JOINT RSS MEETING: MEDICAL SECTION AND PRIMARY HEALTH CARE STUDY GROUP
Statistical Issues in the Design and Analysis of Pilot and Feasibility Studies.
Speakers: Gill Lancaster, Dawn Teare and Sandra Eldridge.
2-5pm, 23rd October 2013
The Pemberton Room, School of Health and Related Research, The University of Sheffield
Regent Court, 30 Regent Street, Sheffield
(http://www.shef.ac.uk/scharr/contacts/howtofindus/index)
The talks will be followed by group discussion. All are welcome and pre-registration is not necessary. However, in order to assist us in planning the event, it would help to email [log in to unmask] if you plan to attend.
Abstracts:
“Feasibility/Pilot Studies in the design, conduct and evaluation of complex interventions”
Gill Lancaster, University of Lancaster
The Medical Research Council’s (2008) guidance document on complex interventions and Campbell et al. (2007) emphasise the importance of thorough groundwork in designing and evaluating complex interventions and stress the importance of contextualising and conceptualising the problem at the development stage. Lancaster et al. (2004) and more recently Arain et al. (2010) have showed that there was and still is a dearth of pilot studies in the literature that stated they were specifically in preparation for a randomised controlled trial, and that give a clear list of key objectives relating to the pilot phase. Feasibility and pilot studies are conducted to assess the feasibility and integrity of the study protocol, but the differences between the two are not clear-cut. This talk will provide an overview of the use of feasibility and pilot studies in the literature, consider statistical issues in study design, and provide useful examples and references.
“How large does a pilot study need to be to estimate the key design parameters for the future RCT?”
Dawn Teare, University of Sheffield
External pilot or feasibility studies can be used to estimate key unknown parameters, to inform the design of the definitive randomised controlled trial (RCT). However, there is little consensus on how large pilot studies need to be and some debate on how to use these estimates derived from the pilot study when designing the definitive RCT. In this talk I will use a simulation approach to visually illustrate the impact of pilot study size on the sampling distribution (in terms of precision and bias) of the standard deviation for continuous outcomes and the event rate for binary outcomes. I also demonstrate the consequences of using imprecise estimates (with and without correction or inflation factors) from the pilot studies on the power and sample sizes required for the definitive RCT. For simplicity I limit attention to pilot/feasibility studies preceding two arm balanced group superiority RCTs. I will show why external pilot studies need at least a total of 70 (35 per group) measured subjects to estimate the pooled standard deviation for a continuous outcome and at least 60 subjects are required per group for event rate estimation. It is generally more efficient to use a larger external pilot study, than to guard against the lack of precision by using one-sided confidence interval inflated estimates.
“Reporting of pilot/feasibility studies”
Sandra Eldridge, Queen Mary, University of London
Pilot/feasibility studies can be an essential part of trial preparation, particularly in planning complex interventions. However, recent research indicates that these studies are often not published and when they are they are generally poorly reported. We are currently producing CONSORT guidelines for feasibility and pilot studies conducted in advance of a full trial. These guidelines include clarification of definitions. In this talk I will present the background to this work, new ideas about the definition of feasibility and pilot studies, a summary of some of the issues that have arisen in trying to construct the guidelines, for example, what should be reported in relation to the main trial, how bias, blinding and multiple objectives should be handled, whether cost-effectiveness analyses are justified in a pilot study, and our current thinking on the shape of the guidelines.
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|