Our study specifically investigated the therapeutic effects of index ECT on major depression, focusing on the temporal changes in symptom dimensions. As with previous studies, there was a rapid improvement of depressive symptoms in the early course of the index ECT course (by about the third session) . Our main findings are as follows: First, the depression dimension had a longer duration of significant improvement than the other dimensions. Second, it took at least six ECT sessions before a significant difference in the amount of improvement between the remitter and non-remitter groups began to appear and continued, only in the depression and anxiety dimensions. Third, the depression dimension continued to show a greater amount of reduction of the T-score from baseline than the other three dimensions through to the twelfth session of ECT.
We used the four-factor classification (depression, anxiety, insomnia, and somatic symptoms) of HDRS-17 suggested by Shafer to identify and compare the trajectories of depressive symptoms throughout the index ECT course  more clearly. Our four-factor structures are consistent with those of previous studies. In the meta-analysis of HDRS factors by Bagby et al , 'general depression factor' including 'depressive mood', 'guilty', and 'suicide' items were identified, which were very similar to the depression dimension in our study. Our study categorized early, middle, and delayed insomnia into the insomnia dimension, which was replicated in other studies . The 'anxiety/agitation factor' composed of agitation, psychic anxiety, and somatic anxiety was also used in a number of studies, and our study regarded this factor as the anxiety dimension . Lastly, our study selected gastrointestinal somatic symptoms, general somatic symptoms, genital symptoms, and weight loss as somatic symptoms dimension to examine the trajectory of somatic symptoms during the index ECT course. This grouping of symptoms is consistent with the meta-analysis by Bagby et al., in which gastrointestinal somatic symptoms, general somatic symptoms, and genital symptoms were the same factor in two studies, and gastrointestinal symptoms and weight loss were the same factor in 11 out of 15 studies . By categorizing the four depressive symptom dimensions, the patterns of change between symptom dimensions in the index ECT could be carefully explored.
In the analysis of the period of significant improvement, the reduction of the HDRS-17 total score was more evident until the sixth ECT session, and the depression dimension continued to ameliorate significantly longer than the other three dimensions. The rapid decrease in the HDRS-17 total score in the early phases of index ECT (about third to sixth session) is consistent with previous studies . Indeed, there have been numerous studies examining changes in the total scores of depression rating scales (e.g., HDRS and MADRS) through ECT, and some studies have recently investigated the trajectories of specific depressive symptoms in the index ECT period [17–19]. However, to date, very few studies have identified the period during which a significant reduction remains in each dimension of depressive symptoms. The T-score of the depression dimension started to decline significantly from the third session, and the improvement lasted longer than the other three dimensions, until the ninth session of the ECT. Even in the non-remitter group, the decrease in the depression dimension persisted longer than the other dimensions until the sixth session.
In addition, we have provided the results of when the difference in degrees of the T-score changes between the remitter and non-remitter groups becomes significant. Only depression and anxiety dimensions showed a significant difference between the two groups from the sixth to ninth sessions of the index ECT, and this difference persisted until the twelfth session of ECT. The HDRS-17 total score also revealed a significant difference between the two groups after the sixth ECT session. Taken together, significant differences between the remitter and non-remitter groups were observed at least from the sixth session of the index ECT course. In addition, our finding that all four dimensions showed significant improvement after at least six sessions of ECT (ninth session for depression dimension) coincides with the previous demonstration that early termination of ECT could increase the likelihood of relapse [40, 41]. Considering the guidelines that have suggested six to twelve treatments in the typical index ECT course, at least six ECT sessions may be required for significant therapeutic benefit [42, 43].
Meanwhile, the insomnia dimension showed no significant difference in change between the remitter and non-remitter groups at times 2, 3, 4, and 5. McCall et al  suggested that ECT might have a modest effect on insomnia. Although an improvement in insomnia was observed in both remitters and non-remitters in our study, no significant difference in the amount of change was observed between the two groups. According to McCall's findings, sleep improvement might also have occurred due to effects other than ECT (e.g., use of hypnotics), but ECT itself might have had less of an effect on insomnia. Therefore, managing residual symptoms, such as insomnia, that may remain after the middle phases of index ECT course (sixth to ninth session) could help improve the overall treatment effect.
