A Validated  New Model for End-stage Liver Disease (NMELD)

Mohammed Ezzat Moemen, MD; Mostafa Mohammed  Ibrahim, MSc;    Howayda Ahmed Othman , MD; Khaled Mohammed El Sayed , MD.

Department of Anaesthesia and Intensive Care, Faculty of Medicine,     Zagazig University, Egypt.

Correspondence:  Professor  Mohammed Ezzat Moemen, Postal address: 2 Gamal  El- Din Wassel,  Nasr City, first  zone  , Cairo,  11371  , Egypt; E-mail:  ezzatmoemen@yahoo.com;   Cellular: (20)124051079.Website:

http://www.ezzatmoemen.com

To download and use the calculator of the New Model for End-stage Liver Disease (NMELD) scroll to the attachment at the bottom of the page.

ABSTRACT

Background The main author and his co-workers published a Pugh-modified score by adding four variables (serum sodium, serum creatinine, white blood count, and arterial/alveolar oxygen tension ratio), as a subjective-objective score. In the present study, we omit the subjective variables (encephalopathy and ascites) to describe a purely objective score. Objectives To correlate the two scores for outcome at 30 days in patients undergoing abdominal surgery. We hypothethised that a strong correlation of the objective score would facilitate a valid new model for end-stage liver disease (NMELD) to predict short-term patient outcome . Design Prospective study. Setting Zagazig University Hospitals, Egypt. Patients The study included a random sample of 100 cirrhotic patients undergoing abdominal operations. Patients with hepatocellular carcinoma, severe primary cardiopulmonary failure, or intrinsic kidney disease were excluded. We had approval of the local ethics committee and consent from all patients was taken.  General anaesthesia was given with 30 days postoperative follow-up. Methods Preoperative evaluation, correction of reversible factors, and calculation of both scores, followed by their correlation, and validation of the NMELD. An automatic calculator at http://www.ezzatmoemen.com website, was used for calculation of the scoring points and disease severity. Areas under the receiver operating characteristic (AUROC) curves for MELD and NMELD were evaluated for accuracy in predicting outcome. Main outcome measures Survival, mortality. Results There was a strong correlation between the two modified scores: observed agreement 0.89; “r” 0.97, P<0.001, and the equation of the NMELD was deduced. Statistical cross-validation: the equation was applied for original 100 patients (study group) and random 50 original patients (test group) respectively. There were high sensitivity, specificity, accuracy, predictive values, and likelihood ratios in both groups with non-significant differences (P > 0.05). Goodness of fit: X² 0.01; degrees of freedom 1; significance 0.94. There were higher statistical values for the NMELD when compared to the MELD in prediction of patient outcome.  The AUROC curves showed NMELD to have more diagnostic accuracy than MELD. Conclusion The NMELD is a simple valid   measure   to predict   30 day   postoperative  patient  outcome.

 Keywords:

Child-Pugh,  Cirrhosis,  MELD.

   

Background:

Cirrhosis is a final common pathway for chronic liver disease of different etiologies. Many patients with chronic liver disease undergo operative procedures during the final two years of their lives. Patients with cirrhosis carry a risk of perioperative mortality related to the extent of their liver dysfunction^(1). Preoperative recognition of occult liver disease is important because the stresses of anaesthesia and surgery can precipitate overt hepatic failure and other complications in these patients. For this reason, many studies have attempted to predict postoperative outcome in cirrhotic patients by assessing different variables. ^(1-7)

The Child-Turcotte scoring system ⁽³⁾ was designed in 1964 as a predictive formula for patients with liver disease undergoing portosystemic shunt surgery. It quantifies a patient's status with respect to serum albumin, serum  bilirubin, ascites, encephalopathy, and nutritional status. In 1973, Pugh et al ⁽⁴⁾ substituted the most subjective variable, nutritional status, by prothrombin time (PT). This scoring system is currently considered the cornerstone in the prognostic evaluation of cirrhotic patients although it was formulated more than three decades ago.

Moemen et al ⁽⁸⁾ in 2004 modified the five parameter Pugh scoring system by adding four variables; serum sodium, serum creatinine, white blood count, and arterial/alveolar oxygen tension ratio, as a subjective-objective scoring system (table 1). This modified score proved to be a more accurate and specific predictor of mortality. ⁽⁹⁾

Table (1): Moemen  subjective-objective  scoring  system ⁽⁸⁾

Scoring points: mild = 9-10, moderate = 11-14, severe = 15-27

Because of the subjective nature of some elements in this score,we hypothesized that a purely objective score would be more valid and valuable. We, therefore, modified the original subjective-objective score into a purely objective one by omitting the two subjective variables -   ascites  and  encephalopathy (table 2).

          Table (2):    Moemen   objective   scoring    system

Scoring points: mild = 7-8, moderate = 9-12, severe = 13-21

The present study aims to correlate Moemen subjective-objective with Moemen objective scoring systems in relation to 30 postoperative days patients' outcome. A positive correlation would facilitate a new model for end-stage liver disease (NMELD), based on purely objective parameters, that will be statistically validated ,and compared with the current model for end-stage liver disease (MELD).

Patients and Methods:

     The present study has been carried out on  100 consecutively randomized cirrhotic patients undergoing elective or urgent abdominal surgical procedures. The study was approved from the ethics committee of Zagazig University Hospitals and written consent was taken from all patients. Patients with hepatocellular carcinoma, severe primary cardiopulmonary failure, or intrinsic kidney disease were excluded. Abdominal surgery was chosen because it is well established that this is associated with the most severe acute decline in hepatic blood flow ⁽¹⁰⁾.

     Preoperative evaluation included full history taking, proper clinical examination, and carrying out a battery of investigations. Subjective clinical variables included ascites and encephalopathy, while objective laboratory variables included complete blood picture, serum albumin, liver function tests, renal function tests, coagulation profiles, electrolytes, and arterial blood gases.

Through preoperative preparation, reversible correctable variables were optimized for patients undergoing elective surgery, and as time allowed for those requiring urgent surgery. Correction included encephalopathy, ascites, anaemia, hypo-albuminemia, coagulopathy, endoscopic retrograde cholangio-pancreatography, mal-nutrition, anti-infection measures, and others. Then, new sets of subjective and objective variables were used to calculate the scoring points for the two systems.

All patients received general anaesthesia by the second author, through a standard anaesthetic technique with adequate perioperative monitoring and management. Postoperatively, patients were followed up to detect 30 days patient outcome.

 Statistical analysis was computed through the Epi-INFO (2004) and SPSS version 11 software packages. For descriptive statistics, the means, standard deviations (SD) and ranges were obtained for quantitative variables. The frequency distribution and percentages were calculated for qualitative variables. The student’s “t” and the Chi-square “X²” tests were used when appropriate.

Correlation between the means of the two modified scoring systems was carried out in relation to patient outcome over 30 postoperative days. When correlation proved strongly positive, we proceeded to elucidate a NMELD, based on Moemen objective scoring system. Formulation of the model equation was achieved by multiple linear regression analysis. The model equation would denote the liver disease severity classes, as mild (median score: ≤ 50%), moderate (third quartile: 51-75%), and severe (forth quartile: > 75%).

The NMELD scoring points and disease severity were calculated using a calculator at http://www.ezzatmoemen.com. Then, the NMELD equation was validated  through  a cross-validation  statistical  technique. The NMELD equation was applied to a study group of 100 patients (group I), and to a test group of 50 patients taken at random from the study group (group II). For cross-validation, comparison of the application of the equation on the two groups in relation to patient outcome by sensitivity, specificity, accuracy, predictive values, and likelihood ratios should exhibit non-significant differences. Goodness of fit was calculated by the Chi-square values. P<0.05 was considered statistically significant.

Comparison of the NMELD and MELD scores for predicting patient outcome over 30 postoperative days was carried out. Then, the diagnostic accuracies of the NMELD and MELD scoring systems were evaluated through the AUROC curves. Using this system a model is said to have diagnostic accuracy if the AUROC curve exceeds 0.70 and an excellent diagnostic accuracy if it exceeds 0.80.

Results:

One hundred cirrhotic patients undergoing abdominal operations under general anaesthesia were included in the study. They were 56 males and 44 females, with a mean age of 47.4 ± 11.9 ranging from 22 to 85 years. Sixty-nine elective procedures included cholecystectomy, splenectomy, and excision for cancer colon and cancer of the head of pancreas. Thirty-one emergency procedures included abdominal explorations for peritonitis or bleeding. Twenty five patients died and 75 survived during the 30 day postoperative period.

