Factors Associated with Mortality in Severe Acute Cholangitis in a Moroccan Intensive Care Unit: A Retrospective Analysis of 140 Cases

Background Severe acute cholangitis is a life-threatening biliary infection, leading to organ dysfunction, septic shock, and naturally death. Mortality has dropped significantly in the past years through improving resuscitation and biliary drainage techniques. The aim of our study is to analyze our daily practice and the factors associated with mortality. Methods A retrospective study including severe acute cholangitis patients admitted to our unit from January 2009 to December 2018. Variables analyzed (univariate then multivariate analysis) were age, sex, history, origin, evolution time, bilirubin, etiology, organ dysfunction, qSOFA, SOFA, TOKYO, biliary drainage timing and technique, shock, antibiotherapy, and resuscitation. Results 140 patients were included in this study. Average age was 61. Sex ratio M/F was 0.59. Lithiasis etiology was dominant (69%). SOFA average score upon admission was 8. Ceftriaxone + metronidazole was the empirical antibiotic used in 87%. Average time to biliary drainage was 1.58 ± 0.89 days. Endoscopic unblocking was the technique used in 76%. Mean duration of ICU stay was 6 days. Mortality rate was 28%. Statistically significant factors for mortality (p < 0.05) were history of taking anticoagulant treatment, use of catecholamines and mechanical ventilation during ICU stay, and delay in consultation and administration of antibiotic therapy. Conclusions Early recognition, antibiotics, resuscitation, and minimally invasive biliary drainage have improved patient outcomes although there is still progress to be made. Moreover, as multiple organ failure is often associated with mortality in severe acute cholangitis, predictive risk factors of organ failure should be more investigated.


Introduction
Acute cholangitis is a bacterial infection of the biliary tract following cholestasis mainly caused by lithiasic or tumoral biliary obstruction. The classic pain-fever-jaundice symptoms, representing the Charcot triad, were first described in 1877 [1]. The severity of cholangitis is due to infection dissemination with risk of septic shock and organ failure. "The Reynolds pentad," described in 1958 as Charcot triadshock-confusion, was associated with high mortality in the absence of adequate treatment [2]. Early diagnosis and recognition of severe presentations are therefore a challenging task for every clinician in order to initiate an early and appropri-ate therapeutic management. This management is continuously progressing in terms of resuscitation, antibiotics, surgical techniques, and less invasive techniques such as interventional endoscopy and radiology. It therefore requires multidisciplinarity (anesthetists, intensivists, hepatobiliary endoscopists, surgeons, radiologists, and microbiologists) and the implementation of standardized protocols. The recommendations of "the Tokyo Guidelines Working Group" are regularly updated from 2007 to 2018 [1,[3][4][5]. Resuscitation, antibiotics, and biliary drainage as an early intervention approach are absolutely essential for survival in severe acute cholangitis. Mortality has certainly decreased from 50% to 10-30% over the past 30 years, but it remains high, and several questions still unanswered regarding the application of these recommendations on a local level. Furthermore, other factors may influence the prognosis. Recent studies consider admission to intensive care as a predictor of mortality [6]. Our aim in this study was to identify the factors associated to mortality in severe acute cholangitis in our context.

Materials and Methods
2.1. Study Design and Setting. In this retrospective observational monocentric study, we evaluated patients of age ≥ 16 years with severe acute cholangitis admitted to our Intensive Care Unit (ICU) between January 2009 and December 2018. The diagnosis of acute cholangitis was based on clinicalbiological and radiological criteria according to the Tokyo guidelines. The severity was assessed by the presence of at least one organ failure and/or an unbalanced comorbidity. The criteria for diagnosis and severity are presented in Tables 1 and 2. Patients with incomplete and/or nonexploitable records were excluded from the analysis. The study setting was a 14-bed medico-surgical ICU in a tertiary university hospital in Morocco (ICU A4-Hassan II University Hospital of Fez). The study was approved by the institutional review board (Comité d'Ethique Hospitalo-Universitaire de Fès) with a waiver of informed consent.

Data
Collection. Study data were collected retrospectively from both paper charts and electronic medical records of patients using HOSIX electronic data capture tools hosted at Hassan II University Hospital of Fez. Variables collected included demographic information, diagnostic parameters, the Sequential Organ Failure Assessment (SOFA) and quick-SOFA scores, and TOKYO grading upon ICU admission, patient comorbidities, therapeutics, and evolution.

