Cortical thickness signature as a predictor of diagnostic transition from major depressive disorder to bipolar disorder in late adolescence and early adulthood: a prospective nested case–control study

Abstract

Background

Bipolar disorder (BD) and major depressive disorder (MDD) are classified as distinct diagnostic categories. However, early identification of BD from MDD is challenging due to their high overlap in clinical features, particularly in late adolescence and early adulthood. This study aimed to explore biomarkers for the early identification of BD from MDD.

Methods

The study consists of 139 BD, 148 unipolar depression (UD), and 128 healthy controls (HC) participants ranging from late adolescence to early adulthood. In addition, an independent group of 62 patients initially diagnosed with MDD at baseline and transitioned to BD during follow-up was identified as initial depressive episode BD (IDE-BD). Cortical thickness and surface area were measured for all participants, along with the associations with clinical symptoms.

Results

IDE-BD shares similar cortical thickness patterns with BD and UD. Compared to HC, cortical thinning in the left inferior temporal cortex was observed across all depressive episode groups. Cortical thickness alterations in the right caudal anterior cingulate cortex (cACC.R) were observed specifically in the BD and IDE-BD but were absent in the UD. Compared to UD, IDE-BD exhibited significantly increased cortical thickness in the cACC.R. Moreover, the increase in cACC.R thickness in the IDE-BD was associated with hypomania and suicide risk scores.

Conclusions

In this cross-diagnostic study focusing on late adolescence and early adulthood, we found evidence suggesting shared and disease-specific cortical thickness patterns of depressive episodes. Our study offers potential insights into the neuropathological mechanisms of mood disorders and may contribute to the early identification of BD from MDD.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12916-025-04466-4.

Background

Mood disorders are currently diagnosed based on clinical symptoms. According to the Diagnostic and Statistical Manual of Mental Disorders (DSM) and the International Classification of Diseases (ICD), bipolar disorder (BD) and major depressive disorder (MDD) are classified as distinct diagnostic categories. BD typically manifests during late adolescence and early adulthood [1], with depressive episodes often presenting as the initial symptoms [2]. Depressive episodes dominate the overall course of the illness of BD [3]. Due to the significant overlap in clinical manifestations, the early identification of BD from MDD is particularly challenging. Approximately, 60% of patients with BD are initially misdiagnosed with MDD during the early stages [4]. There is often a significant delay, sometimes up to 5 years or more, between the onset of BD and its accurate diagnosis [5, 6]. The conversion from MDD to BD primarily occurs in the first few years after the initial mood episode [7], and the annual conversion rate gradually decreases over time [8]. For example, studies show that the conversion rate in the first year after a diagnosis of MDD is 3.9%; it drops to 3.1% in years 1–2 and further decreases to 1.0% between years 2 and 5 [8]. The delay may be associated with a range of adverse consequences, including the inappropriate use of antidepressants, which can increase the risk of suicide and aggressive behavior, as well as trigger manic episodes [9, 10]. Additionally, recurrent episodes may lead to lasting impairments, often referred to as “scarring,” contributing to poor long-term outcomes [11].

Researchers are striving to uncover the neurobiological alterations underlying psychiatric disorders to facilitate the development of biologically informed diagnostic and therapeutic approaches [12]. Structural brain features derived from magnetic resonance imaging (MRI) provide valuable insights into the associated pathophysiological mechanisms. Current evidence suggests that cortical structural abnormalities are closely linked to the onset of mood disorders [13, 14]. Compared to healthy controls (HC), BD exhibits significant and widespread cortical thinning, while MDD is characterized by spatially restricted and less pronounced thinning [15]. Machine learning models have been employed to differentiate BD from MDD based on gray matter structural features, even in the absence of significant univariate group differences [16]. These findings indicate that gray matter characteristics, particularly cortical thickness, can provide valuable information for the differential diagnosis of BD and MDD [17]. However, most studies have focused on patients with established BD who have experienced manic/hypomanic episodes, offering limited direct evidence for the early identification of BD from MDD. Some research suggests that the pathophysiological processes of BD are progressive, with differences in cortical thickness between BD and MDD potentially becoming more pronounced as the disease advances [18, 19]. This underscores the importance of early identification of BD.

