Statistical analysis
We used the Statistica for Windows (StatSoft Inc.), version 8.0 for all analyses. Preliminary descriptive analysis showed that the VEP N1–P1 and P1–N2 peak-to-peak amplitudes of the six blocks and the habituation slopes had a non-normal distribution. After log transformation, all data reached normal distribution (Kolmogorov-Smirnov test, p > 0.2).
A General Linear Model approach was used to analyze the “between-factor” × “within-factors” interaction effect. The between-subject factor was “group” (HV vs. MAtot or HV vs. MA subgroups); the within-subject factor was “blocks”. Two models of repeated measures ANOVA (rm-ANOVA) followed by univariate ANOVAs were employed to investigate the interaction effect. Univariate results were analyzed only if Wilks’ Lambda multivariate significance criterion was achieved. The sphericity of the covariance matrix was verified with the Mauchly Sphericity Test; in the case of violation of the sphericity assumption, Greenhouse-Geisser (G-G) epsilon (ε) adjustment was used. In rm-ANOVA and ANOVA models, partial eta2ηp2 and observed power (op) were used as measures of effect size and power, respectively. To define which comparison(s) contributed to the major effects, post hoc tests were performed with Tukey Honest Significant Difference (HSD) test.
A regression analysis was used to disclose linear trends in VEP amplitude across blocks (slope) in each group. For slope, we employed ANOVA with group factor “group” (HV vs. MAtot or HV vs. MA subgroups), using Tukey test for post hoc analysis. Also for ANOVA partial eta2 and op was used. Statistical significance was set at p < 0.05.
Pearson’s correlation test was used to search for correlations among VEP amplitude slopes and clinical variables (duration of migraine history, attack frequency, attack duration, days since the last migraine attack).