Short-Course Antibiotic Treatment Is Not Inferior to a Long-Course One in Acute Cholangitis: A Systematic Review

A comprehensive search was carried out by two independent authors using three databases (PubMed, EMBASE, and the Cochrane Library). The aim was to gather relevant articles on the topic of ABT in patients with acute cholangitis from inception up to November 11, 2017.

The PICO items were the following: We included studies that examined patients with acute cholangitis diagnosed according to the most recent guidelines available at the time (P), with at least one group of acute cholangitis patients that were treated with antibiotic therapy for a short duration. A short-course ABT was defined as treatment with antibiotics for a shorter duration than that is suggested by the available protocol or guideline at the publication date of the study (I). We imposed no restriction in terms of the etiology of cholangitis. We had no strict requirement for a control group as that criterion would have narrowed the number of articles included even further; when available, we considered the group of patients who received the traditional, long course of ABT as a control group. Long-course ABT was defined as treatment with antibiotics for the traditional length of time (or more) as suggested by the then available guideline (C). In terms of outcomes, we considered the rate of recurrent cholangitis and mortality as our primary outcomes and the length of hospitalization (LOH) and duration of fever after the ERCP procedure as secondary outcomes (O).

We used the following query in all three databases: “Cholangitis AND (Antibiotic OR Antimicrobial) AND (Duration OR Length OR Stop OR Cessation OR Discontinuation).” For a draft of our search strategy, see Supplementary File 1: Our search strategy.

We limited our search to human and English language studies via the appropriate filters when searching in PubMed and EMBASE. The Trials filter was applied when searching for articles in the Cochrane Library. To expand the search, we performed a recursive hand search on the references of relevant articles.

We also searched PROSPERO, an international database for systematic reviews and meta-analyses, for previously completed reviews on the subject.

We excluded the following types of publications: letters, comments, case reports, case studies, conference abstracts, editorials, and reviews.

In our review, we included both observational (cohort, cross-sectional cohort, and case–control) and interventional (randomized and non-randomized, controlled and non-controlled) trials and studies. We included studies with both prospective data collection and retrospective data collection, regardless of their primary objectives. If there were multiple publications on the same group of patients, the latest was always chosen.

After the initial search, all results were imported into a reference management program (EndNote X7, Clarivate Analytics, Philadelphia, PA, USA). The software was used in the process of removing duplicates by searching for articles with overlapping publication year, author, and/or title.

After duplicates were removed, the authors screened the remaining articles against the pre-defined eligibility criteria by title, abstract, and then full text. Two different researchers conducted each step simultaneously. Any disagreements were resolved by consensus.

Numerical data were extracted by two investigators and manually entered on an Excel 2016 sheet (Office 365, Microsoft, Redmond, WA, USA). Data were collected on the first author, year of publication, study design, geographical location, number of patients, and basic demographics (age and sex ratio) in each group. Data were extracted on the definition of the groups and the length of ABT in each. Finally, data were collected on the four outcomes of interest noted previously.

A modified version of the Newcastle–Ottawa Scale (NOS) was used to assess the quality of cohort studies included in the analysis [10]. Supplementary File 2 shows the NOS Quality Assessment Form for Cohort Studies modified to fit the study design of the articles included. The Cochrane Risk of Bias Tool was used to assess the quality of the RCT included [11]. We used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) methodology to rate the quality of evidence as high (level A), moderate (level B), low (level C), or very low (level D). The strength of recommendations was also graded according to the same guidelines as strong (1) or weak (2) [12].

Our search yielded a total of 692 articles, 101 in PubMed, 566 in EMBASE, and 25 in the Cochrane Library. After removing duplicates and checking the titles and abstracts against the eligibility criteria, six articles were assessed by full text and, finally, four were deemed eligible for systematic review [13‐16]. The reason for discarding the last two studies was that they did not include a separately analyzed subgroup of patients with short ABT [17], or the patients were treated with short-course ABT, but only a fraction of them had acute cholangitis (and no subgroup analysis was performed) [18]. Two out of the four eligible studies were prospective [14, 15] and two were retrospective [13, 16]. These included a total of 205 acute cholangitis patients with 137 and 68 of them receiving short- and long-course ABT, respectively (Fig. 1). Collected data and the characteristics of the studies included are shown in Tables 1 and 2.

