Reanalysis of Recent HIV Trials
Author Information
Author(s): Philippe Flandre
Primary Institution: INSERM UMR-S 943, University Pierre and Marie Curie (UPMC) Paris VI and Department of Virology, AP-HP, Pitié-Salpêtrière Hospital, Paris, France
Hypothesis
Does the choice of statistical method change the conclusions of noninferiority trials in HIV drug development?
Conclusion
The statistical methods used to estimate confidence intervals in noninferiority trials have a strong impact on the conclusion of such trials.
Supporting Evidence
- Different statistical methods can lead to different conclusions in noninferiority trials.
- Confidence intervals varied significantly based on the statistical method used.
- Most trials did not change conclusions despite different methods due to point estimates being far from noninferiority margins.
Takeaway
This study looked at how different ways of analyzing data from HIV trials can lead to different conclusions about whether new treatments are as good as existing ones.
Methodology
The study reanalyzed data from 11 HIV noninferiority trials published in 2010 using five different statistical methods to estimate confidence intervals.
Potential Biases
Potential bias due to the choice of statistical methods and the impact on conclusions drawn from the trials.
Limitations
The study did not apply all statistical methods proposed for estimating confidence intervals for differences between independent proportions.
Participant Demographics
Included HIV-infected adult patients over 18 years old.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website