Title: Surrogate Endpoints for Hepatitis B Trials
1Surrogate Endpoints for Hepatitis B Trials
- Greg Soon, Ph.D
- Rafia Bhore, Ph.D.
- Division of Biometrics III/Antiviral
8/7/2002 AC
2Potential Surrogate Endpoints
- Knodell Score (Biopsy, Histology)
- ALT, HBV DNA
- Proportion
- Change from Baseline
- End of treatment
- Time to suppression
- Duration of suppression
- Time to virologic failure
- DAVG or Average change over time
- Others
- HBeAg, HBeAb, HBsAg, HBsAb
- Composition of the above
3Data Source EpivirHBV Submission
4Data Source Adefovir Submission
5HBV DNA Assays
- Lamivudine trials used Abbott Hybridization
assay, which has a lower limit of approximately
500,000 copies/mL - Limiting ability to differentiate patient
responses on viral load - Converted to copies/mL for this presentation
- Adefovir Trials used PCR assay that has a lower
limit of 400 copies/mL
6Overview
- Summary of Efficacy
- Patient-level Correlation
- Trial-level Correlation
- Proportion of Treatment Effect Explained (PTE)
- Summary
7Overview
- Summary of Efficacy
- Patient-level Correlation
- Trial-level Correlation
- Proportion of Treatment Effect Explained (PTE)
- Summary
8Change of Knodell Score vs Baseline
9Convention
- White for placebo arms
- Yellow for LAM 100mg or ADV 10mg
- Orange for LAM 25mg or ADV 30mg
- Red for IFN LAM 100mg
- Green for IFN alone
10(No Transcript)
11(No Transcript)
12(No Transcript)
13Log10 ALT/ULN Over Time
14Study 438
15Study 437
16LAM Studies
17Study 437
18(No Transcript)
19(No Transcript)
20(No Transcript)
21(No Transcript)
22(No Transcript)
23(No Transcript)
24(No Transcript)
25Log10 HBV DNA Over Time
26Study 438
27Study 437
28LAM Studies
29Study 437
30(No Transcript)
31(No Transcript)
32(No Transcript)
33(No Transcript)
34(No Transcript)
35(No Transcript)
36(No Transcript)
37HBeAg Loss Over Time
38Study 437
39LAM Studies
40Transition of HBeAg Status
41Overview
- Summary of Efficacy
- Patient-level Correlation
- Trial-level Correlation
- Proportion of Treatment Effect Explained (PTE)
- Summary
42Lamivudine HBV DNA vs. Knodell
Change of Knodell Score
Year 1 LOG10 HBV DNA
43Lamivudine HBV DNA vs. Knodell
Change of Knodell Score
Year 1 LOG10 HBV DNA
44Lamivudine CorrelationYear 1 HBV DNA vs.
Knodell Score Change
p-value
45Adefovir HBV DNA vs. Knodell
Change of Knodell Score
Year 1 LOG10 HBV DNA
46Adefovir CorrelationYear 1 HBV DNA vs. Knodell
Score Change
p-value
47Lamivudine ALT vs. Knodell
Change of Knodell Score
Change of LOG10 ALT
48Lamivudine ALT vs. Knodell
Change of Knodell Score
Change of LOG10 ALT
49Lamivudine CorrelationALT Change vs. Knodell
Score Change
p-value
50Adefovir ALT vs. Knodell
Change of Knodell Score
Change of LOG10 ALT
51Adefovir CorrelationALT Change vs. Knodell
Score Change
p-value
52DNA, HBeAg Knodell
438
437 Yr 1 eAg
437 Yr 1 eAg Neg
38 68 62
34 13 21
13 6 17
41 10 16
215 10 53
53Joint Prediction
54Predicting Change of Knodell
Excluding adefovir HBeAg negative
study stratified by study and treatment
55(No Transcript)
56Summary
- With near certainty, better response on change of
log10 ALT and Year 1 log10 HBV DNA are associated
with better Knodell improvement - The associations are weak to moderate
- Multiple predictors do not improve much on top of
ALT
57Overview
- Summary of Efficacy
- Patient-level Correlation
- Trial-level Correlation
- Proportion of Treatment Effect Explained (PTE)
- Summary
58Methods
- Studies were divided into smaller trials
according to region and ethnic background to
increase data points for analysis - Size of trials range from 20 to 70.
59HBV DNA vs. Knodell By Trial and Treatment
60ALT vs. Knodell By Trial and Treatment
61HBV DNA vs Knodell Treatment effects by Trial
R225 (0, 60)
62HBV DNA vs Knodell Treatment effects by Trial
R26 (0, 30)
63ALT vs Knodell Treatment effects by Trial
R224 (0, 49)
64ALT vs Knodell Treatment effects by Trial
R233 (8, 57)
65Variability and Correlation
66Variability and Correlation
Assume overall SD for Knodell is 3
67Impact on relationship among treatment effects
68Replication of a Single Trial
r10.74 r00.31 R0.56
69Replication of a Single Trial
r10.50 r00.47 R0.44
70Other considerations
- Variations in trial results are desirable for the
trial-level correlation validation method. Two
trials of different effect sizes are more useful
than two trials of similar effect size
71Overview
- Summary of Efficacy
- Patient-level Correlation
- Trial-level Correlation
- Proportion of Treatment Effect Explained (PTE)
- Summary
72Proportion of Treatment Effect Explained
- Attempts to determine, of the total effect size
observed for the Knodell score in each trial, how
much can be attributed to the effect on the
potential surrogate? - Linear regression model to determine how much
effect remain when there is no effects on
surrogate - The remainder portion of the treatment effect is
considered mediated through surrogate - Long used but also widely debated
73PTE HBV DNA vs Knodell
74PTE ALT vs Knodell
75Overview
- Summary of Efficacy
- Patient-level Correlation
- Trial-level Correlation
- Proportion of Treatment Effect Explained (PTE)
- Summary
76Summary
- Treatment effects and variability for HBV DNA,
ALT, HBeAg and Knodell score. - Baseline Knodell correlates with change
- Effects over time for HBV DNA, ALT, and HBeAg
- Both ALT and HBV DNA are variable over time
77Summary
- Weak individual level correlation with Change of
Knodell score - Year 1 HBV DNA, r 0.3
- Change of ALT, r 0.45
- Not much improvement with multiple predictors
78Summary
- Weak Trial-level correlation
- Biopsy variability may reduce individual
correlation - Trial level correlation will be similar to
individual level correlation if trials are
similar - Adefovir and Epivir trial results are somewhat
similar in effect sizes
79Summary
- PTE for Year 1 HBV DNA and Change of ALT
- For baseline HBeAg negative study, both HBV DNA
and ALT have low PTE - For baseline HBeAg Study, PTE are somewhat
consistent in adefovir or lamivudine vs placebo
comparisons - Interferon-containing comparisons are not
informative
80END