Authors: Lianzi Guan, Yifei Li, Hui Kong, Jie Fang, Jiakuai Yu, Ting Wang, Jiajia Zhu, Daomin Zhu
Categories: Research, Depression, Cognitive function, Functional magnetic resonance imaging, Event-related potentials, Brain function
Source: BMC Psychiatry
Authors: Lianzi Guan, Yifei Li, Hui Kong, Jie Fang, Jiakuai Yu, Ting Wang, Jiajia Zhu, Daomin Zhu
Accumulating evidence shows that cognitive deficits are common in patients with major depressive disorder (MDD). However, the specific differences in cognitive impairment and brain functional alterations between first-episode depression (FED) and recurrent major depression (RMD) remain unclear, as do the relationships among these factors.
A total of 43 RMD and 41 FED patients were included in this study. All the patients underwent examinations of resting-state functional magnetic resonance imaging (fMRI), event-related potential (ERP) measurements, and a series of standardised neuropsychological tests, including event-based (EBPM) and time-based (TBPM) prospective memory tasks, the Semantic Fluency Test (SFT), and the Continuous Performance Task–Identical Pairs (CPT-IP). Two-sample t-tests were used to compare cognitive functioning, ERP parameters, and brain functional indices between FED and RMD groups. Correlation analyses were performed to explore the associations between these variables.
Compared with FED patients, those with RMD displayed poorer CPT-IP performance, lower prospective memory (EBPM) scores, lower SFT performance, and prolonged P300 latency (all P < 0.05). Moreover, neuroimaging data analysis revealed increased regional neural activity in the right inferior temporal gyrus (ITG), alongside decreased interhemispheric functional connectivity in the bilateral ITG in RMD relative to FED. Correlation analyses indicated that these functional changes were significantly associated with the observed cognitive deficits.
Our data demonstrated more pronounced cognitive deficits and brain functional impairments in RMD relative to FED as well as their potential links. These findings not only elucidate the neural mechanisms underlying cognitive deficits in MDD, but also inform future treatment and prevention of cognitive dysfunction in patients suffering from MDD.
Not applicable.
The online version contains supplementary material available at 10.1186/s12888-025-06758-8.
Globally, major depressive disorder (MDD) ranks among the most common and severely disabling mental health conditions, impacting an estimated 340 million people [1]. It is the single largest contributor to non-fatal health loss and is projected to become the primary contributor to global disease burden by 2030^2^. One of the core symptoms of MDD is cognitive deficits, including impairments in memory and psychomotor skills and a marked decline in executive function [3], particularly attention deficits [4]. Up to two-thirds of patients with MDD experience cognitive deficits [5], which significantly burdens individuals, families, and society.
Resting-state functional magnetic resonance imaging (fMRI) is recognized as a valuable technique for capturing blood oxygen level-dependent (BOLD) signals, which are believed to indicate spontaneous neural activity of physiological importance [6, 7]. Regional homogeneity (ReHo) is a method for measuring local functional uniformity [8, 9]. Changes in ReHo in the frontal cortex region have been found to be potential biological markers of MDD [10, 11]. Voxel-mirrored homotopic connectivity (VMHC) assesses the resting-state functional connectivity (RSFC) between any voxel and its mirror in the opposite hemisphere, reflecting the degree of synchronous activity between the two hemispheres at rest [12]. Numerous studies have reported changes in ReHo and VMHC in the control network (CN), attention network (AN), default mode network (DMN), and insula in MDD patients [13–15]. These functional changes are closely related to emotional processing, autonomic regulation, and cognitive dysfunction [16, 17]. However, previous findings are inconsistent across studies [18]. To date, there is a paucity of studies that have explored alterations in ReHo and VMHC between first-episode depression (FED) and recurrent major depression (RMD).
As a component of event-related potential (ERP), the P300 is a positive voltage wave that appears approximately 300 milliseconds post-stimulus onset and is associated with attention allocation and working memory updating [19]. Its latency and amplitudes may reflect the speed of cognitive processing. Different personality traits may influence P300 responses to emotional stimuli [20, 21], potentially reflecting abnormal emotional and cognitive integration in individuals with depression. Several studies have indicated that, compared with healthy controls, MDD patients may exhibit reduced P300 amplitudes [22, 23] and prolonged P300 latencies, suggesting deficits in attention allocation and information processing in MDD [24–26]. However, other studies have not reported statistical differences in P300 parameters between MDD patients and healthy controls [27, 28], reflecting the heterogeneity and complexity of research findings in this area.
