General Information
Conflict of Interest
Request Clinical Trials
Associated Trial(s): What type of data are you looking for?: Individual Participant-Level Data, which includes Full CSR and all supporting documentationRequest Clinical Trials
Data Request Status
Status: Withdrawn/ClosedResearch Proposal
Project Title: Development of Covariate Adjustment Methods to Improve Efficiency of Randomized Clinical Trials for HIV-Infected Patients
Scientific Abstract:
Background: Although covariate adjustment can improve the efficiency of randomized clinical trials by increasing the precision of effect estimation and the power to detect a beneficial effect, covariate adjustment methods for binary outcomes are controversial.
Objective: The aim of this study is to apply the newly developed covariate adjustment method to actual clinical trial data with a binary outcome and compare the results with those obtained using other statistical methods, including unadjusted analysis.
Study Design: This is a methodological study.
Participants: Treatment-nave HIV-1-infected patients least 18 years of age with plasma HIV-1 RNA of at least 5000 copies/mL.
Primary and Secondary Outcome Measures: The primary outcome is the confirmed virologic response (HIV-1 RNA and
Brief Project Background and Statement of Project Significance:
Because developing new therapies takes a lot of time and money, improving the efficiency of clinical trials has become an important issue. One way to improve the efficiency of clinical trials is to reduce the number of patients needed and shorten the duration of trials. Reducing the number of patients at risk is also ethically desirable.
An effective way to reduce the number of patients required is to adjust for baseline covariates. Adjustment for baseline covariates is known to improve the precision of treatment effect estimates and increase the power to detect a beneficial treatment effect compared to unadjusted analyses. The European Medicines Agency (EMA) and the U.S. Food and Drug Administration (FDA) have issued (draft) guidelines on covariate adjustment that support the use of covariate adjustment in randomized clinical trials [1, 2].
Analysis of covariance (ANCOVA) is the established method for covariate adjustment for continuous outcomes, as suggested in the above EMA and FDA guidance. However, covariate adjustment methods for binary outcomes remain controversial. When logistic regression is used for covariate adjustment, it is difficult to interpret the regression coefficients as treatment effects. This is because the regression coefficients depend on model assumptions and are not robust to model misspecification [3]. Logistic regression standardization is mentioned in the FDA draft guidance as a covariate adjustment method that is robust to model misspecification [4], but this method has rarely been used in clinical trials.
We therefore develop a new simple covariate adjustment method for binary outcomes that is equivalent to ANCOVA for continuous outcomes, based on the previous work by Rosenblum et al [5]. This method, called the regression coefficient approach, is robust to model misspecification and provides efficient estimates of treatment effects. The new statistical method has already been developed and its performance has already been evaluated in simulation experiments. It is necessary to evaluate the new statistical method on actual clinical trial data.
Specific Aims of the Project: The aim of this project is to apply the newly developed covariate adjustment method to actual clinical trial data with a binary outcome and compare the results with those obtained using other statistical methods, including unadjusted analysis.
Study Design: Methodological research
What is the purpose of the analysis being proposed? Please select all that apply.:
Software Used:
Data Source and Inclusion/Exclusion Criteria to be used to define the patient sample for your study:
The inclusion and exclusion criteria are the same as the original analyses [6]. The inclusion criteria include: (1) HIV-1-infected patient; (2) treatment-nave; (3) aged at least 18 years; and (4) plasma HIV-1 RNA at least 5000 copies/ml. The exclusion criteria include: (1) active AIDS-defining illness; (2) any clinically significant disease; (3) clinical or laboratory evidence of significantly decreased hepatic function or decompensation; (4) acute viral hepatitis at screening; (5) calculated creatinine clearance less than 70 ml/min; (6) primary HIV infection; (7) pregnant or breastfeeding; (8) grade 3 or 4 laboratory abnormalities (Division of AIDS grading table) with some exceptions (diabetes or asymptomatic glucose, triglyceride or cholesterol elevations).
The modified intention-to-treat (mITT) population is defined as the ITT population (all randomized and treated patients) in the original analyses [6]. The per-protocol population is defined as the per-protocol population in the original analyses (all randomized patients who had received study medication and had not taken disallowed therapy for more than 1 week) [6]. We will primarily use the mITT population.
Primary and Secondary Outcome Measure(s) and how they will be categorized/defined for your study: The primary outcome is the confirmed virologic response (HIV-1 RNA and
Main Predictor/Independent Variable and how it will be categorized/defined for your study: Main predictor/independent variables include treatment group (darunavir/ritonavir or lopinavir/ritonavir), baseline HIV-1 RNA (continuous or dichotomized to
Other Variables of Interest that will be used in your analysis and how they will be categorized/defined for your study: Other variables will not be used.
Statistical Analysis Plan:
Baseline covariates are summarized as frequency (%) for categorical variables and median (interquartile range) for continuous variables.
To estimate marginal risk difference and ratio, unadjusted analysis and covariate-adjusted analysis are conducted. Covariate-adjusted analysis include (1) the newly developed method; (2) likelihood-based linear and log binomial regression; (3) logistic regression standardization. Both non-inferiority and superiority test are conducted. The non-inferiority margin is 0.12 on the risk difference scale, which is equivalent to the original analyses.
Covariate-adjusted analysis use several distinct sets of covariates. The first covariate set include dichotomized HIV-1 RNA and CD4 cell count, which is the same as the original analysis. The second covariate set include continuous HIV-1 RNA and CD4 cell count. The quadratic term of these continuous covariates will be also included. The third covariate set include HIV-1 RNA, CD4 cell count, and other variables listed in main predictor/independent variables.
Narrative Summary: This study compares different statistical methods for analyzing binary outcomes in randomized clinical trials, using data from a trial of HIV-1 treatment. The main outcomes are whether the patients achieved a low viral load after 48 and 96 weeks. We use a new covariate adjustment method that is similar to analysis of covariance for continuous outcomes. We also use unadjusted analysis, linear and log binomial regression, and logistic regression standardization. They report the marginal risk difference and ratio for each method and discuss their advantages and disadvantages.
Project Timeline:
Anticipated project start date: 7/1/2023
Analysis completion date: 8/31/2023
Date manuscript drafted: 9/31/2023
Date first submitted for publication: 10/31/2023
Date results reported back to the YODA Project: 4/31/2023
Dissemination Plan: Target audience is statistician working for clinical trials. The manuscript will be published in Biometrics, Statistics in Medicine, Statistical Methods in Medical Research, Clinical Trials, Contemporary Clinical Trials, Journal of Biopharmaceutical Statistics, or Statistics in Biopharmaceutical Research.
Bibliography:
1. European Medicines Agency. Guideline on adjustment for baseline covariates in clinical trials. 2015.
2. U.S. Food and Drug Administration. Adjusting for covariates in randomized clinical trials for drugs and biological products: guidance for industry (draft guidance). 2021.
3. Freedman DA. Randomization does not justify logistic regression. Stat Sci. 2008;23: 237?249.
4. Moore KL, van der Laan MJ. Covariate adjustment in randomized trials with binary outcomes: targeted maximum likelihood estimation. Stat Med. 2009;28:39?64.
5. Rosenblum M, van der Laan MJ. Simple, efficient estimators of treatment effects in randomized trials using generalized linear models to leverage baseline variables. Int J Biostat. 2010;6:13.
6. Ortiz R, Dejesus E, Khanlou H, et al. Efficacy and safety of once-daily darunavir/ritonavir versus lopinavir/ritonavir in treatment-naive HIV-1-infected patients at week 48. AIDS. 2008;22:1389?1397.