Casaril et al used Cox's proportional hazard regression model to correlate clinical and laboratory findings with prognosis in adult patients admitted with cirrhosis. This can help identify patients who may benefit from more aggressive therapy. The authors are from the University of Verona in Italy.
NOTE: The presentation in the paper is somewhat unclear. There are 2 places where I had to make assumptions about the prognostic score. The score should be carefully validated prior to use.
Study group: Patients with cirrhosis, with patients having hepatocellular carcinoma excluded. The initial study consisted of 98 adults ranging from 23 to 76 years of age (68 males, 30 females). A subsequent control group consisted of 53 adults from 28 to 70 years (39 males, 14 females).
Parameters found important on analysis:
(1) age in years
(2) leukocyte count in 10^9/L
(3) potassium in mEq/L
(4) calcium in mg/dL
(5) previous diagnosis of cirrhosis made > 6 months previously
(6) serum globulins in g/dL
(7) serum total cholesterol in mg/dL
Parameter 
Cutoff Group 
Coefficient 
age 
< 40 
1 

40 – 49 
1.2575 

50 – 59 
1.9658 

> 60 
1.3455 
serum total cholesterol 
< 100 
1 

100 – 150 
1.3167 

151 – 200 
0.3600 

> 200 
0.6559 
from Table 4, page 73
Parameter 
Coefficient 
leukocyte count 
0.7333 
potassium 
0.4421 
calcium 
0.8233 
previous diagnosis 
0.7022 
serum globulins 
1.2459 
A previous history is graded as present (diagnosis of cirrhosis made > 6 months previously = 1) or absent (diagnosis not made > 6 months ago = 0).
Parameter 
Range 
Cutoff Group 
leukocyte count 
<= 4.0 
1 

4.1 – 9.0 
2 

> 9.0 
3 
potassium 
< 3.5 
1 

3.5 – 3.9 
2 

3.9 – 4.4 
3 

> 4.4 
4 
calcium 
< 7.5 
1 

7.5 – 8.5 
2 

8.6 – 9.0 
3 

> 9.0 
4 
serum globulins 
<= 3.0 
1 

3.1 – 4.0 
2 

> 4.0 
3 
where:
• The previous history is a dichotomous variable (0 or 1), while the remainder are quantitative variables.
• No coefficients are listed in Table 4 for age < 40 and total cholesterol < 100, nor is guidance given in the text. There are 2 options: (1) use a coefficient of 1, or (2) use a coefficient of 0. Since the model uses relative risk, I have opted for using a coefficient of 1.
• Table 2, page 70, lists the units for calcium as mEq/L, but the normal range is that for mg/dL. Since the conversion from mg/dL to mmol/L is 0.25, and there are 2 charges per molecule, the values in mEq/L would be much lower than the range shown.
• The values used to separate potassium and calcium into different classes relative to the normal range are of interest.
prognostic score =
= ((value for age) * (coefficient for age)) + ((value for leukocyte count) * (coefficient for leukocyte count)) + ((value for potassium) * (coefficient for potassium)) + ((value for calcium) * (coefficient for calcium)) + ((value for previous diagnosis) * (coefficient for previous diagnosis)) + ((value for serum globulin) * (coefficient for serum globulin)) + ( (value for cholesterol) * (coefficient for cholesterol))
NOTE: There is a gap in the guidance for determining the value to use for each variable when calculating the prognostic score.
• If the true value for each variable is used, then odd results occur, most notably with age and cholesterol. Fairly reasonable results are obtained with only the coefficient used for age and cholesterol, while using the variable full value for the other continuous variables.
• It mentions in the text that the continuous variables were separated into classes (page 69 and Table 2). An alternative approach is to use the class number rather than the variable value, but it is not stated if the classes should be scored 1 to 4 or 0 to 3.
• Since the hazard model uses the exponent for a (value) * (coefficient pair), and since EXP(0) = 1, I opted for the use of classes and the range of 0 to 3.
• It should be fairly easy to modify the spreadsheet for other interpretations.
Interpretation:
• minimum score in study: 0.555 (lower scores possible)
• maximum score in study: 7.577 (higher scores possible)
• A higher score is associated with a worse prognosis.
• The exponential of the difference in the scores for 2 patients is the ratio of their hazard function and represents a measure of the relative risk (page 71, below equation).
Prognostic Score 
Risk Group 
5 Year Survival Rate (initial group) 
< 3.6 
good 
82% 
3.6 to 5.0 
intermediate 
50% 
> 5 
poor 
20% 
after Figure 2, page 74
Limitations:
• Improvements in diagnostic tests and therapeutic interventions have probably improved survival in patients with liver cirrhosis since the data was collected in the mid 1980's.
Specialty: Gastroenterology