Progressive grey matter alterations in bipolar disorder across the life span – A systematic review

To elucidate the relationship between the course of bipolar disorder (BD) and structural brain changes across the life span, we conducted a systematic review of longitudinal imaging studies in adolescent and adult BD patients.


| INTRODUC TI ON
2][3] Initial psychopathological symptoms in later BD patients frequently appear in childhood and adolescence. 4,5In crosssectional studies, both adolescent and adult BD has been associated with structural brain changes. 6,7However, little is known about how these structural brain changes arise and why some patients develop them more markedly than others. 8A better understanding of the factors that contribute to brain changes across the lifespan could contribute to a better understanding of the course of BD.At this stage, longitudinal studies could improve our understanding of the factors that lead to these structural brain changes. 9These longitudinal studies allow us to examine the complex interplay of brain maturation, disease progression (e.g. the number of mood episodes) and treatment history in BD.In contrast to cross-sectional studies, they yield information about the temporal sequence of mood episodes and grey matter loss.In different age groups, in particular, longitudinal analysis can disentangle the effects of ageing and mood episodes on the brain.Here, the effect of mood episodes after disease onset in adolescent patients can be compared to progressed BD in adults and geriatric patients.Considering withinsubject rather than between-subject comparisons may moreover allow a more sensitive and fine-grained characterisation of the impact of disease progression on brain structure.Insights into the dynamics of brain structural alterations and their aetiology could enable earlier detection of BD -a major challenge in treating BD patients.For these reasons, this review provides a systematic summary of the current state of research regarding longitudinal structural brain changes in BD across the lifespan.

| Importance and current state of evidence on brain structural alterations in BD
The importance of longitudinal neuroimaging studies for the understanding of BD is indisputable, yet most studies have investigated structural brain alterations cross-sectionally.Most frequently, crosssectional studies reported decreases in grey matter volume (GMV).
Meta-analyses of voxel-based morphometry (VBM) studies found decreased GMV in the prefrontal cortex (PFC), the temporal cortex, the insula and the anterior cingulate cortex (ACC) in adult BD patients compared to healthy controls. 10,11Decline of GMV in subcortical regions has been detected in the amygdala and the hippocampus. 12The alterations identified in these studies concern key regions of emotional processing and executive function. 13In the most highly powered analysis of MRI studies to date performed by the ENIGMA (Enhancing Neuro Imaging Genetics Through Meta-Analysis) working group it was reported that BD patients show reduced cortical thickness (CT) in frontal, temporal and parietal regions. 6Taken together, cross-sectional studies indicate that individuals with BD show structural abnormalities in limbic and prefrontal brain regions relative to healthy controls.
While for cross-sectional mega-analyses large consortia such as ENIGMA have gathered thousands of patients with BD, 6 a recent longitudinal ENIGMA mega-analysis contained only a very limited number of investigations of structural brain alterations in BD patients. 14wever, these longitudinal studies yield important perspectives: Longitudinal investigations of brain structural changes allow us to study of the impact of events that occur over time, such as disease episodes or disease progression, in conjunction with brain maturation and ageing.Thus, developmental stages (e.g.adolescent vs. adult BD) can be compared to highlight the role of differential brain changes that are associated with disease progression across the lifespan.Here, changes in one brain structure might predominantly show at disease onset, whereas changes in another might rather be associated with repeated episodes.Moreover, the nature of the mood episode (manic vs. depressive) could affect brain structural changes differently.Only longitudinal perspectives that carefully survey disease trajectories can decipher the complex interplay of brain development and the impact of mood episodes during BD progression.

| Effects of ageing on structural brain abnormalities in bipolar disorder
Most MRI studies of BD individuals include adult patients with a long history of disease.However, the course of BD typically starts in adolescence or young adulthood. 15To detect brain structural abnormalities in the early stages of BD, studies in young and early onset patients with BD are required.
Differences in structural brain changes between younger and older people with BD have been described in a recent cross-sectional mega-analysis. 6In this analysis, adolescents and young adults showed increased cortical thinning compared with older patients with BD and healthy controls (HC) of the same age.Thus, this effect could not be explained by the diagnosis of BD alone, indicating that there is an interplay of adolescent brain maturation processes and the pathophysiology of BD. 6 The adolescent brain is still characterised by greater plasticity than the adult brain. 16Neurodevelopment in healthy adolescence has been found to occur both progressively and regressively in areas critical to the reward system and in areas activated by emotional stimuli. 17These reward-and emotion-processing regions have been found to be smaller in adult BD patients, 18 in line with functional impairment of emotion processing and regulation. 19,20Taken together, the evidence suggests that structural brain changes at least partly occur during or before the onset of BD in adolescence.
To elucidate whether structural brain alterations in BD progress differently over the lifespan, we have summarized study findings with regards to different age groups in this review.This may help to separate structural brain alterations that are already present at the onset of BD from those that develop over the later course of the illness.Disentangling the effects of BD onset from regular brain development in adolescent BD will also help inform the debate about earlier detection and treatment of BD. 21

| Effect of illness progression of bipolar disorder on brain structure
Usually, patients in the first stage of BD experience progression of the disease as symptom-free intervals become shorter, and the frequency of mood episodes increases. 22The effects of this disease progression on the brain have been little studied.Initial crosssectional studies report that longer illness duration is associated with a cortical decline in frontal, medial parietal and occipital regions 6 and that the number of manic episodes experienced is related to lower GMV in frontal brain regions. 23,24In contrast, the effects of depressive episodes on brain structure in BD have not been studied.Against this background, the dynamics between the frequency and intensity of BD mood episodes, on the one hand, and structural changes in brain morphology, on the other, still need to be studied in detail in individuals with BD.

