Schlichting et al evaluated survival of patients with cirrhosis enrolled in a clinical trial comparing therapy with prednisone to that with placebo. They used a number of parameters to derive a prognostic index (PI) which correlated with 5 year survival and median survival time (MST). The authors are from multiple medical institutions in Denmark.
"Prognostic" Parameters:
(1) treatment arm in clinical trial (placebo vs prednisone)
(2) gender/sex
(3) age in years
(4) prothrombin time as percent of normal
(5) acetylcholinesterase activity
(6) eosinophils in liver parenchyma
(7) small focal liver cell necrosis
(8) inflammation in the liver connective tissue
(9) efferent veins in parenchymal nodules
"Therapeutic" Parameters:
(10) ascites
(11) antinuclear factor
(12) large piecemeal necrosis (more than 5 hepatocytes)
(13) parenchymal nodules
Both the prednisone and placebo groups share the same regression coefficients for the prognostic parameters, but use different coefficients for the therapeutic ones.
|
Parameter |
Finding |
Points |
|---|---|---|
|
treatment |
placebo |
1 |
|
|
prednisone |
0 |
|
gender/sex |
female |
0 |
|
|
male |
1 |
|
age in years |
|
(age) - 60 |
|
prothrombin time as percent of normal |
|
LN(percent) - 4 |
|
acetylcholinesterase activity in µmole/min/mL |
|
LN(100 * (value)) – 4 |
|
eosinophils in parenchyma |
none |
0 |
|
|
few |
1 |
|
|
moderate |
2 |
|
|
many |
3 |
|
small foci of liver cell necrosis |
none |
0 |
|
|
present |
1 |
|
inflammation in liver connective tissue |
none |
0 |
|
|
slight |
1 |
|
|
moderate |
2 |
|
|
severe |
3 |
|
efferent veins in parenchymal nodules |
none |
0 |
|
|
few |
1 |
|
|
many |
2 |
|
ascites |
none |
0 |
|
|
slight |
1 |
|
|
moderate |
2 |
|
|
marked |
2 |
|
antinuclear factor |
absent |
0 |
|
|
1+ |
1 |
|
|
2+ |
2 |
|
|
3+ |
2 |
|
large piecemeal necrosis |
none |
0 |
|
|
few |
0 |
|
|
moderate |
1 |
|
|
severe |
1 |
|
parenchymal nodules |
none |
0 |
|
|
all nodules smaller than normal lobules |
0 |
|
|
all nodules larger than normal lobules |
1 |
where:
• In Table 4, page 892, the points for age is given as "Age: 60". In the example on page 893 for a 70 year old woman the value used is 10. I am not sure what is done for a person with an age < 60. I will set the value to 0 but this may not be correct.
• Prothrombin time as percent of normal, which is a value used more often in Europe than the US. This can be calculated by doing a series of dilutions of the normal control and measuring the prothrombin time for each in seconds. The data can then be plotted. The patient's PT can be converted to percent of normal by reading the time off the graph.
• I assume antinuclear factor is the same as antinuclear antibody (ANA).
• Parenchymal nodules the same size as normal lobules are not covered in the description. I will combine these with the "larger than" group. I am not sure where "none" fits in for a patient with cirrhosis.
• Parenchymal nodules larger than normal lobules do not have a value assigned in Table 5 on page 892, but is assigned 1 point in the example on page 893.
prognostic index (PI) for a person receiving prednisone =
= (0.150 * (points for treatment)) + (0.317 * (points for gender)) + (0.049 * (points for age)) + (-0.495 * (points for prothrombin time)) + (-0.612 * (points for acetylcholinesterase)) + (0.299 * (points for eosinophils)) + (0.310 * (points for small foci liver necrosis)) + (-0.390 * (points for inflammation in connective tissue)) + (0.258 * (points for efferent veins)) + (0.719 * (points for ascites)) + (-0.124 * (points for antinuclear factor)) + (-0.739 * (points for large piecemeal necrosis)) + (0.729 * (points for parenchymal nodules))
prognostic index (PI) for a person receiving placebo =
= (0.150 * (points for treatment)) + (0.317 * (points for gender)) + (0.049 * (points for age)) + (-0.495 * (points for prothrombin time)) + (-0.612 * (points for acetylcholinesterase)) + (0.299 * (points for eosinophils)) + (0.310 * (points for small foci liver necrosis)) + (-0.390 * (points for inflammation in connective tissue)) + (0.258 * (points for efferent veins)) + (0.105 * (points for ascites)) + (0.311 * (points for antinuclear factor)) + (0.726 * (points for large piecemeal necrosis)) + (-0.607 * (points for parenchymal nodules))
where:
• The regression coefficient for piecemeal necrosis in a person treated with placebo is (-0.726), while the example on page 893 gives it as (+0.726). I have used the +0.726 in the implementation but this may not be correct. However, the use of +0.726 gave a more believable value than use of the (-0.726).
|
Prognostic Index |
Probability of 5 Year Survival |
Median Survival Time (MST) |
|---|---|---|
|
-4 |
97% |
> 3,700 days |
|
-3.5 |
95% |
> 3,700 days |
|
-3 |
92% |
> 3,700 days |
|
-2.5 |
86% |
> 3,700 days |
|
-2 |
78% |
3,700 days |
|
-1.5 |
67% |
2,800 days |
|
-1 |
52% |
2,000 days |
|
-0.5 |
32% |
1,150 days |
|
0 |
17% |
700 days |
|
0.5 |
5% |
400 days |
|
1 |
1% |
200 days |
|
1.5 |
< 1% |
150 days |
|
2.0 |
< 1% |
100 days |
Equations can be derived using JMP to approximate the curves in Figures 3 and 4 on page 893.
median survival time in days for PI from –2 to 0 =
= (242.857 * ((PI)^2)) – (1044 * (PI)) + 661.4286
median survival time in days for PI from 0 to 2 =
= (-66.6667 * ((PI)^3)) + (385.714 * ((PI)^2)) – (804.76 * (PI)) + 702.857
probability of 5 year survival for PI from –4 to -2 =
= (-4.2857 * ((PI)^2)) – (35.11429 * (PI)) + 24.9714
probability of 5 year survival for PI from –2 to 0 =
= (-3.714286 * ((PI)^2)) – (38.82857 * (PI)) + 15.942957
probability of 5 year survival for PI from 0 to 1.5 =
= (-6.4015 * ((PI)^3)) + (25.60227 * ((PI)^2)) – (35.2 * (PI)) + 17
For the 5 year survival I tried a logistic regression equation in the form of (1 / (1 + EXP((-1) * x))), using the 50% survival to determine the coefficient. However, this did not give values corresponding to the data. If you look at the data in Figure 3, it is not symmetrical, with survival dropping off fairly fast with PI's > -0.5.
Limitations:
• The study was done in 1983 and survival has changed since then, However, the score can give a rough idea of survival without liver transplantation or other specialized therapy.
• There are several discrepancies between the data in Tables 4 and 5 and the example on page 893.
Specialty: Gastroenterology
