Upload
uzuri
View
46
Download
0
Tags:
Embed Size (px)
DESCRIPTION
Statistical Comments on Retrospective Analysis. Girish Aras, Ph.D. Jonathan Ma, Ph.D. Center for Drug Evaluation and Research, FDA. Data Analysis Plan. Logistic Regression model - PowerPoint PPT Presentation
Citation preview
Statistical Comments on Retrospective Analysis
Girish Aras, Ph.D.
Jonathan Ma, Ph.D.
Center for Drug Evaluation and Research, FDA
Data Analysis Plan
• Logistic Regression model
• Dependent variable (Virological Failure) is ‘explained’ with independent variables ( baseline covariates ).
Baseline Covariates• Baseline Log (HIV-1 RNA)• New Drug Covariate• Genotypic Measures
– Genotypic Sensitivity Score– No. of PI, NRTI and NNRTI mutations
• Phenotypic Measures– Overall Sensitivity Score– No. of PI, NRTI and NNRTI drugs with
Phenotypic sensitivity
Relative Risk (Risk Ratio, RR)
• P0 = Probability of virological failure when a covariate is at a given level
• P1 = Probability of virological failure when a covariate is at one unit higher than the previous level
• RR = P1/ P0
Odds ratio
• RR is the quantity of interest but typically not estimable from a retrospective study
• OR (Odds Ratio) = P1/(1-P1) ÷ P0/(1-P0)
• OR can be estimated from a retrospective study, but may not of interest in itself
• When P1 and P0 are small,RR OR
Odds Ratios for Various Relative Risks
0
1
2
3
4
5
6
7
8
9
10
0.08 0.1 0.15 0.15 0.2 0.3 0.23 0.3 0.45 0.3 0.4 0.6 0.38 0.5 0.75
— Relative Risk — Odds RatioP1
P0=0.05
P0=0.1P0=0.15
P0=0.2
P0=0.25
Odd
s R
atio
or
Rel
ativ
e R
isk
Odds Ratios for Various Relative Risks
0
1
2
3
4
5
6
7
8
9
10
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
P0
Odd
s R
atio
RR=3
RR=2.5RR=2 RR=1.5
Mutation-Drug Resistance Table
• Table 1 in DAP titled ‘mutations associated with resistance to specific antiretroviral drugs’ is largely based on in vitro data and/or exploratory subgroup analyses
• A general consensus is yet to emerge on this table.
Mutation-Drug Resistance Table
• Most of the studies were available to the resistance collaborative group before the DAP was developed
0 1 2 3 4
Odds Ratio
(E)Stanford (C)
(E)CPCRA GART (C)
(E)VIRADAPT Genotypic (C)
(E)VIRADAPT Control (C)
(E)CNAA2003,CNAB3001-2,CNAB3009 (C)
(E)CNAA2007 (C)
(E)ACTG372 (C)
(E)ACTG364 (C)
.
95% Confidence Interval for Odds Ratio for Overall Genotypic Sensitivity Score:
Model C (Univariate) and E (Multivariate)
1 3 5 7
Odds Ratio
(F)
Stanford (D)
(F)
CPCRA GART (D)
(F)
VIRADAPT Genotypic (D)
(F)
VIRADAPT Control (D)
(F)
CNAA2007 (D)
(F)
ACTG372 (D)
(F)
ACTG333 (D)
.95% confidence interval for Odds ratio for Number of
PI Mutations: Model D and F
0.0 0.5 1.0 1.5
Odds.Ratio
(G)
CNAA2003,CNAB3001-2,3009 (C)
(G)
CNAA2007 (C)
(G)
Mega-HAART (C)
.
95% Confidence Interval for Overall Phenotypic Score:
Model C (Univariate) and Model G (Multivariate)
Conclusions
• These exploratory analyses suggest predictive link between virological failure and genotypic and phenotypic measures at the baseline
• They will provide insight in generating specific hypotheses to be tested in future confirmatory studies