| Effects of treatment on structural brain alterations in bipolar disorder
Clinical neuroimaging studies often include patients who undergo different forms of treatment such as medication, electroconvulsive therapy or psychotherapy.Treatment factors present a notable confound to longitudinal neuroimaging studies in BD patients due to their potential effect on structural brain alterations.For instance, previous findings indicate lithium to be associated with increases in CT, while antiepileptics seem to be associated with decreases in CT in BD patients. 6Increases of total GMV over time have also been associated with response to lithium treatment in BD. 25,26 In contrast to BD, treatment-related brain changes in depressive disorders have been described in more detail recently. 27For instance, there is strong evidence that electroconvulsive therapy is related to GMV increases, while the evidence regarding the effectiveness of cognitive behavioural therapy on structural brain changes is insufficient to draw conclusions. 27Taking these previous indications of differential treatment effects into account, the need for their consideration in BD neuroimaging studies becomes evident.So far, findings from cross-sectional imaging studies demonstrate that it is highly important to consider (1) study sample characteristics (age, age of onset), (2) the clinical course (number of mood episodes, relapses) and (3) treatment factors (medication, electroconvulsive therapy, psychotherapy) when reviewing longitudinal reports of BD patients.

| Longitudinal structural brain alterations in bipolar disorder
Compared to cross-sectional studies, there are only a few reports on structural brain alterations in BD over time.Such longitudinal studies, specifically those with large sample sizes and well-characterised data acquisition methods are rare.However, only imaging studies with follow-up measurements over a longer time span are indispensable to explore the direction of causality between onset and course of BD on the one hand and brain morphological changes on the other.In an earlier review 28 the pattern of structural brain alterations of BD patients with and without psychotic symptoms in comparison to healthy individuals and individuals with schizophrenia (SZC) was explored.The comparison revealed the progressive loss of GMV in the PFC, the ACC and the subgenual regions in association with BD, whereas temporal and subcortical alterations were reported in some instances only. 28A second more recent review on the grey matter changes after a first manic episode in adult BD indicates GMV decreases in the ACC. 29e to the highly inconsistent evidence included in this review, the authors could not draw any final conclusions on associations between mania and progressive grey matter changes.More specifically, some studies included in these reviews 28,29 did not explicitly classify their patient groups into separate diagnoses and compared patients labelled as 'first-episode psychosis' including SCZ, BD and other psychotic patients, with HC in their analysis. 23This lack of classification leads to the problem that BD patients cannot be considered independently.Thus, the structural changes in the brain that have been identified cannot be specifically linked to the affective or psychotic episodes or symptoms.
To tackle these research gaps, our review focuses on dissecting grey matter changes associated with affective episodes in the course of BD from those associated with ageing and maturation.Hence, we included only studies that explicitly examined adolescent and adult individuals with BD without including affective psychosis or first-episode psychosis patients.
A recent mega-analysis that was published after our initial literature search investigated longitudinal structural brain changes in BD. 7 In this study CT, surface area and subcortical volume of 307 BD adults and 925 HC were compared. 7The authors investigated annualised change rates in each imaging phenotype in BD patients compared to HC as well as the effects of mood episodes during the study course.
The mega-analysis included 14 samples with an inter-scan interval of 0.5 years or longer. 14In contrast, we aim to disentangle the effects of disease onset and illness progression on brain structure by including different age groups affected by BD.Unfortunately, we cannot report the actual overlap between the reviewed studies incorporated in our article and the ENIGMA mega-analysis by Abé and colleagues (2022)   because the study does not mention if the included data sets have been analysed and published elsewhere.However, it can be assumed that the overlap is small due to our inclusion criteria and analysis focus.Notably, we included studies also investigating adolescents as well as older BD individuals and included studies on GMV changes using voxel-based morphometry in addition to the volume-based and surface-based analysis provided by Abé et al. 14

| Aims of this review
In this review, we aim to provide a foundational overview of progressive structural brain changes in BD.More specifically, the impact of (1) the occurrence of mood episodes, (2) age-related differences and (3) treatment factors on progressive structural brain alterations will be investigated.By incorporating disease variables in longitudinal studies, we can provide a better understanding of the dynamics of structural changes and their aetiology in relation to the course of BD.Further, a critical discussion of methodological challenges will point out the relevance of future neuroimaging longitudinal research to understand the effects of illness progression on brain structure.

