14800-DvRappard

112 Chapter 7 Statistical analysis We performed statistical analysis (using IBM SPSS V.22) for patients with baseline examinations and the control subjects. We first corrected for age-dependency: for each metabolite, we estimated the parameters a, b and tau based on a mono-exponential fit (a+b·exp(-age / tau)) through the control measures. For both control subjects and patients, we subtracted the predicted concentration at corresponding age from the observed concentration (residuals). 14 A general linear model was used for overall comparison of these age-corrected metabolite concentrations (residuals) at baseline. Post-hoc, we compared differences between all four groups, i.e. patients with good, moderate or poor outcome, and controls, with Dunnett’s T3 (correcting for multiple comparisons in case of small groups with unequal variances). MRI scores and lesion volumes were similarly compared between patients. p<0.05 was considered significant. For patients with baseline examinations, we intended to use logistic regression to determine which (age-corrected) metabolites at baseline explained outcome. However, data showed complete separation, as illustrated by scatterplots of age-corrected metabolite concentrations (shown in Results). Therefore, we distinguished groups based on visual comparison of scatterplots. Spearman’s rank correlations were determined between baseline metabolite concentrations and MLD-GMF at latest follow-up. RESULTS Metabolite concentrations at baseline and clinical outcome In all MLD patients at baseline, there were significant group differences in MRI score and lesion volume (p≤0.001). Patients with poor outcome had significantly larger lesion volumes than patients with moderate or good outcome, p<0.001 (Table 2), such that overall WM represents predominantly lesions in poor outcome patients. We observed significant group differences (p<0.001) for all investigated metabolites (Table 2, Figure 2). Compared with controls, differences were most pronounced in patients with poor outcome, for whomconcentrations of NAA, Glu and Glx were severely reduced, while Ins and Lac were increased (all p<0.01). Patients with moderate outcome had decreased NAA (p<0.01), and patients with good outcome had increased Cr, Cho, Ins and Glx compared with controls (p<0.05). Differences in metabolite concentrations compared with control WM were most prominent in lesions, and less so in NAWM (data not shown).