http://informahealthcare.com/rnf ISSN: 0886-022X (print), 1525-6049 (electronic) Ren Fail, 2014; 36(6): 889–894 ! 2014 Informa Healthcare USA, Inc. DOI: 10.3109/0886022X.2014.894765
CLINICAL STUDY
Bicarbonate can improve the prognostic value of the MELD score for critically ill patients with cirrhosis Cheng-Yi Chen1,2, Chi-Feng Pan1,2, Chih-Jen Wu1,2, Han-Hsiang Chen1,2, and Yu-Wei Chen3 1
Division of Nephrology, Department of Internal Medicine, Mackay Memorial Hospital, Taipei, Taiwan, 2Mackay Junior College of Medicine, Nursing, and Management College, Taipei, Taiwan, and 3Division of Nephrology, Department of Internal Medicine, Ching Chyuan Hospital, Taichung, Taiwan Abstract
Keywords
The prognosis of critically ill patients with cirrhosis is poor. Our aim was to identify an objective variable that can improve the prognostic value of the Model of End-Stage Liver Disease (MELD) score in patients who have cirrhosis and are admitted to the intensive care unit (ICU). This retrospective cohort study included 177 patients who had liver cirrhosis and were admitted to the ICU. Data pertaining to arterial blood gas-related parameters and other variables were obtained on the day of ICU admission. The overall ICU mortality rate was 36.2%. The bicarbonate (HCO3) level was found to be an independent predictor of ICU mortality (odds ratio, 2.3; 95% confidence interval [CI], 1.0–4.8; p ¼ 0.038). A new equation was constructed (MELD-Bicarbonate) by replacing total bilirubin by HCO3 in the original MELD score. The area under the receiver operating characteristic curve for predicting ICU mortality was 0.76 (95% CI, 0.69–0.84) for the MELD-Bicarbonate equation, 0.73 (95% CI, 0.65–0.81) for the MELD score, and 0.71 (95% CI, 0.63–0.80) for the Acute Physiology and Chronic Health Evaluation II score. Bicarbonate level assessment, as an objective and reproducible laboratory test, has significant predictive value in critically ill patients with cirrhosis. In contrast, the predictive value of total bilirubin is not as prominent in this setting. The MELD-Bicarbonate equation, which included three variables (international normalized ratio, creatinine level, and HCO3 level), showed better prognostic value than the original MELD score in critically ill patients with cirrhosis.
Acidosis, bicarbonate, intensive care unit, liver cirrhosis, MELD
Introduction Regardless of the reason for intensive care unit (ICU) admission, cirrhosis independently worsens the prognosis of critically ill patients.1 Data from recent series involving critically ill patients with cirrhosis have shown ICU and 6-month mortality rates of 41 and 70.8%, respectively.2,3 Without definite treatment by way of liver transplantation, the prognosis of patients with cirrhosis is poor.4 Therefore, predicting the outcomes of critically ill patients with cirrhosis is important for medical decision-making and providing prognostic advice to patients. More than 10 different scoring systems are available for predicting the outcome of patients admitted to the ICU.5 Although the scores have an overall good prognostic value for a general population of critically ill patients, some of the prognostic value is diminished when these systems are applied to a specific subpopulation.6,7 Until now, the Acute Physiology and Chronic Health Evaluation II (APACHE II) and Simplified Organ Failure Assessment (SOFA) scores Address correspondence to Dr. Yu-Wei Chen, Division of Nephrology, Department of Internal Medicine, Ching Chyuan Hospital, No. 80, Sector. 4, Yatan Road, Daya Dist, Taichung City 428, Taiwan, R.O.C. Tel: +886-4-25605600; Fax: +886-4-25606162; E-mail: yw.chen. [email protected]
History Received 29 November 2013 Accepted 7 February 2014 Published online 6 March 2014
were most commonly used for assessing prognosis in the general population of patients admitted to the ICU. Although there is a paucity of scoring systems designed for critically ill patients with cirrhosis, several have been developed in a setting other than the ICU. The Child-Pugh score and the Model of End-Stage Liver Disease (MELD) score were initially designed to predict operative risk and mortality, respectively, for patients who have cirrhosis and are on the waiting list for liver transplantation. The MELD score has been proven to have prognostic value in a different setting also in the case of patients with cirrhosis.8,9 Several studies have compared the accuracy of liverspecific scores (Child-Pugh and MELD) to that of general ICU scores (APACHE II and SOFA).2,10–14 These studies suggest that the accuracy of the SOFA score is slightly superior to that of the APACHE II, MELD, and Child-Pugh scores.2,11–13 Therefore, Das et al. suggested that the prognosis of critically ill patients with cirrhosis is not only correlated with the severity of liver cirrhosis but also rather with the severity of organ failure.2 These general ICU scoring systems include complex and often subjective variables. In contrast, the MELD score relies on readily available and fully objective measures and has been proven to possess a significant prognostic value in patients with cirrhosis in the ICU setting.2,13–15 However, apart from renal, liver, and
890
C.-Y. Chen et al.
coagulation dysfunction, the MELD score does not take into account other organ/system dysfunction and has been suggested to be less accurate than the SOFA score. In this study, we hypothesized that data obtained for the arterial blood gas (ABG)-related parameters at the time of admission to the ICU can help predict ICU mortality in critically ill patients with cirrhosis. Furthermore, we incorporated these parameters into the MELD score and compared the prognostic value of the new equations and the original MELD, Child-Pugh, and APACHE II scores.
Patients and methods Ethics This study was conducted in accordance with the Declaration of Helsinki (2000) of the World Medical Association, and the protocol was approved by the institutional review board of Mackay Memorial Hospital and informed consent was waived. Mackay memorial Hospital is a tertiary referral center for liver disease. This study is a single center investigation and all patients of the study were all afferent, directly diagnosed and followed-up in Mackay memorial hospital. Patients This retrospective cohort study included consecutive adult patients who had cirrhosis and had been admitted to the ICU from August 2010 to April 2012. The diagnosis of cirrhosis was either confirmed by biopsy or was based on a combination of laboratory, endoscopic, and imaging findings. The records of 299 patients with cirrhosis from a total of 9735 ICU-admitted patients were reviewed. The exclusion criteria were end-stage renal disease treated with renal replacement therapy (10 cases), previous transplantation (of any tissue or organ), and no ABG examination at the time of ICU admission (112 cases). Of the 299 patients, 177 met the criteria and were included for analysis. Methods Data pertaining to the baseline characteristics of age, gender, etiological factors of liver disease and ICU admission, and underlying comorbidities were collected. Laboratory tests for ABG, liver function, renal function, and electrolytes were performed upon the patient’s admission to the ICU. ABG-related parameters such as pH, arterial partial pressure of carbon dioxide (PaCO2), arterial partial pressure of oxygen (PaO2), bicarbonate (HCO3), arterial oxygen saturation (SaO2), and base excess were recorded. Mortality during the ICU course was recorded as the primary endpoint. The MELD score was calculated using the following formula:16 MELD score ¼ 11.2LN (INR) + 3.78LN (bilirubin) + 9.57LN (creatinine) + 6.43. In this score, laboratory values less than 1 were set to 1. If the serum creatinine level was above 4 mg/dL, it was automatically set at 4 mg/dL. The score was rounded to the nearest whole number and capped at 40. Statistical analysis Continuous variables were summarized as mean ± SEM unless otherwise stated. We initially compared the demographic data
Ren Fail, 2014; 36(6): 889–894
and laboratory variables of survivors and non-survivors using Student’s t-test and Chi-square test. Correlations between variables and ICU mortality were assessed using logistic regression analysis. To formally examine the relationship among parameters of the ABG and the MELD as predictors of ICU mortality, several multivariate models were constructed. PaCO2 and HCO3, then, were incorporate into the MELD score after adjustment with the original MELD equation by logistic regression analysis. Cox proportional-hazard regression models were conducted for investigating the association between the different scoring system and ICU mortality. The results of these analyses were used to construct a receiveroperating characteristic (ROC) curve from which we sought the optimum cut-off point for predicting successful sites. The optimum cut-off point was defined as the point on the ROC curve closest to the point (0,1), where the false-positive rate was zero and the sensitivity was 100%. The area under the curve (AUC) and 95% confidence interval were calculated. Survival curves for critically-ill cirrhotic patients with different scores by the new scoring system were prepared according to the Kaplan–Meier method. A p value of less than 0.05 was considered statistically significant. All statistical analyses were performed using SPSS software version 17.0 (SPSS Inc., Chicago, IL).
