In order to account for specificity of our PPD ALE results we created a dummy dataset (n = 25) consisting of comparable numbers of participants and MNI-coordinates to the PPD dataset. The dummy data can be found in the Supplementary Materials (S9).
Our database search yielded 1048 records for search 1 (PPD) and 13338 for search 2 (MDD). After removing duplicates, we carefully screened 704 and 9399 records and conducted a thorough review of 60 and 916 full-text documents, respectively. We included 45 articles that met our inclusion criteria for PPD [21, 35,36,37, 40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80] and 55 for fMDD ([55, 81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134]; a flow chart is available in supplementary Fig. S1). For a multimodal meta-analysis of both female and male MDD participants (literature review until 2021), please refer to supplementary Table S3 and Fig. S2.
Demographic, methodological and outcome characteristics of the included studies in PPD and fMDD are summarised in Tables 1 and 2 (a comparison of PPD and fMDD studies main characteristics is available in supplementary Table S4 and Figs. S3-S5). PPD reports include data on DTI/DKI (n = 3), structural MRI (n = 7), resting-state (n = 21), and task-based fMRI (n = 8), fNIRS (n = 2), MRS (n = 5) and PET (n = 2). PPD was diagnosed according to standardized diagnostic criteria (e.g., DSM; major depressive episode with peripartum onset) in most studies, except for four studies where cut-off scores from validated self-report questionnaires were used [21, 68, 74, 76]. The majority of reports (96%) focused exclusively on the postpartum period, ranging from early [21, 35,36,37, 41,42,43,44,45, 49, 50, 52, 57,58,59, 61, 64,65,66,67, 69,70,71,72, 77, 80] to late postpartum [62, 68, 73, 75, 76, 78] and unspecified (up to 1 year postpartum; [40, 46,47,48, 51, 55, 56, 60, 63, 79]), with only two studies collecting data antenatally (2nd or 3rd trimester; [53, 54]).
Twelve studies indicated concomitant anxiety disorders/symptoms [42, 45, 50, 57, 58, 67, 69, 70, 72, 73, 75, 78], while the remaining studies did not report any clinical comorbidity. Additionally, several studies reported first episode PPD (history of previous mental disorder excluded, including depression; n = 23; [21, 35,36,37, 40, 41, 43, 44, 47,48,49, 51,52,53,54,55,56, 59,60,61, 64, 65, 71]), although others included participants with a previous history of non-peripartum MDD (n = 8; [42, 57, 58, 67, 70, 76,77,78]) and/or previous history of PPD (n = 5; [42, 57, 58, 70, 76]). Regarding treatment status, participants across studies were either treatment-naive, not undergoing treatment at the time of the experiment or medication-free, while in four studies antidepressants or psychotherapy were accepted [66, 72, 73, 78].
For fMDD, studies used DTI/DKI/DWI (n = 2), structural MRI (n = 12), resting-state (n = 15) and task-based fMRI (n = 19), NIRS (n = 2), MRS (n = 9) and PET (n = 1). The majority of studies did not provide information on the previous history or current pregnancy status of participants, with 24 studies explicitly excluding pregnant and/or breastfeeding participants [81, 85, 86, 90, 92, 96, 98,99,100,101,102,103, 105, 108, 113, 120, 122, 124, 126, 129,130,131, 133, 134]. Eighteen studies included participants in the reproductive stage (18-49 years old; [55, 88, 95, 96, 98, 100, 101, 104, 105, 113, 115, 117, 122, 126, 129,130,131, 133]), while in the remaining studies the age range surpassed 50 years or was unspecified. All studies used standardized diagnostic criteria to assess MDD, except for one study where the method used is unclear [94].
Of the studies reviewed, most reported no comorbid conditions, while in four anxiety disorders or symptoms were present [91, 93, 95, 108]. Nine studies had either unavailable or unclear data regarding clinical comorbidity [83, 90, 109, 112, 113, 115, 118, 124, 130, 131]. Finally, 29 studies were conducted with participants who were antidepressant/medication free or naive [55, 82, 86, 88, 91,92,93, 95,96,97, 99,100,101, 104, 105, 109, 110, 114, 115, 122,123,124, 128,129,130,131,132,133,134], while in 22 studies participants were using antidepressants, undergoing neuromodulation, or receiving psychotherapy [81, 83, 84, 87, 89, 90, 98, 103, 106,107,108, 111, 112, 116,117,118,119,120,121, 125,126,127]. Four studies had unclear or unavailable data regarding current treatment status [85, 94, 102, 113].
Overall, studies ranged from moderate to high quality (supplementary tables S5-S7). Notably, bias in cross-sectional designs primarily stemmed from a lack of detailed descriptions regarding study participants and design (e.g., failing to specify the postpartum timepoint) and an inadequate identification and control for confounding factors. Concerning cohort studies, a prevalent source of bias centred around the adequacy of follow-up, with instances of follow-up rates falling below 80% or lacking sufficient information.
