Recently we have conducted a prospective study to evaluate if BMI is associated with treatment response to a combination DEMARDs therapy. We found an inverse association between BMI and disease activity at baseline. Also, we found an inverse association between BMI and response to treatment (change in DAS28 after 6 months); those with higher BMI had less DAS28 at baseline and also had less changes in DAS28 after 6 months of therapy. When we run a multivariate analysis and consider the baseline DAS28 as a co-founder, the association between BMI and treatment response disappeared.
The question is that:
- Should we conclude that the whole association is confounded by baseline DAS28 and there is no real association between BMI and treatment response?
- How we could find that what amount of the association between BMI and treatment response is confounded by baseline DAS28? and what amount is a real association?
I agree with Audrey,
mulitcollinearity should be no problem, you actually want a relation of your independent and dependent variables. In your case, by including the DAS28 value of the pretest you are able to control for differences before the intervention.
Also plotting is a very good idea! But why not try a repeated measures AN(C)OVA? Have you tried that? With a model like that you would be able to
a) take a look at the interaction of time and BMI (or other independent variables)
b) include other variables such as gender, age, etc. (I'm just guessing, you can include whatever variables you assessed that might have an impact on either BMI or change in DAS28).
You definitely need to control for baseline DAS28, but as far as I can judge from your information you want to take a look at the interactions in the model. I am not sure what you mean by multivariate: did you really define more than one dependent variable or was your outcome only DAS28 after six months?
After controlling the baseline DAS if you dont get any association then i think there exist no significant correlation between BMI and treatment response.
Indeed we conducted a linear regression analysis considering baseline DAS28 and BMI as possible independents and delta-DAS28 after 6 months as dependent.
One potential problem with this approach is that DAS28 and delta-DAS28 might be highly correlated (say r > .7). If this is the case, you may eventually suffer from a problem of multicollinearity in regression analysis. You could prevent this situation by orthogonalization of DAS28 and delta-DAS28 (i.e., you would only go for the change residuals that are not accounted for by high DAS28 baseline levels, for example). Nevertheless, I would strongly encourage you to using a statistical framewok that allows you to model the dependencies of your variables adequately (e.g., multilevel or path analysis).
Hi,
I suggest you an ANCOVA but you will need to transform your BMI covariate into a categorical one.
For the linear regression, you have to verify several assumptions:
- Is your association linear between i)baseline DAS-28 and delta-DAS28 ii) BMI and delta-DAS28.
- Is delta-DAS28 gaussian ?
If not, you have to use a different kind of regression.
I think there is no problem of collinearity as Nils said, because the correlation he talks about is between the dependant and an independant variable, not between two independant ones.
One final point: make boxplots of delta DAS-28 for different combined classes of BMI and baseline DAS-28. You will be able to see if there is an effect of BMI beyond the baseline effect of DAS-28.
I agree with Audrey,
mulitcollinearity should be no problem, you actually want a relation of your independent and dependent variables. In your case, by including the DAS28 value of the pretest you are able to control for differences before the intervention.
Also plotting is a very good idea! But why not try a repeated measures AN(C)OVA? Have you tried that? With a model like that you would be able to
a) take a look at the interaction of time and BMI (or other independent variables)
b) include other variables such as gender, age, etc. (I'm just guessing, you can include whatever variables you assessed that might have an impact on either BMI or change in DAS28).
Converting BMI to a categorical variable throws information away. If you have continuous data, keep it continuous. There is little advantage to categorizing a continuous variable.
The big question still remains, how would you categorize BMI? Is the typical cut off of 30 for obesity meaningful in this population as RA patients tend to have BMIs above the average of the general population? That's clinical question and your selection of categories might drive your results. Keep BMI continuous.
Also be careful about including the baseline and the change in your model, if not collinearity, then loss of power might be an issue.
We have been successful in modeling BMI using restricted cubic splines rather than categorizing it (ala Frank Harrell's suggestions in RMS). Our experience is that BMI's relationship with disease is far from linear, and is often u-shaped, so plotting the relationships will give a much clearer picture.
I agree with Audrey and Sebastian Franke in biostatistics, but I would like to firstly think about "change in DAS28" after treatment, because "change in DAS28" could be expressed as "the value of decrease/increase" or "the percentage change of baseline", more important these two expressions have different meanings in biology and they also could give you different biostatistics results.
Thanks & Regards
Jinbiao
change in DAS28 was defined as the amount (not percentage) of change in DAS28. Considering percent change, there was no difference in regression analysis results.
I would conclude (not final) that (i) the effect of treatment on DAS28 does not depend on BMI, (ii) the relationship between BMI and DAS28, especially at baseline, need to be further investigated. If BMI truly negatively affects the level of DAS28, the mechanisms behind this is not affected by the combination DEMARDs therapy, hence there is potentially new one. Good luck.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics