Efficient and ethical adaptive clinical trial designs to detect treatment-covariate interaction

Seung Won Hyun, Tao Huang, Hongjian Zhu

Research output: Chapter in Book/Report/Conference proceedingChapter

2 Scopus citations


Personalized medicine, which tailors decisions, practices, or products to individual patients, and optimizes preventative and therapeutic care, has attracted tremendous attention due to the availability of many potentially highly informative biomarkers and the observed heterogeneity of patients’ responses to treatments. Take the multicenter randomized trial, Stroke Prevention in Atrial Fibrillation study [23], for example. A difference between the aspirin treatment group and the placebo group on the number of strokes was detected among patients receiving anticoagulation, but not among patients without anticoagulation therapy. Without noticing the interaction effect between the treatment and the anticoagulation status, aspirin would have been recommended for the general population to prevent the occurrence of stroke, and certain patients would have lost opportunities to obtain timely and correct remedies. Another example is that Cetuximab is not helpful for colorectal cancer patients without a tumour bearing the wild-type KRAS gene [14]. All these examples represent the treatment-covariate interactions and demonstrate the importance of personalized medicine. Therefore, clinical trials which are able to involve a variety of covariates or prognostic factors are desirable for personalized medicine.

Original languageEnglish (US)
Title of host publicationModern Adaptive Randomized Clinical Trials
Subtitle of host publicationStatistical and Practical Aspects
PublisherCRC Press
Number of pages18
ISBN (Electronic)9781482239898
ISBN (Print)9781138893948
StatePublished - Jan 1 2015
Externally publishedYes

ASJC Scopus subject areas

  • Mathematics(all)
  • Medicine(all)
  • Pharmacology, Toxicology and Pharmaceutics(all)


Dive into the research topics of 'Efficient and ethical adaptive clinical trial designs to detect treatment-covariate interaction'. Together they form a unique fingerprint.

Cite this