In white matter, PPD has been associated with increased mean diffusivity (MD) in temporo-parietal areas, superior longitudinal fasciculus, corticospinal tract, cingulum, body and splenium of the corpus callosum, external capsule, internal capsule, inferior longitudinal fasciculus, and putamen [41]. Additionally, decreased fractional anisotropy (FA) has been found in the superior longitudinal fasciculus, corticospinal tract, thalamus [41], and in the left anterior limb of the internal capsule [42], along with reduced radial diffusivity (RD) in the cingulum tract [40]. In contrast, increased FA was observed in the right anterior thalamic radiation and cingulum tracts [40].
In grey matter, PPD participants had increased volume (GMV) in the left DLPFC [43, 44], right anterior insula [44] and OFC [43] and reduced GMV in AMY [46]. Significant volumetric differences were also noted in the right ACC and left middle-PCC [45], as well as increased cortical thickness (CT) in the left superior frontal gyrus, cuneus and fusiform gyrus [49], but decreased CT in the right inferior parietal lobule [47]. Additionally, increased surface area was observed in the left superior frontal gyrus, caudal middle frontal gyrus (MFG), middle temporal gyrus (MTG) and insula, along with increased mean curvature in the parietal lobules [47].
Four studies [42,43,44, 47] on 136 participants were included in the structural meta-analysis and no significant clusters were identified.
In the PFC, increased resting-state connectivity was observed between the left DLPFC and right ACC [43], while connectivity was decreased between the dorsomedial PFC and left ventral striatum [35], precuneus and PCC [58] and between the DLPFC, ACC and AMY [57]. Decreased sample entropy was found in the left medial PFC [37], as well as reduced regional homogeneity (ReHo) in left DLPFC [53, 59] and decreased voxel-mirrored homotopic connectivity (VMHC) in bilateral dorsomedial PFC [65]. Increased values in amplitude of low-frequency fluctuations (ALFF) were found in left medial PFC and DLPFC [54]. For the OFC, increased connectivity was observed with the right MFG and left inferior occipital gyrus [43], as well as decreased ALFF [54] and VMHC [65].
Regarding the ACC, its subgenual part (sgACC) showed increased connectivity with the ventral anterior insula [35, 36] and decreased connectivity with the superior and MTG [36]. Within the right hippocampus, degree centrality and ReHo were increased [59, 64], as well as connectivity with the left precuneus and left superior frontal gyrus [59], while connectivity was reduced with the right MFG and left median cingulate and paracingulate gyri [64]. The PCC showed reduced connectivity with the right AMY [50] and with the right paracentral lobule [52] and increased ReHo [61]. Finally, ReHo and VHMC reductions were found in the right insula [53, 54, 59, 63] and AMY [53, 63].
Eighteen studies were included in the resting-state meta-analysis (16 experiments; [35,36,37, 43, 50, 51, 53,54,55,56,57,58,59, 61,62,63,64,65]; 367 participants), which highlighted abnormalities in the left MFG (Fig. 1; Table S8).
Studies using infant stimuli in fMRI report differential emotional processing in PPD. Specifically, Dudin et al. [66] found an increased right AMY response to unfamiliar smiling infants, while Wonch et al. [72] observed a general increase in BOLD response in the right AMY. The latter also reported decreased bilateral AMY-right insular cortex connectivity when participants viewed faces of their own versus other infants. Lenzi et al. [68] found increased deactivation in the orbital and medial PFC and an increase in right AMY reactivity. Finnegan et al. [67] observed a differential response to infant versus non-infant stimuli in brain regions such as the right dorsolateral superior, middle, and inferior frontal gyri, the left inferior and middle temporal lobe, and bilateral angular gyri. Interestingly, the authors found that a history of depressive episodes did not independently impact these neural responses.
For non-infant stimuli, Moses-Kolko et al. [69] observed increased nonlinear attenuation of left ventral striatal activity after reward in PPD. In response to negative stimuli, decreased activation was observed in bilateral OFC, cingulate, putamen, precuneus, DLPFC, ACC [21], right AMY [21, 71], left AMY and left dorsomedial PFC [70], alongside increased activity in the bilateral insula [21]. For positive stimuli, decreased activity was found in striatum, cingulate gyrus, DLPFC and precentral gyrus [21]. Using fNIRS, increased depression severity was found to be associated with decreased connectivity between the temporoparietal junction (TPJ) with lateral PFC and increased connectivity between TPJ with anterior medial PFC [73]. In contrast, Song et al. [74] found no differences in integral or centroid values.
