Although our work fills a gap regarding lacking meta-analytical evidence on WM alterations in AUD by strictly following state-of-the-art guidelines for neuroimaging meta-analyses (for checklist see Supplementary Table S6), it is subject to several limitations. First, even with pre-specified inclusion and exclusion criteria, the integrated studies are heterogeneous in certain aspects (e.g., sample characteristics like sex distribution, see Table 1). Second, in two studies the AUD samples comprised participants who reported co-consumption of other substances (e.g., [13, 56]). Furthermore, there is a high range of overall abstinence durations between the studies included. Therefore, meta-regression could be informative but unfortunately are not feasible in ALE analyses. Third, the ALE method is insensitive to non-significant results [28] and is thus susceptible to potentially unobserved publication bias. To account for these uncertainties, we calculated a Fail-Safe N, modified for ALE analyses [36], for each of the ALE clusters. Respective results indicate that most clusters show stable effects against additional noise studies and are not driven by a few very dominant studies.
