Table 3.

Summary of methodology used for examining correlates of BV and the associated clinical relevance [45]

Evaluation	BV parameter	Patient/disease parameters	Analysis methodology
Baseline correlation	Baseline NBV	Baseline parameters:  Age  Duration of MS since first symptoms  Number of relapses during the year before study  Number of relapses during the 2 years before study  Disability, measured using EDSS score  Disability, measured using MSFC score  Gd-enhancing lesion number  T2 lesion volume  T1-hypointense lesion volume	Pairwise Pearson (or Spearman) correlation: data presented with 95% CIs and p values determined using Fisher z transformation Statistical model selection process to identify the best baseline predictors of NBV: 1. Forward model selection based on multiple regression models and AIC was conducted separately for each study 2. The order of importance was ranked in each study, and ranks were then averaged across studies to weight the studies equally. The best baseline predictors were defined as those with the lowest mean ranks across studies. Only candidate variables that were consistently selected in all studies were considered for inclusion in the final model 3. The analysis was repeated, excluding MSFC as a candidate variable 4. A multiple linear regression model was used to investigate the combined effect of more than one explanatory variable on NBV The final model was fitted to all three studies for parameter estimation
Baseline predictors of on-study change	PBVC on-study	Baseline parameters As baseline correlation (above), plus:  NBV	Pairwise Pearson or Spearman correlation as above Statistical model selection process as steps 1–3 above, then: 4. A final multiple regression model with treatment and the two best predictors was then refitted to the data from each of the three studies to quantify PBVC as a function of the best predictors. No adjustment was made for multiplicity
Longitudinal (on-study) correlation	PBVC on-study	On-study parameters:  Number of confirmed relapses  EDSS score  MSFC score  Cumulative number of Gd-enhancing lesions  T2 lesion volume  Number of new or enlarged T2 lesions  T1-hypointense lesion volume	Pairwise Pearson or Spearman correlation: data were presented with 95% CIs and p values determined using Fisher z transformation

AIC Akaike Information Criterion, BV brain volume, CI confidence interval, EDSS Expanded Disability Status Scale, Gd gadolinium, MS multiple sclerosis, MSFC MS Functional Composite, NBV normalized brain volume, PBVC percentage brain volume change