Introduction
Obstructive jaundice (OJ) (synonyms: subhepatic, obstructive, obstructive) is a syndrome that occurs when the outflow of bile into the duodenum is impaired [1]. Mortality rates among patients with OJ, according to various authors, reach 20–60 % [2, 3].
The main reason for the unfavorable outcome is liver failure (LF) with rapidly developing liver dysfunction, increasing deterioration of biochemical parameters, decompensation of other organs and systems as the process progresses [2, 3]. The main principle of surgical tactics is the implementation of minimally invasive interventions aimed at decompressing the biliary tract in order to relieve OJ and restore the outflow of bile into the duodenum [1, 4]. A quick and accurate assessment of the severity of LF against the background of OJ, along with predicting its further outcome, is a pressing issue in modern intensive care, determining tactical positions in the treatment of the patient [5, 6].
For an objective assessment of the severity of LF against the background OJ use various criteria, scales, the purpose of which is to determine the severity of patological process, assessing the prognosis of the disease and the formation of adequate treatment regimens [5, 6]. However the use of existing techniques does not always make it possible to ability to solve the above problems [7, 8]. Traditionally, to predict the course and outcome of LF against the background of OJ in the intensive care unit, scales are used to assess the severity of the condition of patients with diseases of the hepatobiliary organs: a scoring scale for assessing end-stage liver diseases and prognosis of life in patients with end-stage LF (MELD) , Child-Turcotte-Pugh, Maddrey, Schindl their modifications and a number of others; as well as universal multiparameter scales for assessing the severity of the condition of critically ill patients: the Acute Physiology and Chronic Health Evaluation II (APACHE II) scale, the integral system for assessing organ dysfunction (Sequential Organ Failure Assessment, SOFA) [8, 9]. A number of authors, in their studies, demonstrate their lack of reproducibility and prognostic significance in relation to LF against the background of OJ [8]. Thus, the MELD score was developed in 2002 to predict mortality within three months after a transjugular intrahepatic portosystemic shunt procedure. Subsequently, it was found that it is important in determining the prognosis and prioritizing the tactical position in liver transplantation. In recent years, works have appeared devoted to the use of the MELD scale for predicting postoperative LF after liver resection [8, 10]. The Child-Turcotte-Pugh scale is an assessment technique designed to determine the severity of liver cirrhosis and patient survival, allowing to assess the likelihood of an unfavorable outcome of the disease. The method was first published in 1964 by medical scientists at the University of Michigan C. Child and D. Turcotte, in 1972 and improved by a group of British doctors led by Pugh. In the final version, the Child-Turcotte-Pugh scale includes an assessment of detoxification (bilirubin level, severity of encephalopathy) and synthetic (albumin level and international normalized ratio) liver functions, as well as an assessment of portal hypertension syndrome [11]. One of the most common integrated systems for determining the severity of a patient’s condition is the APACHE II scale, developed and implemented by Knauss W.A. et al. (1981). The disadvantage of this scale is that it can only be used for severely ill patients in the intensive care unit due to fear of overestimating the degree of severity in other patients [12, 13]. The SOFA scale, which consists of a sequential assessment of organ failure and includes six main parameters (respiratory, cardiovascular, nervous system, renal, liver and blood coagulation) was created by a working group of the European Community of Intensive Care Medicine and adopted at the Paris Consensus Conference in 1994. But the above scale is more intended for quickly scoring and describing a number of complications in patients in critical condition than for predicting the outcome of the disease [14, 15].
Thus, in the current literature there is little information about the possibility of using the above scales to predict LF against the background of OJ, and the use of existing criteria does not always provide a complete assessment of the severity of the disease, which affects the amount of treatment provided [5, 8]. Of course, the leading place in treatment tactics for OJ is occupied by decompression of the biliary system using various methods (x-ray, endoscopic, surgical). Intensive therapy includes the use of hepatoprotectors, which play an important role in reducing the level of bilirubin fractions, infusion therapy (aimed at restoring microcirculation, removing toxic endogenous substances from the body) [4]. With the development of such complications of OJ as acute renal failure and LF, extracorporeal treatment methods become especially relevant, which is especially important for patients at risk of an unfavorable outcome [16].
Objective
To determine the prognostic value of multiparametric rating scales and identify predictors of unfavorable outcome in LF against the background of OJ.