Six studies were included in the task-based meta-analysis [21, 68,69,70,71,72]; 67 participants) and no significant clusters were found.
An increase in monoamine oxidase A in the PFC and ACC was found in PPD [79] but no differences in D2/3 receptor binding potential [80]. MRS studies identified a decrease in glutamate-glutamine (Glx) and N-acetylaspartate (NAA) levels in the left DLPFC [78]. Additionally, McEwen et al. [77] found increased glutamate (Glu) levels in the medial PFC in PPD, though other metabolite levels (NAA, creatine [Cr] and choline [Cho]) did not show significant differences. There was also a trend towards decreased cortical gamma-aminobutyric acid (GABA) levels in PPD [76], although Deligiannidis et al. [58] found no significant differences in GABA/Cr concentrations in the pregenual ACC or occipital cortex.
In the pooled meta-analysis of all included studies (25 experiments, 542 participants), women diagnosed with PPD exhibited structural and functional changes in right putamen, right amygdala and left MFG (Fig. 2 and Table S8). Additional exploratory direction of effect analyses results are provided in supplementary Table S9 and Fig. S6.
In the seed-based connectivity analysis using the left DLPFC (MFG) as the seed region, significant positive connectivity was observed across several cortical areas (Fig. 3). Regions with the strongest connectivity (yellow) include areas of the bilateral DLPFC and angular gyrus. Additional activation is seen in adjacent prefrontal regions, as well as posterior parietal areas. Areas with lower but still significant connectivity (red) extend into occipital and temporal cortices.
In white matter, decreased fibre-density was observed in the left and right frontal projections of the corpus callosum, right anterior limb of the internal capsule, tapetum, and right inferior longitudinal fasciculus [82]. Additionally, reductions in FA were widespread in the genu of the corpus callosum, bilateral cerebral peduncles, forceps minor and major, bilateral inferior fronto-occipital fasciculus [82], left bilateral uncinate fasciculi [81, 82], inferior and superior longitudinal fasciculi [81].
For gray matter, reduced GMV was found in bilateral ventral ACC [91], right AMY [91, 93], bilateral caudate extending into the anterior nucleus of the thalamus [87], medial PFC [84], left lingual gyrus extending to the parahippocampal gyrus, cerebellum [92, 93], bilateral insula, bilateral putamen, and caudal middle-frontal region [93]. However, another study did not find GMV differences within the AMY, hippocampus, sgACC or putamen [87]. Reductions in AMY volume were observed across studies [85, 90].
Nine studies were included in the structural meta-analysis ([82, 84, 86,87,88,89, 91,92,93]; 251 participants). Women diagnosed with MDD manifested structural alterations in left putamen gray matter and bilateral sub-lobar extra-nuclear white matter (Table S10; Fig. 1).
Decreased connectivity was observed between: the right AMY and the ventrolateral PFC, bilateral insula, and bilateral putamen [93]; the left hippocampus and temporo-occipital regions, including the bilateral lingual gyrus and fusiform [103]; and the left middle occipital gyrus and the left OFC [104]. Increased connectivity was found between the left middle occipital gyrus and the left medial prefrontal gyrus and the left hippocampus [104] and between the left MFG and bilateral putamen [96]. ALFF reductions were reported in the right putamen, right MTG [106], left middle occipital gyrus [104], right postcentral gyrus [96] and right superior occipital gyrus [105], while increases were observed in the left medial PFC [106], left MFG [102, 106], left precentral gyrus [102] and left temporal pole [55]. There were no observed differences in fractional ALFF in the ACC and insula [98]. ReHo was elevated in the left sgACC and left thalamus [55]. Conversely, BOLD signal variability was reduced in bilateral cerebellum [100] and the DLPFC [101]. Nine studies were included in the resting-state meta-analysis ([55, 84, 96, 100, 102,103,104,105,106], 192 participants). No significant clusters were found.
The ACC showed increased activation during the presentation of positive stimuli [107, 119], emotional approach and withdrawal conditions [109], incongruent conditions [123] and rejection [125], while reduced activity was observed in response to negative stimuli [116]. Connectivity of the ACC was reduced with the AMY during negative stimuli and with the DLPFC during high-attention stimuli [120]. Different patterns emerged in response to positive and fearful stimuli, with increased inverse connectivity between the left-sided sgACC and AMY to happy faces, and increased positive connectivity between the same regions to fearful faces [108].
The DLPFC also showed increased activation during expectation of negative stimuli [107], incongruent versus congruent contrasts [123] and painful stimuli [110]. In contrast, decreased activation was found during low-risk cheating choices [121] and decreased connectivity with the right AMY during high-attention stimuli [120]. Reduced activity was also found in: the dorsal putamen and anterior insula during low-risk cheating choices [121] and in frontoparietal network and salience networks, irrespective of stress [115]; in the right caudate during the recall of positive specific memories. On the other hand, increased activation was noted in the PCC, insula, and thalamus during the recall of negative specific memories [124]. Using fNIRS, a significant correlation between depression scores and changes in oxy-Hb in the right frontal brain region was observed [126], as well as reduced oxy-Hb activation in the DLPFC [127].
