A study on effect of a particular yogic maneuver was carried out in our department. It involved two groups, control and intervention group( n = 50 in each). It so happened that, at the end of study period cardiovascular parameters like BP, HR showed significant fall in the intervention group, at the same time the other group which was not having any intervention ended up showing a significant increase in BP and HR at the end of the study period. Now how to do i account for this? Has any one come across such a problem before?
A significant change in the control group is not a problem in principle. In many cases the control group will change for any one of a number of reasons:
- test sophistication: the second time that people are assessed, they are more familiar with the assessment (less anxious, for example) and with any tasks involved (so they may have learned skills that help with taking the test)
- regression to the mean: trials often enrol participants who meet threshold criteria. For example, they may have a blood pressure greater than a given figure. Or the participants themselves are at the point where they have just sought help for a problem because it's worse than usual. In each case, measuring them at a second time point will give you a different mean value simply because of random fluctuation in the parameter used to define trial entry
- method of measurement may change. I would suspect this one in your case. We had a study where the pre-trial ECGs were read by one doctor, and the post-trial ECGs were read by another (the first doctor died during the trial).The discrepancies were such that we had to have all the pre-trial ECGs re-read by the post-trial doctor.
But the whole point of a control group is that these changes are accounted for in the design. The analysis model is comparing intervention and control, with adjustment for baseline values.
However, a change in an unexpected direction in your control group will require some explanation when you go to publish! Have you any ideas??
This is a very typical problem, although it is rarely correctly handled in medical literature.
You have two factors that influence the response: "time" and "intervention".
The control group shows you the effect of time, the intervention group shows you the sum of both effects. The difference of these two effects (effect in intervention-group minus effect in control group) gives you the corrected intervention effect.
This kind of experimental design is called BACI for Before-After/Control-Impact. You will find a lot in a google search. They key aspect is that this is eventually a two-factorial model where the interaction of the two factors is the relevant parameter.
Hi, I would assume that you randomly allocated your people to each of the groups. Even so, a lack of this would not explain the individuals groups shift. Were the conditions stable under which you assessed both all participants HR and BP at both T1 and T2 that is, across both groups and by the same machine/person. If there was any variance in the time, place or nature of assessment or by the person interpreting the data, this could explain your results. This is a systematic error and would mean the design rather than the results were at fault- as not all conditions were 'the same' across groups, across times.
This is a typical situation situation for a two-way anova. See the link below for the tutorials. One is for SPSS and other is for R.
https://statistics.laerd.com/spss-tutorials/two-way-anova-using-spss-statistics.php
http://courses.statistics.com/software/R/Rtwoway.htm
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis