Transcript Slide 1

Primary Billiary Cirrhosis
STUDY
PubH5483
Correlated Data Analysis
Final Project
Fall, 2004
Soon Young Jang
12/09/04
PBC STUDY
Slide 1 of 18
PBC - Background
• What is Primary Billiary Cirrhosis (PBC) ?
Progressive chronic liver disease leading to cirrhosis and liver failure
Mostly women’s disease : 75 to 90% of Patients
Mostly occurs between fourth and sixth decades of life
Slow and predictable progress : 40-67% symptomatic disease development in
approximately 5 to 7 years
No totally effective treatment in slowing the progress of PBC other than liver
transplantation
• Historical Stages : system by Ludwig et al and Scheuer
based on the results of liver biopsy
I.
II.
III.
IV.
Portal
Periportal
Septal
Cirrhosis
In general, gradual progression over years from stage I to stage IV
12/09/04
PBC STUDY
Slide 2 of 18
PBC - Background
• Prognosis
Range from a few months to 20 years depending upon when the diagnosis is first
made
The most reliable determinants of patient prognosis are :
Mayo risk score
Serum Bilirubin Level (SBL)
• SBL and Prognosis
Constantly above 6 mg/dL, the mean survival rate is 2.1 years

•
Needed to be evaluated for possible liver transplantation
D-pencillamine : Hepatic copper decreasing and immunomodulatory potentials.
• Scientific Questions
1. D-pencillamine impact on slowing the rate of increase in SBL and whether the
effects are same among historical stages
2. The variability on the rate of increase in SBL among patients
12/09/04
PBC STUDY
Slide 3 of 18
PBC - Method
• Data : published by Terry Therneaus (June 10, 1999)
Repeatedly Measured Data : Mayo Clinic trial in PBC of the liver,
originally conducted between January 1974 and May 1984, with
extended follow-up to April 30, 1988.
For 312 patients who consented to participated in the randomized trial
(D-Pencillamine Vs. Placebo), total of 1,945 patients visits.
Including patient’s id, survival time, status (Alive, Transplanted, and
Dead), gender, age, treatment, 12 laboratory results and historical
stage of each visit on the clinic with the days (of visit) from the
enrollment
• Exclusion of Observations with Missing Values
32(10.26%) patients have missing values on either Serum Cholesterol
or Platelet measurement at baseline, total of 264 observations
No evidence against “missing at random” assumption
2
1. By survival status :  2  1 . 88 , pValue  0 . 39
2 . By observed slope : t(df  31 .42 )  - 0 .17 , pValue  0 .86
12/09/04
PBC STUDY
Slide 4 of 18
PBC – Method (Appendix)
Analysis of Missing Observations
Missing Observation By Survival Status
 2  1 . 88 , pValue  0 . 39
2
Non-Missing
Frequency
Alive
Survival
Stat
us
Missing
Percent
Frequency
Total
Percent
126
88.11%
17
11.89%
143
28
96.55%
1
3.45%
29
Dead
126
90.00%
14
10.00%
140
Total
280
89.80%
32
10.20%
312
Transplanted
Observed Mean Slope By Missing Status
Welch Two Sample t-test : t(df=31.42)=-0.17, pValue=0.86
Non-Missing
Missing
N
Mean
Std Dev
N
Mean
Std Dev
255
0.106
0.283
30
0.122
0.480
12/09/04
PBC STUDY
Slide 5 of 18
PBC - Method
• Outcome of Interest
Change of serum bilirubin Level (SBL) from the baseline
• Main Covariates of Interest
Time : Visit in months from the enrollment ( day / 30.5 )
Drug, randomly assigned
Stage determined at the baseline
• Demographic Covariates
age, sex
• Covariates for adjusting purpose
11 laboratory results measured at the baseline
12/09/04
PBC STUDY
Slide 6 of 18
PBC - Method
• Analysis Model : General Linear Mixed Model
• Covariance Structure
Random Intercept + Random Slope
No correlation was assumed between Random Intercept and Random
Slope H :   0
0
01
LRT 
2
 0 .4 ~
1
2
2 
2
1
2
 3 , pvalue  0 . 90
2
• Mean Model
No significant associations found between change in SBL and 11 other
laboratory results at the baseline ( LRT  112  12 . 2 , pValue  0 . 35 )
Final Selected Mean Model
E ( change in SBL )  drug  stage  age  sex  month 
drug * month  stage * month  drug * stage * month
• Diagnostic by residual analysis
• EDA and Diagnostic using R ver1.8.1, Analysis using SAS ver8.2
12/09/04
PBC STUDY
Slide 7 of 18
PBC – Results Part I: Baseline Characteristics
Table 1.
Top: Categorical Baseline Characteristics (Frequency & Percent, chisq test)
Bottom: Continuous Baseline Characteristics (Mean ± Std Dev, t test)
Placebo
(N = 142)
Treatment
(N = 138)
Overall
(N = 280)
Freq (Percent)
Freq (Percent)
Freq (Percent)
Stage
Portal
3 (25.0%)
9 (75.0%)
12 (4.3%)
Periportal
29 (48.3%)
31 (51.7%)
60 (21.4%)
Septal
61 (54.0%)
52 (46.0%)
113 (40.4%)
Cirrhosis
49 (51.6%)
46 (48.4%)
95 (33.9%)
Sex
Male
Female
14 (41.2%)
20 (58.8%)
34 (12.1%)
128 (52.0%)
118 (48.0%)
246 (87.9%)
Test
χ2
pvalue
df
3.32
0.28
3
1.41
0.24
1
(df = 278)
Variable
SBL in mg/dL
Age in months
12/09/04
Mean (SD)
Mean (SD)
Mean (SD)
3.67
5.32
2.85
3.50
3.27
4.52
1.53
0.128
582.33
118.60
612.74
131.29
591.32
125.72
-2.03
0.043
PBC STUDY
t
Pr > |t|
sig
**
Slide 8 of 18
PBC - Results
• Summary of Data
Summary of Individual Regression Lines (Slopes Only)
Increasing rates across the historical stages
Large Standard Deviations
Table 2. Mean of Individual Slope ( ± Std Dev mg/dL/month )
Historical Stage
D
r
u
g
Portal
Periportal
Septal
Cirrhosis
Placebo
0.001
(± 0.002)
0.061
(± 0.142)
0.072
(± 0.231)
0.217
(± 0.508)
D-pencillamine
0.024
(± 0.063)
0.069
(± 0.135)
0.079
(± 0.145)
0.164
(± 0.320)
12/09/04
PBC STUDY
Slide 9 of 18
PBC - Results
• Conclusion of Analysis:
No Significant Impact of D-pencillamine on the rate of increase in SBL
( F (1, 939 )  0 . 0 , pValue  0 . 98 )
No different effects of D-pencillamine on the rate among the historical
stages ( F ( 3 , 939 )  0 . 2 , pValue  0 . 89 )
The rate of increase in SBL across time significant differs by the historical
stages determined at the enrollment (or when the diagnosis was made
on) ( F ( 3 ,939 )  2 . 77 , pValue  0 . 04 )
Very large  1  0 . 02  and highly significant Variability on the rate of
change in SBL among individual patients.
1 2 1 2
2
( LRT   479 ~  1   2 , pValue  0 . 001 )
2
2
2
12/09/04
PBC STUDY
Slide 10 of 18
PBC – Results (Table 3)
Estimated Slope and 95% Confidence Limits (mg/dL/month)
Historical Stage
Portal
lower
Periportal
upper
lower
upper
Placebo
-0.181
0.002
0.185
-0.015
0.088
0.192
D-Pencillamine
-0.177
0.035
0.246
-0.095
0.082
0.258
Drug
Historical Stage
Septal
lower
Cirrhosis
upper
lower
upper
Placebo
-0.010
0.084
0.178
0.103
0.163
0.222
D-Pencillamine
-0.072
0.091
0.253
0.030
0.132
0.235
Drug
12/09/04
PBC STUDY
Slide 11 of 18
PBC – Diagnostic: Predicted Vs. Observed Slope
Figure I (a) Placebo (from Table 2 & 3)
12/09/04
PBC STUDY
Slide 12 of 18
PBC – Diagnostic : Predicted Vs. Observed Slope
Figure I (b) D-Pencillamine (from Table 2 & 3)
12/09/04
PBC STUDY
Slide 13 of 18
PBC – Diagnostic
• Diagnostic Results
Comparison b/w Observed & Predicted Slopes (Figure I)
Outlying Clusters (Age and Stay under the study in years)
--------------------------------------------------ID
Stay Status Drug Stage
Age
Sex
b1
--------------------------------------------------67
7.6 Trans
P
III
51
F
0.61
100
1.5 Trans
P
IV
51
M
0.50
165
3.0 Trans
D
IV
53
M
0.50
---------------------------------------------------
Normality and linearity assumptions checked by residual plots
Relatively good fit of cluster-specific predicted values on observed
change (Figure II)
12/09/04
PBC STUDY
Slide 14 of 18
PBC – Diagnostic (Figure II)
12/09/04
PBC STUDY
Slide 15 of 18
PBC – Discussion (Figure III)
12/09/04
PBC STUDY
Slide 16 of 18
PBC - Discussion
• Prognosis of SBL by Survival Status (Figure III)
• Limitation
Add-hoc analysis : Survival Data
Exclusion of observations with missing values
Main Question : D-Pencillamine effect  Prediction of SBL
• Summary
No benefit of D-pencillamine to lowering the rate of increase in SBL,
other literature indicates even harmful effect (toxic)