 A strongly positive significant correlation between the two scoring systems in relation to postoperative patient outcome was observed, with an observed agreement  = 0.89; “r” = 0.97; and  P<0.001.

The NMELD equation was formulated, based on Moemen objective scoring system, by multiple linear regression analysis. Determination coefficient "R²" = 0.779, and P = 0.00, with a high correlation between the predicted equation and Moemen objective score (r =0.89).  NMELD  = 17.9 + [0.35 x total bilirubin (mg/dl)] – [1.03 x serum albumin (g/dl)] – [0.05 x serum sodium (meq/L)] + [2.07 x INR] + [0.48 x serum creatinine (mg/dl)] + (0.11 x WBC's/10³] – [2.94 x P(a/A)O₂], where INR: international normalized ratio, WBC: white blood count /mm³ , P(a/AO₂): arterial/alveolar oxygen tension ratio.

  The disease  severity  was classified  into mild, moderate, and severe with round ≤ 8,  9 – 15,  and > 15 scoring  points  respectively. Validation of the NMELD was proved by the cross-validation statistical   technique  (table 3).

  Table (3): Cross-validation  of  the NMELD scoring system                                                                                                                                                                                                                                                                                  

Patients	Sensitivity %	Specificity %	PV + %	PV-%	Accuracy %	LR+	LR-
Group I	94.1	88.0	61.5	98.6	89	7.840	0.067
Group II	88.9	87.8	61.5	97.3	88	7.400	0.126
P value	>0.05	>0.05	>0.05	>0.05	>0.05	>0.05	>0.05

Group I: study group, Group II: test group, PV+: positive predictive value, PV-: negative predictive value, LR+: likelihood ratio for a positive result, LR-: likelihood ratio for a negative result.

    Goodness of fit for the NMELD showed X² 0.01, degrees of freedom 1, and significance of 0.94, proving that the NMELD fits the data.

       Compared to the current MELD in predicting patient outcome, the NMELD showed higher statistical significance for sensitivity, specificity, predictive values, accuracy, and likelihood ratios (table 4 ).

Table (4): MELD versus NMELD in predicting patient outcome                                                                                                                                                                                                                                                                                   

Score	Sensitivity %	Specificity %	PV + %	PV-  %	Accuracy %	LR+	LR-
MELD	84	30.7	28.8	85.2	44	1.2	0.52
NMELD	96	66.7	49	98	74	1.56	0.25
P value	<0.001	<0.001	<0.01	<0.001	<0.001	<0.01	<0.01

MELD: model for end-stage liver disease; NMELD: new model for end-stage liver disease; PV+: positive predictive value; PV-: negative predictive value; LR+: likelihood   ratio  for  a positive  result; LR-: likelihood   ratio  for  a  negative  result.

    The AUROC curve  for MELD showed diagnostic accuracy, while the  AUROC curve for NMELD showed excellent diagnostic accuracy ( 0.78  and  0.85 , respectively).

    Discussion:

   The Child-Turcotte ⁽³⁾ score was formulated in 1964, to evaluate hepatic patients undergoing portosystemic shunt surgery. It quantifies patient status with respect to serum albumin, serum bilirubin, ascites, encephalopathy, and nutritional status. Consisting of two laboratory tests and three clinical observations, the score provides a rough estimate of the hepatic reserve, a reality which should be weighed before undergoing surgery in patients with seemingly marginal liver function ⁽¹¹⁾

    Pugh et al⁽⁴⁾, in 1973, substituted the most subjective variable, nutritional status, by the prothrombin time. However, it still includes two subjective variables; ascites and encephalopathy; a main drawback of the score. Again, grading the disease severity into A, B, and C classes (5-6, 7-9, and 10-15 scoring points, respectively) represents a ceiling effect for group C patients - another drawback of  the  score.

   Although the Pugh score was originally formulated by a purely empiric observation to predict outcomes of 38 hepatic patients with bleeding varices undergoing palliative transection of the oesophagus, it was subsequently used beyond its original intended scope, by different authors to assess hepatic patients before any surgery, and by the United Network for Organ Sharing (UNOS) to prioritize patients on the waiting list for liver transplantation.