Statistical Analysis.
The statistical analysis of the parameters was performed using the SPSS 20 software in the epidemiology laboratory of the Faculty of Medicine and Pharmacy of Fez. Factors associated with mortality were analyzed using univariate and multivariate analysis. Descriptive statistics were used to summarize baseline patient characteristics. The results were expressed in numbers and percentages for the qualitative variables and in means ± standard deviations (SD) for the quantitative variables. Comparison of the quantitative and qualitative variables was based, respectively, on the Student's t-test and the chi-2 test (χ2) through univariate analysis. Both p values and odds ratio (OR) with corresponding 95% confidence interval (CI) were reported for qualitative (categorical) variables while only p values were presented for quantitative (continuous) variables. Multivariable logistic regression was used to identify variables associated with ICU mortality as the outcome variable of interest. Multiple logistic regression models were fitted by regressing mortality status on multiple clinical variables. Considering the large number of clinical variables included in the study, we chose a stepwise regression approach using backward elimination. Order of elimination was based on clinical relevance and statistical strength. The statistical significance threshold was determined at p = 0:05. Results of the multivar-iate analysis were shown as odds ratio (OR) and corresponding 95% confidence interval (CI). Most of the significant variables at the univariate analysis were entered in the multivariate analysis: antithrombotic therapy, time to hospital consultation, tumor origin, quick-SOFA ≥ 2 on admission, SOFA on admission, TOKYO Grade III on admission, time to biliary decompression, mechanical ventilation, use of catecholamines, septic shock during stay, persistence of hematological failure after decompression, and persistence of renal failure after decompression.
(A) Systemic inflammation (A-1) Fever (temperature > 38°C) and/or shaking chills (A-2) Evidence of Inflammatory Response. White blood cells (WBC) count < 4000 or >10000/mm 3 , C-reactive protein ≥ 10 mg/l.     Table 3). The variables entered into the logistic regression model were antithrombotic therapy, time to hospital consultation (days), tumor origin, severity assessment upon admission (quick-SOFA ≥ 2, SOFA, TOKYO Grade III), organ support therapies during ICU stay (use of catecholamines, mechanical ventilation), persistence of renal failure after decompression, persistence of hematological failure after decompression, and septic shock during ICU stay. Variables considered but not retained in the final model were creatinine, total bilirubin, GCS, respiratory failure as they are included in the severity assessment scores upon admission, dialysis as it is highly correlated to persistence of renal failure after decompression, and admission from medical ward as we considered it is not clinically relevant. Multivariate analysis showed that mortality in our ICU population was significantly associated with the history of taking antithrombotic treatment (OR = 10:146), the time to hospital consultation (OR = 1:137), and the use of catecholamines (OR = 5:819) and mechanical ventilation (OR = 13:649) during ICU stay (Table 4).