No recommendation of the use of D-pencillamine for the
patients with PBC
Large variability of the rate of increase in SBL among patients

12/09/04
No practicality of using the model to predict the progress
of SBL for each individual patients
PBC STUDY
Slide 17 of 18
PBC - Reference
1. Dickson, et al., “Prognosis in Primary Biliary Cirrhosis: Model for Decision
Making”, Hepatology 10:1-7 (1989)
2. Markus, et al., “Efficacy of Liver Transplantation in Patients with Primary
Biliary Cirrhosis”, New England Journal of Medicine 320:1709-13 (1989)
3. Murtugh, et al., “Primary Biliary Cirrhosis: Prediction of Short-term Survival
Based on Repeated Patient Visits”, Hepatology 20(1.1):126-34 (1994)
4. Bonnand, et al., “Clinical Significance of Serum Bilirubin Levels Under
Ursodeoxycholic Acid Therapy in Patients With Primary Biliary Cirrhosis”,
Hepatology, Hepatology 29(1):39-43 (1999)
5. Kim, et al., “Adaptation of the Mayo Primary Biliary Cirrhosis Natural History
Model for Application in Liver Transplant Candidates”, Liver Transplantation
6(4):489-94 (2000)
6. Therneau, Terry, “Primary Biliary Cirrhosis, sequential data”,
http://www.mayo.edu/hsr/people/therneau/book/data/pbcseq.html
7. Worman, H.J., “What is Primary Biliary Cirrhosis (PBC)?”,
http://cpmcnet.columbia.edu/dept/gi/PBC.html
8. Pyrsopoulos, N.T., “Primary Biliary Cirrhosis”,
http://www.emedicine.com/med/topic223.htm
12/09/04
PBC STUDY
Slide 18 of 18