| MATERIAL S AND ME THODS
We applied the PRISMA guidelines for our systematic review. 30ticles had to fulfil the following PICOS (Participants, Intervention, Comparator, Outcome, Study Design) inclusion criteria to be included: We developed a coding system ('Quality score', QS) on study characteristics based on the one used by Böhnlein et al. 31 Based on this coding system, we aimed to evaluate the quality and risk of bias of included studies (see Table S1).
Despite our careful risk of bias assessments, risk of bias cannot be completely ruled out.A publication bias should be considered probable, yet longitudinal studies tend to be published also with negative findings as they are highly costly and time-consuming.By following the PRISMA protocol and assigning quality scores, we hoped to provide the best possible coverage for the risk of bias.

| RE SULTS
A total of n = 11 studies were identified fulfilling all criteria.For the summarised process of literature selection, see the PRISMA flow chart in Figure 1.Two studies by Abé et al 32,33 were included in the review given their different focus sets in the analysis, although some patients (overlap of 34%) were examined in both studies.One study by Abé et al 32 applied a region of interest (ROI) approach of the DLPFC and the inferior cortex in BD type I patients (n = 31), while the other examined the influence of genetic risk and lithium use of both BD type I and BD type II patients on CT with a bigger sample (n = 90). 33Due to the small overlap, both studies were included in our review.One ROI analysis of the amygdala 29 was not included in favour of another study that used the exact same sample to conduct a whole-brain analysis. 30

| Study sample characteristics
Sample characteristics of each study are presented in Table 1.In total, n = 329 BD patients (BD type I = 236, BD type II = 80, 13 = not specified) were included in the literature review.There were only two studies with sample sizes of n > 50 patients. 33,346][37][38][39][40] Different age groups were examined in the studies: Four studies assessed adolescents as their sample or as a subset, 34,35,38,41 and one study was performed on late-life BD patients. 36The exclusion criteria of the studies differed widely regarding comorbidities.While there was the strict exclusion of patients currently diagnosed with any further DSM-IV Axis-I disorder in some studies, 34,36,37 two studies excluded only a history of or current substance abuse. 35,39[42] Another study by Bitter et al 35 excluded patients who had a history of hospitalisation due to a mood episode which possibly led to the exclusion of more severe cases.Most studies have included patients regardless of their current mood states resulting in samples with mixed mood states at the time the MRI scans were performed.
Various mood states at both baseline and follow-up were reported in five studies. 34,35,37,38In four studies, patients were required to be euthymic at baseline assessment. 32,33,36,39One study included various mood states at baseline but did not report the follow-up mood state. 40One study sample comprised BD patients in the recovery from their first manic episode. 42

| Instruments of assessment
See supplements including Table S2 for details.

| Survey of clinical course and treatment
See supplements including Table S3 for details.

| MRI acquisition
See Supplements including Table S4 for details.b BD patient sample.

| Main results
Inconsistencies of reported sample characteristics, methodological differences and different dependent variables yielded heterogeneous study findings.The main results of each study are summarized in Table 2. Structural brain alterations over time according to brain areas are further illustrated in Figure 2. The studies reviewed to provide a differentiated view on the occurrence of mood episodes during the follow-up period, on the effects of ageing in patients with BD and on possible treatment effects.

| Relationship between structural brain alterations and mood episodes
Decrease overtime in frontal brain GMV and CT was the most prominent result, reported in five studies. 32,33,38,40,42Four of them linked greater progressive volume or thickness decline to the experience of mood episodes in the follow-up interval. 32,33,40,42In one study, the loss in prefrontal regions was greater in young adult patients who experienced recurrence versus patients who remained well after their first manic episode. 42Adult BD patients who experienced at least one manic episode in the follow-up interval showed greater cortical thinning in the left inferior frontal cortex than did HC or BD patients without another manic episode.The pattern was similar in BD type II patients experiencing hypomanic episodes. 33In another study, the occurrence of manic episodes was strongly correlated with a greater decrease in the DLPFC and the inferior frontal cortex even for patients that experienced one single manic episode. 32In a different study, depressive episodes were found to be associated with greater thinning in prefrontal cortices in BD. 40 Patients with many depressive episodes (≥2) had greater thinning in the ventromedial prefrontal cortices over time.These effects did not survive correction for multiple comparisons, yet this finding can be seen as preliminary evidence linking depressive episodes to greater prefrontal cortical thinning. 39 TA B L E 2 Main findings and follow-up interval.