Fourteen studies were included in the task-based meta-analysis ([107, 110,111,112, 116,117,118,119,120,121,122,123,124,125]; 225 participants). Women diagnosed with MDD manifested alterations in the left ACC (Fig. 1; Table S10).
In the medial PFC, there were no significant differences in Glu, GABA, or Glx levels [106, 130]. Similarly, Glu levels in the left DLPFC were not significantly different, although there was a reduction in GABA+ levels and in the GABA+ to Glu ratio in this region [131]. In the ACC, there was a reduction in GABA levels [115] and in NAA to phosphocreatine plus creatine (NAA/PCr+Cr) ratio [133]. In the ventral prefrontal white matter, reduced Cho/Cr ratios were observed bilaterally, while NAA/Cr levels showed no significant differences [132]. In the hippocampus, there were increased total Cho levels [103] and decreased taurine concentration [129]. The only PET imaging study revealed a global reduction in regionally normalized cerebral blood flow in the DLPFC [134].
In the pooled meta-analysis of all included studies (32 experiments, 652 participants, [107, 110,111,112, 116,117,118,119,120,121,122,123,124,125]), women diagnosed with MDD exhibited changes in right putamen and amygdala (Fig. 2). Direction of effect analyses are provided in supplementary Table S11 and Fig. S7. In order to investigate if there was a potential effect of age in the fMDD sample, we calculated an exploratory subgroup analysis for fMDD studies that was age-matched to PPD samples. The general pattern of ALE results was maintained in this age-matched exploratory analysis (supplementary Fig. S8).
The PPD derived network (Fig. 3) was marked by involvement of bilateral DLPFC and angular gyrus.
The fMDD derived right amygdala seed region (Fig. 3), showed connectivity increase in the somatosensory and motor cortices and anterior temporal lobes. Further connectivity changes extend into the posterior temporal cortices and insulae.
Both PPD and fMDD show widespread reductions in FA in several tracts, including the superior longitudinal fasciculus [41, 81]. PPD is associated with increased GMV in the right insula and right precentral gyrus [43, 44], while MDD exhibits reduced GMV in the same areas [89, 93]. Both conditions show reduced AMY volumes [46, 91, 93].
PDD and fMDD present both shared and distinct patterns of altered brain function. Shared findings include increased ALFF in left medial PFC [54, 106], increased ReHo in thalamus [55] and decreased activity in the ACC and cingulate in response to negative stimuli [21, 116]. Additionally, there is reduced activity in the cingulate gyrus in response to positive stimuli [21, 116], and reduced AMY-insula connectivity [72, 93]. However, PPD is characterized by decreased activity in AMY and precuneus, along with increased activity in the insula, in response to negative stimuli [21, 71]. Conversely, fMDD has increased activity in the AMY and precuneus and decreased activity in the insula [107, 116].
One study directly compared PPD, fMDD and HC to identity shared and different resting-state neural circuits [55]. Both PPD and fMDD groups showed higher fALFF in the left temporal pole (vs HC). The fMDD group showed a specifically increased FC in the right cerebellum, whereas PPD had specifically decreased fALFF in the left supplementary motor area and the posterior MTG, and reduced posterior MTG-precuneus and left-right sgACC connectivity. Additionally, there were significant ReHo differences in the left thalamus and left sgACC (PPD > fMDD > HC).
In the medial PFC, PPD shows increased Glu levels [77]. In fMDD, however, there were no significant differences in Glu, GABA, or Glx concentrations in this region [106, 130].
The conjunction analysis (57 experiments, 1194 participants) identified regions in the right insula, left ventral lateral nucleus (thalamus), left caudate, right amygdala, left cingulate gyrus and bilateral putamina demonstrating convergent brain changes in both PPD and fMDD (Table S12; Fig. 4). Contrast analyses revealed that PPD was associated with more prominent alterations in temporal lobes and somatosensory cortices, while fMDD showed stronger involvement of the DLPFC and ACC (Fig. 4).
Dice coefficients for PPD and fMDD were low for both symptom networks: 0.0858 (PPD/anxiosomatic), 0.0924 (PPD/dysphoric), 0.0845 (fMDD/anxiosomatic) and 0.0995 (fMDD/dysphoric).
The dummy dataset of randomly assigned coordinates did not result in any significant cluster accounting for strict thresholding. This suggests that our meta-analytical results on PPD are not merely the consequence of compiling heterogeneous study results for analysis.