Materials and methods
This prospective observational single-center study of a continuous sample included 53 patients with LF associated with benign OJ. Personal data of patients is anonymized. The diagnosis of the disease was established based on the results of clinical, laboratory and instrumental research methods using clinical recommendations developed by the working group on OJ of the Russian Society of Surgeons (2018) [1]. Etiologically, the patient population is represented by: choledocholithiasis in 66 cases (66 %); in 30 cases, cicatricial strictures of the bile ducts (30 %); intraoperative damage to the common bile duct in 4 cases (4 %). According to the classification of the severity of OJ by E.I. Halperin [5], 11 patients (21 %) were classified as class C, 28 (53 %) as class B and 14 (26 %) patients as class A.
In accordance with the clinical recommendations for OJ developed by the working group of the Russian Society of Surgeons (2018), all patients, for the purpose of decompression of the biliary tract, underwent various options of antegrade minimally invasive interventions under the control of ultrasound and X-ray television against the background of pathogenetically based complex conservative therapy in the conditions intensive care units.
Inclusion criteria for the study: patients over 18 years of age with benign liver lesions who underwent antegrade decompression of the biliary tract to resolve OJ during the 1st day of hospitalization; absence of chronic inflammatory liver diseases. Exclusion criteria from the study: age under 18 years, inability to assess factors included in the study, decompensated concomitant pathology; chronic inflammatory liver diseases.
Monitoring the development of impaired functional state of the liver based on clinical and laboratory data with assessment of the severity of the patients’ condition using the APACHE II, SOFA, MELD, Child-Turcotte-Pugh scales [2], carried out on the day of hospitalization, on the 3rd, 7th and 11th day after decompression intervention. To characterize the severity of the condition, the following laboratory parameters were analyzed: the level of platelets, leukocytes, and serum hematocrit (automatic hematology analyzer “UniCel DxH800” Beckman Coulter, USA); level of bilirubin, albumin, creatinine, blood electrolytes (biochemical analyzer “AU480” Beckman Coulter, USA); markers of the hemostasis system — international normalized ratio (automatic coagulometer “AMAX DESTINY-plus”, Ireland); gas composition, acid-base state and lactate level in blood serum (automated analyzer “ABL-800 Fleх” Radiometer, Denmark). Hemodynamic parameters were monitored using the Mindray BeneView T1 bedside monitor from Mindray, China. The severity of ascites was assessed using the SonoScape S11 ultrasound apparatus from SonoScape, China.
Statistical processing of data was carried out using the statistical software package for analyzing social science data IBM SPSS Statistics 22. Presentation of indicators included calculation of the median (Me), 25th and 75th percentiles (p25 % and p75 %, respectively). To predict the risk of death, methods of logistic regression and ROC analysis, MedCalc and SPSS 17 statistical packages were used. The area under the operating characteristic curve (AUROC, Area Under Receiver Operator Curve) shows how well the model differentiates patients with a favorable and unfavorable prognosis. For models predicting mortality, reliable resolution (as measured by AUROC) was greater than 0.9. In order to verify the accuracy of the results obtained, in this study, methods recommended for small samples were used, such as two-factor nonparametric (rank) Friedman analysis of variance, Kruskal-Wallace H-test for nonparametric (rank) one-way analysis of variance, Mann-Whitney U-test. Statistically significant results were considered at p < 0.05.
Results
The study included 53 patients: 26 (49 %) men and 27 (51 %) women, average age 60 ± 3.8 years. Depending on the outcome of the disease, patients were divided into two categories: group 1 with a favorable outcome (27/51 %) and group 2 with an unfavorable outcome (26/49 %) of the disease. The assessment was based on the results of 30-day mortality. The causes of unfavorable outcome were LF (8/31 %), multiple organ failure (6/26 %), cholangitis (4/15 %), renal failure (4/15 %), sepsis (3/12 %), gastrointestinal bleeding (1/4 %). In the presented study, the groups of patients were comparable not only in gender and age, but also did not have statistically significant differences in the main rating scales (Table 1).