MDD has a high recurrence rate, with studies suggesting that approximately 60–85% of patients may experience multiple episodes during their lifetime [29, 30]. In recent years, increasing research has focused on factors influencing depression recurrence and has identified marked differences between FED and RMD [31, 32]. Compared with first-episode patients, patients with recurrent depression have smaller hippocampal volumes [33]. Some studies have also shown that cognitive deficits in MDD patients may persist during remission [34, 35], potentially reflecting the “scarring” effects of depression. Patients with RMD perform worse than those with FED in terms of multiple cognitive domains, including information processing speed, working memory, learning ability, immediate and delayed recall, executive function, and verbal fluency [36, 37]. Cognitive function encompasses a range of distinguishable domains mediated by different neural circuits [38], and it is reasonable to speculate that the differences in cognitive performance between individuals with first-episode and recurrent depression are related to specific neural changes. The profound adverse effects of cognitive impairments on individuals with MDD necessitate the identification of neurobiological indicators that capture alterations in cognitive symptoms and traits related to depressive episodes. This is essential for improving diagnostic accuracy and refining treatment approaches. Nevertheless, research specifically addressing cognitive function in FED and RMD remains limited. This study aims to examine the cognitive differences between FED and RMD as well as their underlying mechanisms by integrating cognitive assessments, ERP, and fMRI. We hypothesised that patients with RMD would present more severe cognitive impairments and that these deficits would be related to functional alterations in specific brain regions.
Patients with MDD were recruited from the Sleep Disorders Department of Hefei Fourth People’s Hospital. All participants met the ICD-10 criteria for depressive episodes, with diagnoses confirmed by two psychiatrists. Ultimately, 41 patients with FED and 43 patients with RMD were included. The inclusion criteria
A diagnosis of either first-episode or recurrent major depressive disorder;An age range of 18 to 60 years, following the limits established in our previous study [39]; and.Completion of at least primary school education and sufficient cognitive ability (based on a brief clinical interview) to understand and follow the test instructions, ensuring no severe intellectual disability, agnosia, or aphasia.
Exclusion criteria
Any history of major neurological conditions, particularly epilepsy, brain injury, or loss of consciousness;Having severe psychiatric comorbidities (e.g., schizophrenia, intellectual disability, bipolar disorder, or dementia); and.Pregnancy or any condition that contraindicates MRI.
Most participants had been receiving antidepressant treatment prior to enrollment. To ensure comparability across individuals, these dosages were converted to fluoxetine equivalents [40]. Participants who had not taken any antidepressants for more than two weeks were considered to have completed a medication washout period and were therefore classified as medication-free. The age distributions in each group were as 18–30 FED (22 participants), RMD (10 participants); 30–45 FED (13 participants), RMD (14 participants); 45–57 FED (8 participants), RMD (19 participants). The Hefei Fourth People’s Hospital Ethics Committee granted approval for this study. Prior to participating, each individual provided their written informed consent, ensuring that they fully understood the study’s purpose, procedures, and any potential risks involved.
Informed by prior studies investigating cognitive function in first-episode versus recurrent depression [41], we employed GPower software to estimate the sample size for our independent samples t-test [42]. We assumed an effect size (Cohen’s d) of 0.71, set the statistical power (1–β) at 0.80, and used a significance level (α) of 0.05. Under these parameters, a minimum of 33 participants per group would be required to detect a difference of this magnitude with adequate power.
Each participant’s level of depressive and anxious symptoms was meticulously assessed using standardized assessment tools. Specifically, the Hamilton Depression Rating Scale (HAMD) was employed to evaluate the severity of depressive symptoms, while the Hamilton Anxiety Rating Scale (HAMA) was utilized to measure the intensity of anxiety-related symptoms. These validated scales provided a comprehensive and quantitative analysis of each individual’s emotional and psychological state, ensuring accurate and reliable measurements of their depressive and anxious symptomatology [43–45].
Prospective memory refers to the capacity to carry out tasks that have been planned in advance at a predetermined time or upon encountering certain conditions. Prospective memory is generally divided into time-based prospective memory (TBPM) and event-based prospective memory (EBPM) [46]. Detailed methodologies and protocols for the EBPM and TBPM assessments employed in this study are comprehensively described in earlier published works [47, 48]. In the EBPM task, participants were shown a total of 30 cards, each containing 12 Chinese words. The participants were asked to identify words of the same category across the cards, and if the same category was animals, they were instructed to tap the table (with a total of 6 sets of animal target words). Participants were awarded 2 points for every accurate response and were subsequently asked to recall their phone number at the conclusion of the task, receiving an additional 2 points for correct memorization, for a maximum of 8 points—higher scores indicating better performance. In the TBPM task, individuals were shown 100 cards in succession, each containing 20 numbers. They were instructed to identify the highest and lowest numbers on each card within a 17-minute period. During the task, participants were also instructed to remember to tap the table or inform the examiner when 5, 10, and 15 min had passed. Responses made within 30 s before or after the target time earned 1 point, while those within 15 s earned 2 points, up to a maximum of 6 points, with higher scores reflecting superior task execution.