References Follow-up Main findings
Abé et al 33 6 years BD I showed greater cortical thinning in bilateral middle temporal cortices over time.Cortical thinning in frontal cortices was greater in BD who experienced at least one manic episode.BD subjects showed an unexpected increase in CT in the visual-somatosensory network, which was partly influenced by Lithium use and genetic factors.Abnormal reduction in frontal brain areas between BD and HC remained stable over time.No differences between BD subtypes.
Abé et al 32 6 years Reduction of fronto-cortical volume (DLPFC and inferior cortex) over time in BD patients who experienced a manic episode.No changes in patients who remained well.
Adleman et al 41 2 years Abnormal development in children with BD vs. HC and children with SMD.Increase in GMV of BD children over time in the right superior inferior parietal lobe and the inferior parietal lobe.
Bitter et al 35 1 year Increase in left amygdala volume in HC and ADHD adolescents but not in BD adolescents over time.
Baseline left and right amygdala volume was greater in BD patients who achieved symptom recovery vs. patients who relapsed.
Delaloye et al 36 2 years No differences at baseline between euthymic older BD and HC.Decreases in the hippocampus, entorhinal cortex and amygdala both in BD and HC.
Ferro et al 37 2.4 years Parietal lobe grey matter reduction in BD persists over time.Both BD and HC showed a decline from baseline to follow up.
Kalmar et al 38 2.3 years Volume decline was greater in BD compared to HC in the bilateral PFC (superior medial frontal gyri extended to rostral ACC and medial frontal gyri) and cerebellar regions.Amygdala reduction in BD was persistent over time.
Kozicky et al 42 1 year BD patients who experienced recurrence showed greater GMV loss in several prefrontal and temporal regions.Patients who remained well showed changes over time.
Lisy et al 34 3-34 months Baseline GMV in BD vs. HC reduced in temporal structures (temporal gyrus and right amygdala).More increase in PFC, medial temporal structures, cingulate gyrus and subcortical structures in BD compared to HC over the course of ≥12 months.GMV changes differed between adult and adolescent patients (e.g.decrease of right superior frontal gyrus volume in adult BD).Occurrence of depressive episodes was associated with volume decline in the left cerebellar vermis.
Moorhead et al 39 4 years Greater tissue loss in left hippocampus, ACC, fusiform gyrus and cerebellum in BD vs. HC.There was a correlation between both depressive and (hypo-)manic episodes occurring during the follow-up interval and grey matter loss in the cerebellar and temporal lobe.Regarding the temporal lobe, brain structural decline linked to mood episodes was found in three studies 39,40,42 : One study observed a correlation between the number of (hypo-)manic episodes during the follow-up and GMV loss in the temporal lobe and a similar correlation for depressive episodes. 36The experience of multiple depressive episodes (>2) led to greater cortical thinning in temporal cortices (left posterior and inferior temporal cortices). 40Additionally, one study found that the recurrence of manic episodes was associated with greater temporal volume loss in BD patients. 42The same pattern was observed in the left inferior parietal lobe and the ACC, where greater decreases were found in patients who experienced recurrence compared to patients who achieved recovery. 42Changes in the amygdala have not been linked to the occurrence of mood episodes directly, yet one study found greater amygdala volume at baseline in adolescent BD patients who achieved symptom recovery compared to those who experienced recurrence. 35Decreased GMV in the left cerebellar vermis and in the thalamus was shown to correlate with the number of depressive episodes. 34 summary, the occurrence of mood episodes is associated with decreases in GMV and CT in various brain regions, such as prefrontal and temporal brain regions.These findings are evident for both depressive 34,39,40 and manic episodes. 32,33,39,42

| Structural brain alterations in BD across the lifespan
40,42 In frontal regions, mostly decreases have been reported, as described previously.Two studies found reduced frontal CT in BD patients compared to HC, both at baseline and follow-up, but no differences in decline over time. 33,40e study observed reduced GMV in the right superior frontal gyrus exclusively in adult BD but not in adolescent BD patients.
In the same study, baseline GMV in temporal structures (including the temporal gyri and right amygdala) was initially reduced in adult BD compared to HC.Over the illness course, patients' volume of these limbic subcortical and medial temporal structures increased. 34Others reported greater tissue loss in BD adults compared to HC in the left hippocampus, the ACC, the fusiform gyrus and the cerebellum. 39One study detected persistent GMV reduction in the parietal lobe of BD adults but found the decline over time not to differ from HC. 37 In a sample of adolescents and young adults, structural brain alterations were progressive with BD patients having a greater volume decrease of ventral and rostral PFC at follow-up compared to the baseline measurement. 38Others found an increase over time in the right ventrolateral PFC in adolescents but not in adults with BD. 34 Persistent amygdala GMV reduction has been revealed in a sample of BD adolescents over time but has not been seen in HC. 38 Another study did not observe abnormalities in the amygdala at baseline yet showed an impaired development of the amygdala in adolescents after experiencing their first manic episode.While the HC showed increased amygdala volume over the study course, this increase was not observed in BD adolescents. 35 the parietal lobe, adolescents with BD showed an increase in GMV in the right superior/inferior parietal lobe, whereas healthy children were having a decrease in the same region over time. 41 a study investigating a sample of elderly patients with BD, the patient group showed GMV reduction in the hippocampus, the amygdala and the entorhinal cortex that was similar to GMV decline in HC. 36 Taken together, structural alterations observed in longitudinal studies differed between age groups with older BD patients showing a volume decrease comparable with normal brain ageing. 36Results in BD adults were mostly heterogenous showing both greater decline in CT, 32,33,40 GMV loss 34,42 and increases in GMV 34 depending on the ROI.Normal brain development in the first years of life is marked by GMV and CT increases in childhood followed by decreases in CT and GMV during adolescence. 16Adolescents with BD showed abnormal GMV increase over time compared to HC. 41 Others found adolescent BD patients lacking an increase in volume compared to HC. 32 BD adolescents also had persistent reductions or greater decline in GMV, 38,42 and GMV increases. 34

