Commentary
on Spreadsheets
for Analysis of Controlled Trials Alan M Batterham Sportscience 10, 54, 2006
(sportsci.org/2006/amb.htm) |
The updating of the
spreadsheets for analysing controlled trials to include adjustment for a
single covariate represents a genuine advance in Will Hopkins’ mission to
provide robust yet user-friendly analysis tools for the non-expert. Even in a
relatively large randomised controlled trial (RCT), there may be chance
imbalances across the trial arms for an important covariate. This potential
problem applies also to chance imbalances at baseline for the primary outcome
variable. Hence, in pretest-posttest RCTs a spreadsheet that permits the
inclusion of the pre-test score as the covariate is a valuable tool. (As
Hopkins points out, differences between intervention and control groups in
the mean value of a covariate may be due also to poor randomization or
selective drop-out of participants.) Further, Hopkins makes a key point that
is often under-appreciated in the analysis of RCTs: when the covariate
interacts with the treatment, including the covariate in the analysis may
improve the precision of estimation of the mean intervention effect, even in
the absence of substantial differences in the mean for the covariate between
trial arms. All of the
modifications to the spreadsheets detailed in the article enhance the usability
of the analysis tool. Among these enhancements, the correction of the standardized
effects for the small sample bias of the standard deviation stands out as an
important advance. This correction uses an appropriate modification of a
formula presented by Becker. A further highlight is the inclusion of
qualitative inferences based on the width of the confidence interval for the
experimental effect against the thresholds defined a priori for the minimum
clinically or practically important difference (for benefit and harm). In the article the
obvious limitations of the spreadsheet are acknowledged. Clearly, the tool
does not offer the flexibility and analytical power of a sophisticated
software package like SAS. However, the resource implications of using the
latter (primarily the degree of technical and statistical expertise required)
indicate the genuine need for simple yet conceptually and analytically robust
tools like these spreadsheets. In sum, the Hopkins has provided a very
valuable addition to the data analysis armoury of sport and exercise and
other scientists. Published Dec 2006. Back to article/homepage |