In recent years, several studies have utilized longitudinal follow-up designs to investigate biomarkers for the early identification of BD from MDD [20, 21]. However, these studies have not sufficiently accounted for the heterogeneity of BD, which may limit the clinical applicability of their findings. Late adolescence and early adulthood represent the peak period for the onset of BD. Evidence indicates that the age of onset for BD follows either a bimodal or trimodal distribution. The bimodal distribution suggests that approximately 66% of BD cases begin in late adolescence or early adulthood, while the trimodal distribution shows that up to 80% of BD cases have their onset within this period [1]. The distribution of age at onset may reflect subgroups of BD with a more evenly distributed pattern [22]. Additionally, the conversion rate from MDD to BD varies widely depending on study design and participant characteristics, with rates reported between 5 and 20% [23]. Higher rates are typically associated with younger patients. A study in young people in the UK found that the rate of transition from MDD to BD has been increasing since adolescence [24]. Another age-stratified survey showed that most transitions from MDD to BD occur before the age of 30 [25]. These studies highlight that the transition from MDD to BD during late adolescence and early adulthood is an important area of research.

This study aims to explore biomarkers for the early identification of BD from MDD using structural MRI (sMRI). We hypothesize that there are distinct cortical structural abnormalities between BD and unipolar depression (UD), particularly differences in cortical thickness. To address the limitations of previous studies, we focus on late adolescence and early adulthood. Additionally, we identified a cohort of patients with depressive episodes whose nature was found to be BD through follow-up. We quantified cortical morphological alterations in patients with depressive episodes across different diagnostic categories to identify specific cortical structural markers for diagnosing BD, with a particular focus on its early stages.

Methods

Participants

The sample for this study was derived from the research cohort at Nanjing Brain Hospital, which has been enrolling individuals with mood disorders since 2012. Detailed information regarding the research cohort is available in the Additional file 1: Methods. All patients were assessed by two experienced psychiatrists using the Chinese version of the Mini-International Neuropsychiatric Interview (M.I.N.I) and diagnosed based on the Diagnostic and Statistical Manual of Mental Disorders-IV/V (DSM-IV/V). The study identified individuals diagnosed with MDD or BD who were in a depressive episode, termed patients with depressive episodes (PDE). All enrolled patients had to meet the following inclusion criteria: (1) 17-item Hamilton Depression Scale (HAMD-17) scores > 7, (2) 15–30 years old, (3) right-handed and native Chinese speaker, and (4) no treatment such as electroconvulsive therapy or repetitive transcranial magnetic stimulation or psychotherapy within the last 6 months. The exclusion criteria were as follows: (1) Severe physical diseases, including severe cardiac, hepatic, or renal insufficiency, or autoimmune disorders, (2) organic brain disorders, (3) psychoactive substance or drug abuse, (4) history of other psychiatric disorders, (5) contraindications to MRI scanning, and (6) special populations, including pregnant or lactating women. Since the primary endpoint of the research was the occurrence of manic/hypomanic episodes to assess the diagnostic transition from MDD to BD, any participants with past or present manic/hypomanic episodes were carefully screened and excluded from the MDD group. During the same period, HC were recruited from the local community, schools, and hospitals. The criteria of this study for HC were as follows: (1) aged 15–30 years, (2) right-handed and native Chinese speaker, (3) no mental disorders and severe illness, (4) no family history of mental disorders in first-degree relatives, and (5) no contraindications to MRI scanning. All participants underwent comprehensive data collection (see Additional file 1: Methods [26–29]). In addition, the Medication Load Index was calculated to quantify psychotropic medication exposure (see Additional file 1: Methods [30–33]).

Follow-up diagnostic information was collected through face-to-face interviews, telephone interviews, and reviews of medical records. Based on these results, we identified a group of patients with initial depressive episode BD (IDE-BD). IDE-BD refers to individuals initially diagnosed with MDD at the baseline who were later diagnosed with BD during follow-up due to manic/hypomanic episodes. In other words, IDE-BD participants had no history of manic/hypomanic episodes at the time of enrollment. Meanwhile, MDD patients whose diagnosis had remained unchanged for at least five years since their first depressive episode were included in the UD group. Further details about the follow-up process can be found in the Additional file 1: Methods. The sample comprised 144 BD participants, 64 IDE-BD participants, 155 UD participants, and 133 HC.

Ethics approval and consent to participate

The study adhered to the principles of the Declaration of Helsinki and was approved by the Ethics Committee of the Affiliated Nanjing Brain Hospital at Nanjing Medical University (2011KY027). All participants were thoroughly informed about the research procedures and provided written informed consent prior to participation. For patients with depressive episodes, a professional psychiatrist assessed their capacity to understand the research content and its potential risks, ensuring they were able to make an informed decision to consent. For minors, consent was obtained not only from the minors themselves but also from their parents or legal guardians, who also signed the written informed consent forms. All participants were informed that they could withdraw from the study at any time without any negative consequences.