Sustained attention was assessed using a computerized continuous performance task-identical pairs (CPT-IP) [49], which reflects participants’ sustained attention ability. In this test, participants were required to observe whether two consecutive numbers from a continuous sequence of 2, 3, or 4 randomly displayed digits were the same, and if so, they clicked the left mouse button. The sensitivity index (d’) was calculated for each digit-length condition (CPTIP-2, CPTIP-3, and CPTIP-4), where greater d’ values signified better attentional performance.
The semantic fluency test (SFT) was used to measure semantic memory and vocabulary knowledge [50]. In this test, participants were asked to generate as many words as possible within 60 s from 6 given semantic categories. The participants’ responses were recorded, after which repeated or unrelated words were excluded and the average number of correct words was calculated. A higher count of correct words indicated stronger semantic fluency performance.
The digit span (forward and backward) tests were used to assess short-term memory, working memory, attention, and executive functioning [51, 52]. During this procedure, the examiner recited progressively lengthened sequences of digits, and participants were required to repeat them either in the original order or in reverse. Scores were based on the number of correctly repeated digits (with total scores ranging from 2 to 9), and higher scores denoted better cognitive performance.
Electroencephalography (EEG) was recorded using the Fpz electrode as the reference electrode and the Cz electrode as the recording electrode. The auditory oddball paradigm was employed to evoke P300 activity. The oddball paradigm consisted of 30–36 target stimuli, with 80% standard stimuli (60 dB) and 20% deviant stimuli (80 dB). The participants were required to count the number of deviant stimuli. A professional processed and recorded the P300 parameters.
Resting-state fMRI was carried out on 3.0T GE scanner (Discovery MR750w, General Electric, Milwaukee, WI, USA). To minimize motion artifacts, each participant was instructed to stay as still as possible during the scanning procedure, provided with earplugs to reduce acoustic disturbances, and had the head stabilized using foam pads. BOLD signals were gathered using a gradient-echo single-shot echo planar imaging (GRE-SS-EPI) sequence with the following repetition time (TR) = 2000 ms, echo time (TE) = 30 ms, flip angle (FA) = 90°, field of view (FOV) = 220 × 220 mm, matrix dimensions = 64 × 64, slice thickness = 3 mm, inter-slice gap = 1 mm, a total of 35 axial slices, and 185 volumes, resulting in 370 s of data acquisition. High-resolution T1-weighted structural images were obtained using a BRAVO (brain volume) sequence with the following TR = 8.5 ms, TE = 3.2 ms, inversion time (TI) = 450 ms, FA = 12°, FOV = 256 × 256 mm, matrix size = 256 × 256, slice thickness = 1 mm with no gap, covering 188 sagittal slices over 296 s of scanning.
Data preprocessing and statistical evaluations for the resting-state fMRI were conducted in DPABI (Data Processing & Analysis of Brain Imaging, http://rfmri.org/dpabi)^53^. The initial 10 time points of each acquisition run were discarded to allow participants to acclimate to the scanning environment and reduce any transient noise. The remaining data underwent the following (1) slice timing correction; (2) realignment, in which the data at all time points were spatially aligned with the first time point to correct for head motion; (3) linear detrending to remove signal drifts; (4) regression of nuisance signals, including Friston 24 head motion parameters, global signal, white matter signal, and cerebrospinal fluid signal. (5) spatial normalization to the Montreal Neurological Institute (MNI) space using a 3 × 3 × 3 mm voxel resolution; (6) band-pass filtering (0.01–0.1 Hz) to retain relevant frequency bands; For image quality checks, we performed thorough structural image checks by segmenting T1-weighted images into gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF). Any cases with severe segmentation errors (e.g., misclassification artifacts) were excluded. In addition, we visually inspected the coregistration between each participant’s mean functional image and T1-weighted structural image, and any dataset with significant misalignment was either reprocessed or excluded. For motion assessment, after each scan we calculated framewise displacement (FD) for each volume to quantify volume-to-volume head motion, confirming that FD remained under 2 mm and that the images adequately covered the full cerebral cortex. We also visually inspected realignment plots and the raw functional time series for severe motion artifacts (e.g., ringing, ghosting, or abrupt head shifts). Datasets failing these criteria were either reprocessed or excluded to ensure acceptable fMRI data quality.