| Characteristic | 1st group (n = 27) | 2st group (n = 27) | p |
|---|---|---|---|
| Patient age avg. (minimum−maximum), years (AA) |
63.5 (37–85) | 61.9 (35–88) | < 0.09 |
| Men/women, % (AA) | 14/13 (51.9/48.1) | 12/14 (48.3/51.7) | < 0.05 |
| SOFA, Me points (IR) | 7.4 (4.0–9.0) | 8.8 (6.0–10.0) | < 0.05 |
| APACHE II, Me points (IR) | 21.1 (9.0–28.0) | 23.8 (12.0–32.0) | < 0.05 |
| MELD, Me points (IR) | 16.8 (10.0–23.0) | 18.2 (13.0–28.0) | < 0.04 |
| Child-Turcotte-Pugh, Me points (IR) | 12.1 (11.0–14.0) | 14.2 (12.0–15.0) | < 0.04 |
At the first stage of the study, multiparameter rating scales were analyzed in relation to the population of patients with LF against the background of OJ, traditionally used in the intensive care unit. All severity rating scales analyzed (SOFA, MELD, APACHE II and Child-Turcotte-Pugh) showed very good predictive accuracy regarding the development of an unfavorable outcome of LF due to OJ. The SOFA score, when analyzed at four study time points, demonstrated a sensitivity of 78.9 % (73.1–84.6 %), specificity of 77.8 % (74.1–81.5 %), area under the ROC curve (AUC) 0.862 (0.820–0.904) [95 % CI 0.707–0.823 ... 0.933–0.985], p = 0.04, cut-off point more than 8 points. The MELD scale sensitivity in this study was 80.8 % (73.1–88.5 %), specificity — 79.7 % (74.1–85.2 %), area under the ROC curve (AUC) 0.882 (0.872–0.892) [95 % CI 0.779–0.806 ... 0.965–0.984], p = 0.01, cut-off point 15 points. For the APACHE II scale, sensitivity was 57.7 %, specificity — 88.9 %, area under the ROC curve (AUC) 0.864 [95 % CI 0.550–0.777], p = 0.04, cut-off point more than 30 points. The area under the operational characteristic curve (AUROC) for the Child-Turcotte-Pugh score was 0.810 (0.758–0.863) [95 % CI 0.629–0.966], p = 0.03, cut-off point greater than 12 points with sensitivity and specificity values of 75.1 % (69.2–80.8 %) and 70.9 % (63.0–88.9 %), respectively.
The above results of statistical analysis of predicting outcome in PN against the background of breast syndrome for the APACHE II and Child-Turcotte-Pugh, SOFA and MELD scales are presented in Table 2.
| Day | Area under the ROC curve | Standard error | Value (p)* | Asymptotic 95 % confidence interval | Se, % | Sp, % | |
|---|---|---|---|---|---|---|---|
| Bottom line | Upper limit | ||||||
| APACHE II scale | |||||||
| 1 | 0.864 | 0.058 | 0.04 | 0.750 | 0.977 | 57.7 | 88.9 |
| Child-Turcotte-Pugh scale | |||||||
| 1 | 0.867 | 0.050 | 0.05 | 0.769 | 0.964 | 69.2 | 88.2 |
| 3 | 0.758 | 0.066 | 0.01 | 0.629 | 0.887 | 80.8 | 63.0 |
| 7 | 0.863 | 0.052 | 0.05 | 0.761 | 0.966 | 73,1 | 78.9 |
| 11 | 0.769 | 0.067 | 0.01 | 0.638 | 0.900 | 69.2 | 77.8 |
| MELD scale | |||||||
| 1 | 0.887 | 0.049 | 0.00 | 0.791 | 0.984 | 88.5 | 85.2 |
| 3 | 0.892 | 0.044 | 0.00 | 0.806 | 0.978 | 73.1 | 77.8 |
| 7 | 0.872 | 0.048 | 0.01 | 0.779 | 0.965 | 76.9 | 74.1 |
| 11 | 0.876 | 0.046 | 0.01 | 0.785 | 0.967 | 80.8 | 85.2 |
| SOFA scale | |||||||
| 1 | 0.820 | 0.058 | 0.05 | 0.707 | 0.933 | 73.1 | 74.1 |
| 3 | 0.904 | 0.041 | 0.04 | 0.823 | 0.985 | 76.9 | 76.9 |
| 7 | 0.890 | 0.046 | 0.04 | 0.801 | 0.980 | 80.8 | 81.5 |
| 11 | 0.890 | 0.046 | 0.04 | 0.800 | 0.979 | 84.6 | 81.5 |
ROC curves of logistic models with the considered predictors (SOFA and MELD scale) for predicting an unfavorable outcome of LF in OJ syndrome are presented in Figure 1 and 2.
Fig. 1. ROC-curves predicting adverse outcome of liver failure due to obstructive jaundice at four time points with “SOFA score”
Fig. 2. ROC-curves predicting adverse outcome of liver failure due to obstructive jaundice at four time points with “MELD score” predictor
The results obtained indicate the high predictive ability of multiparameter severity rating scales, but insufficient sensitivity and specificity values do not allow their use as an independent predictor of an unfavorable outcome of LF in OJ syndrome, which determines the need to search for additional predictors.
At the next stage of the study, the statistical significance of laboratory data routinely used in practice was tested as independent predictors of an unfavorable outcome of the disease. From this perspective, regression analysis revealed only two laboratory biomarkers: platelets and albumin.