Results Baseline demographic, clinical, and laboratory characteristics of the 177 enrolled patients are shown in Table 1. The overall ICU mortality rate was 36.2%. Patients were grouped into survivors (113 patients) and non-survivors (64 patients); of these patients, 131 (74%) received intubation with mechanical ventilator support (72 cases in the survivor group and 59 cases in the non-survivor group). The most common cause of ICU admission overall, that is, gastrointestinal bleeding, was also the most common cause of admission in the survivor group. The ABG-related parameters that showed a significant difference between survivors and non-survivors were pH, PaCO2, HCO3, SaO2, and base excess. Interestingly, the serum sodium level showed no significant difference between these two groups. Univariate logistic regression analysis was initially used to evaluate the association between the parameters and ICU mortality, and those with significant prognostic values are listed in Table 2. It is noteworthy that the total bilirubin level, one of the three variables of the MELD score, showed no significant difference between the two groups. Subsequent multivariate logistic regression analysis showed that both PaCO2 and HCO3 were independent predictors of ICU mortality. Table 3 shows several multivariate models for predicting ICU mortality by Cox regression analysis. Model 1, which was included in the original MELD score only, was a significant predictor. When the three parameters of the MELD score were included, the HCO3 level remained significant (model 3) but PaCO2 did not (model 2). As with the previous results, the total bilirubin level was insignificant in model 2 and model 3, and it was excluded subsequently in model 4 and model 5. All the parameters in model 4 and model 5 were significant, and both models were better than the original MELD score (model 1) with respect to the values of the
MELD score in the ICU
DOI: 10.3109/0886022X.2014.894765
891
Table 1. Demographic, clinical, and laboratory data for 177 critically ill patients with cirrhosis. Parameter
All patients (N ¼ 177)
ICU survivals (N ¼ 113)
ICU non-survivals (N ¼ 64)
p
58.1 ± 1.1 135:42 51 79.1 ± 1.4 20.8 ± 0.6 131
58.0 ± 1.3 88:25 35 83.0 ± 1.7 18.6 ± 0.7 72
58.2 ± 1.9 47:17 16 72.4 ± 2.3 24.6 ± 1.1 59
NS (0.937) NS (0.505) NS (0.399) 50.001 50.001 50.001 NS (0.051)
58 63 18 38
30 44 10 29
28 19 8 9
70 67 26 14 115 90 44 9.7 ± 0.2 22.8 ± 0.6 294.3 ± 77.4 2.6 ± 0.0 6.0 ± 0.5 1.7 ± 0.0 42.8 ± 2.6 2.6 ± 0.2 135.6 ± 0.5 4.2 ± 0.1 4.6 ± 0.3 7.4 ± 0.0 29.4 ± 0.7 130.5 ± 4.8 18.1 ± 0.5 97.2 ± 0.2 5.7 ± 0.6
43 46 18 6 68 54 27 9.2 ± 0.2 20.4 ± 0.7 160.8 ± 38.2 2.7 ± 0.1 5.4 ± 0.6 1.5 ± 0.0 36.8 ± 2.9 2.1 ± 0.1 135.8 ± 0.6 4.2 ± 0.1 3.7 ± 0.2 7.4 ± 0.0 30.9 ± 0.8 129.4 ± 4.8 20.3 ± 0.6 97.7 ± 0.2 3.0 ± 0.6
27 21 8 8 47 36 17 10.7 ± 0.3 27.1 ± 1.0 537.7 ± 204.7 2.3 ± 0.1 7.2 ± 0.9 2.0 ± 0.1 53.5 ± 4.8 3.5 ± 0.4 135.4 ± 1.0 4.3 ± 0.1 5.9 ± 0.5 7.3 ± 0.0 26.8 ± 1.2 132.4 ± 10.3 14.3 ± 0.9 96.3 ± 0.6 10.6 ± 1.1
Age (years) Male: Female Diabetes mellitus Mean arterial pressure (mm Hg) APACHE II Mechanical ventilation Etiology of ICU admission Septic shock Gastrointestinal bleeding Isolated acute respiratory failure Others Etiology of liver cirrhosis Viral hepatitis Alcoholic Alcohol with viral hepatitis Cryptogenic Ascites Hepatic encephalopathy Hepatoma Child-Pugh score MELD score GOT (15–41 IU/L) Albumin (3.5–5 g/dL) Total bilirubin (0.3–1.2 mg/dL) INR BUN (8–12 mg/dL) Creatinine (0.4–1.2 mg/dL) Sodium (136–144 mEq/L) Potassium (3.5–5.1 mEq/L) Phosphate (2.7–4.5 mg/dL) pH (7.35–7.45) PaCO2 (32–45 mm Hg) PaO2 (75–100 mm Hg) HCO3 (20–26 mmol/L) SaO2 (%) Base excess (2 to +2 mmol/L)
NS (0.278)
NS (0.076) NS (0.279) NS (0.693) 50.001 50.001 NS (0.075) 50.001 NS (0.087) 50.001 0.003 0.001 NS (0.742) NS (0.260) 50.001 50.001 0.003 NS (0.771) 50.001 0.019 50.001
Values were expressed as mean ± SEM unless otherwise defined. Statistical comparison was performed with student’s t test or chi-square test. Abbreviation: APACHE II, acute physiology and chronic health evaluation II; ICU, intensive care unit; MELD, model for end-stage liver disease; INR, international normalized ratio; BUN, blood urea nitrogen; NS, not significant.
likelihood ratio test and 2 log likelihood. After adjustment with the original MELD equation by logistic regression analysis (bilirubin and international normalized ratio [INR] were set at 1.0 for values less than 1.0), new equations from model 3, model 4, and model 5 were constructed and are listed below: MELD HCO3 ¼ 8:38LN ðbilirubinÞ þ 24:59LN ðINRÞ þ 21:57LN ðCreatinineÞ þ ð0:13ÞLN ðHCO3 Þ þ 7:08 MELD PaCO2 HCO3 ¼ 37:16LN ðINRÞ þ 21:07LN ðCreatinineÞ þ ð2:68ÞLN ðPaCO2 Þ þ 1:49LN ðHCO3 Þ þ 11:40 MELD Bicarbonate ¼ 37:13LN ðINRÞ þ 21:09LN ðCreatinineÞ þ 0:03LN ðHCO3 Þ þ 9:29 The values obtained from these new equations were rounded off to the nearest integer for easy use. Unlike the
original MELD equation, there was no predefined upper limit in these new equations. On performing univariate Cox regression analysis (Table 4), all the scoring systems were found to have significant value in predicting ICU mortality. Three scoring systems remained significant (MELD-PaCO2-HCO3, MELDBicarbonate, and APACHE II) in further multivariate Cox regression analysis. In the subsequent ROC curve analysis, the MELD-Bicarbonate equation showed better value than the original MELD, Child-Pugh, and APACHE II scores (Table 5). When the patients were divided into two groups according to the cutoff point of the MELD-Bicarbonate equation in the ROC curve analysis, the survival curve showed that critically ill patients who had cirrhosis and MELD-Bicarbonate scores 23 had significant ICU mortality (p50.001; Figure 1).
Discussion This retrospective study evaluated the clinical value of ABG parameters in critically ill patients with cirrhosis. We found that both PaCO2 and HCO3 were independent predictors of ICU mortality. We subsequently incorporated PaCO2 and HCO3 into the original MELD score and three new equations
892
C.-Y. Chen et al.
Ren Fail, 2014; 36(6): 889–894
were constructed. The MELD-Bicarbonate score (including INR, creatinine, and HCO3) provided better results pertaining to the ROC curve than the original MELD, Child-Pugh, and APACHE II scores. Table 2. Parameters that associated with ICU mortality. Parameter
Odds ratios 95% CI
Univariate logistic regression analysis Mean arterial pressure (mm Hg) 0.97 APACHE II 1.11 Mechanical ventilation 6.72 Albumin (3.5–5 g/dL) 0.27 INR 4.29 Child-Pugh score 1.39 MELD score 1.13 BUN (8–12 mg/dL) 1.01 Creatinine (0.4–1.2 mg/dL) 1.36 GOT (15–41 IU/L) 1.00 Phosphate (2.7–4.5 mg/dL) 1.43 PaCO2 (32–45 mm Hg) 0.95 HCO3 (20–26 mmol/L) 0.87 SaO2 (%) 0.86 Base excess (2 to +2 mmol/L) 0.88 Multivariate logistic regression analysis Mean arterial pressure (mm Hg) 1.01 APACHE II 1.07 Mechanical ventilation 9.91 Albumin (3.5–5 g/dL) 0.39 INR 3.13 Child-Pugh score 0.72 MELD score 1.11 BUN (8–12 mg/dL) 1.03 Creatinine (0.4–1.2 mg/dL) 0.75 GOT (15–41 IU/L) 1.00 Phosphate (2.7–4.5 mg/dL) 1.24 PaCO2 (32–45 mm Hg) 0.71 HCO3 (20–26 mmol/L) 2.25 SaO2 (%) 0.79 Base excess (2 to +2 mmol/L) 0.60
p
0.9–1.0 1.1–1.2 2.5–18.1 0.1–0.5 2.1–8.6 1.2–1.6 1.1–1.2 1.0–1.0 1.1–1.6 1.0–1.0 1.2–1.7 0.9–1.0 0.8–0.9 0.8–1.0 0.8–0.9
50.001 50.001 50.001 50.001 50.001 50.001 50.001 0.003 0.001 0.033 50.001 0.004 50.001 0.011 50.001
1.0–1.0 1.0–1.2 1.7–56.4 0.1–1.4 0.5–20.3 0.5–1.2 1.0–1.3 1.0–1.1 0.5–1.2 1.0–1.0 0.9–1.8 0.7–0.9 1.0–4.8 0.6–1.0 0.4–1.0
NS (0.760) NS (0.134) 0.010 NS (0.151) NS (0.232) NS (0.170) NS (0.198) 0.048 NS (0.267) NS (0.372) NS (0.221) 0.004 0.038 NS (0.062) NS (0.053)
Statistical analysis was performed with univariate and multivariate logistic regression analysis. Abbreviation: ICU, intensive care unit; CI, confidence interval; APACHE II, acute physiology and chronic health evaluation II; MELD, model for end-stage liver disease; INR, international normalized ratio; BUN, blood urea nitrogen; NS, not significant.
