published by Pucher et al in 2018,²³ which included randomized clinical trials (RCT) and prospective and retrospective nonrandomized comparative and observational studies, overall encompassing over 500,000 patients, the pooled rate for BDI was 0.52%, or 52 per 10,000. In this same meta-analysis, the pooled rate for conversions from laparoscopic to open surgery was 620 per 10,000.

Studies have shown that the main cause of BDI and conversions to open surgery in patients undergoing LC is inadequate visualization of essential extra-hepatic biliary structures, such as the cystic duct.26, 27, 28 During the past decade, new technology has emerged to facilitate the visualization of such structures, via the preoperative intravenous injection of indocyanine green (ICG), followed by the intraoperative use of near-infrared fluorescence imaging, a combined process called near-infrared fluorescent cholangiography (NIFC). In a recently published RCT that compared 321 laparoscopic cholecystectomy procedures during which NIFC was used against 318 procedures during which it was not, NIFC dramatically increased predissection visualization rates for all 7 extra-hepatic biliary structures that were assessed: cystic duct, right hepatic duct, common hepatic duct, common bile duct, cystic CBD junction, cystic gallbladder junction, and accessory ducts, with strongly statistically significant odds ratios ranging from 2.3 to 3.6.²⁹ Five of these 7 structures also were statistically more often visible postdissection, with odds ratios ranging from 2.4 to 3.3. To date, this just-mentioned study is the only published RCT comparing NIFC versus standard white light during LC. Although there were 2 BDI in the control group and none in the NIFC group, and 4 conversions in the control group versus 1 in the NIFC group, the study lacked the statistical power for these to be statistically significant differences (P = .25 and 0.17, respectively). In fact, assuming a 0.4% BDI event rate²³ and a 1:1 ratio of cases and controls, even detecting a 50% decrease between groups with 95% confidence and 80% power would require a study with almost 12,000 patients per treatment arm.³⁰ With an assumed 5.0% event rate for conversions to open surgery, such a study would require over 2,000 patients per group.

The primary purpose of this article is to report the results of a systematic literature review and meta-analysis conducted to compare both BDI and conversion-to-open–surgery rates in patients undergoing minimally invasive cholecystectomy with versus without NIFC. Given the recent steady rise in the percentage of cholecystectomy procedures being performed robotically (albeit still only accounting for under 10% of all cholecystectomies),³¹ a decision was made to include both laparoscopic cholecystectomy (LC) and robotic cholecystectomy (RC) studies in the analysis. Secondary purposes were to compare the effects of NIFC on BDI and open conversion rates in LC and RC individually and evaluate all studies for the risk of bias.

A thorough search of the medical literature was conducted to identify all studies, published in full print or open access form, for which the incidence of BDI or open conversion was reported, assessing the effectiveness of NIFC during LC; and in which NIFC was not used during LC.

To be eligible for analysis, articles had to meet the following eligibility criteria: (1) report absolute numbers of patients who had undergone LC or RC, either with or without the use of NIFC; (2) either report on the absolute number of BDI or conversions to open surgery or provide percentages to at least 1 decimal place, so those absolute numbers could be accurately calculated; (3) be published from 2013 onward because the only NIFC studies we were able to identify were from 2013 onward; (4) have virtually all data collected from 2010 onward, to similarly reduce the use of old data published late; (5) have study cohorts consisting of at least 100 patients either receiving or not receiving NIFC, to reduce the risk of positive publication bias; (6) be available in print or full open-access form, so a full risk of bias assessment could be performed; (7) be deemed at low to, at most, moderate risk of bias, specifically related to the rate of BDI and conversions to open surgery (thereby, not concerned with the level of bias for other outcomes); and (8) not include patients with cholelithiasis, biliary atresia, or cancer as the indication for surgery. Non-English articles only were excluded if they could not be translated or otherwise interpreted by at least one of the authors.