     In 2004, we introduced a modified-Pugh score⁽⁸⁾, through a prospective study by adding four variables to the original score, for better patient assessment. The variables were serum sodium, serum creatinine, white blood count, and arterial/alveolar oxygen tension ratio. The score proved to be more accurate and specific  than the original Pugh et al score⁽⁹⁾. However, this subjective-objective score still carries the  drawbacks  of the original Pugh score. In the present study, we report the effect of changing it into a purely objective score, by omitting the two subjective variables; ascites and encephalopathy.

    The present study proved a strong correlation between the original subjective-objective and the purely objective score; with an observed agreement of 0.89, r = 0.97, and P<0.001. Again, there was a high correlation between the predicted NMELD equation and Moemen objective  score (''r''=0.89),  with  determination coefficient of 0.779 and p =0.00. Unavoidable variability in assessing ascites and encephalopathy makes a prognostic system based purely on objective laboratory parameters an obvious asset, specially when cirrhotic patients appear in the emergency room for urgent surgery ⁽¹²⁾ .

   In 1999, the UNOS formulated the model for end-stage liver disease (MELD) as an objective assessment tool. It was developed by physicians at the Mayo Clinic for patients undergoing transjuglar-intrahepatic- portosystemic-shunt-surgery. It is based on serum bilirubin, serum creatinine, prothrombin time expressed as international normalized ratio, and the cause of liver disease. ⁽¹³⁾ However, in 2000, the cause of liver disease was dropped because it proved to be prognostically  insignificant ⁽¹⁴⁾. The score can readily be estimated through a calculator downloaded from the Mayo Clinic website http://www.mayoclinic.com. This score cannot, however, be calculated at the bedside as it is complex, sophisticated, and includes logarithmic transformations and multiplication by several factors ⁽¹⁵⁾, but it does not have either of the drawbacks of the Pugh score. The MELD score has gained acceptance over the Pugh score, as a more reliable measure of short-term mortality  risk  in  patients with  end-stage  liver  disease ⁽¹⁵⁾. On the other hand, Botta et al ⁽¹⁶⁾ proved that Pugh and MELD scores have similar accuracy, and recommended their complementary use in evaluating the prognostic worsening of the clinical condition of the cirrhotic patient. Papatheodoridis et al ⁽¹⁷⁾ showed that neither of the scores had excellent diagnostic accuracy. It deserves mentioning that mono-ethyl-glycine-xylidide (MEGX), as a quantitative test , has been combined with the Pugh score to create a composite score in an attempt to improve it's prognostic accuracy ⁽¹⁸⁾ .

    Our modified Pugh scores represent a step in the right direction for the assessment of hepatic patients undergoing surgery. Angermayr et al ⁽¹⁹⁾ added serum creatinine as a sixth categorial variable to the original Pugh score, and  Ruf et al ⁽²⁰⁾ added serum sodium into the MELD score. Both serum creatinine and serum sodium are two out of the four variables we added to the original Pugh  score ⁽⁸⁾ .Recently, Papatheodoridis  et  al ⁽¹⁷⁾ hold the idea that if two or more scoring systems offer similar accuracy in predicting survival, then other parameters should be taken into account for adopting one of them into clinical practice. Again, Francois Durand ⁽²¹⁾ introduces an inquiry: if scores have to be complex (such as MELD score is) with the need of computation, why not to take into account more specific variables shown to weigh heavily on a given event?

    The present study proved that the NMELD offers higher sensitivity, specificity, predictive values, accuracy, and likelihood ratios than the current MELD score in predicting short-term patient outcome. It is more dynamic than the MELD. It is simple and easier to calculate at the bedside by using a manual calculator, not necessarily a computer program, since it is devoid of logarithmic  transformations  and  multiplications.

     The NMELD scoring system showed excellent diagnostic accuracy with an AUROC curve of 0.85, denoting  superior discrimination    than  the  MELD  scoring  system.

     We conclude that the NMELD scoring system is a simple and valid measure to predict 30 day patient outcome, with higher diagnostic accuracy compared to the MELD scoring system. Further studies on  much  larger  surgical  hepatic  populations  are  recommended.