Our finding that the core depressive symptoms ('depression' dimension), including depressed mood and loss of interest, have a greater degree of improvement than other dimensions has been repeated in previous studies. Huang et al  used the GEE method to analyze longitudinal data, as in our study, and reported that the improvement of this core factor subscale was consistently greater than that of the anxiety/somatization subscale of the HDRS-17 during the index ECT course. Wade et al  also suggested that the degree of improvement of core mood/anhedonia factor was larger than that of insomnia and somatic disturbance factors in remitters. In addition, Veltman et al  demonstrated that the mood dimension improved faster than the melancholic and suicidality dimensions in the early course of ECT (after one week). Our study repeatedly assessed the HDRS-17, the most used scale, every three ECT sessions, and compared depression, anxiety, insomnia, and somatic symptom dimensions.
Based on these results, understanding the disproportionate effects of ECT on core mood symptom dimension compared with other dimensions plays an important role in expanding our knowledge of the therapeutic mechanisms of ECT in depression. First, at the neurotransmitter level, depressive mood, loss of interest, and psychomotor retardation, which belong to the 'depression' dimension of our study, might be caused by dopamine deficiency or reduced reactivity of the post-synaptic dopamine receptor . It is known that the changes in D1 and D2 receptor function in the early phases of index ECT, especially the downregulation of D2 receptors, can lead to symptom improvement. These mechanisms may affect the amelioration of core depressive symptoms in the early stages of ECT [46, 47]. In addition to dopamine, serotonin-related effects induced by ECT can affect mood symptoms and suicidality. The 5-HT2A receptor is thought to be a major receptor involved in the pathophysiology of depression. There have been several studies which have suggested that upregulation of this receptor is linked to suicide, reduction of intracellular signaling pathways, and melancholic depression [48–50]. Thus, we considered the possibility that the 5-HT2A receptor and emotion-related symptoms might be correlated. Based on the results that ECT leads to downregulation of 5-HT2A receptors in humans, we suggest that the downregulation of 5-HT2A receptors by ECT may have affected suicidality and depressed mood [47, 51].
Second, we considered the association between the brain structural changes induced by ECT and depression dimensions. Referring to the meta-analysis of Wilkinson et al , which suggested that ECT increases the volume of the hippocampus, and the study of Zhu et al , which studied the association between the hippocampus and emotional processing, we considered that the volume increase of the hippocampus induced by ECT can contribute to the improvement of core mood symptoms. However, there is a multi-site study in which there was no positive correlation between hippocampal volume changes and clinical outcomes. Therefore, the correlation between ECT-induced hippocampal volume changes and core mood symptoms requires further studies .
A change in the activity of the subgenual anterior cingulate cortex (sgACC) could be another hypothesis for the disproportionate reduction mechanism of depressive symptoms following ECT. The sgACC, a part of the anterior cingulate cortex, plays a major role in unpleasant emotion regulation. The relationship between sgACC and depression has been a prominent topic in mood disorder and neuromodulation studies [55, 56]. Although the effect of ECT on sgACC is estimated to influence the reduction of the depression dimension, the direction of this effect is inconsistent. Liu et al  reported increased regional activity of the sgACC and connectivity of the sgACC with other brain regions (i.e., the parahippocampal gyrus, middle temporal gyrus, and orbitofrontal codex) with eight ECT sessions. In contrast, Hamani et al  suggested increased sgACC activity in depressed individuals which was reversed by antidepressant therapy. Taken together, although the effect of ECT on sgACC has not been clearly identified, it could be a possible mechanism for the specific improvement in core mood symptoms by ECT.
Our study has several limitations. First, as a retrospective study, it is possible that medication use might have functioned as a confounding variable because patients were given psychopharmacologic treatment during the index ECT process. However, there are many combinations of ECT and pharmacotherapy in contemporary clinical settings, and it is not easy to discontinue all medications abruptly . Indeed, this was similar to other studies in that the comparison of remitters and non-remitters was performed under the same condition where all patients received medication. This study, using concomitant pharmacotherapy, may be used to determine the specific effect of index ECT on various depressive symptoms in actual clinical practice. The second limitation of this study was that some patients had missing data. For the analysis including missing data, we used the GEE method, which maintains relatively high power in longitudinal measurements with missing data and has the advantage of identifying overall average effects. Third, our study did not assess anxiety, insomnia, and somatic symptoms with additional validated scales (e.g., Pittsburgh Sleep Quality Index, Beck Anxiety Inventory, and Depression and Somatic Symptoms Scale) in addition to the HDRS-17. However, the HDRS is the most widely used assessment measure in depression studies. In this study, standardization was performed with a T-score for comparison between each depressive symptom dimension. In future, it will be necessary to apply an assessment scale that can more precisely evaluate each symptom.