The identification of relevant studies was conducted in 2 stages. In stage 1, a PubMed search was conducted, from April through October 2020, looking for relevant search terms, from which all abstracts were reviewed, seemingly pertinent articles read to completion, and final eligibility determined. In stage 2, all tables and bibliographies listing other pertinent studies in read articles were reviewed to identify additional articles that might be eligible for inclusion, then these articles read to determine their eligibility.

For NIFC studies, we used the following combined search terms: “fluorescence” and “cholecystectomy” (n = 128 abstracts); “indocyanine green” and “cholecystectomy” (n = 114); “robotic” and “cholecystectomy” (n = 407); “indocyanine green” and “bile duct injury” (n = 30); and “indocyanine green” and “conversion” (n = 173; total n = 852; Fig 1). For non-NIFC studies, the combined search terms used were “cholecystectomy” and “bile duct injury” (n = 1,253); “cholecystectomy” and “conversion” (n = 2,478), and “cholecystectomy” and “robotic” (n = 407; total n = 4,138). To be eligible for inclusion, studies had to also report the absolute numbers of BDI or conversions to open surgery and the number of patients treated by LC. When studies also included patients who underwent open surgery, the number of patients who had their surgery performed laparoscopically or robotically and number of BDI or conversions specifically among those patients had to be reported, accompanied by approach-specific outcomes.

Variables of interest were first author name; year of article publication; period of data collection, country where the study was conducted; surgical approach (LC versus RC); study design, objectives, and treatment arms; number of patients in each treatment arm; treatments rendered; and absolute number of BDI and conversions to open surgery.

Each study was assessed for bias risk using either the Cochrane Risk of Bias tool³² for RCTs and quasi-randomized trials or the Cochrane ROBINS I tool³³ for non-RCTs. Using the Cochrane Risk of Bias tool,³² each RCT was evaluated for the 6 of 8 potential sources of bias deemed relevant (excluding mortality and long-term outcomes): random sequence generation and allocation concealment (both for selection bias), blinding of patients and personnel (performance bias), blinding of outcome assessments (detection bias), completeness of short-term data (attrition bias), and selective reporting (selection bias). Using the Cochrane ROBINS I tool,³³ studies were assessed for potential confounding, subject selection, intervention classification, unintended differences and contamination between treatment arms, missing data, measurement bias, and selective reporting. The risk of bias for each study was rated as very low, low, moderate, or high, specifically pertaining to the determination of BDI and conversion-to-open–surgery rates, as follows: very low, if the study was judged at low risk across all domains; low, if no more than one domain was judged at moderate risk, and no domain at high risk; moderate, if judged to be at moderate risk in more than one domain, but at high risk in none; high, if any domain was considered at high risk. Domains for which data were not clearly reported were automatically deemed a source of moderate risk. Studies considered at high risk were excluded from additional analysis.

The 2 primary outcomes of interest were the incidence of BDI and incidence of conversions to open surgery, each expressed as the number of events per 10,000 patients, weighted for the number of subjects per study. Unweighted rates also were calculated, by averaging incidence rates over all studies, to adjust for large variations in study size. To account for acute cholecystitis, the number of patients who underwent LC or RC for acute cholecystitis was extracted from each study to allow for calculation of the overall percentage of acute cholecystitis patients in each of the 2 patient cohorts (NIFC versus non-NIFC), after which BDI and open-conversion rate estimates were multiplied by a product of this percentage difference and the estimated difference in BDI or conversion rate from other studies (outside the current analysis) restricted to acute cholecystitis patients. To account for potential differences between LC and RC, rates for BDI and open conversions were estimated combining LC and RC patients, and with LC and RC patients analyzed separately, after which weighted estimates were generated using the same method described above for acute cholecystitis.

Forest plots of the weighted data were generated for BDI and conversions for each of 3 comparisons: (1) all patients, LC and RC, who underwent LC with NIFC versus no NIFC; (2) NIFC versus non-NIFC just among LC patients; and (3) NIFC versus non-NIFC just among RC patients.