Focus on Alternative and Complementary Therapies
www.pharmpress.com/fact
Focus Alternat Complement Ther©2005 Pharmaceutical Press
Focus Altern Complement Ther 2003; 8: 476
To develop methods for inference to deal with the problem that late-phase research criteria are often inappropriately applied to early-phase CAM designed studies.
Development and testing of statistical methods, using mathematical analysis, simulation, and application to existing data.
(1) Design-adaptive allocation is generally far superior to randomisation in balancing influential variables across treatment groups. With appropriate computer support the method is quite feasible, and creates no problems for subsequent statistical inference. (2) Participant-centred analysis is a collection of techniques for analysing and displaying study results at the level of the individual participant, instead of summary group statistics. It focuses on the question ‘Who benefited, who was harmed, and for whom can’t we tell?’ rather than ‘Which group had the higher average response?’ (3) Separation tests use the original Neyman–Pearson approach to hypothesis testing, in place of the later (and currently dominant) Fisher approach. Each analysis can have a guaranteed bound of 0.05 on both type I and type II errors, while the strength of the assertion about the group difference depends on the variability in the data. The sample size that can be justified will typically be 70% less than using customary procedures.
Combining the three methods presented here can substantially strengthen the role of early-phase studies in CAM research. Design-adaptive allocation produces more accurate estimation of effects, intensive data acquisition on individual participants humanises the research process, and separation tests are more appropriate to the limited aims of early-phase research.