For ReHo analysis, Kendall’s coefficient of concordance (KCC) was calculated for each voxel in comparison with its 26 nearest neighbors. Meanwhile, VMHC was computed by determining the Pearson correlation between signals in symmetrical voxel pairs across the two hemispheres. In order to enhance bilateral symmetry, these functional images were coregistered into a symmetric template. Subsequent to this, the ReHo and VMHC maps were each transformed using Fisher’s r-to-z method to approximate normal distributions, then smoothed using a 4 mm full-width at half maximum (FWHM) Gaussian kernel to reduce noise before statistical modeling.
Statistical tests were carried out using SPSS 27.0 (SPSS, Chicago, IL, USA). Two-sample t-tests were applied to assess between-group differences in age, years of education, body mass index (BMI), HAMD scores, HAMA scores, age at illness onset, total duration of illness, overall cognitive scores, and P300 measures. Sex distribution was evaluated using chi-square analysis. A significance threshold of P < 0.05 (two-tailed) was adopted. For the MRI measures, group comparisons of ReHo and VMHC values between FED and RMD were conducted with two-sample t-tests in DPABI, including age, sex, education, and FD as nuisance regressors. Corrections for multiple comparisons were performed with a Gaussian random field (GRF) approach using a voxel threshold of P = 0.001 and a cluster threshold of P < 0.05 (corrected). The z-scores derived from ReHo and VMHC in regions showing significant differences were then extracted for correlation assessments with clinical and cognitive measures, adjusting for sex, age, and education. To derive 95% confidence intervals for significant correlations, we used 5,000 bootstrap resamples. Given that multiple cognitive tests and correlation analyses between brain function indices and cognitive performance were conducted, we performed a Bonferroni correction for multiple comparisons [54].
There were no statistically significant variations between the FED and RMD groups concerning several factors, including years of education (t = 1.369, P = 0.18), sex distribution (χ² = 0.715, P = 0.40), BMI (t = -1.002, P = 0.32), HAMD (t = 0.559, P = 0.58), HAMA (t = 1.699, P = 0.09), age at onset (t = -0.113, P = 0.91), current episode duration (t = 0.892, P = 0.38), medication dosages (t = -0.374, P = 0.71) and duration of medication use (t = -1.361, P = 0.18), as detailed in Table 1. However, a marked distinction was noted in age (t = -3.081, P < 0.01) and in total illness duration (t = -5.440, P < 0.01).
Table 1Demographic and clinical characteristics of all participantsVariableFED group (n = 41)RMD group (n = 43)StatisticP valueAge32.37 ± 12.3640.56 ± 11.98t = -3.081< 0.01Sex (female/male)24/1729/14X^2^ = 0.7150.40Years of education12.67 ± 3.9111.51 ± 3.86t = 1.3690.18BMI (kg/m^2^)22.21 ± 3.4423.13 ± 4.86t = -1.0020.32HAMD34.63 ± 7.0133.78 ± 6.88t = 0.5590.58HAMA21.60 ± 6.5419.12 ± 6.85t = 1.6990.09Age at onset (years)31.35 ± 12.1431.63 ± 10.84t =-0.1130.91Total illness duration (years)1.26 ± 1.6911.25 ± 11.92t = -5.440< 0.01Current Episode Duration (years)0.61 ± 0.650.46 ± 0.89t = 0.8920.38Antidepressant medicationsSSRIs1422 -
- SNRIs125 -
- NaSSA31 -
- Medication-free1215 -
- Medication dosages (mg)21.07 ± 21.0022.73 ± 19.80t =-0.3740.71Duration of Medication Use (days)55.14 ± 124.60111.53 ± 239.11t =-1.3610.18Employment statusEmployed2238 -
- Unemployed92 - -Students103 - -Marital statusMarried2126 - -Divorced12 - -Data are presented as means ± standard deviations. Medication dosages were calculated in fluoxetine equivalents. Patients who had discontinued medication for more than 2 weeks were considered to have undergone a washout period and were treated as medication-free. Abbreviations: FED, First-Episode Depression; RMD, Recurrent Major Depression; BMI, Body Mass Index; HAMD, Hamilton Rating Scale for Depression; HAMA, Hamilton Rating Scale for Anxiety; SSRIs, selective serotonin reuptake inhibitors; SNRIs, serotonin norepinephrine reuptake inhibitors; NaSSA, noradrenergic and specific serotonergic antidepressant
As presented in Table 2, RMD patients presented poorer cognitive functions than FED patients. Compared with FED patients, those with RMD had significantly lower scores in EBPM (t = 2.248, P = 0.03, Cohen’s d = 0.49), CPT-IP-2 (t = 3.205, P < 0.01, Cohen’s d = 0.71), and SFT (t = 2.251, P = 0.03, Cohen’s d = 0.49) scores. No significant differences were observed between the two groups in TBPM and digit span scores. Moreover, RMD patients exhibited longer P300 latency (t = -2.105, P = 0.03, Cohen’s d = -0.46), while no significant differences were observed in P300 amplitude. After applying Bonferroni correction, only the difference in CPT-IP-2 (P = 0.02) between the two groups remained significant.