| Treatment-related structural brain alterations
One study found an increase in CT in the visual-somatosensory network in the medial occipital cortex and central gyrus in BD patients while HC showed age-related cortical thinning.Further analysis revealed lithium as a contributor to this effect.Increases were greater in BD individuals who started to use lithium during the study interval versus patients already receiving lithium at baseline. 33Elsewhere, atypical antipsychotics and antiepileptics had an impact on GMV.BD patients receiving atypical antipsychotics or antiepileptics showed increased GMV in the left medial frontal gyrus as well as increased GMV in the right cerebellum over time. 34All further studies assessed medication but did not report significant effects on brain changes.

| DISCUSS ION
The aim of this work was to systematically review longitudinal studies of structural brain alterations in BD patients.Progressive brain changes in different age groups and the relationship with the occurrence of mood episodes as well as treatment effects are reported.
Due to considerable differences in study sample characteristics, instruments of assessment, statistical analyses of the clinical course and imaging data acquisition, we also observed large heterogeneity among the findings.

| Mood episodes during study interval and their impact on structural brain alterations
All seven studies assessing the relationship between mood episodes and brain morphological changes reported significant effects of both manic and depressive episodes on CT and GMV, most notably representing a decline in frontal brain regions.This finding is in line with previous results of cross-sectional studies showing an inverse correlation between the patients' history of mood episodes and volume in frontal brain areas 23,24,43 hence indicating similar adverse effects of manic and depressive episodes.Surprisingly, study findings indicate that significant grey matter changes already occur after a single episode during the study interval. 32,34,42Nevertheless, changes were greater in BD groups that experienced a larger number of mood episodes.Most of the research has been limited to the occurrence of mood episodes, and there is limited evidence on the effect of duration of illness.Only one study reported the association of illness duration and depressive episodes with increased GMV in subcortical and limbic structures, while three studies reported no effects of duration of illness on brain structure. 32,36,40,42 Currently, the reviewed study findings of longitudinal neuroimaging in BD indicate that particularly the frequency of occurring mood episodes is associated with structural brain alterations.It remains unclear how the type of occurring mood episode -manic or depressive -induces distinct structural alterations.Some studies found exclusively (hypo-)manic episodes to be associated with cortical thinning, even if patients had more depressive episodes. 33nversely, others found only the effects of depressive episodes on structural brain alterations. 34,40In line with these findings, a longitudinal study of unipolar depression (UD) individuals also revealed associations of larger GMV decline with a higher number of depressive episodes during the study course. 44Yet other studies investigating the effects of depressive episodes on structural brain changes in UD found no evidence. 45Most longitudinal studies of UD patients seem to examine the effects of any relapse during the follow-up interval.
The reason for this could be that the number of possible episodes in short study intervals is limited. 46,47Hence, longitudinal MRI studies with longer follow-up intervals could yield more information on the effect of the number of manic and/or depressive episodes on brain structure.Moreover, evidence on the associations of duration of illness in addition to the occurrence of episodes on brain structural changes in affective disorders could provide further information: We know from cross-sectional studies that illness duration is associated with greater fronto-insular GMV decline as well as an increase in GMV in the subgenual ACC, amygdala and basal ganglia in chronic patients with BD. 11 In the reviewed studies on BD patients, only one study reported significant effects of illness duration as well as depressive episodes in GMV increase in limbic and subcortical structures. 34Three studies have not found an association between illness duration 32,36,37 and structural brain alterations while most of the other studies did not include illness length in their analysis of brain alterations.Thus, conclusive evidence regarding the effects of illness duration on brain structure in BD patients is lacking.
In the reviewed studies, methodological inconsistencies regarding the data collection on disease trajectories hampered the interpretation of study results.Especially, the assessment of mood episodes and models of statistical control varied.Some divided the patient sample groups by single and multiple episodes, respectively 32,33,40,42 ; other studies investigated the total number. 34Not all studies clarified whether an episode of mania or depression occurred as some only reported relapses. 35,42Therefore, specific structural brain alterations can yet not be linked conclusively to manic or depressive episodes.However, a recently published study of the ENIGMA consortium found greater cortical thinning associated with more frequent manic episodes, 14 further corroborating the findings of our review.In the studies that were included in our review, there is a lack of detailed information on the severity and duration of symptoms and more in-depth research is needed to illuminate the effects of illness course on brain morphology.
Besides the occurrence of mood episodes, we aimed to explore how the current mood state (manic, depressive or euthymic) and its change over the study interval affects structural brain alterations.
By doing so, the extent to which there are specific effects of mood changes on structural brain alterations or if such structural alterations are trajectories of more general effects of chronicity could be approached.A review of longitudinal imaging studies on individuals with UD found that changes in depressive state (acute or remitted) had an influence on structural brain alterations.More severe brain atrophy was observed in patients who experienced relapse, whereas smaller structural brain alterations or normalisation of GMV were found in remitted patients. 48In the reviewed studies of BD individuals, most studies assessed the patient's mood state at the time of their initial and final scan.Yet, none of them investigated how this affected the structural brain alterations measured.Due to the different focuses of longitudinal imaging studies of individuals with UD and individuals with BD, the comparison of study findings is limited.