Sample size

We calculated the sample size required to estimate cortical morphological differences across the four groups. The parameters were set to a medium effect size (0.25), a two-tailed significance level of α = 0.05, and a target statistical power (1–β) of 0.80 based on Cohen’s effect size classification criteria [34]. The F-test was selected, and the total sample size required to meet these conditions was determined to be 180, as calculated using G*Power software.

Image acquisition and processing

MRI scans were conducted within 3 days of enrollment using a 3.0T Siemens Verio scanner (Erlangen, Germany) with an eight-channel radiofrequency coil. Detailed scanning procedures and parameters are available in the Additional file 1: Methods. T1-weighted images underwent visual quality control (QC), and those with poor quality—such as incomplete brain coverage, low signal-to-noise ratio, visible artifacts (e.g., ringing or blurring), or organic abnormalities—were excluded. Cortical thickness was calculated using FreeSurfer [35] version 7.3 for the 68 regions of the Desikan-Killiany (DK) atlas [36] (34 regions in the left hemisphere and 34 in the right hemisphere), as well as for the entire left hemisphere, right hemisphere, and the global mean cortical thickness. Cortical surface area was also assessed for the 68 regions of the DK atlas, along with the total surface area of the left and right hemispheres and the overall cortical surface area. QC was performed through visual inspection of the segmentations and parcellations. Cases with poor segmentations or parcellations were excluded. As a result, five BD participants, two IDE-BD participants, seven UD participants, and five HC participants were excluded from the analysis due to poor image quality and segmentation.

Statistical analysis

Data analysis was conducted using SPSS version 27. In the analysis of demographic and general clinical data, continuous variables are presented as means ± standard deviation or median (interquartile range), depending on the data distribution, and statistical analysis is performed using the F-test or K–W test. Categorical variables are presented as frequencies and analyzed using the chi-square test. Multiple comparisons between groups were adjusted using the Bonferroni correction. Missing data were imputed using multiple imputation by chained equations.

Effect sizes (Cohen’s d) was calculated to assess the differences in cortical thickness and surface area between each PDE group (BD, IDE-BD, and UD) and HC. In Analysis A, 71 regions were analyzed, with age, gender, and education level included as covariates. For Analysis B, global mean cortical thickness or total cortical surface area was included as an extra covariate to evaluate regional specificity. The significance threshold was set at a false discovery rate (FDR) q < 0.05. Analysis A involved 71 comparisons, including 68 DK atlas regions, the left and right hemispheres, and global cortical measures (either global mean cortical thickness or total surface area). Analysis B focused on 68 comparisons within the DK atlas regions. The Benjamini–Hochberg FDR method was applied to correct for multiple comparisons in both analyses. We assessed the similarity of regional cortical thickness differences across PDE groups, using absolute value of effect sizes derived from 68 regions in comparisons with the HC group. The effect sizes for each of the 68 regions were represented as 68-dimensional vectors (e.g., a for disorder a and b for disorder b). The inner product of these vectors (a∙b) was normalized to a range of − 1 to 1 to calculate the cosine similarity (cos θ = $\frac{a\bullet b}{||a||||b||}$) for cortical thickness. A cosine similarity close to 1 indicated similar patterns of cortical thickness, while a value near − 1 suggested divergent patterns. A similar method has been utilized in a previous study [37].

The analysis of cortical thickness differences between the IDE-BD and UD groups was also conducted to investigate disease-specific changes. In the primary imaging analyses, illness duration and number of episodes were not included as covariates. To address potential residual confounding, a sensitivity analysis was performed in which total illness duration and number of episodes were additionally included as covariates in the IDE-BD vs. UD comparisons. Further, partial correlation analysis was performed between cortical thickness and the scores of the 32-item Hypomania Checklist (HCL-32) [28] and Nurses’ Global Assessment of Suicide Risk (NGASR) [29] in IDE-BD group, with age, gender, and education level included as covariates.

Results

Demographic and clinical characteristics

After rigorous diagnostic evaluation and image QC, a total of 477 participants were included in the analysis, consisting of 139 BD patients, 62 IDE-BD patients, 148 UD patients, and 128 HC individuals. There were no significant differences in gender or age between the paired groups. Compared to UD, BD exhibited a higher prevalence of family history of mental disorder and an earlier age of onset, aligning with findings from the previous study [8]. Detailed demographic and clinical characteristics (Table 1) as well as medication usage, including the Medication Load Index (Additional file 1: Table 1), are provided.