Table 2Comparison of FED and RD in cognitive variablesVariableFED group (n = 41)RMD group (n = 43)StatisticP valueP-adj*EBPM5.23 ± 2.663.93 ± 2.662.2480.030.27TBPM4.65 ± 2.014.17 ± 2.111.0690.291CPT-IP-22.57 ± 0.662.01 ± 0.943.205< 0.010.02CPT-IP-31.73 ± 0.761.65 ± 0.750.4910.631CPT-IP-41.22 ± 0.571.05 ± 0.551.3170.191Digit span forward8.05 ± 1.177.85 ± 1.320.7100.481Digit span backward5.51 ± 1.524.98 ± 1.311.7330.090.87SFT11.48 ± 2.4210.27 ± 2.492.2510.030.27P300 latency344.10 ± 20.08354.08 ± 18.97-2.1050.030.38P300(ms)Amplitudes(uv)8.30 ± 5.846.68 ± 4.141.4660.151Data are presented as means ± standard deviations. *P-values were adjusted using the Bonferroni correction by multiplying each value by the total number of comparisons; adjusted P-values greater than 1 were uniformly reported as 1. Abbreviations: FED, First-Episode Depression; RMD, Recurrent Major Depression; EBPM, Event-Based Prospective Memory; TBPM, Time-Based Prospective Memory; CPT-IP, Continuous Performance Task-Identical Pairs; SFT, Semantic Fluency Test
In contrast to individuals diagnosed with FED, patients experiencing RMD showed a pronounced elevation of ReHo in the right inferior temporal gyrus (R-ITG). This region comprised a 66-voxel cluster at MNI coordinates (x = 45, y = 9, z = -45), with a peak t-value of 7.006 (GRF voxel-level P = 0.001, cluster-level P < 0.05; see Fig. 1A). Meanwhile, VMHC measurements were substantially lower in both left and right inferior temporal gyrus. Specifically, the left side included a 52-voxel cluster at peak MNI (x = -57, y = -51, z = -12) with a peak t-value of -6.421, and the right side featured an equivalent 52-voxel cluster at peak MNI (x = 57, y=-51, z=-12) showing a peak t-value of -6.421(GRF voxel-level P = 0.001, cluster-level P < 0.05; see Fig. 1B).
Fig. 1Brain regions showing ReHo (A) and VMHC (B) differences between the FED and RMD groups (p < 0.01). (A) Compared with first-episode depression (FED), recurrent major depression (RMD) demonstrated increased regional homogeneity (ReHo) in the right inferior temporal gyrus (R-ITG). The warm-coloured clusters denote regions where ReHo is significantly elevated in the RMD group. The corresponding t-values are indicated by the colour bar, with brighter shades reflecting higher t-values. The box plot illustrates mean ReHo values; error bars represent the standard error of the mean (SEM). (B) Compared with FED, the RMD group exhibited reduced voxel-mirrored homotopic connectivity (VMHC) in the bilateral inferior temporal gyri (ITG). The cool-coloured clusters represent areas of decreased VMHC. Brighter shades in the colour bar denote more pronounced t-values. The box plot depicts mean VMHC values (SEM error bars)
Correlation analyses demonstrated that EBPM (r = -0.254, P = 0.02, 95% CI = -0.430, -0.074) (Fig. 2A) was inversely related to ReHo in the R-ITG. Additionally, VMHC within the bilateral ITG was significantly associated with SFT (r = 0.255, P = 0.02, 95% CI = 0.034, 0.465) (Fig. 2B), P300 amplitudes (r = 0.263, P = 0.02, 95% CI = 0.019, 0.457) (Fig. 2C), and CPT-IP-2 scores (r = 0.223, P = 0.046, 95% CI = 0.001, 0.435) (Fig. 2D). However, after performing Bonferroni correction, none of these correlations remained statistically significant.
