I'm going to give a talk tomorrow about confidence intervals, and I will be
using examples from a paper on exercise
Dunn AL et al (1999) Comparison of Lifestyle and Structured Interventions to
Increase Physical Activity and Cardiorespiratory Fitness, JAMA 281(4)
327-334.
I chose this paper because it doesn't have words like "hyperbilirubinemia"
that I always mispronounce and barely understand.
The basic conclusion of this paper is that both a traditional program of
structured exercise and a newer program that also includes lifestyle changes
are effective in increasing physical activity and fitness, but that there is
no substantive difference between the two interventions.
I want to get a discussion going about whether the size of the changes
indicated by confidence intervals in this paper are clinically relevant.
Perhaps some members of this list would be willing to share their expertise
on what size difference they would hope to see and why.
The primary outcome variable is energy expenditure (kcal/kg per day). In the
methods section, the authors imply that an improvement of 2 units would be
considered relevant. Although both interventions showed an improvement, the
confidence limits were well below this target (0.42 to 1.25 and 0.25 to
1.12). In the discussion section, the authors nicely point out that "even
though the mean increases were statistically significant, some may not
consider them to be practically significant."
They then mention that changes which may seem small from an individual
viewpoint may still be important from a public health context.
So the question becomes, is it worthwhile to find a change of about 0.5 to 1
kcal/kg per day over 24 months for a targeted intervention of structured
exercise and/or lifestyle changes?
Some of the secondary endpoints are interesting also. How much of an
improvement would be considered clinically relevant for the following
outcomes:
Walking (min/day)
VO2peak (ml/kg per minute)
Submaximal heart rate (beats/min)
Body fat (percentage)
Total cholesterol (mg/dL)
Systolic and Diastolic blood pressure (mm Hg)
I'm very interested in HOW you arrived at the decision about what you would
consider clinically relevant for these endpoints.
Any comments received today (Wednesday, February 24) would be greatly
appreciated, but late comments will also be helpful as I hope to repeat this
talk for other groups.
Steve Simon, [log in to unmask], Standard Disclaimer.
STATS - Steve's Attempt to Teach Statistics: http://www.cmh.edu/stats
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|