All of the four articles [13‐16] assessed the rate of recurrent cholangitis during the follow-up and were included in the systematic review. In van Lent et al., 23.7% (n = 19) of all the patients (n = 80) developed recurrent cholangitis, with 26.8% (n = 11) in the short ABT group (ABT for three or fewer days), 21.1% (n = 4) in the medium-length ABT group (ABT for 4–5 days), and 20.0% (n = 4) in the long ABT group (ABT for more than five days). The p value (0.8) showed no statistically significant difference across groups [13]. In Kogure et al. and Limmathurotsakul et al., none of the 34 patients examined in these studies had a recurrent episode of cholangitis during their follow-up period [14, 15]. In Uno et al., the recurrence rate was 4.4% (n = 4) in 91 patients, with 0.0% (n = 0) and 13.3% (n = 4) in the short and long ABT groups, respectively. This difference appeared to be significant (p = 0.036) [16].

All of the four articles included [13‐16] assessed mortality. Van Lent et al. reported a 11.3% (n = 9) overall mortality among the patients included, with 14.6, 10.5, and 5% (n = 6, 2, 1) in the short, medium, and long ABT groups, respectively. This difference between groups did not prove to be significant [13]. In both Kogure et al. and Limmathurotsakul et al., there were no deaths registered among the 34 patients included in total [14, 15]. In Uno et al., two deaths were observed among the 91 patients (2.2%), both of which occurred in the long ABT group, accounting for a 5.7% mortality in this group, with no significant difference between the short and long ABT groups (p = 0.179) [16].

LOH was only assessed in one study. Uno et al. [16] reported a 14-day median LOH (IQR: 10.0–17.0 days) for the short ABT group and a 17.5-day median LOH (IQR: 16–22.5 days) in the long ABT group with a significant difference in between (p < 0.001).

Three out of the four articles included data on the duration of fever after the ERCP procedure. Van Lent et al. only reported a median of one day of fever (range 0–17 days) duration after ERCP for all patients in the study [13]. Kogure et al. found the median duration of fever to be two days across all patients included [14]. Limmathurotsakul et al. reported the duration to be 1.8 h (SD 1.8) in the short ABT group and 1.2 h (SD 0.7) in the long ABT group. The authors of this study found no significant difference between the two groups in terms of fever duration (p = 0.467) [15].

Figure 2 and Table 3 show the results of the quality assessment of the studies included. With regard to the RCT (by Limmathurotsakul et al.), the Cochrane Risk of Bias Tool indicated an overall poor quality, since three points (blinding of participants and personnel; blinding of outcome assessment; and selective reporting) were evaluated as having a potentially high risk of bias (see Table 3).

Applying the GRADE approach to each of the outcomes assessed above resulted in a weak (2) recommendation and a very low (D) level of evidence for every outcome [12].

During the process of conducting our systematic review, we came across several potential limitations that may impair the strength of our findings. In terms of general characteristics, the low number of articles (n = 4) that could be included in the final review and the low combined patient number (n = 205, with n = 137 cases and n = 68 controls) are insufficient to provide basis for strong recommendations. Moreover, only one of these studies is randomized and controlled, and it only involves 16 patients (n = 8 cases and n = 8 controls) [15]. We also included an article without a control (comparison, short-course ABT) group [14], and two out of the four studies utilized retrospective data collection [13, 16].

In addition, the inclusion criteria were not the same in the articles. Uno et al. [16] only included cholangitis patients with positive blood culture, and there was a difference between the severities of cholangitis in the inclusion criteria across the studies analyzed. Furthermore, the type and regimen of antibiotics, as well as the definition and timing of successful biliary drainage procedures, also varied greatly. The causes of acute cholangitis were not consistent across the studies. For example, Limmathurotsakul et al. [15] excluded all causes except for common bile duct stones; however, the other studies were not so restrictive in terms of etiology.

With regard to the methodology in the studies, the follow-up period ranged from one to six months in the four studies included. This inconsistency may have an influence on outcomes, such as mortality and recurrent cholangitis rate during follow-up, especially when taking into consideration that the median time for developing recurrent cholangitis is 18 weeks after biliary drainage [27]. The criteria or reasons for the cessation of antibiotic therapy were not clearly stated in all the studies. In Kogure et al. and Limmathurotsakul et al., antibiotic treatment was stopped after the patient’s temperature dropped below a certain threshold (37 °C and 37.8 °C, respectively) for a period of time (24 h and 72 h, respectively). Neither of these studies specifies the protocol of the temperature measurement used, and only Kogure et al. [14] provide data on the method of measurement. The other two articles contain no information on the protocol for stopping antibiotic therapy. Supplementary File 3 highlights some of the major differences between studies that contribute to the limitations of our review.