Fig. 2Correlations between brain functional changes and cognitive variables. Correlation analyses revealed that ReHo in the R-ITG was negatively correlated with EBPM (r = -0.254, P = 0.02, 95% CI = -0.430, -0.074) (A). Meanwhile, VMHC within the bilateral ITG was significantly and positively associated with SFT performance (r = 0.255, P = 0.02, 95% CI = 0.034, 0.465) (B), CPT-IP-2 scores (r = 0.223, P = 0.046, 95% CI = 0.001, 0.435) (C), and P300 amplitude (r = 0.263, P = 0.02, 95% CI = 0.019, 0.457) (D). Abbreviations: ReHo, Regional Homogeneity; VMHC, Voxel-Mirrored Homotopic Connectivity; R-ITG, Right Inferior Temporal Gyrus; EBPM, Event-Based Prospective Memory; TBPM, Time-Based Prospective Memory; CPT-IP, Continuous Performance Task-Identical Pairs; SFT, Semantic Fluency Test
To evaluate the robustness of these results, we conducted three additional sensitivity analyses, controlling for potential confounders and examining a subgroup matched on age. In all cases, the primary findings concerning ReHo increases in the right ITG and VMHC decreases in the bilateral ITG in RMD versus FED generally remained consistent. A detailed description of each analysis, including the specific models, covariates, and the correlation results, is provided in the Supplementary Materials (see Supplementary Figures S1–S3).
Our research demonstrated that individuals with RMD exhibited more pronounced impairments in cognitive abilities, specifically in areas such as attention, prospective memory, and verbal fluency, in addition to experiencing longer P300 latency. In contrast to those experiencing their FED, the RMD cohort displayed heightened ReHo within the right ITG and diminished VMHC across bilateral ITG. Moreover, these cognitive shortcomings in MDD participants were correlated with functional alterations in the brain. The associations between EBPM and ReHo, as well as between SFT and VMHC, were notably stronger. These findings support the hypothesis that multiple depressive episodes contribute to progressive declines in cognitive and neural functions. Integrating data from fMRI and ERP, the results suggest that while FED individuals may preserve certain aspects of cognitive control and neural flexibility, those with RMD suffer significant cognitive deterioration, potentially due to the accumulated effects of recurrent depressive episodes.
Numerous studies have demonstrated that individuals with depression often exhibit impairments in sustained attention and prospective memory. Recent review studies have shown that MDD patients perform worse on tasks requiring sustained attention, such as list learning and free recall tasks [3]. Studies on adolescent depression have found that performance on the CPT-IP-2 and CPT-IP-4 tasks is significantly worse than in healthy control groups [55]. Our study showed that RMD patients scored lower on the CPT-IP, with significant results for CPT-IP-2, which aligns with prior research findings. Numerous studies have demonstrated significant impairments in attentional functions among MDD patients [56, 57]. A recent meta-analysis of 23 studies also shows that, compared with healthy individuals, MDD patients exhibit moderate deficits in overall attention performance [58]. In this study, we observed that patients with RMD scored lower specifically on the CPT-IP-2 subtest. We speculate that this may be due to the fact that basic sustained attention or “low-load rapid judgement” abilities are more susceptible to impairment, whereas more complex tasks (e.g., those involving three- or four-digit sequences) rely on multiple strategies (such as reviewing, comparing, or pausing to think). As a result, differences between FED and RMD may be masked in these more complex tasks. It is also possible that at higher levels of difficulty, all participants encountered greater challenges in recognition and vigilance, making their overall performance relatively similar. By contrast, the two-digit condition may have avoided the “floor effect” triggered by excessive difficulty. This interpretation is further supported by the fact that average scores on the CPT-IP-3 and CPT-IP-4 subtests were even lower in the RMD group.
Prospective memory is the ability to execute previously planned tasks at a specific future time or when a particular situation arises [59]. Its successful execution requires the active integration of attention monitoring, recognition of contextual cues, timely memory retrieval, and action initiation. Such impairments are frequently observed in individuals with MDD and are strongly linked to deficits in working memory and executive function [60, 61]. Moreover, these deficits are not limited to depression; they can also manifest in conditions such as schizophrenia and bipolar disorder, thus representing a potential transdiagnostic ‘intermediate phenotype’ warranting more targeted cross-disorder comparisons [62]. Depressive mood may weaken prospective memory capacity [63], a finding consistent with the “emotion–cognition interaction” theory, whereby depressive affect diminishes one’s ability to actively regulate attentional resources during tasks, ultimately hindering memory performance [64]. EBPM relies more on specific environmental cues to trigger memory and action, while TBPM depends on internal time-monitoring mechanisms [65].