 

Acknowledgment:

       The  authors  would  like  to   thank  doctor  Hosnia    Mohammed El-Sayed, professor  of  community,  faculty  of  medicine, Zagazig university, Egypt; who  provided  the  statistical   advice.

References:

1.         Doberneck, R.C.; Sterling, W.A. and Laaison, D.C.: Morbidity and mortality after operations in non-bleeding cirrhotic patients. Am J Surg  1983; 146: 306-9.

2.         Cayer, D. and Sohmer, M.F.: Surgery in patients with cirrhosis. Arch Surg 1955; 71: 828-38.

3.         Child, C.G. and Turcotte, J.G.: Surgery and portal hypertension. In: Child CG (editor). The liver and portal hypertension. Philadelphia, WB Saunders; 1964; 34-45.

4.         Pugh, R.N.H.; Murray-Lyon, I.M. and Dawson, J.L. et al.: Transection of the oesophagus for bleeding oesophageal varices. Br J Surg 1973; 60: 646-9.

5.         Wirthlin, L.S.; Urk, H.V. and Malt, R.B. et al.: Predictors of surgical mortality in patients with cirrhosis and non-variceal gastroduodenal bleeding. Surg  Gynecol  Obstet 1974; 139: 65-8.

6.         Garrison, R.N.; Cryer, H.M. and Howard, D.A. et al.: Classification of risk factors for abdominal operations in patients with hepatic cirrhosis. Ann. Surg. 1984 ; 199: 648-55.

7.         Kamath, P.S, Russel H.W., Malinchoc,M., et al. A model to predict survival in patients with end-stage liver disease. Hepatology 2001 ; 33: 464-70.

8.         Moemen, M.E.; Gaafar, T.Y. and Farag, W.A. et al.: Prognostic categorization in cirrhotic patients undergoing abdominal surgery: A randomized trial. Eg. J. Anaesth 2004; 20: 7-14.

9.         Webster, N.R.: Scoring systems and surgery in cirrhotic patients. Eg. J. Anaesth.2004 ; 20: 1-3.

10.      Gelman SI. (1976): Disturbances in hepatic blood flow during anaesthesia and surgery. Arch Surg  1976; 111: 881-3.

11.     Jeremiah G Turcotte, Matthew J Lambert. Variceal hemorrhage, hepatic cirrhosis, and  portocaval  shunts . Surgery 1973; 73 : 810-17.

12.     Farnsworth BA, Fagan SP, Berger DH. Child-Turcotte-Pugh versus MELD score as a predictor of outcome after elective and emergency surgery in cirrhotic patients. Amer  J  Surg  2004; 188: 580-3.

13.     Kamath PS, Wiesner RH, Malinchoc M, el al. A model to predict survival in patients with end-stage liver disease. Hepatology 2001; 33: 464-70.

14.     Pagliaro L.MELD: the end of Child-Pugh classification J Hepatology 2002; 36: 141-2.

15.     Forman L, Lucey M. Predicting the prognosis of chronic liver disease: an evaluation from CHILD to MELD. Hepatology 2001; 33:473-5.

16.     Botta F, Giannini E, Romagnoli P, et al. MELD scoring system is useful for predicting prognosis in patients with liver cirrhosis and is correlated with residual liver function: a Euopean study.GUT 2003;52: 134-9.

17.     Papatheodorids GV, Cholongitas E, Dimitriadou E, et al. MELD  versus Child-Pugh and creatinine-modified Child-Pugh scores for predicting survival in patients with decompensated cirrhosis. World Gastroenterol  2005; 11 : 3099-104.

18.     Testa R,Valente U, Risso D, et al. Can the MEGX test and serum bile acids improve the prognostic ability of Child-Pugh's score in liver cirrhosis ?  Europ J Gastroenterol  Hepatol 1999; 11: 559-63.

19.     Angermayr B, Koenig F, Cejna M, et al. Creatinine –modified Child-Pugh score (CPSC) compared with MELD-score to predict survival in patients undergoing TIPS. Hepatology 2002; 36: 378A.

20.     Ruf AE, Kremers WK, Chaves LL, et al. Addition of serum sodium into the MELD score predicts waiting list mortality better than MELD. Transplantation 2005; 11: 336-43.

21.     Francois Durand. Risk scores in cirrhotic patients: From non-transplant surgery to transplantation and back. J Hepatology 2006; 44: 620-1.