The “platelet” indicator demonstrated statistically significant results for the integral indicator of the quality of predicting an unfavorable outcome of the disease (area under the ROC curve (AUC) 0.816 (0.796–0.835) [95 % CI 0.967–0.991 ... 0.978–0.994], p = 0.01), but low values of sensitivity and specificity (71.15 % (65.4–76.9 %) and 72.3 % (70.4–74.1 %), respectively) with a cut-off point of 115 × 109/l.
The “albumin” indicator, as an independent predictor of an unfavorable outcome of the disease, using the logistic regression method demonstrated sensitivity of 88.6 % (84.6–92.3 %), specificity of 85.2 % (77.8–92.6 %), area under the ROC curve (AUC) 0.909 (0.891–0.927) [95 % CI 0.434–0.759 ... 0.579–0.841], p = 0.01 with a cut-off point of 28.3 g/l.
The results of statistical analysis of the “platelet” and “albumin” indicators from the standpoint of predicting an unfavorable outcome of PN in breast syndrome are presented in Table 3.
| Biochemistry indicator | Odds ratio* | Asymptotic 95 % confidence interval | р** | Se, % | Sp, % | Area under the ROC curve | Cut off threshold |
|---|---|---|---|---|---|---|---|
| When patients are admitted to the hospital | |||||||
| Platelets | 0.984 | 0.975–0.993 | 0.01 | 65.4 | 70.4 | 0.806 | 114 × 109/l |
| Albumen | 0.635 | 0.498–0.810 | 0.01 | 92.3 | 81.5 | 0.925 | 25.5 g/l |
| On the 3rd day of treatment in hospital | |||||||
| Platelets | 0.986 | 0.978–0.994 | 0.01 | 73.1 | 74.1 | 0.796 | 115 × 109/l |
| Albumen | 0.698 | 0.579–0.841 | 0.01 | 84.6 | 81.5 | 0.891 | 28.0 g/l |
| On the 7rd day of treatment in hospital | |||||||
| Platelets | 0.979 | 0.967–0.991 | 0.01 | 76.9 | 70.4 | 0.835 | 110 × 109/l |
| Albumen | 0.574 | 0.434–0.759 | 0.01 | 88.5 | 92.6 | 0.927 | 26.0 g/l |
| On the 11rd day of treatment in hospital | |||||||
| Platelets | 0.981 | 0.971–0.992 | 0.01 | 76.9 | 70.4 | 0.830 | 105 × 109/l |
| Albumen | 0.592 | 0.452–0.775 | 0.00 | 84.6 | 77.8 | 0.910 | 26.5 g/l |
ROC curves of logistic models with the considered predictors of predicting an unfavorable outcome of LF in OJ syndrome are presented in Figure 3 and 4.
Fig. 3. ROC curves predicting adverse outcome of liver failure due to obstructive jaundice at four time points with predictor “platelets”
Fig. 4. ROC curves predicting adverse outcome of liver failure due to obstructive jaundice at four time points with predictor “albumin”
The results obtained indicate the high predictive ability of albumin, but the low values of sensitivity and specificity at some points of the study (days 3 and 11) do not allow us to consider them as reliable predictors of an unfavorable outcome of LF against the background of OJ syndrome.
Discussion
The number of patients with diseases of the hepatopancreatoduodenal organs, complicated by OJ syndrome, continues to increase, maintaining high postoperative mortality, reaching 12.5 % [2, 13, 17]. The main reason leading to an unfavorable outcome and deterioration of the results of surgical treatment for OJ syndrome is progressive LF, the problem of treatment of which remains serious and relevant today [4].
Assessing the severity of the condition of patients with LF due to OJ syndrome who are in the intensive care unit is an integral part of the diagnostic and treatment process. All kinds of criteria and scales are used, the purpose of which is to timely determine the severity of the pathological process with the subsequent formation of adequate treatment regimens [6, 9, 18, 19]. However, to date, there is very little data in the literature on their prognostic ability regarding the outcome of LF against the background of OJ syndrome [9, 20].
In this study, patients were analyzed, divided into two groups depending on the outcome of the disease, with a diagnosis of LF against the background of benign OJ syndrome. In the presented study, the groups of patients were comparable not only in gender and age, but also did not have statistically significant differences in the main rating scales.