Acute kidney injury (AKI) is an ominous and common condition in patients with cirrhosis,17,18 and it has been reported that ICU patients who have cirrhosis and require acute renal replacement therapy have a 94% hospital mortality rate.19 Similarly, Cholangitis et al. reported that patients who had cirrhosis and were admitted to the ICU with three or more failing organ systems had a 90% mortality rate.13 Recently, Das et al. also showed that the occurrence of three nonhematologic organ failures after 3 days was associated with a mortality rate of close to 90%.2 Therefore, it has been suggested that the number and severity of extrahepatic organ failures are the most important risk factors for mortality in critically ill patients with cirrhosis.2,20 Consequently, it is not surprising that the SOFA score showed that a better prognostic value than the MELD or Child-Pugh score, since the general ICU scores are more accurate at grading multiple organ failure.2,11–13 It is noteworthy that even when critically ill patients with cirrhosis develop multiple organ failure, the liver is central to the outcome; thus, assessment of liver function in the general ICU scores might be inappropriate. For instance, the SOFA score evaluates markers of neurologic, cardiovascular, renal, respiratory, hematologic, and hepatic dysfunction and includes two variables, namely, creatinine and bilirubin levels, which are also included in the MELD score; however, the weight given to creatinine and bilirubin is not the same for the two scoring systems. Second, INR, which represents coagulation and is a pivotal marker of liver dysfunction, is not included in the SOFA score. Thus, the prognostic value of general ICU scoring systems may diminish in specific subpopulations of critically ill patients (e.g., those with cirrhosis). The MELD score had good discrimination in predicting inhospital death in critically ill patients with cirrhosis.2,13–15 The MELD score is suggested to act as an organ failurespecific score rather than as a liver disease-specific score in the setting of ICU admission.2 Indeed, the MELD system does evaluate organ failure with the three main organ systems: renal, hepatic, and coagulation systems. There is an overwhelming body of evidence showing that renal dysfunction is associated with worse outcomes in patients with
Table 3. Relationship between PaCO2, HCO3, total bilirubin, INR, and MELD score as a predictor of ICU mortality. Models
Variables
Model 1 Model 2
MELD Total bilirubin INR Creatinine PaCO2 Total bilirubin INR Creatinine HCO3 INR Creatinine HCO3 INR Creatinine PaCO2 HCO3
Model 3
Model 4 Model 5
Hazard ratios (95% CI) 1.11 1.38 47.79 7.81 0.18 1.34 37.70 4.18 0.06 43.16 4.11 0.05 28.76 3.89 69.22 0.00
(1.1–1.1) (0.8–2.3) (9.2–248.4) (2.3–26.3) (0.0–1.1) (0.8–2.3) (7.1–201.5) (1.2–14.4) (0.0–0.2) (8.9–210.5) (1.2–14.2) (0.0–0.2) (6.0–137.8) (1.1–13.3) (4.9–969.8) (0.0–0.0)
p
Likelihood ratio test (p value)
2 Log Likelihood
50.001 NS (0.243) 50.001 0.001 NS (0.058) NS (0.276) 50.001 0.024 50.001 50.001 0.025 50.001 50.001 0.030 0.002 50.001
36.1 (50.001) 45.4 (50.001)
601.2 591.9
63.5 (50.001)
573.8
63.1 (50.001)
575.1
72.4 (50.001)
565.8
Statistical comparison was performed with Cox proportional-hazard regression analysis. Abbreviation: INR, international normalized ratio; MELD, model for end-stage liver disease; ICU, intensive care unit; CI, confidence interval; NS, not significant.
MELD score in the ICU
DOI: 10.3109/0886022X.2014.894765
Table 4. The comparisons between the new equations and existing scoring system in predicting ICU mortality. Variables
Hazard ratios
Univariate Cox regression analysis APACHE II 1.08 Child-Pugh score 1.32 MELD score 1.11 MELD–HCO3 1.11 MELD–PaCO2–HCO3 1.11 MELD–Bicarbonate 1.11 Multivariate Cox regression analysis APACHE II 1.08 Child-Pugh score 1.17 MELD score 0.93 MELD–HCO3 1.11 MELD–PaCO2–HCO3 0.35 MELD–Bicarbonate 3.07
95% CI
p
1.1–1.1 1.2–1.5 1.1–1.1 1.1–1.1 1.1–1.1 1.1–1.1
50.001 50.001 50.001 50.001 50.001 50.001
1.0–1.1 1.0–1.4 0.8–1.1 0.9–1.4 0.2–0.8 1.3–7.0
50.001 NS (0.051) NS (0.495) NS (0.306) 0.011 0.008
Statistical comparison was performed with univariate and multivariate Cox proportional-hazard regression analysis. Abbreviation: ICU, intensive care unit; CI, confidence interval; APACHE II, acute physiology and chronic health evaluation II; MELD, model for end-stage liver disease; NS, not significant. Table 5. Receiver operating characteristic curve between the new equations and existing scoring systems. Cut-off point Sensitivity Specificity
Source of curve
AUC
95% CI
APACHE II Child-Pugh score MELD score MELD–HCO3 MELD–PaCO2 –HCO3 MELD–Bicarbonate
0.71 0.68 0.73 0.73 0.76
0.63–0.80 0.60–0.76 0.65–0.81 0.66–0.81 0.69–0.84
23 11 25 25 25
0.61 0.55 0.63 0.63 0.66
0.73 0.70 0.73 0.73 0.78
0.76 0.69–0.84
23
0.75
0.78
Abbreviation: AUC, area under curve; CI, confidence interval; APACHE II, acute physiology and chronic health evaluation II; MELD, model for end-stage liver disease.
Figure 1. Survival curve of the two groups, which were divided according to the cutoff point of the MELD-Bicarbonate equation. ICU, intensive care unit; MELD, model for end-stage liver disease.
cirrhosis.13,17–19 Similarly, disseminated intravascular coagulation is suggested to be common in critically ill patients and is associated with worse outcomes.21 In contrast, different results have been obtained regarding the role of total bilirubin.
