Design and Analysis for Clinical Trials: taking account of patient and clinician preferences
Impartit per: Stephen D. Walter, Clinical Epidemiology and Biostatistics, McMaster University, Canada.
Llengua del curs: Anglès.
Dates i horaris del curs: 25 a 28 de juny, de 9:30 a 12 h.
Lloc: aula 101
Tipus d'activitat i càrrega lectiva: Curs de 10 hores.
1,5 ECTS com a assignatura optativa per als estudiants del MEIO UPC-UB, com a assignaturas de lliure elecció per a estudiants de màster, grau i llicenciatura i com a crèdits de doctorat.
Data de matrícula: 29 d'abril a 10 de maig.
Descripció: This course would be very suitable for students who are in the early stages of research careers (for instance, if they are currently doing a research thesis or project). The topics to be covered would also be of interest to anyone wanting to expand their knowledge of clinical trial methodology.
Students should preferably have had some previous exposure to the basic concepts of randomised trial designs, and the associated statistical methodology for data analysis. This course will deal with several selected, newer design and analysis issues in the area of clinical trials. These topics are being actively researched by our group and elsewhere, and they have been the subject of methodology publications in the past few years.
The course topics have an overall theme of attempting to move beyond the conventional parallel group randomised trial design, and take account of patient and clinician preferences for treatments. These preferences are common, and can be importantly predictive of study outcomes. These effects typically cannot be estimated in a conventional design, which highlights the need to consider alternative approaches, as we will discuss during the course.
- Topic 1: As an introduction to the area, we will review various designs for clinical trials that allow for patient and / or clinician preferences in assigning treatments. Some will involve a mixture of observational and randomised assignments, and other require the use of recorded data on preferences in fully randomised trials of various kinds. Comparisons will be made between these designs with respect to feasibility of execution, ethics, data interpretation and statistical efficiency.
- Topic 2: We will explore in more detail several aspects of the two-stage randomised design. This approach allows some participants (or their physicians) to choose their treatment, and through an appropriate analysis, permits an examination of selection and preference effects on study outcomes, in addition to the usual treatment effect. These former effects can be important, even in situations where the treatment effect is small, and should be of considerable interest to investigators. Recent research has examined optimisation of this design for statistical efficiency, and the question of how to deal with undecided participants. If time allows, we may also more briefly discuss the partially randomised and fully randomised preference designs. Questions such as optimisation, sample size and power calculations are being actively researched for these various designs.
- Topic 3: We will discuss design and analysis issues for Expertise-Based trials. The expertise-based design has been proposed to avoid so-called expertise bias, whereby newer, experimental treatments may be disadvantaged because of less familiarity and skill of clinicians with its administration, compared to standard and better-known techniques. Clinicians in this type of trial can be relieved of the necessity to administer both treatments being compared, but instead can be involved in only one treatment group, corresponding to their area of greater clinical experience, expertise or comfort. There are a number of open issues on the statistical aspects of these trials, and on how expertise should be defined and measured.
- Topic 4: Preference-based analysis of parallel group trials: by observing treatment crossover and refusal patterns in conventional trials, it is possible to adopt a preference-based approach to the analysis, and to examine the treatment effect according to preference profiles of participants. One can also contrast these treatment effect estimates to a number of alternatives, such as ITT (intention-to-treat),” as treated”, and “per protocol” approaches to the analysis.
Evaluation will be based on one or more of:
- Getting students to extract data from papers that contain applications of the trial designs we are discussing, and to calculate certain quantities such as the treatment and selection effects.
- Commenting on strengths and weaknesses of alternative designs.
- Outlining alternative analytic approaches to a given study design, and comparing their strengths and weaknesses.
- Doing a brief review (1 page maximum) of selected papers giving examples of a particular trial design.