This preview has intentionally blurred sections. Sign up to view the full version.View Full Document
Unformatted text preview: More Complex Analysis for Comparative Clinical Trials This lecture focuses on some more complex analytic issues that arise in clinical trials. We start with who gets analyzed. Will also talk about more complicated designs that include comparison of multiple factors, analysis of more complex designs, and ways to report the strength of the finding. Treatment non-adherence and statistical analysis The WHI trial reported only 60-80% compliance on the calcium and vitamin D portion. Some women took only part of the pills, some quit taking altogether. This raises several big questions for analysis. • Who should be included in the treatment? • “Intent to treat” (ITT) includes everyone who was randomized to treatment. • Under ITT, they are analyzed as being in the assigned treatment group, even if they never got any treatment, quit taking, or did not take as per protocol. • One alternative is to exclude non-adherent patients. • How do you define non-adherent? Never took, stopped taking, took too little or too much? • A second alternative is to include everyone but adjust the analysis to reflect treatment actually received (TR). • How do you adjust the analysis? Most clinical trials prescribe analysis of the ITT data set for efficacy. There are several reasons for this. • The results of ITT reflect what is likely to happen in clinical practice. • The results if people who drop out are omitted tend to be biased. (Why?) • The benefits of randomization are at risk if you do as-treated analysis (Why?). • Studies reported in Piantadosi Chapter 15 suggest that in many cases, TR analyses are biased in favor of the alternative hypothesis. Example: VA Cooperative Study of Coronary Artery Bypass Surgery (CABG) randomly assigned about 700 patients to either bypass surgery or medical treatment between 1972 and 1974. There was quite a lot of non-adherence. • Of 352 assigned to surgery, 312 actually underwent surgery. • About 25% subsequently had repeat surgery. • Of 354 assigned to medical treatment, 160 later crossed over and had surgery. • Of these, 21% had reoperation. The primary analysis proposed was ITT, but some people argue for TR because effect of surgery may be diluted by including those who did not get it or treating as medical those who actually had surgery. The Peduzzi study of coronary artery bypass compared ITT versus four other approaches: 1. Exclude non-adherent from analysis....
View Full Document
This note was uploaded on 12/30/2010 for the course BST 252 taught by Professor Tsodikov during the Winter '06 term at UC Davis.
- Winter '06