Discussion
This is a single center retrospective study analyzing the factors related to mortality in patients with severe acute cholangitis in ICU. Four variables, including the history of taking antithrombotic treatment, the time to hospital consultation, and the use of catecholamines and mechanical ventilation during ICU stay, were associated with the mortality of ICU patients with severe acute cholangitis. Since the first surgical biliary decompression attempt was only described in 1903, mortality associated with acute cholangitis treated without biliary drainage was close to 100%. Despite surgical advances and the introduction of antibiotics over the following decades, mortality remained at 50%. But it has decreased to 10-30% since 1980, with the development of endoscopic biliary drainage techniques and the various associated therapeutic modalities [1,7,8]. Current mortality rates vary between 9.6% and 37% [6,[9][10][11], and death is most often due to a multiorgan failure related to a refractory septic shock [1]. This is consistent with our results which showed a mortality rate of 28% and refractory septic shock as the leading cause of death.
There are few studies on the prognostic factors for severe acute cholangitis. Although the TG18/TG13 severity criteria are currently widespread and very precise for the diagnosis and assessing the severity, they are based on expert opinions and therefore requiring additional validation in clinical practice [12]. In Morocco, to our knowledge, there has been no study to investigate prognostic factors in critically ill patients with severe acute cholangitis. In this study, multivariate analysis identified four independent risk factors for mortality: history of taking antithrombotic treatment, time to hospital consultation, and the use of catecholamines and mechanical ventilation during ICU stay. Table 5 summarizes the different prognostic factors found in different studies compared to those in our study.
Severity assessment scores such as TOKYO grading as well as the classic quick-SOFA and SOFA did not stand out as prognostic factors in our study. There are two possible reasons: (1) the inclusion of patients in the study was based on these scores and (2) admission to ICU is considered in itself as a prognostic factor [6]. On the other hand, use of catecholamines and mechanical ventilation (both reflects of hemodynamic and respiratory failures) did stand out as prognostic factors. Both criteria are included in the SOFA and the TOKYO grading. Furthermore, mechanical ventilation is also associated with its own complications. Organ failure is therefore the main prognostic factor and the detection of patients at risk of progression to organ failure by scores such as quick-SOFA, SOFA, or TOKYO would improve the prognosis. In a recent study [13], quick-SOFA was associated with high specificity but decreased sensitivity to predict severity (97% vs. 43%) and admission to intensive care (96% vs. 60%). However, it is an easily reproducible clinical tool in  5 Gastroenterology Research and Practice the emergency room. Particular attention is to be addressed to the elderly and the immune suppressed patients. Often these patients do not have clear and definite clinical symptoms to guide the diagnosis and are also more likely to deteriorate rapidly due to their limited physiological reserve.
In this study, mean time to hospital consultation was of 10 days. This delay in consultation, and therefore the delay in the administration of antibiotics and biliary drainage, will make infection control more difficult with poor treatment results, especially in patients with comorbidities [6,14]. The use of traditional therapies (fire points, etc.) and difficult access to care in remote areas is still a real problem to be taken into account in local and national policies to raise awareness and provide care.
Early antibiotic therapy is as important as appropriate antibiotic therapy. Several studies have also reported inappropriate probabilistic antibiotic therapy as a prognostic factor [15], which underlines the importance of bile samples and blood cultures for secondary adaptation as well as knowledge of local ecology. Moreover, other study [10] has reported that the presence of ESBL organism was significantly associated with organ failure in bacteria cholangitis. The initial empiric antibiotherapy dose did not show significant association with mortality in this study, but further analysis based on bile sample results is needed. We did not study this etiological factor since the bile sample results were missing in some cases.
No other study, to our knowledge, has reported taking antithrombotics as a prognostic factor. A history of anti-thrombotic therapy may reflect the severity of the underlying comorbidity, can worsen or precipitate hematological failure, and may delay or complicate biliary drainage. This factor is included as prognosis factor in some ICU scores such as for trauma patients. Further studies are needed, but we should pay particular attention to these patients.
Emergency biliary decompression is the primary treatment for severe acute cholangitis. Increasing bile pressure promotes biliary sepsis dissemination and prevents biliary penetration of antibiotics. Antibiotic therapy, which main purpose is to limit the infection spread while waiting for biliary decompression, remains essential but is alone insufficient. Biliary drainage is then recommended in acute cholangitis regardless of severity, with the exception of a few nonserious cases progressing spontaneously under antibiotic therapy and initial resuscitation measures [16]. 92% of our patients underwent biliary decompression. Three patients died before, and nine patients' conditions improved while waiting for decompression and were transferred. Endoscopic drainage, less invasive and compatible with a better quality of life (no bile leak compared to external drainage), is the first choice technique despite the risk of pancreatitis post-ERCP [17]. In our study, endoscopic drainage was the golden standard. Surgical treatment was the first alternative if endoscopic treatment (hepatobiliary endoscopists and/or equipment) was unavailable. Percutaneous drainage, performed in 2% of cases, is a technique recently introduced to our center. Early biliary drainage is associated with less intrahospital mortality, 30-day mortality, and hospital costs, particularly within 48 hours of admission and regardless of severity [18,19]. Procalcitonin has been proposed by some authors as a predictive indicator of severity and therefore of urgent biliary decompression [20], but this needs to be validated by more well-designed studies. According to previous studies, biliary drainage is the most important factors associated with mortality [9][10][11]. Time to biliary decompression did not stand out as a prognosis factor in our series. Our population study is different: only ICU patients with severe cholangitis are included. Plus, most of our patients (87%) underwent early biliary drainage, within 48 hours, associated with early aggressive resuscitation. Furthermore, Lee et al. [21] reported that drainage beyond 48 hours is associated with persistent organ failure and prolonged length of stay but does not significantly affect mortality.
RDW (Red Cell Distribution Width) is traditionally high in cases of ineffective erythropoiesis or excessive destruction of erythrocytes. The prognostic value of RDW in different acute or chronic inflammatory circumstances (cardiovascular, traumatic, neurological, septic…) and especially in critically patients is recently raised by several studies [22]. RDW is associated to admission to intensive care and bilirubin levels within a predictive mortality score [6] but needs to be validated by more studies.
Our results may not be generalized as they are limited by the retrospective design of our study and by the monocentric recruitment, but this is the first large study of ICU patients with severe cholangitis identifying local prognostic factors. As with any other observational study, even after adjusting for clinically and statistically significant prognostic factors, other unmeasured factors may have contributed to patient mortality. All retrospective series on this question are limited by significant confounding factors. The key to a successful logistic regression model is to choose the correct variables to enter into the model. While it is tempting to include as many input variables as possible, this can dilute true associations and lead to large standard errors with wide and imprecise confidence intervals, or, conversely, identify spurious associations [23]. We first run the univariate analyses and then used only the variables meeting a preset cutoff for significance of p < 0:05 to run a multivariable model. We also considered the scientific plausibility and the clinical meaningfulness of the association while trying to avoid the use of highly correlated variables.
As multiple organ failure is often associated with mortality in severe acute cholangitis; predictive risk factors for organ failure should be further investigated. Finally, a local contextualized protocol emerged from this work and is being validated in practice. These procedures are provided in Supplementary Materials (available here).

Conclusions
History of taking antithrombotic treatment, the time to hospital consultation, and the use of catecholamines and mechanical ventilation during ICU stay were associated with the mortality of ICU patients with severe acute cholangitis in this study. Predictive risk factors for organ failure should be further investigated.

Data Availability
All data and tables used to support the findings of this study are included within the article.

Conflicts of Interest
The authors declare that they have no conflicts of interest with the contents of this article.