Far East Journal of Theoretical Statistics
Volume 16, Issue 2, Pages 353 - 361
(July 2005)
|
|
GENERALIZABILITY OF REGRESSION MODELS FOR THE FEV1-MEASUREMENTS IN ASTHMATIC CHILDREN
Melissa Naylor (U. S. A.), Benjamin Raby (U. S. A.), Scott T. Weiss (U. S. A.) and Christoph Lange (U. S. A.)
|
Abstract: Large
scale asthma studies often rely on
measurements of lung function such as the
forced expiratory volume in one second (FEV1)
to assess severity of asthma. However, it is
generally accepted that prediction equations
derived in populations from one ethnic group
are not generalizable to other self designated
ethnic populations (reference PubMed ID
1952453). It is therefore possible that using
percent of predicted FEV1 as a
metric of lung function may not be optimal for
population-based studies. We examined this
formally using data from the Childhood Asthma
Management Program (CAMP), a multicenter,
randomized, clinical trial including 1,041
children with mild to moderate asthma. New
prediction equations for FEV1 were
developed for each of three self designated
ethnicities (Caucasian, African American, and
Hispanic).
The correlations between residuals for
each of the new regression models and the
standard percent of predicted FEV1
based on normal self designated Caucasians
were examined. There was a strong correlation
among the residuals of all of the percent
predicted models based on this data set;
however, correlation between the standard
percent of predicted FEV1 based on
normal subjects and all ethnic specific models
was low. Therefore, percent of predicted FEV1
based on normal subjects is not an optimal
phenotype for this study.
Since FEV1 depends on a
variety of factors such as height, weight,
environmental exposures, genetics, etc.,
direct adjustment for the relevant covariates
rather than indirect adjustment using percent
of predicted FEV1 may be the most
useful measurement in large scale asthma
studies. |
Keywords and phrases: regression analysis, confounding, model building, asthma. |
|
Number of Downloads: 271 | Number of Views: 574 |
|