At this point, it is not possible to determine the extent to which each affective disorder causes specific structural brain alterations over time or if there are general effects of chronicity that occur across affective disorders.
The study by Lisy et al 34 shows the importance of regarding the mood state on scan days.In this study, GMV increases were reported in BD patients in frontal, temporal and subcortical structures over time, whereas most other studies found GMV or CT decreases in respective areas.The current mood state of BD patients in the study varied both at baseline and follow-up scan.Yet the largest group of BD individuals changed mood state from manic/mixed to euthymic over the study course.We assume this symptom recovery in a substantial group of patients in the study by Lisy et al to be reflected in GMV increase, as GMV increase is widely associated with symptom improvement or recovery in other mental disorders. 27,47Such effects of symptom recovery on structural brain normalisation rather support the assumption that there are effects that are independent of the specific disorder.Future longitudinal studies should assess and control for mood state at the time of the scan and assess the effects of chronicity on brain structure across (affective) disorders.
We can only speculate about the functional role of grey matter loss as no study has found clear evidence of such connections yet.Grey matter decline in prefrontal and temporal regions might represent a result from reduced inhibitory capacity toward negative emotional states and thereby elevate emotional reactivity resulting in emotional regulation difficulties. 20Individuals with BD show deficient activation in the ventrolateral prefrontal cortex during an affective face task even in euthymic episodes. 49,50Moreover, hypoactivity in the prefrontal cortex is a main finding from functional neuroimaging studies in BD patients. 51Whether the clinical course is also associated with functional brain changes during emotion regulation should be investigated in future studies.Such an effect of illness progression on functional brain activation has already been shown in individuals with UD.Here, more frequent and longer hospitalization were linked to lower brain activity in regions associated with emotion regulation, such as the insula and hippocampus.Since these confounding factors are less present in young individuals, the relationship between brain morphological changes and BD can be inferred more reliably. 53Studies in young patients with BD (e.g.adolescent BD) are rare, as most patients with BD are misdiagnosed at illness onset and remain undiagnosed and untreated for approximately 7.5 years. 54Specifically, early onset BD is tough to diagnose due to mostly mixed symptom profiles characterised by irritability. 55,56Hence, a wait-and-see approach has been suggested for the diagnostics and treatment of adolescent BD. 57 In contrast, other experts have indicated the high risks of a wait-and-see approach, such as an increased risk of mortality due to delayed treatment and the negative effects of untreated illness progression. 21ngitudinal structural neuroimaging approaches could disentangle illness progression from treatment effects.Thus, results may inform the debate by providing details on the effects of delayed treatment on brain structure in early onset BD.
Our review indicates that healthy adolescent brain development is characterised by growth in subcortical temporal lobe structures, notably the amygdala and the hippocampus.Adolescents who recently experienced their first manic episode did not differ from HC in amygdala volume at baseline.However, they also did not show an increase in amygdala volume over time, which is observed during healthy development in adolescence. 35This indicates that decreased amygdala volume is not present prior to illness onset, but impaired development leads to a relative volume reduction in the course of disease.In line with this finding, amygdala volume reduction was observed early in the course of BD adolescents and persisted over time. 38A further study of adolescents with BD suggested that changes in amygdala volume are not present in patients with established illness. 39In summary, these studies suggest that amygdala atrophy develops early after illness onset but does not represent a premorbid condition.More support for this idea comes from a study showing that amygdala and hippocampal volume alterations were absent in high-risk individuals with a family history of BD that were either unaffected or in an early stage of disease. 58e findings of structural brain alterations during the first years of the disorder also include cortical regions.Here, adult BD patients showed a reduction in frontal volume at both baseline and follow-up scans but no progressive volume loss compared to HC in frontal CT 33,40 and parietal lobe GMV. 37In contrast, younger adults who recently experienced their first manic episode showed significant grey matter loss over time. 42Further high-risk approaches and longitudinal studies of BD patients before and after the onset of disease are needed to clarify the impact of age-related neurodevelopmental factors and illness-related effects reflected in structural alterations during disease.
Age-related brain atrophy is known to accelerate in older adults independent of BD. 56 We did not find late-life BD patients to differ from HC.Both groups showed the same pattern of normal age-related GMV decrease in subcortical regions. 36It remains unclear whether this result suggests that progressive structural brain changes no longer occur in older patients.The absence of structural brain changes compared to HC may also be due to the absence of depressive and manic symptoms.Whether this applies here cannot be stated as in the study of Delaloye et al, 36 BD patients had been euthymic for 2 months before the study at baseline and follow-up assessment.The exclusion criteria of the study were thus very strict, unlike in most other studies that included acutely depressed and manic patients.Consistent inclusion and exclusion criteria are needed for valid interpretations of longitudinal study findings comparing different age groups.Due to the lack of methodological standards in the studies presented, no final conclusions can yet be drawn concerning the differences in structural changes in BD patients of different age groups with the notable exception of the amygdala.
Here, the structure of the amygdala appears to change after disease onset in adolescents and could hence be a correlate that could inform the debate on early onset BD.In adult BD patients with a longer disease progression, this structural correlate is not observable, rendering the amygdala as a correlate of adolescent disease onset.

