Use this URL to cite or link to this record in EThOS:
Title: Analysis of treatment effect in crossover designs with missing data
Author: Rodgers, Lauren Rebecca
ISNI:       0000 0001 3530 2760
Awarding Body: Newcastle University
Current Institution: University of Newcastle upon Tyne
Date of Award: 2008
Availability of Full Text:
Access from EThOS:
In the analysis of clinical trials it is well-known that the omission of subjects randomized to treatments from the analysis can lead to bias in the final inference. A remedy that is widely adopted is to use the dictum of analysis by intention-to-treat (ITT), in which the groups as randomized are compared, even if the subjects have received treatments other than those prescribed in the protocol or have otherwise deviated from the prescribed regimen. A greater difficulty arises when subjects default and no response can be observed. This approach can often be rationalised in terms of comparing appropriate treatment policies. Crossover trials, in which subjects receive more than one of the trial treatments, present special problems. It is no longer clear that the use of ITT is appropriate. This work provides an overview of current approaches to missing or off-protocol responses in crossover trials and considers the underlying principle of what should be estimated from the data. Using simulation studies a rationale for the handling of missing data in crossover trials is developed. The model for the simple AB/BA crossover design is looked at in the simulation studies and analysed in the presence of ignorable or non-ignorable missing data and in the situation of off-protocol responses. This model is then extended to higher order crossover designs. From the outcome of these simulation studies we introduce a simple method for handling missing data in crossover trials. The motivation of this work lies in the data from Frank et al. (2008) where we have an AB/BA crossover trial with a high proportion of subjects who withdraw before the end of the trial due to side effects of their treatment.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available