Table 1.

The demographic and clinical characteristics of the participants

Variables	BD (n = 139)	IDE-BD (n = 62)	UD (n = 148)	HC (n = 128)	Statistic	P
Gender (M/F)	51/88	22/40	68/80	45/83	4.430	0.219a
Age at recruitment (year)	21.97 ± 3.95	21.65 ± 3.67	22.68 ± 3.85	21.80 ± 2.31	2.037	0.108b
Education (year)	15 (12–16)	15 (12–16)	15 (12–16)	15 (14–15)	1.488	0.685c
Family history of mental disease, Y/N	42/97	13/49	29/119	-	4.821	0.090a, 0.037d
Onset age (year)	17.91 ± 4.12	19.47 ± 3.92	19.10 ± 4.27	-	4.296	0.014b,e,f
Number of episodes	3 (2–3)	1 (1–2.25)	2 (1–3)	-	44.951	< 0.001c,g,h
Current episode duration (month)	3 (1–8)	6 (3–11)	5 (2–12)	-	18.195	< 0.001c,g,h
Total illness duration (month)	36 (12–72)	19 (6–36)	30 (7.25–60)	-	14.797	< 0.001c,g
HAMD-17	20.94 ± 5.83	21.60 ± 6.01	21.84 ± 5.81	-	0.878	0.417b
HAMA	16.04 ± 6.87	16.90 ± 6.82	17.23 ± 6.25	-	1.193	0.305b
HCL-32	18.92 ± 6.42	16.52 ± 7.05	14.64 ± 7.17	-	14.450	< 0.001b,f
NGASR	9.06 ± 3.42	9.89 ± 4.11	8.63 ± 3.03	-	2.285	0.103b

Multiple comparisons among groups were corrected using Bonferroni correction. Pairwise comparisons only display the p-values for variables with significant differences (p < 0.05). 

Abbreviations: BD bipolar disorder, IDE-BD initial depressive episode BD, HC health control, UD unipolar disorder, HAMD-17 17-item Hamilton Depression Rating Scale, HAMA Hamilton Anxiety Scale, HCL-32 32-item Hypomania Checklist, NGASR Nurses’ Global Assessment of Suicide Risk

^aThe p-value was obtained by the Pearson chi-square test across all groups

^bThe p-value was obtained by one-way ANOVA. ^cThe p-value was obtained by Kruskal–Wallis. ^dThe p-value was obtained by Pearson chi-square test in comparison between BD and UD

^eThe p-values were obtained by post hoc t-test in comparison between BD and IDE-BD

^fThe p-values were obtained by post hoc t-test in comparison between BD and UD

^gThe p-values were obtained by the Mann–Whitney U-test in comparison between BD and IDE-BD

^hThe p-values were obtained by Mann–Whitney U-test in comparison between BD and UD

Global analysis of cortical thickness and surface area

Compared to the HC group, the BD group exhibited a significant reduction in mean cortical thickness (global mean, d = − 0.245, p = 0.024; left hemisphere, d = − 0.234, p = 0.030; right hemisphere, d = − 0.248, p = 0.022; FDR correction). No significant changes in mean cortical thickness were observed in the IDE-BD and UD groups compared to the HC group. Additionally, no significant differences in total cortical surface area were found between each PDE group and the HC group. The results of the global cortical thickness and surface area comparison between groups of PDE and the HC are presented in Fig. 1 and Additional file 1: Tables 2, 3, 4, and 5.

Fig. 1.

Effect sizes for global comparisons in mean cortical thickness and total surface area (Analysis A) between groups of PDE (BD, IDE-BD, and UD) and HC. The red bar indicates the right hemisphere, and the blue bar indicates the left hemisphere. FDR q < 0.05 is marked with an asterisk. BD, bipolar disorder; IDE-BD, initial depressive episode bipolar disorder; UD, unipolar disorder; thickness, mean cortical thickness; surface, total cortical surface area

Cosine similarity was used to evaluate the cortical thickness pattern similarity across groups of PDE. The similarity score ranges from − 1 to 1, with values closer to 1 indicating greater similarity. In Analysis A, the similarity of cortical thickness pattern was 0.739 between the IDE-BD and UD groups and 0.777 between the IDE-BD and BD groups. In Analysis B, the similarity of cortical thickness pattern was 0.820 between the IDE-BD and UD groups and 0.860 between the IDE-BD and BD groups.