| Treatment factors
Lithium is widely hypothesised to have neuroprotective effects in BD and is frequently associated with increases in GMV. 25,59In the studies reviewed here, only one found lithium intake to be linked to increased medial occipital CT in BD patients. 33It is striking to find so little lithium effects in studies with a follow-up after more than 1 year compared to short-term studies 25,59 and cross-sectional studies. 6,60It can be speculated that lithium only exerts acute effects on brain structure, which may be reduced when patients take lithium over a longer period.
The remaining studies controlled for medication, yet no further medication effects were detected.One further study reported an association of atypical antipsychotics and antiepileptic drug intake with GMV increase in the medial frontal cortex and a portion of the cerebellum but did not find an effect for lithium.However, the authors also pointed out that the small number of patients receiving lithium which could have contributed to this null result. 34 further assessment of treatment-related factors except for medication was provided by any of the reviewed studies.In future studies, measuring other treatment methods for BD could be insightful.For example, psychotherapy has shown neuroprotective effects in various mental disorders such as schizophrenia 61 and borderline personality disorder. 62Since psychotherapeutic treatment has not been considered an influencing variable in the reviewed studies, its role in mediating the observed brain changes cannot be resolved.In general, it is important to first understand the longitudinal natural history of BD and how it manifests in structural brain changes.This will allow treatment effects to be addressed directly.

| Methodological limitations across the reviewed studies
In our review, inconsistencies in inclusion criteria resulted in heterogeneous patient groups.Some studies had strict inclusion criteria, protocols also deviated between studies.For these reasons, study findings need to be treated with caution and only preliminary conclusions can be drawn.We can also not rule out that study findings were biased by the selection of specific ROI, even though most studies included in this review performed whole-brain analysis.

| Future perspectives
The presented evidence cannot conclusively answer the question of how the clinical course of BD is reflected in structural brain alterations across the lifespan.There is strong evidence that depressive and manic episodes lead to decreases in frontal and temporal regions.Future studies should not only investigate how the occurrence of mood episodes is related to brain changes but also how the severity of mood episodes, the total duration of episodes and the current mood state affect brain structure.Accelerated longitudinal designs could be used for this, especially to delineate effects in different age cohorts. 63e to the progressive nature of BD, early, correct diagnosis, and Another way to reduce recall bias is an ecological momentary assessment that uses technologies like digital diaries or physiological sensors that allow symptoms to be recorded in real time. 66Both the NIMH-LCM-p and ecological momentary assessment techniques can support the accurate assessment of disease progression in BD.
The progressive nature of BD can be represented in clinical staging models that gained more scientific and clinical attention recently. 67

| CON CLUS ION
In this review, we summarised evidence from longitudinal imaging studies indicating that affective episodes in BD are associated with localised reduction in cortical thickness or grey matter.
In particular, disease onset in adolescence was associated with amygdala volume reduction, which was not persistent throughout adulthood, suggesting the amygdala as a correlate of adolescent disease onset.In the entire sample, grey matter reductions have been observed after single episodes, indicating that minimising pathological mood swings in BD patients may prevent further grey matter decline.The appropriate treatment potentially may reduce adverse effects of mood episodes or even revert structural brain alterations, specifically after disease onset in adolescent BD.The current clinical practice should take the high rates of mood episodes in BD patients into account and provide accurate survey tools to detect and prevent affective episodes appropriate to the age group at hand.

ACK N OWLED G EM ENTS
Open Access funding enabled and organized by Projekt DEAL.
(a) Participants had to be humans affected with BD.The patient sample had a BD diagnosis based on the Diagnostic and Statistical Manual of Mental Disorders (DSM) criteria according to objective and standardized measures or interviews.As there are no indications of BD subtypes to show distinct structural brain alterations, 6 both BD type I and BD type II studies were included.Aiming to provide a comprehensive overview of progressive brain changes in BD, we excluded all studies that had BD patients in one group mixed with other affective disorders with psychotic symptoms.(b) Regarding the intervention, we only included observational studies that examined BD patients after BD onset.Almost all patients with BD are undergoing some form of treatment, usually medication.No studies on the effects of specific treatments in BD were included.(c) Concerning the comparator, we included studies that measured within-subject structural brain alterations.As the comparison group, studies included HC or divided the BD patient group into those who did and those who did not have a mood episode between baseline and follow-up scan.(d) The outcome had to be a direct measure of grey matter morphometric changes (GMV, CT and/or surface area).Reported structural brain alterations during the follow-up period in BD patients compared to HC were reviewed by brain regions.(e) Regarding the study design, included studies were longitudinal with at least two time points of MRI measurement.The interval between the two time points of MRI scan had to be ≥1 year.No case studies or reviews were included.The independent variables were required to be a potential influencing factor such as age, treatment or mood episodes during the study course.The dependent variable was brain morphological change over time.Research was limited to articles published in English before 9 December 2020.The total number of results yielded n = 2094 studies.Reference lists and citing articles of identified studies were examined in order to check for further relevant studies not found by the initial search.RH screened the studies identified by the search terms for their suitability based on title and abstract.Eligibility of the studies was then assessed according to the PICOS criteria in a full-text review by RH and KF.Inter-rater reliability was high with Pearson's r = 0.936 (p < 0.001).