893
In a study that included 312 critically ill patients with cirrhosis, Cholangitis et al. found that bilirubin, but not INR or creatinine, is an independent predictor of short-term mortality.13 In another study that included 441 critically ill patients with cirrhosis, Cavallazzi et al. found that bilirubin is not an independent predictor of death.14 Our results also showed that bilirubin was not a significant predictor of ICU mortality and lost its significance in a different model involving Cox regression analysis (Table 3). A possible explanation is that the total bilirubin level, as a marker of liver function, may not represent all aspects of liver dysfunction in the ICU setting. This indicates that directly applying the original MELD score for critically ill patients with cirrhosis is insufficient. In this study, an effort had been made to identify objective and reproducible variables that can improve the prognostic value of the MELD score in the ICU population, while keeping it simple and reproducible. We found that incorporating the HCO3 level into the MELD system to replace the bilirubin level resulted in improvement in predicting ICU mortality. The development of metabolic acidosis is a common occurrence during critical illness.22 The association between acidosis and increase in multiple organ dysfunction and mortality in critically ill patients has been long known. Therefore, clinicians have developed the concept of endpoints of therapy or resuscitation,23 and one of the most commonly used endpoints is arterial base deficit/excess (BD/E). BD/E is directly calculated from the blood gas analysis results for PaCO2, pH, and the serum HCO3 level. However, BD/E determination generally requires an arterial puncture, and the procedure is painful, invasive, and costly and can be associated with complications. The serum HCO3 level, which can be routinely assessed as a part of the chemistry panel on ICU admission, has been shown to decrease in linear fashion with increasing acid load.24 In a retrospective study including 3102 patients, Martin et al. found that the serum HCO3 level was significantly correlated with arterial BD levels both at admission (p50.01) and throughout the ICU stay (p50.01). The authors concluded that the serum HCO3 level predicted the presence of significant metabolic acidosis more reliably (p50.01) than pH, the anion gap, and the lactate level. The serum HCO3 level at admission can also predict ICU mortality as accurately as the admission arterial BD and more accurately than either admission pH or the anion gap.25,26 It is noteworthy that arterial pH is stabilized by excretion or retention of acid or alkali through control of the plasma bicarbonate level by the kidney or the control of PaCO2 by the central nervous system and respiratory system; and AKI, usually a result of metabolic acidosis, is a common event in critically ill patients with cirrhosis.27 Serum sodium concentration has been recognized as an important prognostic factor in patients with cirrhosis and is associated with the hepatorenal syndrome, ascites, and death from liver disease.28 The MELDNa score, which incorporates serum sodium into the original MELD score, provides better calibration and discrimination of the risk of death among candidates for liver transplantation.29 However, its prognostic value is not better than the original MELD score in critically ill patients with cirrhosis (AUC: 0.77 vs. 0.77, p ¼ 0.35).14 Our results also showed that the serum sodium level was not a
894
C.-Y. Chen et al.
significant predictor of ICU mortality; therefore, it may not be a suitable parameter for incorporation into the MELD score in the ICU setting. This study was limited by its retrospective nature; therefore, the treatment strategy and subsequent outcomes could not be validated here. Furthermore, the SOFA score was not included in the database and could not be compared with the new equations here. In addition, we used ICU mortality, that is, short-term mortality, as our primary outcome, and a longer-term outcome, such as 3-month mortality, should be tested. A further prospective study is warranted to test and verify our conclusions. In summary, the prognosis of critically ill patients with cirrhosis is poor, and the number of organ failures is an important predictor of mortality. Although the MELD score is significant with respect to predicting the short-term mortality of critically ill patients with cirrhosis, its prognostic value is not better than the general ICU scoring systems. We found that HCO3 level assessment, an objective and reproducible laboratory test, has significant predictive value in critically ill patients with cirrhosis. In contrast, the predictive value of the total bilirubin level is not as prominent in this setting. The MELD-Bicarbonate equation, which includes three variables (INR, creatinine, and HCO3), showed better prognostic value than the original MELD score in critically ill patients with cirrhosis.
Declaration of interest The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.
References 1. Gines P, Fernandez J, Durand F, Saliba F. Management of critically-ill cirrhotic patients. J Hepatol. 2012; 56(Suppl 1):S13–S24. 2. Das V, Boelle PY, Galbois A, et al. Cirrhotic patients in the medical intensive care unit: early prognosis and long-term survival. Crit Care Med. 2010;38(11):2108–2116. 3. Tu KH, Jenq CC, Tsai MH, et al. Outcome scoring systems for short-term prognosis in critically ill cirrhotic patients. Shock. 2011; 36(5):445–450. 4. Planas R, Montoliu S, Balleste B, et al. Natural history of patients hospitalized for management of cirrhotic ascites. Clin Gastroenterol Hepatol. 2006;4(11):1385–1394. 5. Vincent JL, Moreno R. Clinical review: scoring systems in the critically ill. Crit Care. 2010;14(2):207. 6. Metnitz PG, Lang T, Vesely H, Valentin A, Le Gall JR. Ratios of observed to expected mortality are affected by differences in case mix and quality of care. Intensive Care Med. 2000; 26(10):1466–1472. 7. Murphy-Filkins R, Teres D, Lemeshow S, Hosmer DW. Effect of changing patient mix on the performance of an intensive care unit severity-of-illness model: how to distinguish a general from a specialty intensive care unit. Crit Care Med. 1996; 24(12):1968–1973. 8. Chen YW, Chen HH, Wang TE, Chang CW, Chang CW, Wu CJ. Difference between CKD-EPI and MDRD equations in calculating glomerular filtration rate in patients with cirrhosis. World J Gastroenterol. 2011;17(40):4532–4538. 9. Kamath PS, Kim WR, Advanced Liver Disease Study G. The model for end-stage liver disease (MELD). Hepatology. 2007; 45(3):797–805.
Ren Fail, 2014; 36(6): 889–894
10. Rabe C, Schmitz V, Paashaus M, et al. Does intubation really equal death in cirrhotic patients? Factors influencing outcome in patients with liver cirrhosis requiring mechanical ventilation. Intensive Care Med. 2004;30(8):1564–1571. 11. Wehler M, Kokoska J, Reulbach U, Hahn EG, Strauss R. Short-term prognosis in critically ill patients with cirrhosis assessed by prognostic scoring systems. Hepatology. 2001;34(2):255–261. 12. Chen YC, Tian YC, Liu NJ, et al. Prospective cohort study comparing sequential organ failure assessment and acute physiology, age, chronic health evaluation III scoring systems for hospital mortality prediction in critically ill cirrhotic patients. Int J Clin Pract. 2006;60(2):160–166. 13. Cholangitis E, Senzolo M, Patch D, et al. Risk factors, sequential organ failure assessment and model for end-stage liver disease scores for predicting short term mortality in cirrhotic patients admitted to intensive care unit. Aliment Pharmacol Ther. 2006; 23(7):883–893. 14. Cavallazzi R, Awe OO, Vasu TS, et al. Model for end-stage liver disease score for predicting outcome in critically ill medical patients with liver cirrhosis. J Crit Care. 2012;27(4):424 e421–e426. 15. Thabut D, Massard J, Gangloff A, et al. Model for end-stage liver disease score and systemic inflammatory response are major prognostic factors in patients with cirrhosis and acute functional renal failure. Hepatology. 2007;46(6):1872–1882. 16. Kamath PS, Wiesner RH, Malinchoc M, et al. A model to predict survival in patients with end-stage liver disease. Hepatology. 2001; 33(2):464–470. 17. Chen YW, Wu CJ, Wang TE, Chang CW, Chang CW, Chen HH. The mortality survey of older patients with cirrhosis in Taiwan—a single-center experience. J Am Geriatr Soc. 2010; 58(11):2230–2232. 18. Chen YW, Wu CJ, Chang CW, Lee SY, Sun FJ, Chen HH. Renal function in patients with liver cirrhosis. Nephron Clin Pract. 2011;118(2):c195–c203. 19. Mackle IJ, Swann DG, Cook B. One year outcome of intensive care patients with decompensated alcoholic liver disease. Br J Anaesth. 2006;97(4):496–498. 20. Warrillow SJ. Predictions and outcomes for the critically ill patient with cirrhosis: is it time to settle on the SOFA and let jaundiced views on outcome MELD away? Crit Care Med. 2010; 38(11):2259–2260. 21. Gando S, Saitoh D, Ogura H, et al. Natural history of disseminated intravascular coagulation diagnosed based on the newly established diagnostic criteria for critically ill patients: results of a multicenter, prospective survey. Crit Care Med. 2008;36(1):145–150. 22. DuBose TD. Metabolic acidosis and alkalosis. In: Fink MP, Abraham E, Vincent JL, Kochanek PM, eds. Textbook of Critical Care. Philadelphia: Elsvier Saunders; 2005:1069–1083. 23. Rivers EP, Kruse JA, Jacobsen G, et al. The influence of early hemodynamic optimization on biomarker patterns of severe sepsis and septic shock. Crit Care Med. 2007;35(9):2016–2024. 24. Wiederseiner JM, Muser J, Lutz T, Hulter HN, Krapf R. Acute metabolic acidosis: characterization and diagnosis of the disorder and the plasma potassium response. J Am Soc Nephrol. 2004; 15(6):1589–1596. 25. Martin MJ, FitzSullivan E, Salim A, Berne TV, Towfigh S. Use of serum bicarbonate measurement in place of arterial base deficit in the surgical intensive care unit. Arch Surg. 2005;140(8):745–751. 26. FitzSullivan E, Salim A, Demetriades D, Asensio J, Martin MJ. Serum bicarbonate may replace the arterial base deficit in the trauma intensive care unit. Am J Surg. 2005;190(6):941–946. 27. Garcia-Tsao G, Parikh CR, Viola A. Acute kidney injury in cirrhosis. Hepatology. 2008;48(6):2064–2077. 28. Arroyo V, Colmenero J. Ascites and hepatorenal syndrome in cirrhosis: pathophysiological basis of therapy and current management. J Hepatol. 2003;38(Suppl 1):S69–S89. 29. Kim WR, Biggins SW, Kremers WK, et al. Hyponatremia and mortality among patients on the liver-transplant waiting list. N Engl J Med. 2008;359(10):1018–1026.