Individuals with RMD exhibited more pronounced impairments in EBPM, whereas no significant distinctions in TBPM were observed between those experiencing their first episode and those with recurrent episodes. This may suggest that following frequent or repeated depressive episodes, individuals with RMD experience a further weakening in their ability to monitor target cues in the external environment, making them more prone to errors or omissions when executing future plans. Due to the lack of a salient triggering signal in TBPM tasks, individuals must allocate considerably more cognitive resources to actively monitor time and self-initiate intentions, thereby increasing task difficulty [66]. Conversely, EBPM relies on the automatic triggering of responses by environmental events, which requires relatively less involvement of the prefrontal cortex [67, 68]. Previous investigations have underscored the pivotal involvement of the Brodmann area 10, in supporting prospective memory functions. This region is situated adjacent to and maintains connections with the temporal lobe, highlighting its integral role in cognitive processes related to memory and time management [69, 70]. Overall, our findings are largely consistent with previous research on executive functions in depression and provide unique insight into the cumulative effect of cognitive impairments across depression episodes.
The temporal lobe plays a pivotal role in visual perception and semantic processing [71]. In line with this function, our study revealed that patients with RMD exhibited lower scores on the semantic fluency test. In a longitudinal study of over 1,000 individuals, researchers found that those with higher levels of loneliness scored lower on verbal fluency and backward digit span cognitive functions, demonstrating the impact of long-term mood disruption on cognitive decline [72]. A small-sample study that strictly matched for age and education levels revealed that depression patients exhibited decreases in attention shifting and verbal fluency, but did not differentiate between first-episode and recurrent patients [73]. A recent review of depression research reported that structural abnormalities in the superior and inferior temporal gyri may be linked to childhood trauma among patients with MDD [74]. Concurrently, decreased grey matter volume (GMV) in the superior and middle temporal gyri has been consistently reported in depressed individuals [75, 76]. In a study involving 44 MDD patients and 44 healthy control participants matched for age, sex, and educational level, those in their first depressive episode exhibited significantly reduced GMV in the right inferior temporal gyrus [77]. These findings highlight the distinctive role of the temporal lobe in depressive episodes. White matter imaging studies also suggest that the ITG may be linked to the frontal lobe via the arcuate and uncinate fasciculi, implying its pivotal position in emotion regulation, language, and executive functioning [78]. Given that the ITG is indispensable for advanced visual processing, shape or letter recognition, and semantic interpretation, a decline in interhemispheric coordination is likely to compromise more complex cognitive tasks, including language and attention. Our findings show that patients with RMD experience diminished synchronous activity in the bilateral ITG, corresponding with reduced verbal fluency—supporting the notion that ITG dysfunction is tightly interwoven with language impairment in MDD patients. A separate study also revealed a significant positive correlation between semantic fluency test scores and glucose metabolism in the temporal cortex [79], reinforcing, from a metabolic standpoint, the ITG’s vital role in language. Furthermore, a quantitative synaptic density study of 27 patients with differing cognitive statuses reported a robust relationship between synapse density in the ITG and verbal fluency [80], suggesting that techniques such as diffusion tensor imaging (DTI) may help clarify the anatomical basis of this dysfunction in future research.
