We read with interest the meta-analysis by Cheng and colleagues entitled “Neural systems underlying autobiographical memory dysregulations in depression: A neuroimaging meta-analysis” (Cheng CCC, Tsang MHL, Han M, et al. Psychological Medicine. 2026;56:e1. doi:10.1017/S0033291725102833). The authors preregistered the study, which is an important step toward methodological transparency. However, we have several concerns about the validity and robustness of the reported findings.

First, the authors did not define how multigroup studies were handled if no pairwise comparisons were reported. Five studies included designs with three to four groups and reported F-test results but no post-hoc comparisons. This would, however, be needed to derive the required pairwise statistics (t-values/effect sizes) for the seed-based d mapping approach. The authors did not state whether – nor how – pairwise statistics were approximated. In addition, they did not explain how contradictory findings were handled in the analysis. For example, in studies by Young and colleagues (e.g. Young, Bellgowan, Bodurka, & Drevets, Reference Young, Bellgowan, Bodurka and Drevets2013, Reference Young, Bellgowan, Bodurka and Drevets2014; Young, Bodurka, & Drevets, Reference Young, Bodurka and Drevets2016), activity did not follow a graded pattern for some regions (i.e. lowest activity in healthy participants, highest in patients with depression, and at-risk individuals falling in between). In contrast, the highest activity was observed in individuals at risk for depression.

Second, the selection of contrasts is not clear. The authors included several studies that reported multiple whole-brain contrasts from the same participant sample, but they did not specify which contrast was entered into the meta-analysis. Table 1 lists the tasks used in each study but does not indicate the specific contrasts either (e.g. Foland-Ross et al., Reference Foland-Ross, Cooney, Joormann, Henry and Gotlib2014; Gillard, Werner-Seidler, Dalgleish, & Stretton, Reference Gillard, Werner-Seidler, Dalgleish and Stretton2023; Parlar et al., Reference Parlar, Densmore, Hall, Lanius and McKinnon2018; Young et al., Reference Young, Bellgowan, Bodurka and Drevets2013, Reference Young, Bellgowan, Bodurka and Drevets2014; Young, Bellgowan, Bodurka, & Drevets, Reference Young, Bellgowan, Bodurka and Drevets2015). This lack of clarity raises concern that multiple contrasts from the same sample may have been included in the meta-analysis. If so, this would violate the assumption of independence by effectively counting the same data more than once, potentially inflating the meta-analytic effects.

Third, the authors included two studies (Keedwell et al., Reference Keedwell, Andrew, Williams, Brammer and Phillips2005; Macdonald et al., Reference Macdonald, Murray, Moutsiana, Fearon, Cooper, Halligan and Johnstone2016) that reported center-of-mass coordinates rather than peak voxel coordinates. A seed-based d mapping meta-analysis requires peak voxel coordinates, as center-of-mass coordinates may be displaced from the location of the maximum effect, thereby reducing spatial precision and accuracy. The authors may have used a pragmatic approach to maximize study inclusion, but this should have been acknowledged. Presenting analyses both with and without these studies would clarify whether the findings are influenced by their methodological choice.

Fourth, at least two included studies shared the same participant pool (Young et al., Reference Young, Bellgowan, Bodurka and Drevets2013, Reference Young, Bodurka and Drevets2016). The authors did not explain how this was handled in their analysis. Seed-based d mapping assumes independence between studies, and treating overlapping samples as separate studies may inflate effect sizes.

Fifth, the definition of who counts as depressed, and therefore should be included in comparisons with healthy individuals, is unclear. In the introduction and the inclusion criteria, it reads that “fMRI studies on depression and at-risk of depression” are included, while table 1 also lists patients remitted from depression. The text later refers only to individuals at risk of depression or patients with depression. If they included patients with remitted depression, a sensitivity analysis that excludes them would be missing. The authors do report a sensitivity analysis that excluded individuals at risk of depression. They say in the discussion that this does not change the results. Supplementary table S3, however, shows a significant difference between patients with depression and healthy individuals in the right medial cingulate but not in the right precuneus, the right middle temporal gyrus, or the right insula.

Sixt, iSixt, in the main results, the authors report that negative autobiographical memory recall is associated with increased activity in the cingulate/paracingulate cortex and the right precuneus. However, in figure 2b, which intends to illustrate the main results, one region is labeled as ‘corpus callosum’, which is not mentioned in the text. Table 2 lists the ‘corpus callosum’ at coordinates (18, −62, 44). These coordinates are unlikely to correspond to the corpus callosum, suggesting a possible mislabeling in both the figure and the table. There is also an inconsistency in the reported number of included studies. The main text refers to 13 studies with 12 unique participant pools, table 1 lists 15 studies with 13 unique participant pools, and figure 1 indicates 14 studies.

Then, in the discussion, the authors interpret increased activation in the dorsal anterior cingulate/paracingulate cortex in patients with depression as reflecting enhanced involvement of the salience network. However, reduced activity in the anterior insula was also reported. Given that the anterior insula is a core node of the salience network, it is unclear why this apparent inconsistency was neither acknowledged nor discussed. We would also like to note that the insula was described as hypoactive in the results section but as hyperactive in the discussion.