Copyright of Renal Failure is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
CLINICAL STUDY
Bicarbonate can improve the prognostic value of the MELD score for critically ill patients with cirrhosis Cheng-Yi Chen1,2, Chi-Feng Pan1,2, Chih-Jen Wu1,2, Han-Hsiang Chen1,2, and Yu-Wei Chen3 1
Division of Nephrology, Department of Internal Medicine, Mackay Memorial Hospital, Taipei, Taiwan, 2Mackay Junior College of Medicine, Nursing, and Management College, Taipei, Taiwan, and 3Division of Nephrology, Department of Internal Medicine, Ching Chyuan Hospital, Taichung, Taiwan Abstract
Keywords
The prognosis of critically ill patients with cirrhosis is poor. Our aim was to identify an objective variable that can improve the prognostic value of the Model of End-Stage Liver Disease (MELD) score in patients who have cirrhosis and are admitted to the intensive care unit (ICU). This retrospective cohort study included 177 patients who had liver cirrhosis and were admitted to the ICU. Data pertaining to arterial blood gas-related parameters and other variables were obtained on the day of ICU admission. The overall ICU mortality rate was 36.2%. The bicarbonate (HCO3) level was found to be an independent predictor of ICU mortality (odds ratio, 2.3; 95% confidence interval [CI], 1.0–4.8; p ¼ 0.038). A new equation was constructed (MELD-Bicarbonate) by replacing total bilirubin by HCO3 in the original MELD score. The area under the receiver operating characteristic curve for predicting ICU mortality was 0.76 (95% CI, 0.69–0.84) for the MELD-Bicarbonate equation, 0.73 (95% CI, 0.65–0.81) for the MELD score, and 0.71 (95% CI, 0.63–0.80) for the Acute Physiology and Chronic Health Evaluation II score. Bicarbonate level assessment, as an objective and reproducible laboratory test, has significant predictive value in critically ill patients with cirrhosis. In contrast, the predictive value of total bilirubin is not as prominent in this setting. The MELD-Bicarbonate equation, which included three variables (international normalized ratio, creatinine level, and HCO3 level), showed better prognostic value than the original MELD score in critically ill patients with cirrhosis.
Acidosis, bicarbonate, intensive care unit, liver cirrhosis, MELD
Introduction Regardless of the reason for intensive care unit (ICU) admission, cirrhosis independently worsens the prognosis of critically ill patients.1 Data from recent series involving critically ill patients with cirrhosis have shown ICU and 6-month mortality rates of 41 and 70.8%, respectively.2,3 Without definite treatment by way of liver transplantation, the prognosis of patients with cirrhosis is poor.4 Therefore, predicting the outcomes of critically ill patients with cirrhosis is important for medical decision-making and providing prognostic advice to patients. More than 10 different scoring systems are available for predicting the outcome of patients admitted to the ICU.5 Although the scores have an overall good prognostic value for a general population of critically ill patients, some of the prognostic value is diminished when these systems are applied to a specific subpopulation.6,7 Until now, the Acute Physiology and Chronic Health Evaluation II (APACHE II) and Simplified Organ Failure Assessment (SOFA) scores Address correspondence to Dr. Yu-Wei Chen, Division of Nephrology, Department of Internal Medicine, Ching Chyuan Hospital, No. 80, Sector. 4, Yatan Road, Daya Dist, Taichung City 428, Taiwan, R.O.C. Tel: +886-4-25605600; Fax: +886-4-25606162; E-mail: yw.chen. [email protected]
History Received 29 November 2013 Accepted 7 February 2014 Published online 6 March 2014
were most commonly used for assessing prognosis in the general population of patients admitted to the ICU. Although there is a paucity of scoring systems designed for critically ill patients with cirrhosis, several have been developed in a setting other than the ICU. The Child-Pugh score and the Model of End-Stage Liver Disease (MELD) score were initially designed to predict operative risk and mortality, respectively, for patients who have cirrhosis and are on the waiting list for liver transplantation. The MELD score has been proven to have prognostic value in a different setting also in the case of patients with cirrhosis.8,9 Several studies have compared the accuracy of liverspecific scores (Child-Pugh and MELD) to that of general ICU scores (APACHE II and SOFA).2,10–14 These studies suggest that the accuracy of the SOFA score is slightly superior to that of the APACHE II, MELD, and Child-Pugh scores.2,11–13 Therefore, Das et al. suggested that the prognosis of critically ill patients with cirrhosis is not only correlated with the severity of liver cirrhosis but also rather with the severity of organ failure.2 These general ICU scoring systems include complex and often subjective variables. In contrast, the MELD score relies on readily available and fully objective measures and has been proven to possess a significant prognostic value in patients with cirrhosis in the ICU setting.2,13–15 However, apart from renal, liver, and
890
C.-Y. Chen et al.
coagulation dysfunction, the MELD score does not take into account other organ/system dysfunction and has been suggested to be less accurate than the SOFA score. In this study, we hypothesized that data obtained for the arterial blood gas (ABG)-related parameters at the time of admission to the ICU can help predict ICU mortality in critically ill patients with cirrhosis. Furthermore, we incorporated these parameters into the MELD score and compared the prognostic value of the new equations and the original MELD, Child-Pugh, and APACHE II scores.
Patients and methods Ethics This study was conducted in accordance with the Declaration of Helsinki (2000) of the World Medical Association, and the protocol was approved by the institutional review board of Mackay Memorial Hospital and informed consent was waived. Mackay memorial Hospital is a tertiary referral center for liver disease. This study is a single center investigation and all patients of the study were all afferent, directly diagnosed and followed-up in Mackay memorial hospital. Patients This retrospective cohort study included consecutive adult patients who had cirrhosis and had been admitted to the ICU from August 2010 to April 2012. The diagnosis of cirrhosis was either confirmed by biopsy or was based on a combination of laboratory, endoscopic, and imaging findings. The records of 299 patients with cirrhosis from a total of 9735 ICU-admitted patients were reviewed. The exclusion criteria were end-stage renal disease treated with renal replacement therapy (10 cases), previous transplantation (of any tissue or organ), and no ABG examination at the time of ICU admission (112 cases). Of the 299 patients, 177 met the criteria and were included for analysis. Methods Data pertaining to the baseline characteristics of age, gender, etiological factors of liver disease and ICU admission, and underlying comorbidities were collected. Laboratory tests for ABG, liver function, renal function, and electrolytes were performed upon the patient’s admission to the ICU. ABG-related parameters such as pH, arterial partial pressure of carbon dioxide (PaCO2), arterial partial pressure of oxygen (PaO2), bicarbonate (HCO3), arterial oxygen saturation (SaO2), and base excess were recorded. Mortality during the ICU course was recorded as the primary endpoint. The MELD score was calculated using the following formula:16 MELD score ¼ 11.2LN (INR) + 3.78LN (bilirubin) + 9.57LN (creatinine) + 6.43. In this score, laboratory values less than 1 were set to 1. If the serum creatinine level was above 4 mg/dL, it was automatically set at 4 mg/dL. The score was rounded to the nearest whole number and capped at 40. Statistical analysis Continuous variables were summarized as mean ± SEM unless otherwise stated. We initially compared the demographic data
Ren Fail, 2014; 36(6): 889–894
and laboratory variables of survivors and non-survivors using Student’s t-test and Chi-square test. Correlations between variables and ICU mortality were assessed using logistic regression analysis. To formally examine the relationship among parameters of the ABG and the MELD as predictors of ICU mortality, several multivariate models were constructed. PaCO2 and HCO3, then, were incorporate into the MELD score after adjustment with the original MELD equation by logistic regression analysis. Cox proportional-hazard regression models were conducted for investigating the association between the different scoring system and ICU mortality. The results of these analyses were used to construct a receiveroperating characteristic (ROC) curve from which we sought the optimum cut-off point for predicting successful sites. The optimum cut-off point was defined as the point on the ROC curve closest to the point (0,1), where the false-positive rate was zero and the sensitivity was 100%. The area under the curve (AUC) and 95% confidence interval were calculated. Survival curves for critically-ill cirrhotic patients with different scores by the new scoring system were prepared according to the Kaplan–Meier method. A p value of less than 0.05 was considered statistically significant. All statistical analyses were performed using SPSS software version 17.0 (SPSS Inc., Chicago, IL).