Regional comparison of cortical thickness and surface area between groups of PDE and HC

In Analysis A, the BD group exhibited significant cortical thickness alterations in 36 regions compared to the HC group, with 34 regions showing cortical thinning and 2 regions showing cortical thickening. The region with the most pronounced effect size was the left inferior temporal cortex (ITC.L; d = − 0.993, p < 0.001; FDR correction). The IDE-BD group demonstrated significant cortical thickness changes in six regions compared to the HC group, including one region with cortical thinning and five regions with cortical thickening. The region with the largest effect size was also ITC.L (d = − 0.432, p = 0.021; FDR correction). The UD group exhibited significant cortical thickness alteration in only one region compared to the HC group, specifically showing cortical thinning in the ITC.L (d = − 0.331, p = 0.030; FDR correction). Additionally, compared to the HC group, significant cortical thickness alterations in the right caudal anterior cingulate cortex (cACC.R) were observed in both the IDE-BD (d = 0.381, p = 0.024; FDR correction) and BD (d = − 0.403, p < 0.001; FDR correction) groups, while no such changes were detected in the UD group (d = − 0.040, p = 0.816; FDR correction). Regional comparisons of cortical thickness between each PDE group and the HC group in Analysis A are presented in Fig. 2 and Additional file 1: Tables 2, 3, and 4. Significant alterations in cortical surface area were observed exclusively in the BD group compared to the HC group (Additional file 1: Table 5), with no such alterations identified in the IDE-BD and UD groups.

Fig. 2.

Effect sizes for cortical thickness comparison between groups of PDE and HC, with age, gender, and education as covariates (Analysis A). The red bar indicates the right hemisphere, and the blue bar indicates the left hemisphere. FDR q < 0.05 is marked with an asterisk. A BD, bipolar disorder. B IDE-BD, initial depressive episode bipolar disorder. C UD, unipolar depression

In Analysis B, either global mean cortical thickness or total surface area was also included as a covariate. Compared to the HC group, the BD group exhibited significant cortical thickness alterations in 31 regions, with 19 regions showing cortical thinning and 12 regions showing cortical thickening. The region with the largest effect size was also the ITC.L (d = − 0.584, p < 0.001; FDR correction). The IDE-BD group showed significant cortical thickness changes in 12 regions compared to the HC group, with 4 showing thinning and 8 showing thickening. The largest effect size was observed in the ITC.L (d = − 0.524, p < 0.001; FDR correction). Cortical thinning was only observed in the ITC.L (d = − 0.324, p = 0.037; FDR correction) in the UD group compared to the HC group. Similarly, significant cortical thickening in the cACC.R was observed in the IDE-BD (d = 0.381, p = 0.024; FDR correction). Decreased cortical thickness in the cACC.R was observed in the BD group, though this difference did not reach significance after correction (d = − 0.204; p = 0.033, before correction; p = 0.064, FDR correction). While no such changes in the cACC.R were detected in the UD group (d = − 0.010, p = 0.978, FDR correction). Regional comparisons of cortical thickness between each PDE group and the HC group in Analysis A are presented in Fig. 3 and Additional file 1: Tables 6, 7, and 8. Significant surface area changes were only observed in the BD group compared to the HC group (Additional file 1: Table 9), with no such changes identified in the IDE-BD or UD groups.

Fig. 3.

Effect sizes for cortical thickness comparison between groups of PDE and HC, with age, gender, education, and global mean cortical thickness as covariates (Analysis B). The red bar indicates the right hemisphere, and the blue bar indicates the left hemisphere. FDR q < 0.05 is marked with an asterisk. A BD, bipolar disorder. B IDE-BD, initial depressive episode bipolar disorder. C UD, unipolar depression

Regional comparison of cortical thickness between IDE-BD and UD

In Analysis A, IDE-BD exhibited significant increased cortical thickness in the cACC.R (d = 0.469, p = 0.022, FDR correction) and right parahippocampal cortex (PHC.R, d = 0.421, p = 0.041, FDR correction) relative to UD (Fig. 4; Additional file 1: Table 10). Similarly, in Analysis B, compared to UD, IDE-BD also exhibited a significant increase in cortical thickness in the cACC.R (left, d = 0.445, p = 0.042, FDR correction) (Fig. 4; Additional file 1: Table 11). To further account for potential residual confounding, a sensitivity analysis was conducted by additionally including total illness duration and number of episodes as covariates. In this model, the group differences in cACC.R and PHC.R were attenuated to marginal significance after FDR correction, while the effect-size directions remained consistent (see Additional file 1: Results and Table 12). Further correlation analysis on the IDE-BD patients showed that the cortical thickness of cACC.R was positively correlated with the NGASR (r₍₆₂₎ = 0.305, p = 0.019) and HCL-32 (r₍₆₂₎ = 0.265, p = 0.042).

