Interestingly, the baseline sample tree based on the clustering of significant pathways was very similar to the baseline intrinsic subset assignment results from Fig. 1.
When the effect of group was adjusted for the clustering of participants within communities, baseline weight, age, and sex, the magnitude of the association remained strong at 3 (P = 0.004) and 6 (P = 0.004) months in favor of FF.
During the winter, when there were the most baseline visits, a cluster of size 10 was about one-seventh the size of the standard deviation for number of weekly visits.
For continuous outcomes, analyses will adjust for baseline by inclusion of the cluster mean of the outcome in question as a covariate in statistical models.
At baseline, the mean cluster size was 12.7 (range 1 56).
We compared differences in two year moderate and severe exacerbation counts by using the negative binomial model, 30 correcting for age, sex, MRC score above 2, exacerbation rate in the year before baseline, and clustering of patients per practice.
A univariate regression analysis was performed for each outcome variable for all 250 imputed cases, adjusted for baseline values and the clustering effect of school.
Between group differences at 12 months were analysed using linear regression on complete data, with the variable of interest at 12 months as the outcome variable adjusted for baseline values and the clustering effect of schools.
This formed the baseline data of the cluster randomised control study described above.
A significant association was also found between blood SDF-1α and cortisol baseline, and clusters of CXCR4-expressing cells scattered were observed within the adrenal corticosteroid productive area.