Results Baseline demographic, clinical, and laboratory characteristics of the 177 enrolled patients are shown in Table 1. The overall ICU mortality rate was 36.2%. Patients were grouped into survivors (113 patients) and non-survivors (64 patients); of these patients, 131 (74%) received intubation with mechanical ventilator support (72 cases in the survivor group and 59 cases in the non-survivor group). The most common cause of ICU admission overall, that is, gastrointestinal bleeding, was also the most common cause of admission in the survivor group. The ABG-related parameters that showed a significant difference between survivors and non-survivors were pH, PaCO2, HCO3, SaO2, and base excess. Interestingly, the serum sodium level showed no significant difference between these two groups. Univariate logistic regression analysis was initially used to evaluate the association between the parameters and ICU mortality, and those with significant prognostic values are listed in Table 2. It is noteworthy that the total bilirubin level, one of the three variables of the MELD score, showed no significant difference between the two groups. Subsequent multivariate logistic regression analysis showed that both PaCO2 and HCO3 were independent predictors of ICU mortality. Table 3 shows several multivariate models for predicting ICU mortality by Cox regression analysis. Model 1, which was included in the original MELD score only, was a significant predictor. When the three parameters of the MELD score were included, the HCO3 level remained significant (model 3) but PaCO2 did not (model 2). As with the previous results, the total bilirubin level was insignificant in model 2 and model 3, and it was excluded subsequently in model 4 and model 5. All the parameters in model 4 and model 5 were significant, and both models were better than the original MELD score (model 1) with respect to the values of the
MELD score in the ICU
DOI: 10.3109/0886022X.2014.894765
891
Table 1. Demographic, clinical, and laboratory data for 177 critically ill patients with cirrhosis. Parameter
All patients (N ¼ 177)
ICU survivals (N ¼ 113)
ICU non-survivals (N ¼ 64)
p
58.1 ± 1.1 135:42 51 79.1 ± 1.4 20.8 ± 0.6 131
58.0 ± 1.3 88:25 35 83.0 ± 1.7 18.6 ± 0.7 72
58.2 ± 1.9 47:17 16 72.4 ± 2.3 24.6 ± 1.1 59
NS (0.937) NS (0.505) NS (0.399) 50.001 50.001 50.001 NS (0.051)
58 63 18 38
30 44 10 29
28 19 8 9
70 67 26 14 115 90 44 9.7 ± 0.2 22.8 ± 0.6 294.3 ± 77.4 2.6 ± 0.0 6.0 ± 0.5 1.7 ± 0.0 42.8 ± 2.6 2.6 ± 0.2 135.6 ± 0.5 4.2 ± 0.1 4.6 ± 0.3 7.4 ± 0.0 29.4 ± 0.7 130.5 ± 4.8 18.1 ± 0.5 97.2 ± 0.2 5.7 ± 0.6
43 46 18 6 68 54 27 9.2 ± 0.2 20.4 ± 0.7 160.8 ± 38.2 2.7 ± 0.1 5.4 ± 0.6 1.5 ± 0.0 36.8 ± 2.9 2.1 ± 0.1 135.8 ± 0.6 4.2 ± 0.1 3.7 ± 0.2 7.4 ± 0.0 30.9 ± 0.8 129.4 ± 4.8 20.3 ± 0.6 97.7 ± 0.2 3.0 ± 0.6
27 21 8 8 47 36 17 10.7 ± 0.3 27.1 ± 1.0 537.7 ± 204.7 2.3 ± 0.1 7.2 ± 0.9 2.0 ± 0.1 53.5 ± 4.8 3.5 ± 0.4 135.4 ± 1.0 4.3 ± 0.1 5.9 ± 0.5 7.3 ± 0.0 26.8 ± 1.2 132.4 ± 10.3 14.3 ± 0.9 96.3 ± 0.6 10.6 ± 1.1
Age (years) Male: Female Diabetes mellitus Mean arterial pressure (mm Hg) APACHE II Mechanical ventilation Etiology of ICU admission Septic shock Gastrointestinal bleeding Isolated acute respiratory failure Others Etiology of liver cirrhosis Viral hepatitis Alcoholic Alcohol with viral hepatitis Cryptogenic Ascites Hepatic encephalopathy Hepatoma Child-Pugh score MELD score GOT (15–41 IU/L) Albumin (3.5–5 g/dL) Total bilirubin (0.3–1.2 mg/dL) INR BUN (8–12 mg/dL) Creatinine (0.4–1.2 mg/dL) Sodium (136–144 mEq/L) Potassium (3.5–5.1 mEq/L) Phosphate (2.7–4.5 mg/dL) pH (7.35–7.45) PaCO2 (32–45 mm Hg) PaO2 (75–100 mm Hg) HCO3 (20–26 mmol/L) SaO2 (%) Base excess (2 to +2 mmol/L)
NS (0.278)
NS (0.076) NS (0.279) NS (0.693) 50.001 50.001 NS (0.075) 50.001 NS (0.087) 50.001 0.003 0.001 NS (0.742) NS (0.260) 50.001 50.001 0.003 NS (0.771) 50.001 0.019 50.001
Values were expressed as mean ± SEM unless otherwise defined. Statistical comparison was performed with student’s t test or chi-square test. Abbreviation: APACHE II, acute physiology and chronic health evaluation II; ICU, intensive care unit; MELD, model for end-stage liver disease; INR, international normalized ratio; BUN, blood urea nitrogen; NS, not significant.
likelihood ratio test and 2 log likelihood. After adjustment with the original MELD equation by logistic regression analysis (bilirubin and international normalized ratio [INR] were set at 1.0 for values less than 1.0), new equations from model 3, model 4, and model 5 were constructed and are listed below: MELD HCO3 ¼ 8:38LN ðbilirubinÞ þ 24:59LN ðINRÞ þ 21:57LN ðCreatinineÞ þ ð0:13ÞLN ðHCO3 Þ þ 7:08 MELD PaCO2 HCO3 ¼ 37:16LN ðINRÞ þ 21:07LN ðCreatinineÞ þ ð2:68ÞLN ðPaCO2 Þ þ 1:49LN ðHCO3 Þ þ 11:40 MELD Bicarbonate ¼ 37:13LN ðINRÞ þ 21:09LN ðCreatinineÞ þ 0:03LN ðHCO3 Þ þ 9:29 The values obtained from these new equations were rounded off to the nearest integer for easy use. Unlike the
original MELD equation, there was no predefined upper limit in these new equations. On performing univariate Cox regression analysis (Table 4), all the scoring systems were found to have significant value in predicting ICU mortality. Three scoring systems remained significant (MELD-PaCO2-HCO3, MELDBicarbonate, and APACHE II) in further multivariate Cox regression analysis. In the subsequent ROC curve analysis, the MELD-Bicarbonate equation showed better value than the original MELD, Child-Pugh, and APACHE II scores (Table 5). When the patients were divided into two groups according to the cutoff point of the MELD-Bicarbonate equation in the ROC curve analysis, the survival curve showed that critically ill patients who had cirrhosis and MELD-Bicarbonate scores 23 had significant ICU mortality (p50.001; Figure 1).
Discussion This retrospective study evaluated the clinical value of ABG parameters in critically ill patients with cirrhosis. We found that both PaCO2 and HCO3 were independent predictors of ICU mortality. We subsequently incorporated PaCO2 and HCO3 into the original MELD score and three new equations
892
C.-Y. Chen et al.
Ren Fail, 2014; 36(6): 889–894
were constructed. The MELD-Bicarbonate score (including INR, creatinine, and HCO3) provided better results pertaining to the ROC curve than the original MELD, Child-Pugh, and APACHE II scores. Table 2. Parameters that associated with ICU mortality. Parameter
Odds ratios 95% CI
Univariate logistic regression analysis Mean arterial pressure (mm Hg) 0.97 APACHE II 1.11 Mechanical ventilation 6.72 Albumin (3.5–5 g/dL) 0.27 INR 4.29 Child-Pugh score 1.39 MELD score 1.13 BUN (8–12 mg/dL) 1.01 Creatinine (0.4–1.2 mg/dL) 1.36 GOT (15–41 IU/L) 1.00 Phosphate (2.7–4.5 mg/dL) 1.43 PaCO2 (32–45 mm Hg) 0.95 HCO3 (20–26 mmol/L) 0.87 SaO2 (%) 0.86 Base excess (2 to +2 mmol/L) 0.88 Multivariate logistic regression analysis Mean arterial pressure (mm Hg) 1.01 APACHE II 1.07 Mechanical ventilation 9.91 Albumin (3.5–5 g/dL) 0.39 INR 3.13 Child-Pugh score 0.72 MELD score 1.11 BUN (8–12 mg/dL) 1.03 Creatinine (0.4–1.2 mg/dL) 0.75 GOT (15–41 IU/L) 1.00 Phosphate (2.7–4.5 mg/dL) 1.24 PaCO2 (32–45 mm Hg) 0.71 HCO3 (20–26 mmol/L) 2.25 SaO2 (%) 0.79 Base excess (2 to +2 mmol/L) 0.60
p
0.9–1.0 1.1–1.2 2.5–18.1 0.1–0.5 2.1–8.6 1.2–1.6 1.1–1.2 1.0–1.0 1.1–1.6 1.0–1.0 1.2–1.7 0.9–1.0 0.8–0.9 0.8–1.0 0.8–0.9
50.001 50.001 50.001 50.001 50.001 50.001 50.001 0.003 0.001 0.033 50.001 0.004 50.001 0.011 50.001
1.0–1.0 1.0–1.2 1.7–56.4 0.1–1.4 0.5–20.3 0.5–1.2 1.0–1.3 1.0–1.1 0.5–1.2 1.0–1.0 0.9–1.8 0.7–0.9 1.0–4.8 0.6–1.0 0.4–1.0
NS (0.760) NS (0.134) 0.010 NS (0.151) NS (0.232) NS (0.170) NS (0.198) 0.048 NS (0.267) NS (0.372) NS (0.221) 0.004 0.038 NS (0.062) NS (0.053)
Statistical analysis was performed with univariate and multivariate logistic regression analysis. Abbreviation: ICU, intensive care unit; CI, confidence interval; APACHE II, acute physiology and chronic health evaluation II; MELD, model for end-stage liver disease; INR, international normalized ratio; BUN, blood urea nitrogen; NS, not significant.