Fig. 4.

Effect sizes for cortical thickness comparison between IDE-BD and UD. Analysis A included age, gender, and education as covariates; Analysis B included age, gender, education, and global mean cortical thickness as covariates. The red bar indicates the right hemisphere, and the blue bar indicates the left hemisphere. FDR q < 0.05 is marked with an asterisk. IDE-BD, initial depressive episode bipolar disorder; UD, unipolar disorder

Discussion

This study investigated the brain structural characteristics of BD and UD, as well as biomarkers for the early identification of BD from MDD. Our study addresses several limitations of previous research by focusing on individuals in late adolescence and early adulthood, with all patients being in a depressive episode at the time of recruitment. Furthermore, the prospective design holds significant clinical implications. Through longitudinal follow-up, we identified a group of patients initially diagnosed with MDD who subsequently transitioned to BD, termed IDE-BD. We assessed structural neuroimaging measurements across PDE groups. We found that IDE-BD shares similar cortical thickness patterns with BD and UD. Cortical thinning in the ITC.L was observed across PDE groups, with the most prominent effect size. Cortical thickness alterations in cACC.R were observed specifically in the BD and IDE-BD but absent in the UD. Additionally, the increase in cACC.R thickness in the IDE group was associated with NGASR and HCL-32.

Distinguishing between BD and UD is particularly challenging due to the substantial overlap in their clinical presentations and neurobiological features, especially in the early stage [5]. In this study, we employed cosine similarity to quantify the degree of overlap in cortical thickness patterns between individuals with IDE-BD and UD. Cosine similarity values range from − 1 (completely dissimilar) to + 1 (identical), with 0 indicating no similarity. Although there is currently no universally accepted threshold for cosine similarity in neuroimaging research, values above 0.7 are generally considered indicative of substantial similarity, and values around 0.5 reflect moderate similarity, as supported by previous studies [38]. The high cosine similarity observed between IDE-BD and UD in the present study further supports the notion of a highly overlapping cortical morphological substrate in the early stages of BD and UD, providing a neurobiological explanation for the difficulty in distinguishing these conditions early on. This finding is approximately consistent with the results of a previous case–control study, despite differences in age ranges between the studies [15]. Furthermore, genetic studies have shown a strong correlation between BD and MDD [39, 40]. Cortical thickness is shaped by processes such as myelination, branching, pruning, and other developmental changes occurring after the mid-fetal stage [41]. Therefore, the cortical thickness similarity pattern may reflect a parallel trend to the genetic correlation between BD and MDD.

We found that BD showed extensive cortical thinning relative to HC, involving both global and regional measurements. This finding aligns with previous studies [15, 42]. However, it is important to note that widespread cortical thinning is not exclusive to BD. Similar findings of extensive cortical thinning have been consistently reported in severe psychotic disorders, such as schizophrenia [43, 44]. While total cortical surface area did not differ across groups, we detected focal surface area alterations confined to BD—bilateral frontal pole and left temporal pole—whereas IDE-BD and UD showed no differences. This dissociation fits two complementary accounts. Developmentally, surface area is shaped largely by early arealization processes (e.g., radial unit proliferation, tangential migration), is highly heritable, and is relatively stable across adolescence [45–48]. Thus, BD-linked surface area deviations likely index a trait-like liability that remains absent or subthreshold in UD and IDE-BD at baseline. Progression-wise, surface area differences may require cumulative illness burden—earlier onset, greater episode count, and persistent allostatic load—to become measurable, conditions that characterize BD but not IDE-BD at baseline [49–52]. The frontal and temporal poles, transmodal association cortices with protracted maturation, are plausible loci where developmental biases and neuroprogressive influences converge [53, 54]. Overall, selective frontotemporal surface-area alterations seem BD-specific and progression-related; longitudinal cohorts across the conversion are needed for confirmation.