RMD patients exhibited increased ReHo in the right ITG and decreased VMHC in both hemispheres of the ITG compared to those with FED. Emerging evidence indicates that disrupted interhemispheric information integration is a key feature of the pathophysiology of depression [81]. Additionally, increased VMHC in the middle cingulate cortex has been correlated with depression severity, suggesting a potential compensatory mechanism in the disease process. Furthermore, altered interhemispheric functional connectivity in the amygdala may influence emotional processing and cognitive function, reinforcing the critical role of disrupted interhemispheric communication in the pathophysiology of depression [82]. Our study further demonstrated that RMD patients exhibited reduced synchronous activity in the bilateral ITG, which was associated with diminished verbal fluency. This suggests that dysfunctions within the ITG are intricately linked to the language impairments observed in individuals with depression. Attention, which is essential for selective, sustained, and categorical cognitive processes, was also adversely affected in RMD patients, as indicated by the positive correlations between VMHC and CPT-IP-2 scores. Prior studies have shown that older MDD patients exhibit significantly reduced VMHC in multiple brain regions, such as the superior frontal gyrus, posterior cerebellar lobe and precentral gyrus, which correlates with impaired cognitive flexibility [83]. However, that research did not focus on patients with RMD. Notably, heightened ReHo in the right superior temporal gyrus also demonstrates bilateral functional asymmetry and is markedly associated with a decline in prospective memory. In MDD patients, the right hemisphere often demonstrates overactivity, whereas the left hemisphere shows relatively attenuated activity [84], especially in extensive right-hemisphere areas like the temporal pole, parietal cortex, and thalamus [85]. This suggests that abnormal activation in the right ITG may be intimately linked with language disturbances in depressed individuals, and that reduced bilateral ITG synchronisation may further exacerbate these impairments. We did not observe corresponding functional alterations in the left ITG, indicating that, as depression becomes more chronic or recurrent, disturbances in local functional coordination of the right ITG—leading to a pronounced desynchronisation between the hemispheres—could represent a key neural mechanism driving progressive cognitive decline.
In addition, we found that patients with RMD exhibited prolonged P300 latency, which is generally regarded as an indicator of slowed attentional resource allocation [86, 87]. This aligns with our conclusion about insufficient attentional resource distribution in the CPT-IP-2 task, suggesting that after multiple episodes, patients experience marked delays in rapid alertness and information-processing efficiency. Overall, our results concur with previous research on executive dysfunction in depression, while also revealing the unique pattern of cumulative cognitive impairment associated with recurrent depressive episodes. Furthermore, the positive correlation between VMHC and P300 amplitude implies that reduced functional symmetry in the bilateral ITG may influence P300 responses. Consistent with our findings, prior studies have identified significantly lower VMHC in several brain regions of MDD [83].
Our study’s strength lies in the use of multiple cognitive indicators—including sustained attention, prospective memory, short-term memory, verbal function, and ERP—combined with objective fMRI, allowing for a comprehensive comparison of cognitive and brain functional differences between RMD and FED patients. To ensure that our conclusions were not driven by confounding variables, we performed three separate sensitivity analyses (Supplementary Materials, Figures S1–S3). These analyses included additional covariates such as medication use and duration of the current episode, as well as an age-matched subgroup comparison. Across all models, the primary differences in ReHo and VMHC between RMD and FED remained statistically significant, supporting the robustness of our results. Nonetheless, our study has several limitations. Firstly, as a cross-sectional investigation, it cannot establish a causal relationship. Secondly, nearly all patients were receiving antidepressant treatment, so we cannot rule out the potential influence of these medications on our findings. Thirdly, despite our efforts to control for major confounders such as age, sex, and educational background, individual differences in baseline cognitive function remain possible, especially given that recurrent patients tend to be older. Finally, after applying Bonferroni correction for multiple comparisons, only the group difference in CPT-IP-2 remained significant, and none of the correlation results survived this stringent threshold. This underscores the exploratory nature of our findings and highlights the need for replication with larger samples. Nonetheless, our consistent patterns across sensitivity analyses suggest that the observed trends are unlikely to be mere artifacts of statistical noise. However, the consistent findings across the sensitivity analyses suggest that the results are robust and unlikely to be driven solely by statistical noise.
Overall, our results indicate that FED and RMD patients exhibit distinct cognitive and brain functional changes, underscoring the need for future research to treat recurrent and first-episode populations as distinct cohorts. Multiple factors contribute to cognitive deficits in depression, such as lipid metabolism [88], inflammation [89], and the gut microbiome [90], with each patient potentially exhibiting different cognitive impairments. Therefore, future research should focus on designing intervention experiments to identify which factors play key roles in cognitive impairment in depression patients with continuing episodes.
Our findings demonstrated that RMD patients exhibit more pronounced cognitive impairments than FED patients, accompanied by distinctive alterations in the ITG. Notably, increased ReHo in the right ITG and decreased VMHC in the bilateral ITG were closely linked to deficits in attention, prospective memory, and verbal fluency. These results highlight the ITG as a potential neural substrate or stimulation target for novel therapeutic strategies. In particular, neuromodulation techniques—such as repetitive transcranial magnetic stimulation—tailored to this temporal region could hold promise for mitigating or preventing further cognitive deterioration in recurrent depression. Future research should concentrate on longitudinal, large-scale intervention studies to validate the efficacy of targeting the ITG, aiming to optimize treatment outcomes and curb relapse-related cognitive decline in MDD.
Below is the link to the electronic supplementary material.
Supplementary Material 1