Acute kidney injury (AKI) is an ominous and common condition in patients with cirrhosis,17,18 and it has been reported that ICU patients who have cirrhosis and require acute renal replacement therapy have a 94% hospital mortality rate.19 Similarly, Cholangitis et al. reported that patients who had cirrhosis and were admitted to the ICU with three or more failing organ systems had a 90% mortality rate.13 Recently, Das et al. also showed that the occurrence of three nonhematologic organ failures after 3 days was associated with a mortality rate of close to 90%.2 Therefore, it has been suggested that the number and severity of extrahepatic organ failures are the most important risk factors for mortality in critically ill patients with cirrhosis.2,20 Consequently, it is not surprising that the SOFA score showed that a better prognostic value than the MELD or Child-Pugh score, since the general ICU scores are more accurate at grading multiple organ failure.2,11–13 It is noteworthy that even when critically ill patients with cirrhosis develop multiple organ failure, the liver is central to the outcome; thus, assessment of liver function in the general ICU scores might be inappropriate. For instance, the SOFA score evaluates markers of neurologic, cardiovascular, renal, respiratory, hematologic, and hepatic dysfunction and includes two variables, namely, creatinine and bilirubin levels, which are also included in the MELD score; however, the weight given to creatinine and bilirubin is not the same for the two scoring systems. Second, INR, which represents coagulation and is a pivotal marker of liver dysfunction, is not included in the SOFA score. Thus, the prognostic value of general ICU scoring systems may diminish in specific subpopulations of critically ill patients (e.g., those with cirrhosis). The MELD score had good discrimination in predicting inhospital death in critically ill patients with cirrhosis.2,13–15 The MELD score is suggested to act as an organ failurespecific score rather than as a liver disease-specific score in the setting of ICU admission.2 Indeed, the MELD system does evaluate organ failure with the three main organ systems: renal, hepatic, and coagulation systems. There is an overwhelming body of evidence showing that renal dysfunction is associated with worse outcomes in patients with
Table 3. Relationship between PaCO2, HCO3, total bilirubin, INR, and MELD score as a predictor of ICU mortality. Models
Variables
Model 1 Model 2
MELD Total bilirubin INR Creatinine PaCO2 Total bilirubin INR Creatinine HCO3 INR Creatinine HCO3 INR Creatinine PaCO2 HCO3
Model 3
Model 4 Model 5
Hazard ratios (95% CI) 1.11 1.38 47.79 7.81 0.18 1.34 37.70 4.18 0.06 43.16 4.11 0.05 28.76 3.89 69.22 0.00
(1.1–1.1) (0.8–2.3) (9.2–248.4) (2.3–26.3) (0.0–1.1) (0.8–2.3) (7.1–201.5) (1.2–14.4) (0.0–0.2) (8.9–210.5) (1.2–14.2) (0.0–0.2) (6.0–137.8) (1.1–13.3) (4.9–969.8) (0.0–0.0)
p
Likelihood ratio test (p value)
2 Log Likelihood
50.001 NS (0.243) 50.001 0.001 NS (0.058) NS (0.276) 50.001 0.024 50.001 50.001 0.025 50.001 50.001 0.030 0.002 50.001
36.1 (50.001) 45.4 (50.001)
601.2 591.9
63.5 (50.001)
573.8
63.1 (50.001)
575.1
72.4 (50.001)
565.8
Statistical comparison was performed with Cox proportional-hazard regression analysis. Abbreviation: INR, international normalized ratio; MELD, model for end-stage liver disease; ICU, intensive care unit; CI, confidence interval; NS, not significant.
MELD score in the ICU
DOI: 10.3109/0886022X.2014.894765
Table 4. The comparisons between the new equations and existing scoring system in predicting ICU mortality. Variables
Hazard ratios
Univariate Cox regression analysis APACHE II 1.08 Child-Pugh score 1.32 MELD score 1.11 MELD–HCO3 1.11 MELD–PaCO2–HCO3 1.11 MELD–Bicarbonate 1.11 Multivariate Cox regression analysis APACHE II 1.08 Child-Pugh score 1.17 MELD score 0.93 MELD–HCO3 1.11 MELD–PaCO2–HCO3 0.35 MELD–Bicarbonate 3.07
95% CI
p
1.1–1.1 1.2–1.5 1.1–1.1 1.1–1.1 1.1–1.1 1.1–1.1
50.001 50.001 50.001 50.001 50.001 50.001
1.0–1.1 1.0–1.4 0.8–1.1 0.9–1.4 0.2–0.8 1.3–7.0
50.001 NS (0.051) NS (0.495) NS (0.306) 0.011 0.008
Statistical comparison was performed with univariate and multivariate Cox proportional-hazard regression analysis. Abbreviation: ICU, intensive care unit; CI, confidence interval; APACHE II, acute physiology and chronic health evaluation II; MELD, model for end-stage liver disease; NS, not significant. Table 5. Receiver operating characteristic curve between the new equations and existing scoring systems. Cut-off point Sensitivity Specificity
Source of curve
AUC
95% CI
APACHE II Child-Pugh score MELD score MELD–HCO3 MELD–PaCO2 –HCO3 MELD–Bicarbonate
0.71 0.68 0.73 0.73 0.76
0.63–0.80 0.60–0.76 0.65–0.81 0.66–0.81 0.69–0.84
23 11 25 25 25
0.61 0.55 0.63 0.63 0.66
0.73 0.70 0.73 0.73 0.78
0.76 0.69–0.84
23
0.75
0.78
Abbreviation: AUC, area under curve; CI, confidence interval; APACHE II, acute physiology and chronic health evaluation II; MELD, model for end-stage liver disease.
Figure 1. Survival curve of the two groups, which were divided according to the cutoff point of the MELD-Bicarbonate equation. ICU, intensive care unit; MELD, model for end-stage liver disease.
cirrhosis.13,17–19 Similarly, disseminated intravascular coagulation is suggested to be common in critically ill patients and is associated with worse outcomes.21 In contrast, different results have been obtained regarding the role of total bilirubin.
893
In a study that included 312 critically ill patients with cirrhosis, Cholangitis et al. found that bilirubin, but not INR or creatinine, is an independent predictor of short-term mortality.13 In another study that included 441 critically ill patients with cirrhosis, Cavallazzi et al. found that bilirubin is not an independent predictor of death.14 Our results also showed that bilirubin was not a significant predictor of ICU mortality and lost its significance in a different model involving Cox regression analysis (Table 3). A possible explanation is that the total bilirubin level, as a marker of liver function, may not represent all aspects of liver dysfunction in the ICU setting. This indicates that directly applying the original MELD score for critically ill patients with cirrhosis is insufficient. In this study, an effort had been made to identify objective and reproducible variables that can improve the prognostic value of the MELD score in the ICU population, while keeping it simple and reproducible. We found that incorporating the HCO3 level into the MELD system to replace the bilirubin level resulted in improvement in predicting ICU mortality. The development of metabolic acidosis is a common occurrence during critical illness.22 The association between acidosis and increase in multiple organ dysfunction and mortality in critically ill patients has been long known. Therefore, clinicians have developed the concept of endpoints of therapy or resuscitation,23 and one of the most commonly used endpoints is arterial base deficit/excess (BD/E). BD/E is directly calculated from the blood gas analysis results for PaCO2, pH, and the serum HCO3 level. However, BD/E determination generally requires an arterial puncture, and the procedure is painful, invasive, and costly and can be associated with complications. The serum HCO3 level, which can be routinely assessed as a part of the chemistry panel on ICU admission, has been shown to decrease in linear fashion with increasing acid load.24 In a retrospective study including 3102 patients, Martin et al. found that the serum HCO3 level was significantly correlated with arterial BD levels both at admission (p50.01) and throughout the ICU stay (p50.01). The authors concluded that the serum HCO3 level predicted the presence of significant metabolic acidosis more reliably (p50.01) than pH, the anion gap, and the lactate level. The serum HCO3 level at admission can also predict ICU mortality as accurately as the admission arterial BD and more accurately than either admission pH or the anion gap.25,26 It is noteworthy that arterial pH is stabilized by excretion or retention of acid or alkali through control of the plasma bicarbonate level by the kidney or the control of PaCO2 by the central nervous system and respiratory system; and AKI, usually a result of metabolic acidosis, is a common event in critically ill patients with cirrhosis.27 Serum sodium concentration has been recognized as an important prognostic factor in patients with cirrhosis and is associated with the hepatorenal syndrome, ascites, and death from liver disease.28 The MELDNa score, which incorporates serum sodium into the original MELD score, provides better calibration and discrimination of the risk of death among candidates for liver transplantation.29 However, its prognostic value is not better than the original MELD score in critically ill patients with cirrhosis (AUC: 0.77 vs. 0.77, p ¼ 0.35).14 Our results also showed that the serum sodium level was not a
894
C.-Y. Chen et al.
significant predictor of ICU mortality; therefore, it may not be a suitable parameter for incorporation into the MELD score in the ICU setting. This study was limited by its retrospective nature; therefore, the treatment strategy and subsequent outcomes could not be validated here. Furthermore, the SOFA score was not included in the database and could not be compared with the new equations here. In addition, we used ICU mortality, that is, short-term mortality, as our primary outcome, and a longer-term outcome, such as 3-month mortality, should be tested. A further prospective study is warranted to test and verify our conclusions. In summary, the prognosis of critically ill patients with cirrhosis is poor, and the number of organ failures is an important predictor of mortality. Although the MELD score is significant with respect to predicting the short-term mortality of critically ill patients with cirrhosis, its prognostic value is not better than the general ICU scoring systems. We found that HCO3 level assessment, an objective and reproducible laboratory test, has significant predictive value in critically ill patients with cirrhosis. In contrast, the predictive value of the total bilirubin level is not as prominent in this setting. The MELD-Bicarbonate equation, which includes three variables (INR, creatinine, and HCO3), showed better prognostic value than the original MELD score in critically ill patients with cirrhosis.
Declaration of interest The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.
References 1. Gines P, Fernandez J, Durand F, Saliba F. Management of critically-ill cirrhotic patients. J Hepatol. 2012; 56(Suppl 1):S13–S24. 2. Das V, Boelle PY, Galbois A, et al. Cirrhotic patients in the medical intensive care unit: early prognosis and long-term survival. Crit Care Med. 2010;38(11):2108–2116. 3. Tu KH, Jenq CC, Tsai MH, et al. Outcome scoring systems for short-term prognosis in critically ill cirrhotic patients. Shock. 2011; 36(5):445–450. 4. Planas R, Montoliu S, Balleste B, et al. Natural history of patients hospitalized for management of cirrhotic ascites. Clin Gastroenterol Hepatol. 2006;4(11):1385–1394. 5. Vincent JL, Moreno R. Clinical review: scoring systems in the critically ill. Crit Care. 2010;14(2):207. 6. Metnitz PG, Lang T, Vesely H, Valentin A, Le Gall JR. Ratios of observed to expected mortality are affected by differences in case mix and quality of care. Intensive Care Med. 2000; 26(10):1466–1472. 7. Murphy-Filkins R, Teres D, Lemeshow S, Hosmer DW. Effect of changing patient mix on the performance of an intensive care unit severity-of-illness model: how to distinguish a general from a specialty intensive care unit. Crit Care Med. 1996; 24(12):1968–1973. 8. Chen YW, Chen HH, Wang TE, Chang CW, Chang CW, Wu CJ. Difference between CKD-EPI and MDRD equations in calculating glomerular filtration rate in patients with cirrhosis. World J Gastroenterol. 2011;17(40):4532–4538. 9. Kamath PS, Kim WR, Advanced Liver Disease Study G. The model for end-stage liver disease (MELD). Hepatology. 2007; 45(3):797–805.
Ren Fail, 2014; 36(6): 889–894
10. Rabe C, Schmitz V, Paashaus M, et al. Does intubation really equal death in cirrhotic patients? Factors influencing outcome in patients with liver cirrhosis requiring mechanical ventilation. Intensive Care Med. 2004;30(8):1564–1571. 11. Wehler M, Kokoska J, Reulbach U, Hahn EG, Strauss R. Short-term prognosis in critically ill patients with cirrhosis assessed by prognostic scoring systems. Hepatology. 2001;34(2):255–261. 12. Chen YC, Tian YC, Liu NJ, et al. Prospective cohort study comparing sequential organ failure assessment and acute physiology, age, chronic health evaluation III scoring systems for hospital mortality prediction in critically ill cirrhotic patients. Int J Clin Pract. 2006;60(2):160–166. 13. Cholangitis E, Senzolo M, Patch D, et al. Risk factors, sequential organ failure assessment and model for end-stage liver disease scores for predicting short term mortality in cirrhotic patients admitted to intensive care unit. Aliment Pharmacol Ther. 2006; 23(7):883–893. 14. Cavallazzi R, Awe OO, Vasu TS, et al. Model for end-stage liver disease score for predicting outcome in critically ill medical patients with liver cirrhosis. J Crit Care. 2012;27(4):424 e421–e426. 15. Thabut D, Massard J, Gangloff A, et al. Model for end-stage liver disease score and systemic inflammatory response are major prognostic factors in patients with cirrhosis and acute functional renal failure. Hepatology. 2007;46(6):1872–1882. 16. Kamath PS, Wiesner RH, Malinchoc M, et al. A model to predict survival in patients with end-stage liver disease. Hepatology. 2001; 33(2):464–470. 17. Chen YW, Wu CJ, Wang TE, Chang CW, Chang CW, Chen HH. The mortality survey of older patients with cirrhosis in Taiwan—a single-center experience. J Am Geriatr Soc. 2010; 58(11):2230–2232. 18. Chen YW, Wu CJ, Chang CW, Lee SY, Sun FJ, Chen HH. Renal function in patients with liver cirrhosis. Nephron Clin Pract. 2011;118(2):c195–c203. 19. Mackle IJ, Swann DG, Cook B. One year outcome of intensive care patients with decompensated alcoholic liver disease. Br J Anaesth. 2006;97(4):496–498. 20. Warrillow SJ. Predictions and outcomes for the critically ill patient with cirrhosis: is it time to settle on the SOFA and let jaundiced views on outcome MELD away? Crit Care Med. 2010; 38(11):2259–2260. 21. Gando S, Saitoh D, Ogura H, et al. Natural history of disseminated intravascular coagulation diagnosed based on the newly established diagnostic criteria for critically ill patients: results of a multicenter, prospective survey. Crit Care Med. 2008;36(1):145–150. 22. DuBose TD. Metabolic acidosis and alkalosis. In: Fink MP, Abraham E, Vincent JL, Kochanek PM, eds. Textbook of Critical Care. Philadelphia: Elsvier Saunders; 2005:1069–1083. 23. Rivers EP, Kruse JA, Jacobsen G, et al. The influence of early hemodynamic optimization on biomarker patterns of severe sepsis and septic shock. Crit Care Med. 2007;35(9):2016–2024. 24. Wiederseiner JM, Muser J, Lutz T, Hulter HN, Krapf R. Acute metabolic acidosis: characterization and diagnosis of the disorder and the plasma potassium response. J Am Soc Nephrol. 2004; 15(6):1589–1596. 25. Martin MJ, FitzSullivan E, Salim A, Berne TV, Towfigh S. Use of serum bicarbonate measurement in place of arterial base deficit in the surgical intensive care unit. Arch Surg. 2005;140(8):745–751. 26. FitzSullivan E, Salim A, Demetriades D, Asensio J, Martin MJ. Serum bicarbonate may replace the arterial base deficit in the trauma intensive care unit. Am J Surg. 2005;190(6):941–946. 27. Garcia-Tsao G, Parikh CR, Viola A. Acute kidney injury in cirrhosis. Hepatology. 2008;48(6):2064–2077. 28. Arroyo V, Colmenero J. Ascites and hepatorenal syndrome in cirrhosis: pathophysiological basis of therapy and current management. J Hepatol. 2003;38(Suppl 1):S69–S89. 29. Kim WR, Biggins SW, Kremers WK, et al. Hyponatremia and mortality among patients on the liver-transplant waiting list. N Engl J Med. 2008;359(10):1018–1026.
Copyright of Renal Failure is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