A strength of this study is the comparison within the same study, rather than comparing effect sizes across multiple previous studies. In this cross-diagnostic study, a common finding was that compared to HC, all groups of PDE exhibited decreased cortical thickness in the ITC.L, with the largest effect size observed. This suggests that ITC.L may serve as a state marker of depressive episodes. The ITC is primarily involved in the processing of visual information, playing a key role in object recognition, face recognition, and the perception of complex visual stimuli [55]. It integrates input from the ventral visual stream for object and face recognition and further projects to the orbitofrontal cortex to process reward value and emotions [56, 57]. Abnormalities in the ITC may contribute to the underlying pathology of mood disorders. Studies have reported cortical thinning in the ITC in patients with depression, alongside a bias in recognizing emotional faces that could exacerbate social and emotional difficulties [58, 59]. Our previous magnetoencephalography study also revealed the association between functional impairment of the ITC.L and the bias in emotional face processing in MDD [60]. Another study showed that patients with remitted psychotic depression exhibited cortical thinning in the inferior frontotemporal cortex [61]. Studies have also reported the decreased cortical thickness in the ITC.L in BD and the association with cognitive function [42, 62, 63]. Cross-diagnostic research is not unprecedented. A large cross-diagnostic study also reported cortical thinning in the ITC.L in both BD and MDD [15], consistent with the findings of this study. However, it is important to note that all participants in the current study were scanned during depressive episodes, and therefore the observed cortical thinning in the ITC.L may reflect a state marker of depression. Consequently, these findings should be interpreted with caution.

Our study revealed that compared to HC, IDE-BD exhibited significant cortical thickening in the cACC.R, while the BD group showed significant cortical thinning in the same region. The cACC is involved in both emotional and cognitive processes, including emotion regulation, conflict monitoring, attention modulation, and error detection [64]. Studies have shown structural and functional abnormalities in the cACC in patients with BD [15, 65]. These abnormalities may result in impairments in the integration of emotion and cognition, leading to excessive emotional reactivity or insufficient regulation, particularly during emotional conflict and decision-making processes [66, 67]. Cortical thickness alterations in BD may be progressive. A longitudinal study showed that cortical thickness in patients with BD undergoes abnormal thinning as the disease progresses [18]. Although the study did not specifically identify changes in the cACC, it seems to reflect a noticeable trend. Different from our current study, the study was conducted in individuals with established BD and had a different age group. Our study seems to indicate that thickness changes in the cACC may be associated with the progression of BD. Mania/hypomania episodes could serve as a critical point for cACC thickness changes, with cACC thickening observed prior to manic/hypomanic episodes and thinning in BD who have suffered manic/hypomanic episodes. However, it is important to emphasize that given the cross-sectional nature of our imaging data, we can only observe disease-related changes at a specific point in time and cannot draw definitive conclusions about the progression of cortical thickness changes over time. Therefore, while we speculate that the thickness changes in the cACC may be associated with the progression of BD, further longitudinal studies are required to confirm these trends and gain a deeper understanding of the temporal dynamics of cortical thickness changes in BD.

Our results suggest that abnormal changes in cACC.R cortical thickness appear to be disease-specific. We did not observe significant changes in cACC.R cortical thickness in the UD group. Furthermore, we found that IDE-BD patients exhibited cortical thickening in the cACC.R compared to UD. This finding reinforces previous reports, which also identified cACC.R cortical thinning exclusively in the BD group, with no such changes observed in the UD group [15]. Moreover, a study on the offspring of BD found that abnormal functional activity in the cACC predicted future emotional instability [68]. Taken together, these studies highlight the potential role of cACC.R cortical thickness as a marker of diagnostic transition from MDD to BD. In our sensitivity analysis, the differences between IDE-BD and UD in cACC.R and PHC.R were attenuated after adjusting for illness duration and number of episodes, suggesting that clinical course variables may partially contribute to the observed structural alterations. Nevertheless, the effect-size directions remained consistent, indicating that these regions may serve as potential neuroimaging markers for the early identification of BD. Overall, these findings suggest that the observed changes may arise from both early BD risk and the accumulation of illness course, emphasizing the necessity of larger, well-balanced, and longitudinal cohorts to distinguish effects intrinsic to the disorder from those driven by illness burden.

An important caveat is the diagnostic precision within the IDE-BD group. Due to incomplete subtype information, we could not reliably distinguish between BD-I and BD-II. Such heterogeneity may inflate within-group variance and obscure subtype-specific cortical morphological effects [69]. Previous studies have indicated that BD-I and BD-II may differ in cortical morphology, including both the magnitude and spatial distribution of thickness alterations [70, 71]. Therefore, the IDE-BD effects reported here are likely to reflect mixed bipolar-spectrum features rather than subtype-specific signatures. Future studies should stratify by subtype at conversion and formally test group × subtype interactions in longitudinal designs that include scans both before and after diagnostic transition.