In relation to the population of patients with LF on the background of OJ syndrome, scales that allow assessing the severity of the condition of patients with diseases of the hepatobiliary organs (MELD, Child-Turcotte-Pugh) and universal multiparameter scales for assessing the severity of the condition of critically ill patients (APACHE II and SOFA) were analyzed. All analyzed rating scales demonstrated very good predictive accuracy regarding the unfavorable outcome of LF due to OJ syndrome, but insufficient sensitivity and specificity values. Thus, the area under the operational characteristic curve (AUROC) for the SOFA scale is 0.862 (0.820–0.904), p = 0.04, the cut-off point is more than 8 points, which indicates the highest risk of death when registering above the specified; sensitivity 78.9 % (73.1–84.6 %), specificity 77.8 % (74.1–81.5 %). Unfortunately, there values is no data in the literature on the prognostic ability of this scale for predicting the outcome of LF in the setting of OJ cancer.
For the APACHE II scale, the area under the ROC curve (AUROC) was 0.864 [95 % CI 0.550–0.777], p = 0.04, cut-off point more than 30 points, with little significant sensitivity (57.7 %) and specificity (81.9 %). A similar result was obtained in the work of Kądziołka I. et al., which showed a significant underestimation of the severity of patients with OJ syndrome using the APACHE II scale compared to other scoring systems, which does not allow it to be recommended for predicting the course and outcome of the disease [12].
The sensitivity of the MELD scale, throughout the study, was in the range of 73.1–88.5 % (80.8 %), specificity 74.1–85.2 % (79.7 %), area under the ROC curve (AUC) 0.882 (0.872–0.892) [95 % CI 0.779–0.806 ... 0.965–0.984], p < 0.01, cut-off point more than 15 points. The American College of Surgeons National Surgical Quality Improvement Program study found the MELD score to be good at predicting 30-day mortality for patients with scores between 10 and 22, but underestimated the risk for patients with higher scores [21]. This, obviously, can explain the insufficient predictive ability of the scale under consideration in relation to the population of patients with LF against the background of OJ syndrome.
The Child-Turcotte-Pugh scale in our study has an area under the ROC curve (AUC) of 0.813 (0.758–0.867) [95 % CI 0.0629–0.966], p = 0.03, cut-off point more than 12 points, sensitivity 75.1 % (69.2–80.8 %) and specificity 70.9 % (63.0–78.9 %). In their work, Mohamed Elmeliegy et al. pointed out the limitations of using the Child-Turcotte-Pugh scale for the category of patients with oncological pathology, since hypoalbuminemia, encephalopathy and ascites may be associated with the course of the oncological process or metastases, and not with impaired liver function. This does not allow us to recommend the use of the Child-Turcotte-Pugh scale other than its main purpose [22], which obviously explains the result obtained in our study.
Thus, the analyzed multivariate rating scales have a high, but still insufficient predictive ability regarding the outcome of LF against the background of OJ syndrome (at some points in the study, sensitivity and specificity values are less than 80 %), which determines the need to search for additional predictors.
A number of authors indicate that the main predictors of an unfavorable outcome of LF in OJ syndrome are age over 50 years, bilirubin levels over 340 µmol/l and creatinine over 132 µmol/l, prothrombin time over 35 seconds [23, 24]. Our study found that two laboratory biomarkers had statistically significant changes reflecting the likelihood of an unfavorable outcome of the disease: platelets (area under the ROC curve (AUC) 0.816 (0.796–0.835) [95 % CI 0.967−0.991 ... 0.978−0.994], p = 0.01, with sensitivity 71.15 % (65.4 %–76.9 %) and specificity 72.3 % (70.4–74.1 %)) and albumin (area under the ROC curve (AUC) 0.909 (0.891−0.927) [95 % CI 0.434−0.759 ... 0.579−0.841], p = 0.01, with sensitivity 84.6–92.3 % and specificity 77.8–92.6 %).
The results obtained indicate a good predictive accuracy of platelets and excellent for albumin, however, the values of sensitivity and specificity at some study points also do not allow us to consider them as reliable independent predictors of an unfavorable outcome of LF against the background of OJ syndrome.
Conclusion
Thus, the multiparametric assessment scales of the severity of the condition SOFA, APACHE II, Child-Turcotte-Pugh, MELD demonstrated good, and the candidate biomarker “albumin” excellent predictive ability regarding the development of an unfavorable outcome in patients with LF against the background of OJ syndrome. However, insufficient sensitivity and specificity values (less than 80 %) at some study points do not allow us to consider them as independent predictors of an unfavorable outcome. This requires further search for factors that predict the development of LF in OJ syndrome.
Disclosure. The authors declare no competing interests.
Author contribution. All authors according to the ICMJE criteria participated in the development of the concept of the article, obtaining and analyzing factual data, writing and editing the text of the article, checking and approving the text of the article.
Ethics approval. Ethic approval by the local Ethical Committee is not required for current study design.
Data Availability Statement. Data that support the findings of this study are available from the corresponding author, upon reasonable request.
