(i) In multivariate models, after accounting for the protective effects of HLA-B*57 alleles, we find that the HCP5-G appears to have detrimental effects on both viral load and disease progression.
However, in multivariate models, after partitioning out the protective effects of B*57, the HCP5-G allele associates with disease-acceleration and enhanced viral replication; these associations for HCP5-G are otherwise obscured because of the very strong LD between this allele and a subset of protective B*57 alleles.
Table 3 shows the final multivariate model after backwards removal of statistically weak associations.
In the multivariate model, after the adjustment for confounders, the two associations remained statistically significant.
Variables selected more than 5 times out of 10 replications were included in the multivariate model after multiple imputations.
We included alcohol consumption as a covariate in the multivariate model after excluding women without alcohol information, restricting the sample size to 114,804 women.
For count variables, negative binomial regression was used in bivariate and multivariate models after rejecting the hypothesis of equivalence in the mean and variance of the count.
In the multivariate model, after backwards elimination, only local transducer tenderness, a thickened appendix and peri-appendiceal fat infiltration were significant at the level 0.05 (Table 1).
Two of seven key species (Prevotella spp. and A. vaginae) remained significantly associated with BV in a multivariate model after adjusting for other bacterial species.
In the full multivariate model, after adjustment for neurologic treatment status, the median parametric survival estimate for the entire population was 13.6 days (95% CI 12.4 14.7).