This study has several limitations. First, it is a single-center study with samples recruited from eastern China, and all MRI data were acquired using the same scanner. Therefore, the results require further external validation. Second, this preliminary study involved only a single MRI scan. As mentioned earlier, we were unable to capture progressive brain structural changes associated with disease progression. Third, many patients in this study were on medication, making it difficult to determine whether the observed effects stem from the disorder, the treatment, or a combination of both. Since patients were in the acute phase and undergoing medication adjustment, the instability of medication status could have introduced confounding effects on cortical thickness changes. Future research should prioritize drug-free participants or, ideally, those who are drug-naïve. Additionally, although we tracked symptom progression in MDD patients, we cannot rule out the possibility that the MDD patients in our study might transition to BD. Long-term follow-up is necessary to improve diagnostic reliability. Lastly, BD encompasses clinically meaningful subtypes (e.g., BD-I, BD-II) that are known to show differences in cortical profiles. In our cohort, subtype adjudication at follow-up was incomplete, preventing us from determining the distribution of BD-I versus BD-II within the IDE-BD group. This limitation reduces diagnostic precision and may have influenced the observed effects. Replication in larger, well-balanced, and subtype-resolved longitudinal cohorts will be essential to clarify whether the cACC-based signature preferentially tracks the transition to BD-I or BD-II.

Conclusions

In conclusion, this cross-diagnostic study focusing on late adolescence and early adulthood suggests that IDE-BD may share similar cortical thickness patterns with BD and UD. Cortical thinning in the ITC.L could potentially represent a shared pathological alteration in mood disorders, while alterations in cACC.R cortical thickness appear to be more specific to BD. Additionally, cACC.R cortical thickening might be associated with the diagnostic transition from MDD to BD. These findings provide some support for the early identification of BD, although further research is needed to confirm these relationships.

Abbreviations

BD

Bipolar disorder

cACC.R

Right caudal anterior cingulate cortex

DK

Desikan-Killiany

DSM

Diagnostic and Statistical Manual of Mental Disorders

ECT

Electroconvulsive therapy

FDR

False discovery rate

HAMD-17

17-Item Hamilton Depression Scale

HC

Healthy controls

HCL-32

32-Item Hypomania Checklist

ICD

International Classification of Diseases

IDE-BD

Initial depressive episode bipolar disorder

ITC.L

Left inferior temporal cortex

M.I.N.I

Mini-International Neuropsychiatric Interview

MDD

Major depressive disorder

MRI

Magnetic resonance imaging

NGASR

Nurses’ Global Assessment of Suicide Risk

PDE

Patients with depressive episodes

PHC.R

Right parahippocampal cortex

QC

Quality control

sMRI

Structural magnetic resonance imaging

UD

Unipolar depression

Authors’ contributions

**HS and NS:** Conceptualization, Methodology, Formal analysis, Visualization, Writing-Original Draft, Writing-Reviewing and Editing. **ZLC, JLS, RY, HWZ, ST, YWW and XQW:** Investigation and Data Curation. **TW:** Visualization. **LX:** Formal analysis. **ZJY and QL:** Supervision, Conceptualization, Writing-Reviewing and Editing, Funding Acquisition. All authors read and approved the final manuscript.

Funding

This study was supported by grants of the National Natural Science Foundation of China (82271568, 82151315, 82301718, 82101573, 82401788); the Jiangsu Medical Innovation Center for Mental Illness (CXZX202226); the Jiangsu Province Science and Technology Plan Project (BE2023667); the Key Project of Science and Technology Innovation for Social Development in Suzhou (2022SS04); Jiangsu Provincial Natural Science Youth Fund (BK20230154, BK20241120); the General Project of Nanjing Science and Technology Development Plan(YKK23136). The funding agency did not participate in the study design or data analysis.

Data availability

The datasets generated during the current study are not publicly available due to the subjects’ privacy, but are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study adhered to the principles of the Declaration of Helsinki and was approved by the Ethics Committee of the Affiliated Nanjing Brain Hospital at Nanjing Medical University (2011KY027). All participants were thoroughly informed about the research procedures and provided written informed consent prior to participation. For patients with depressive episodes, a professional psychiatrist assessed their capacity to understand the research content and its potential risks, ensuring they were able to make an informed decision to consent. For minors, consent was obtained not only from the minors themselves but also from their parents or legal guardians, who also signed the written informed consent forms. All participants were informed that they could withdraw from the study at any time without any negative consequences.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.