F I G U R E 1
Flow chart of study inclusion based on PRISMA criteria.excluded on basis of title and abstract (n = 2060) Full-text articles assessed for eligibility (n = 48) Excluded after full-text review for following reasons: -no follow-up MRI scan -follow-up interval <12 months no BD diagnosis at baseline (n = 37) Studies included in qualitative synthesis (n = 11) TA B L E 1 Study sample characteristics including patient demographics and mood states at scan days and Quality Score indicate quality differences concerning study sample and confounder control between the included studies.

4. 2 |
Differences in progressive structural changes across age groupsDue to the methodological heterogeneity across most of the studies, it is difficult to draw final conclusions about the course of BD in different age groups across the lifespan.Some insight may be provided by the results in adolescent BD patients and late-life BD patients regarding structural alterations in subcortical brain regions.In general, it should be considered that adolescents typically have a shorter disease history and show less exposure to psychopharmacological treatment compared to older BD patients.
excluding patients with prior hospitalisation, substance abuse and acute symptom load, whereas others included patients with current depression or mania and various comorbidities.Variations in inclusion criteria are likely to result in different grades of illness severity across studies.In turn, the interpretation of results is hampered as the absence of structural brain alterations in some cases might be due to a milder form of BD.BD patients further differed in their mood state at baseline and follow-up scans.Effects of mood state may have led to the previously described inconsistencies.The assessment of clinical variables such as the occurrence of mood episodes poses a general problem to clinical research.It is typically based on retrospective self-reports and therefore prone to recall bias and imprecision.Beyond that, it must be noted that different methods of MRI data acquisition were employed.Field strengths of MRI scanners operated at 1.5 Tesla in six studies, while the remaining five studies used scanners operating at 3.0 and 4.0 Tesla.Differences in field strengths reduce the comparability of study results and should be taken into critical account.There was one study using two scanners of different field strengths.Though the respective study controlled for potential scanner effects through adjusted scanner-type analyses, the authors acknowledged the possibility of discrepancies in ROI volumes. 35Morphometric analyses were commonly performed by volume-based grey matter measurements like VBM, surfacebased or CT measures by Freesurfer resulting in different dependent variables (thickness, volume, surface).Underlying segmentation appropriate treatment are vitally needed.To help patients recall the longitudinal course of affective episodes, Post et al 64 propose methods for illustrating the course of disease in a graph form, including the severity of episodes, treatment response and the relationship with external factors.The prospective National Institute of MentalHealth Life Chart Method (NIMH-LCM-p) was developed as a corresponding instrument whose validity and usability was confirmed in clinical practice.65 Such staging models classify patients into clusters based on severity scores in domains such as illness course, physical health and cognition using life chart data.Staging models have shown construct and longitudinal validity and may contribute to a more comprehensive understanding and personalised treatment.[67][68][69]These comprehensive diagnostic tools, incorporating the progressive course of the disease also yield crucial information to allow for an early diagnosis and treatment of young patients with BD.Medication as a confounder of these investigations needs to be controlled in future studies.Information on dose and duration of treatment should be included.Further assessment and statistical control of psychotherapeutic treatment, that is missing completely in the reviewed data, should be factored in as well.In addition to prospective studies of the naturalistic disease course, research on how therapy success may reverse structural abnormalities is needed.Besides conventional psychopharmacological medication, studies should capture the effects of psychotherapy, electroconvulsive therapy and ketamine therapy70 on structural brain alterations.Research projects by the ENIGMA working group and others allow for a confident perspective on handling major challenges in clinical neuroimaging studies though large-scale collaborative studies.Standardised multimodal measures are being applied to provide a better understanding of BD pathophysiology.7It is desirable that further research of progressive structural brain alterations in BD is delineated in the longitudinal course over the entire life span.

FU
N D I N G I N FO R M ATI O N Philipp Kanske was supported by the German Research Foundation (KA 4412/2-1, KA 4412/4-1, KA 4412/5-1, CRC940/C07).Katharina Förster was supported by a grant from the Else Kröner-Fresenius-Stiftung (2022_EKEA.102) and the Graduate Academy of the TU Dresden with funds from the Federal Ministry of Education and